Sediment Erosion in Hydro Turbines 1

Sediment Erosion in Hydro

Turbines
Hari Prasad Neopane
This thesis is submitted to

Faculty of Engineering Science and Technology,
Norwegian University of Science and Technology (NTNU)
in partial fulfilment of the requirement of the degree of Doctor of Philosophy
Trondheim
Norway
March 2010
Summary
SUMMARY
Sediment erosion is caused by the dynamic action of sediment flowing along with water
impacting against a solid surface. Hydraulic turbine components operating in sediment‐
laden water are subject to abrasive and erosive wear. This wear not only reduces the
efficiency and the life of the turbine but also causes problems in operation and
maintenance, which ultimately leads to economic losses. This is a global operation and
maintenance problem of hydropower plants. The high sediment concentration combined
with high percentage of quartz content in water causes severe damage to hydraulic
turbine components. Withdrawal of clean water from the river for power production is
expensive due to design, construction and operation of sediment settling basins. Even
with the settling basins, 100 % removal of fine sediments is almost impossible and
uneconomical.
A number of factors can influence the process of sediment erosion damage in hydro
turbine components. The erosion intensity depends on the sediment type and its
characteristics (shape, size, hardness, concentration etc.), hydraulic design and operating
conditions of turbine (flow rate, head, rotational speed, velocity, acceleration, turbulence,
impingement angle etc.), and material used for the turbine components. All these factors
are needed to be considered for predicting the erosion. Therefore, dealing with sediment
erosion problems requires a multidisciplinary approach. More research and development
is needed to investigate the relationship between the particle movement and erosion
inside a turbine and to establish the operating strategy for the turbine operating in
sediment‐laden water.
In order to achieve the main objective of this PhD study, the overall research
methodology adopted for this work ‘sediment erosion in hydro turbines’ include;
experimental studies, numerical simulation, and field studies. This research work is
based on result from laboratory experiment, and numerical simulation.
A previously made test rig (Thapa, 2004), was reviewed and modified to create a strong
swirl flow in curved path. This flow was found similar to the flow between the guide
vane outlet and the runner inlet of a Francis turbine. The flow in the guide vane cascade
was simulated in order to verify the particle separation process and to investigate the
relation of the velocity and the drag coefficient with different shape and size of the
particle. There was a provision to introduce particles, with sizes ranging from 1 to 10
mm, and to observe the motion of the particles from Plexiglas windows located on the
cover of the tank using a high‐speed digital camera. When a particle is flowing in swirl
flow, drag force and centrifugal force are two major forces influencing the particle
equilibrium. The equilibrium of these two forces provides a critical diameter of the
particle. While, a particle larger than the critical diameter move away from the centre and
hit the wall, a particle smaller than the critical diameter flows along with the water, and
ultimately sinks. For critical diameter, the particle continues to rotate in the turbine.
Different shapes and sizes of particles were tested with the same operating conditions
i HPN PhD Thesis, NTNU

Summary
and found that triangularly shaped particles were more likely to hit the suction side of
the guide vane cascade. Furthermore, this study supports the concept of separation of
particles from streamlines inside the test rig, which led to the development of an
operating strategy for a Francis turbine processing sediment‐laden water. This study also
permitted experimental verification of the size and the shape of a particle as it orbits in
the turbine, until either the velocity components are changed or the particle became
smaller.
The steady state numerical simulations were carried out on the Cahua power plant
Francis turbine design, mainly at two operating conditions with varying particle size,
shape, and concentration using ANSYS CFX. The predictions of erosion, based on the
Lagrangian calculation of particle paths in a viscous flow, are described for stay vanes,
guide vanes, and runner vanes of a Francis turbine, for which the results of the field tests
have been available for verification. The flow simulation was obtained through use of a
commercially available computational fluids dynamics (CFD) code, namely ANSYS CFX.
The code utilizes a finite‐volume, multi‐block approach to solve the governing equations
of fluid motion numerically on a user‐defined computational grid. The flow solution
procedure first generates the computational grid. A pre‐processor is available in the
software to perform this task. Second, the solution option such as inlet and boundary
conditions, turbulence model, and discretization scheme, are specified. The final step is
running the flow solver to generate the actual flow field simulation.
Sediment erosion analysis of a Francis turbine gives an indication of relative erosion
intensity and critical zones of erosion damage of the turbine components. The most
realistic numerical prediction of erosion is found on a turbine blade. The highest
velocities and accelerations occurred at outlet of the runner blade and more erosion was
predicted especially at the pressure side of the blade outlet and at the lower cover.
Furthermore, unexpected sediment erosion was found at the suction side of the guide
vane where concept of critical diameter can be utilized. It has been concluded that if the
particle size in the water is more than critical particle sizes, the turbine should not be
operated at low guide vane opening.
The numerically obtained erosion pattern and the field test observation and inspection at
Cahua Francis turbine components are in good qualitative agreement. The encouraging
agreement shows that, for this application, numerical simulation really can be used in a
predictive manner. This information may serve as an input in an early stage of turbine
design process to identify the regions where special surface treatment is necessary in
order to increase the lifetime of the components for new hydropower projects involving
risks of sediment erosion.
The size of a particle is inversely proportional to the velocity of the particle, and it was
determined that spherically shaped particles had higher settling velocities than particles
with other shapes. However, non‐spherical shape of the particles will tend to have lower
settling velocities because both decreases in spheroid and increases in angularity tend to
decrease velocities. Moreover, larger cross‐sectional areas tend to be directed
perpendicular to the transport path. As a result, higher coefficient of drag, higher
HPN PhD Thesis, NTNU ii

Summary
rotational motion and more separation of flow are likely to occur and hence more erosion
rate was predicted. The roles played by the shape of the particle significantly affect
erosion rate prediction inside the Francis turbine components.
Furthermore, it has been found that the erosion process is strongly dependent on the
particle size, shape, concentration, and operating conditions of the turbine. The reduction
of the erosion is not only linked to the reduction of particle velocity but also is linked to
the reduction of separation of flow, which further depends on shape, size, and
concentration of the particle. The significant reduction of erosion rate can be achieved by
operating turbine at best efficiency point. The full load operation reduced efficiency,
increased turbulence, and increased relative velocity of flow at outlet of the blades.
The present knowledge and findings, although may not be enough to deal with this
problem completely, can be utilised to achieve one major step forward in sediment
erosion prediction and prevention.
iii HPN PhD Thesis, NTNU

Preface and Acknowledgements
PREFACE AND ACKNOWLEDGEMENTS
The research presented in this thesis was mainly carried out at Waterpower Laboratory,
Department of Energy and Process Engineering, Norwegian University of Science and
Technology (NTNU), Norway, during the period from February 2007 to March 2010.
Professor Ole Gunnar Dahlhaug has been my main supervisor. His theoretical and
practical knowledge in the field of hydropower has been huge source of help and
inspiration during the course of study. He inspired me to start my academic and
professional carrier in the field of hydropower in general, and sediment erosion in
particular. I am indebted to him, who has not only been an advisor to me but also a
guardian putting tireless efforts to guide, arrange financial support, field work, and
ensuring a conducive environment for my study at Waterpower Laboratory.
Furthermore, I would like to thank him for making this study possible, for valuable
discussions, and for guiding me in the right direction at the time when I was lost.
I would like to acknowledge sincerely to Professor Hermod Brekke, Professor Arne
Kjølle, Professor Torbjørn Nielsen, and Associate Professor Morten Kjeldsen, for their
valuable suggestions, support, and cooperation during my study.
I would like to thank staffs of the Waterpower Laboratory Bård Brandåstrø, Joar Grilstad,
Ellef Bakken, Trygve Opland, Halvor Haukvik, Idun Schjerven Østgård, for their help to
find out and set up earlier test rig at the laboratory, to modify the test rig and finally to
conduct the experiments. Similarly, I would like to thank Wenche Johansen for her help
not only to provide all the necessary logistic support to complete my work but also to
keep track of all my deadlines and others administrative challenges. In addition, her
cooperation and friendly behaviour created a homely environment at Waterpower
Laboratory. I would also like to thank Eugen Uthaug for helping and providing the
relevant computer software throughout my study.
My sincere thanks go to the Kathmandu University and then the Department of
Mechanical Engineering for allowing me to pursue this study at NTNU. I would like to
thank especially Dr. Bhola Thapa, Professor and dean, school of engineering, for his help,
cooperation and guidance throughout the study. He is the one who motivated me to
continue the research in this field. I have used his PhD experimental set up at the
beginning of my experimental work and then modified the same set up for the rest of my
experiment. It has been a pleasure working with this experimental set up at Waterpower
Laboratory. Moreover, he has always been available with me for discussion and guidance
to accomplish the objective of this study.
In addition, I would like to thank Dr. Bhupendra Bimal Chhetri, Associate Professor, at
Department of Electrical and Electronics Engineering, KU, for his help to correct
language mistakes during final phase of my writing. It has been pleasure to have his
comments on right time.
HPN PhD Thesis, NTNU iv

Preface and Acknowledgements
The discussions and interactions with the colleagues at the Waterpower Laboratory have
been very fruitful. Many thanks go to PhD fellows, Pål‐Tore Selbo Storli, Jørgen Ramdal,
Einar Kobro, Håkon Hjort Francke, and Lars Frøyd. I would also like to thank, Mette
Eltvik, former master degree student and present research assistant at the Waterpower
Laboratory, for her help especially in Turbo Grid generation and CFD simulation. It has
been always pleasure to discuss with her regarding sediment erosion issues and their
challenge.
I wish to express my heartfelt gratitude to the State Education Loan Fund (Lånekassen),
Norway, for partly supporting my living expenses during my stay at Trondheim. I would
like to thank Anette Moen, Coordinator of Quota Programme, Ragnhild Brakstad, Turid
Bræk, and Gro Johnsen, staffs at the office of international relations, NTNU for their
administrative arrangements for loan fund and residence permit issues for me and my
family. I am equally indebted to Energy Norway (former EBL) for their additional
financial support during my study. I would like to thank and acknowledge all Nepalese
colleagues those who were here during my study period for supporting me, and for my
family.
I declare that this study is my own work and outside inputs are referenced at the relevant
places. I am sure that there will be some errors, for which I take absolute responsibility.
This is a small step towards achieving sediment erosion prediction and prevention in
hydraulic turbine components in hydropower plants. I do believe that this work will
encourage many researchers to put their knowledge and efforts in order to achieve the
significant contribution against sediment erosion problem.
I wish to thank my wife Radhika and two lovely sons Sushant and Hardik, whose
presence, patience, smiles and moral support always encouraged me to complete my
work on time. I would like to thank my brother Tulsi for his cooperation and continued
support towards my home responsibilities during my absence in Nepal.
Last but not the least, I dedicate this work to my loving parents, Aama and Baba. I am
always grateful to them and now I am here in this position because of their hard work,
continued support, encouragement and blessing.

Hari Prasad Neopane

PhD Candidate
Waterpower Laboratory
Department of Energy and Process Engineering (EPT)
Faculty of Engineering Science and Technology (IVT)
Norwegian University of Science and Technology(NTNU)
Trondheim
Norway
March, 2010

v HPN PhD Thesis, NTNU

Table of Contents
TABLE OF CONTENTS
Summary......................................................................................................................................i
Preface and Acknowledgements................................................................................................iv
Table of contents .......................................................................................................................vi
List of figures..............................................................................................................................ix
List of tables................................................................................................................................x
Nomenclature.............................................................................................................................xi
CHAPTER 1 INTRODUCTION .......................................................................................... 1-1
1.1 Background.................................................................................................................... 1-1
1.1.1 General............................................................................................................... 1-1
1.1.2 Sediment erosion problems ................................................................................ 1-3
1.2 Motivation for this research ............................................................................................ 1-4
1.3 Objective of this study .................................................................................................... 1-6
1.4 Study Methodology ........................................................................................................ 1-6
1.5 Scope of study ............................................................................................................... 1-7
1.6 Outline of the thesis ....................................................................................................... 1-7
CHAPTER 2 LITERATURE REVIEW ................................................................................. 2-1
2.1 Introduction .................................................................................................................... 2-1
2.2 Mechanisms of solid particle erosion ............................................................................. 2-1
2.3 Abrasive and Erosive Wear............................................................................................ 2-2
2.3.1 Mechanism of abrasive wear .............................................................................. 2-3
2.3.2 Mechanisms of erosive wear .............................................................................. 2-4
2.4 Controlling parameters for erosive wear ........................................................................ 2-5
2.5 Sediment........................................................................................................................ 2-7
2.6 Forces acting on the particles ........................................................................................ 2-8
2.6.1 Drag and lift forces.............................................................................................. 2-8
2.6.2 The buoyancy force ............................................................................................ 2-9
2.6.3 Rotation force ..................................................................................................... 2-9
2.6.4 Turbulence force................................................................................................. 2-9
2.7 Factors responsible for sediment erosion ...................................................................... 2-9
2.7.1 Characteristics of the sediment......................................................................... 2-10
2.7.2 Characteristics of fluids..................................................................................... 2-15
2.7.3 Characteristics of the base material.................................................................. 2-21
2.8 Wear theory ................................................................................................................. 2-25
2.9 Erosion Models ............................................................................................................ 2-27
2.9.1 General erosion model ..................................................................................... 2-27
2.9.2 Erosion models for hydraulic machinery ........................................................... 2-28
2.10 Conclusion ................................................................................................................... 2-29
CHAPTER 3 SEDIMENT EROSION IN HYDRAULIC MACHINERY ......................................... 3-1
3.1 Introduction .................................................................................................................... 3-1
3.2 Impulse turbine: Pelton .................................................................................................. 3-3
3.2.1 Inlet system ........................................................................................................ 3-3
3.2.2 Nozzle system ................................................................................................... 3-4
3.2.3 Pelton turbine runner .......................................................................................... 3-5
3.2.4 Criteria for Pelton turbine design ........................................................................ 3-6
3.3 Reaction turbine: Francis ............................................................................................... 3-7
3.3.1 Inlet system: Stay vane....................................................................................... 3-7
3.3.2 Guide vane system ............................................................................................. 3-8
3.3.3 Runner.............................................................................................................. 3-10
3.3.4 Labyrinth seals ................................................................................................. 3-11
3.3.5 Draft tube .......................................................................................................... 3-11
3.3.6 Shaft seal.......................................................................................................... 3-11
HPN PhD Thesis, NTNU vi

Table of Contents
3.4 Turbine Design .............................................................................................................3-11

3.4.1 Hydraulic design of turbine ................................................................................3-12
3.4.2 Mechanical design of turbine .............................................................................3-13
3.4.3 Operation of turbine...........................................................................................3-14
3.5 Alternative design of Francis Turbine ...........................................................................3-15
3.6 Conclusion....................................................................................................................3-16
CHAPTER 4 PARTICLE VELOCITY MEASUREMENT IN SWIRL FLOW, LABORATORY STUDIES4-1
4.1 Background ................................................................................................................... 4-1
4.1.1 General............................................................................................................... 4-1
4.1.2 Characterization of non spherical particles......................................................... 4-3
4.2 Objective of experiment ................................................................................................. 4-3
4.2.1 Description of test rig and test procedure ........................................................... 4-3
4.2.2 Measurement of particle velocity ........................................................................ 4-6
4.2.3 Visualization of particle motion ........................................................................... 4-7
4.2.4 Particle in swirl flow ............................................................................................ 4-7
4.2.5 Drag coefficient for particles ..............................................................................4-11
4.2.6 Uncertainties for the measurements..................................................................4-13
4.3 Results and Discussions...............................................................................................4-14
CHAPTER 5 COMPUTATIONAL FLUID DYNAMICS THEORY............................................... 5-1
5.1 Introduction.................................................................................................................... 5-1
5.2 Governing Equations ..................................................................................................... 5-2
5.2.1 Two equation turbulence closure models ........................................................... 5-3
5.3 Particle equation of motion ............................................................................................ 5-3
5.3.1 Drag force........................................................................................................... 5-4
5.3.2 Buoyancy force................................................................................................... 5-5
5.3.3 Rotation force ..................................................................................................... 5-5
5.3.4 Virtual or added mass force................................................................................ 5-5
5.3.5 Pressure gradient force ...................................................................................... 5-6
5.4 Restitution coefficient for particles ................................................................................. 5-6
5.5 Basic Erosion Model ...................................................................................................... 5-7
5.5.1 Model of Finnie ................................................................................................... 5-7
5.5.2 Model of Tabakoff and Grant .............................................................................. 5-8
CHAPTER 6 CFD SIMULATION MODEL .......................................................................... 6-1
6.1 CAHUA Hydropower Plant (HPP) .................................................................................. 6-1
6.2 Description of computational model............................................................................... 6-2
6.2.1 Flow model ......................................................................................................... 6-3
6.2.2 Mesh generation................................................................................................. 6-3
6.2.3 Properties of sand .............................................................................................. 6-5
6.2.4 Boundary conditions ........................................................................................... 6-6
CHAPTER 7 SIMULATION RESULT AND DISCUSSIONS..................................................... 7-1
7.1 Hydraulic performance of turbine................................................................................... 7-1
7.2 Sediment erosion on turbine components...................................................................... 7-2
7.2.1 Stay vane ........................................................................................................... 7-2
7.2.2 Guide vane ......................................................................................................... 7-3
7.2.3 Runner blade...................................................................................................... 7-4
7.3 Effect of sediment concentration on erosion .................................................................. 7-5
7.4 Effect of Sediment size on erosion ................................................................................ 7-7
7.5 Effect of sediment shape factor on erosion.................................................................... 7-8
7.6 Effect of operating condition on erosion......................................................................... 7-9
7.7 Validation of numerical simulations ..............................................................................7-10
CHAPTER 8 CONCLUSIONS AND RECOMMENDATIONS FOR FURTHER WORK ................... 8-1
8.1 Conclusions ................................................................................................................... 8-1
8.2 Recomandations for further work................................................................................... 8-2
References ............................................................................................................................. R-1
vii HPN PhD Thesis, NTNU

Table of Contents
APPENDIX A HYDRAULIC TURBINE ROTOR REPORT AT BEP………………………………...A-1

APPENDIX B HYDRAULIC TURBINE ROTOR REPORT AT FULL LOAD………………………....B-1
APPENDIX C PAPER I………………………………………………………………………...C-1

APPENDIX D PAPER II………………………………………………………………………..D-1
APPENDIX E PAPER III………………………………………………………………………..E-1
HPN PhD Thesis, NTNU viii

List of Figures and Tables
LIST OF FIGURES
Figure 1-1 Sediment erosion at Jhimruk hydropower plant, Nepal...................................................................... 1-3
Figure 1-2 Sediment erosion at Cahua hydropower plant, Peru ........................................................................ 1-4
Figure 2-1 Mechanisms of solid particle erosion ................................................................................................. 2-2
Figure 2-2 Mechanisms of abrasive wear (Stachowiak and Batchelor, 1993)..................................................... 2-3
Figure 2-3 Mechanisms of erosive wear (Stachowiak and Batchelor, 1993)....................................................... 2-4
Figure 2-4 Physical and material parameters for controlling erosive wear (Batchelor et al., 2002) .................... 2-6
Figure 2-5 Contrast in dependence on impingement angle for brittle and ductile modes of wear (Stachowiak and
Batchelor, 2006) .................................................................................................................................................. 2-7
Figure 2-6 Effect of particle size on mode and rates of erosive wear (Hojo et al., 1986) .................................. 2-11
Figure 2-7 Effect of medium on impingement angle by erosive particles (Stachowiak and Batchelor, 2006) ... 2-18
Figure 2-8 Example of particle trajectory analysis to predict erosive wear (Stachowiak and Batchelor, 2006). 2-19
Figure 2-9 Effect of temperature on the erosive wear rate of stainless steel (Stachowiak and Batchelor, 2006).. 2-
20
Figure 2-10 Relationship between mechanical properties of materials and erosion rate at elevated temperature
(Y. Shida et al., 1985):....................................................................................................................................... 2-20
Figure 2-11 Effect of flow on erosive wear (Stachowiak and Batchelor, 2006) ................................................. 2-21
Figure 2-12 Effect of primary material characteristics and erosion parameters on erosive wear rate (Zum et al.,
1987) ................................................................................................................................................................. 2-23
Figure 2-13 Comparison of the high and low elastic modulus modes of erosive wear protection (Stachowiak and
Batchelor, 1993). ............................................................................................................................................... 2-23
Figure 3-1 Sediment erosion at Pelton turbine nozzle and needle...................................................................... 3-4
Figure 3-2 Sediment erosion at Runner buckets ................................................................................................. 3-5
Figure 3-3 Illustration of separation of particle in a Pelton bucket (Thapa and Brekke, 2004) ............................ 3-6
Figure 3-4 Erosion at stay vane at Cahua power plant ....................................................................................... 3-8
Figure 3-5 Erosion at guide vane and facing plates at Cahua power plant ......................................................... 3-9
Figure 3-6 Erosion at runner at Cahua power plant .......................................................................................... 3-10
Figure 3-7 Variation of diameter for reduction of erosion .................................................................................. 3-16
Figure 4-1 Photographs of test rig ....................................................................................................................... 4-4
Figure 4-2 Schematic diagram of experimental set up ........................................................................................ 4-5
Figure 4-3 Pitot tube for measuring the velocity of flow....................................................................................... 4-6
Figure 4-4 Photograph of middle plate with radial and angular markings ........................................................... 4-6
Figure 4-5 Illustration of particle flow in spiral swirl ............................................................................................. 4-8
Figure 4-6 Erosion damage of suction side of Tokke guide vane by large particles ........................................... 4-9
Figure 4-7 Critical diameter relation based upon size of runner and drag coefficient ....................................... 4-10
Figure 4-8 Standard drag curve for motion of particle in a fluid......................................................................... 4-11
Figure 4-9 Drag coefficient of the sphere (Spurk, 1997) ................................................................................... 4-12
Figure 4-10 Particle velocity and head relation for different size and shape of particle .................................... 4-14
Figure 4-11 Particle velocity for different sizes and same size with different shape ......................................... 4-14
ix HPN PhD Thesis, NTNU

List of Figures and Tables
Figure 4-12 Particle velocity relation for same/different particle with different shape ....................................... 4-15
Figure 4-13 Rotation radius for particles of the same size but different shape ................................................. 4-15
Figure 4-14 Drag coefficient relation for same /different size of particle ........................................................... 4-16
Figure 4-15 Drag coefficient relation for regular and irregular shape of same/different particle ....................... 4-16
Figure 6-1 Cahua hydropower plant ................................................................................................................... 6-1
Figure 6-2 A single blade cascade computational model for three vanes .......................................................... 6-4
Figure 6-3 A single blade cascade mesh model for three vanes ........................................................................ 6-5
Figure 7-1 Hydraulic performance of turbine ...................................................................................................... 7-1
Figure 7-2 Predicted erosion pattern on stay vanes ........................................................................................... 7-2
Figure 7-3 Predicted erosion pattern on guide vanes ......................................................................................... 7-3
Figure 7-4 Velocity of sand particle inside a turbine ........................................................................................... 7-3
Figure 7-5 Predicted erosion pattern on turbine blades...................................................................................... 7-5
Figure 7-6 Effect of concentration rate on erosion rate density of turbine blade................................................. 7-6
Figure 7-7 Variation of relative erosion rate density with concentration ............................................................. 7-6
Figure 7-8 Effect of sediment size on erosion rate density of turbine blade ....................................................... 7-7
Figure 7-9 Variation of relative erosion rate density with sediment size ............................................................. 7-8
Figure 7-10 Effect of sediment shape factor on erosion rate density of turbine blade ........................................ 7-8
Figure 7-11 Effect of operating conditions on erosion rate density of turbine blade ......................................... 7-10
Figure 7-12 Particle velocity for different size around turbine blade ................................................................. 7-11
Figure 7-13 Particle velocity, shape factor and erosion rate around turbine blade ........................................... 7-12
LIST OF TABLES
Table 1-1 Hydropower potential in different river basins in Nepal (MoWR, 2003) .............................................. 1-2
Table 2-1 Classification of river sediment (Lysne et al., 2003) ........................................................................... 2-8
Table 2-2 Relative qualities of erosive wear resistant materials (Batchelor et.al., 1993).................................. 2-24
Table 3-1 Turbine erosion categories (Duan et al., 2002) .................................................................................. 3-2
Table 3-2 Classification of erosion (Matsumura and Chen, 2002)...................................................................... 3-2
Table 3-3 Variable input parameters................................................................................................................ 3-15
Table 4-1 The guide vane maximum angle at full load condition ...................................................................... 4-10
Table 4-2 Reynolds number ranges for single particle drag coefficient correlations ....................................... 4-12
Table 5-1 Coefficients for some materials using the Tabakoff erosion model ................................................... 5-8
Table 6-1 Sediment load of Cahua power plant (Ole et al., 2009) ..................................................................... 6-2
Table 6-2 Mesh statistics of numerical models .................................................................................................. 6-4
Table 6-3 Boundary conditions .......................................................................................................................... 6-7
HPN PhD Thesis, NTNU x

Nomenclature
NOMENCLATURE
Symbols
Symbol Description Unit Symbol Description Unit
FD Drag force N S1 Coefficient of sediment concentration ‐
CD Drag coefficient ‐ S2 Coefficient of sediment hardness ‐
Ap Projected area m2 S3 Coefficient of sediment particle size ‐
U Peripheral velocity m/s S4 Coefficient of sediment particle shape ‐
FB Buoyancy force N Mr Coefficient of wear resistance of base ‐
material
FC Centrifugal force N W Erosion rate mm/yr
FR Rotation force N Wc Cutting wear mm/yr
FE External force N WC Corrosion rate mm/yr
m Mass of the particle kg WD Deformation wear mm/yr
dc Critical dia. of particle m Wt Total wear (cutting and deformation) mm/yr
dp Diameter of particle m WT Total wear rate (erosion and corrosion) mm/yr
g Gravity m/s 2 R Radius of curvature of surface m
r Radius m C Velocity of particle m/s
r p Radius of particle m V Volume of particle m3
Cm Meridional velocity m/s N Number of abrasive particles ‐
component
Vp Particle impact velocity m/s Kmat Material constant ‐
P Plastic flow stress N/m2 Kenv Environmental constant ‐
M Total mass of material kg KT Maximum particle velocity m/s
S Abrasive depth mm Km Material factor ‐
h Net head m k Coefficient of discharge ‐
ns Specific speed rpm Cu Peripheral velocity component m/s
htot Total enthalpy J Y+ Dimensionless distance from the wall ‐
B Sum of body force N Q Flow rate m3/s
P Pressure Pa US Slip velocity m/s
FP Pressure gradient force N F VM Force due to virtual mass N
UF Under relaxation factor ‐ Cμ Constant ‐

n Speed rpm D Characteristic dimension of machine m
U
Mean velocity m/s f (α ) Function of impingement angle ‐
U U + u m/s f (VPN ) Function of velocity of particle ‐
U Reduced peripheral m/s Re , p Particle Reynolds’s number ‐

velocity
k1 Shape constant ‐ k2 Hardness constant ‐
c Concentration of sediment ppm k3 Abrasive resistance coefficient ‐
xi HPN PhD Thesis, NTNU

Nomenclature
Greek Symbols
ρ Density of fluid kg/m3 λ Turbine coefficient ‐
ω Angular velocity red/s ϕ Diffusion of turbulence stress W/kg
ηh Hydraulic efficiency ‐ φ
Viscous stress N/m2
σ Normal stress N/m2 τ Shear stress N/m2
ε Turbulence eddy dissipation m2/s3 Ω Speed number ‐
η Efficiency ‐ β Blade angle degree
γ Impact angle rad γ 0 Maximum Impact angle rad
μ Dynamic viscosity Kg/N‐m μ eff Effective viscosity kg/N‐m

ν Kinematic viscosity m2/s μt Turbulent viscosity kg/N‐m
π Constant ‐ κ Turbulence kinetic energy m2/s2
α0 Maximum Guide vane angle degree α Guide vane/ Impingement degree

angle
Sub-symbols
m Refers to meridional direction • Refers to vector dot product
n Refers to net value ∇ Refers to del vector operator
p Refers to particle ⊗ Refers to dyadic operator
u Refers to the peripheral direction ∝ Refers to directly proportional
r Relative value h Refers to hydraulic
* Refers to best efficiency point of turbine a Refers to average grain size
° Refers to the full load of the turbine x Concentration exponent
1 Refers to the inlet of the turbine runner y Size exponent
2 Refers to the outlet of the runner z Characteristic velocity exponent
f Refers to fluid ∂ Refers to partial derivative
HPN PhD Thesis, NTNU xii

Nomenclature
Abbreviations
AISI American Iron and Steel Institute
ADB Asian Development Bank
BPC Butwal Power Company Limited
BEP Best Efficiency Point
BHEL Bharat Heavy Electricals Limited
CFD Computational Fluids Dynamics
CPU Central processing unit
CAT Catchment Area Treatment
DNS Direct Numerical Simulation
EPT Department of Energy and Process Engineering
EDA Effective desilting arrangements
FL Full Load
GoN Government of Nepal
HPP Hydropower Plant
HB Brinell Hardness Number
IEC International Electro‐Technical Commission
IVT Faculty of Engineering Science and Technology
JHP Jhimruk Hydropower Plant
KU Kathmandu University
MUSIG Multiple Size Group
MSDM Micro‐Scale Dynamic Model
MoWR Ministry of Water Resources
NTNU Norwegian University of Science and Technology
N‐S Navier‐Stokes
NEA Nepal Electricity Authority
PL Particle load
PPM Parts per million
RANS Reynolds Averaged Navier‐Stokes
RMS Root mean square
RoR Run‐ of ‐River
SST Shear Stress Transport
SN‐Power Statkarft Norfund Power Invest
SRE Silt resistant equipment
URANS Unsteady Reynolds Averaged Navier‐Stokes equations
VHN Vickers Hardness Number
xiii HPN PhD Thesis, NTNU

Chapter 1
Introduction
1.1 BACKGROUND
1.1.1 General
Nepal, where the author comes from, is situated in South Asia on the lap of mighty
Himalayas, and possesses enormous water resources, which is one of the major resources
for technological, social and economic development of the nation. The energy supply and
demand characteristics have a great role to play in order to attain sustainable
development in the country. Thus, meeting Nepal’s current energy demand would help
foster higher economic growth. On the other hand, energy demand of the country
increases not only due to the increases in population but also due to the gradual
industrial development and modernization of society. However, Nepal has one of the
lowest per‐capita energy consumption rates in the world. Less than one third of the
country’s population has access to electricity. In the rural areas, where most people live,
has even less access to the electricity. According to recent report of Asian Development
Bank (ADB), the demand for power in Nepal has increased steadily with an annual
average growth rate of 8.5 % over the past decade, and it is estimated to grow at least by
10 % annually until 2020. Moreover, the current energy generation trend shows that
meeting this projected energy demand is possible due to huge water resources available
for hydropower development.
The total estimated hydropower potential of the nation is about 83,000 MW and out of
which about 43,000 MW is estimated to be economically feasible (MoWR, 2003). So far,
Nepal has been able to exploit about 2 % of the economically feasible hydropower
potential. This shows a huge scope for hydropower development in the country. The total
installed capacity of the country is about 680 MW, of which about 84 % is produced from
Run‐of‐River (RoR) hydropower plants. Therefore, RoR hydropower plants are the main
sources of electricity in the country. Among the different sources of energy, hydropower
is considered economical, non‐polluting and environmental friendly renewable source of
energy.

1-1 HPN PhD Thesis, NTNU

Chapter 1 Introduction
The overview of hydropower potential in different River basins in Nepal is presented in
Table 1‐1.
Table 1‐1 Hydropower potential in different river basins in Nepal (MoWR, 2003)
Main river Small river Major river Total power Economically

basins power potential power potential feasible power
( GW) potential (GW) (GW) potential (GW)
Sapta Kosi 3.6 18.75 22.35 10.86
Sapta Gandaki 2.7 16.95 20.65 5.27
Karnali and
3.5 32.68 36.18 25.1
Mahakali
Southern 1.04 3.07 4.11 0.88
Country Total 10.84 72.45 83.29 42.13
Furthermore, there is a huge potential for hydropower development in Himalayan region
in general. However, there are also technical challenges for hydropower development
due to erosion and sedimentation problem. The climatic and physical conditions are
highly responsible for the erosion and sedimentation problem in the region. The tropical
climate, immature geology, and intense seasonal rainfall, are the main reasons for this
problem. As a result, the Rivers in this region transport substantial amount of sediments
during the monsoon.
According to Naidu, 1999, about 20 billion tones of earth material are carried to the sea
each year by Rivers and Streams throughout the world, of which nearly 6 billion tones is
from the Indian subcontinent alone. Southeast Asia contributes two thirds of the worldʹs
total sediment transport to oceans (http://www.gemswater.org/atlas‐gwq/solids‐e.html)
[accessed on, 31 August 2009]. Handling sediments in hydropower projects has therefore
been a problem.
The management of the hydropower projects for achieving higher efficiency of hydraulic
turbines is an important factor. Hence, this problem has become primary concern for the
safety, reliability and longer life of the RoR hydropower projects. Many of the Himalayas
sites in Asia and South America’s sites in Andes, are examples of large regions where, the
problem of erosion and sedimentation is challenging. From the global perspective, the
RoR hydropower projects are becoming popular due to relatively less environmental
impacts than reservoir type of projects. Recent research has indicated that the benefits of
Run‐of‐River projects can be valued much more highly than they used to be considered,
because of environmental consequences.

HPN PhD Thesis, NTNU 1-2

Introduction Chapter 1
Both types of hydropower projects (storage as well as Run‐of‐River) suffer from sediment
erosion problems, however the nature of problem is different. The storage projects suffer
from reservoir capacity depletion over time due to sediment deposition. On the contrary,
the RoR projects, which do not have room for storage of sediments, suffer from excessive
wear of the turbines from the first year of operation itself. The sediment erosion of
turbines causes losses in revenue due to losses in energy generation and increase in
maintenance costs. Dealing with sediment has, therefore been a great challenge while
developing hydropower projects in sediment‐loaded rivers (Bishwakarma, 2008).

1.1.2 Sediment erosion problems

Sediment erosion problem is a global operation and maintenance problem of hydropower
plants. Hydraulic turbine components operating in sediment‐laden water are subject to
abrasive and erosive wear. This wear not only reduces efficiency and life of the turbine
but also causes problems in operation and maintenance, and ultimately leads to economic
losses. Many hydropower plants built on sediment‐loaded rivers have faced serious
problems of sediment erosion during the first years of operation itself. The 12 MW
Jhimruk hydropower plant (JHP) built in Nepal can be taken as an example.

(a) Guide vanes (b) Turbine runner
Figure 1‐1 Sediment erosion at Jhimruk hydropower plant, Nepal

The photographs presented in Figure 1‐1 illustrate the extent of sediment erosion in
guide vane cover and the runner blades after operating during a single monsoon. The
sediment study conducted in this power plant indicated that, the sediment concentration
exceeds 4,000 ppm for about 15 % of the monsoon. The average content of quartz in the
sediment is found to be above 60 % (Basnyat, 1999).

Similarly, 22 MW Cahua hydropower plant built in Peru can be taken as another example
of excessive sediment erosion. The photographs presented in Figure 1‐2, also illustrate the
extent of sediment erosion in guide vanes and turbine runner. These photographs were
taken during inspection visit from NTNU in April 2009. The sediment study conducted in
this power plant indicated that, the sediment concentration exceeded 120,000 tons of
sediment only after six weeks of operation. The average content of quartz particles in the
sediment was found to about 35 % and feldspar was found about 30 %.

(a) Guide vanes (b) Turbine runner
Figure 1‐2 Sediment erosion at Cahua hydropower plant, Peru

The high sediment concentration combined with high percentage of quartz and feldspar,
which are harder than the turbine material (hardness 6‐7 in Moh’s scale), was the main
causes of excessive sediment erosion in the hydraulic machinery operating in these
power plants. During monsoon season, a large amount of sediment (as high as 50,000
ppm) is carried by Rivers and it becomes difficult to remove all these sediments before
passing through the turbine, and causes severe damage to the turbine components.

1.2 MOTIVATION FOR THIS RESEARCH
There are huge technical challenges to develop new hydropower projects involving risks
of sediment erosion. The declining performance of hydro turbines has become one of the
major technical issues in the development of hydropower plants. Sediment transport
from the rivers is a natural phenomenon, it neither can be completely controlled, it nor
can be completely avoided; it should however be managed. Withdrawal of the clean
water from the river for power production is expensive due to design, construction and

operation of sediment settling basins. Even with the settling basins, 100 % removal of fine
sediments is impossible and uneconomical (Thapa, 2004). A 3‐dimentioanal approach is
desirable to encounter sediment problems in hydropower plants (Naidu, 1999), namely,
catchment area treatment (CAT) for reduction of sediment load, effective desilting
arrangements (EDA) for prevention of sediment, and sediment resistant equipment (SRE)
for withstanding the sediment.
The sediment erosion of turbine runners is a complex phenomenon, which depends upon
different parameters such as silt size, hardness and concentration, velocity of water, and
base material properties. The efficiency of the turbine decreases with the increase in the
sediment wear and finally breakdown of hydro turbines results. Various researchers have
conducted experiments to study the effect of these parameters on sediment wear, but
most of these experiments are on small‐size samples in different types of test rigs to
simulate the flow conditions in the turbine. However, actual flow conditions and the
phenomenon of sediment wear are too complex to simulate (Padhy et al., 2008)
Extensive research has been done to develop a wear model in terms of the material
properties involved but little attention has been given to clarify the influence of fluid
motion, especially in the turbulent flow regime. Over the years, many models have been
proposed for many different situations. Unfortunately, most models are co‐relational in
nature and therefore system specific. They only work for the particular material pair,
contact geometry, operating condition and range, and the particular environment and
lubricant. The available equations are so confusing that few designers can use any of
them to predict product life with confidence. The complexity of wear and the large
number of parameters affecting the outcome are the primary reasons for this situation.
Furthermore, several attempts have been made to minimize the loss of material due to
erosion. While they have proved successful elsewhere, they have not much success in
Himalayan region. Development in the areas of control of sediment from reaching the
turbine and protective coating on the surfaces, prone to erosion, has been the main area
of research. Very little has been done for developing erosion resistant flow passage
design. The problem blows multi‐folds as the sediment mainly consists of quartz (60 – 90
%) which is extremely hard and causes severe erosion damage.
A study carried out at the Department of Mechanical Engineering, Kathmandu
University, Nepal, shows that all rivers that flow through Nepal have sediment load and
high concentration of quartz (Thapa, 2004). This means that all turbines installed in these
rivers will experience sediment erosion. This study also revealed that the intensity of
erosion rate is no longer directly proportional to the percentage of quartz content of the
sediment. As a result, a few instances have been noticed where the intensity of damages
at different power plants were not identical. While components at a particular power
plant eroded very fast, damages to components at other power plants were insignificant.
This leads one to believe that large numbers of factors influences the process of erosion
damage (Naidu, 1999). In addition, they are jointly responsible for predicting sediment
erosion damage in the turbine components. There is strong need to study and identify the
effect of different shape and size of the sediment particle to predict the erosion rate

correctly. Therefore, dealing with sediment erosion problems requires a multidisciplinary
approach. More research and development is needed to investigate the relationship
between the particle movement and the erosion inside the turbine.

1.3 OBJECTIVE OF THIS STUDY
The main objective of this study is to investigate the relationship between the particle
movement and erosion inside the Francis turbine and to establish the operating strategy
for Francis turbine operating in sediment‐laden water. In order to achieve the main
objective the following specific activities have been carried out during this study period:
• Develop and build a laboratory test rig that can facilitate the testing of particle to
study the separation process in fluid flow in curved path
• Conduct experiment of the particle in the laboratory test rig to study the forces
acting on particle in rotational motion and to investigate the drag coefficient and
velocity relation with respect to different size, and shape of the particles
• Perform numerical simulation on a Cahua power plant Francis turbine design
and describe the prediction of the erosion for stay vanes, guide vanes, and runner
vanes for which the results of the field tests are available for verification
• Identify critical zones of erosion and investigate empirical relations based upon
particle shape, size, concentration, and operating conditions

1.4 STUDY METHODOLOGY
The overall study methodology adopted for this research work includes experimental
studies, numerical simulation, and field studies. A previously made test rig (Thapa,
2004), was reviewed and modified to create a strong swirl flow in curved path. This flow
was found similar to the flow between the guide vane outlet and the runner inlet of a
Francis turbine. The flow in the guide vane cascade was simulated in order to verify the
particles separation process and to investigate the velocity and the drag coefficient
relations based upon different shape and size of the particle. Then, numerical simulations
on a Cahua Francis turbine design were carried out, mainly at two operating conditions
with varying particle size, shape, and concentration in ANSYS CFX 11. The predictions of
erosion, based on the Lagrangian calculation of particle paths in a viscous flow, have
been described for different components of the Francis turbine. The numerically
obtained erosion patterns were compared with the field studies at Cahua hydropower
plant, in Peru. The experimentally obtained velocities and drag coefficients relations were
utilised for the validations of numerical simulations.

1.5 SCOPE OF STUDY
The scope of this research is not only to improve the Francis turbine components
operating in sediment‐laden water subject to sediment erosion through design
modification but also to identify the regions where special surface treatment is necessary
in order to increase the lifetime of the components for new hydropower projects
involving risks of sediment erosion. This is pursued by achieving the objectives
mentioned in the previous section. This research focuses only on the role of sediment
characteristics in fluid dynamics, even though the erosion primarily depends on both
sediment characteristics and material used for the turbine components.
This sediment erosion research in hydraulic machinery in general and the Francis turbine
in particular, is aimed to contribute the knowledge for:
• operation strategy of hydropower plants
• selection and design of turbine
• selection of appropriate material for turbine construction and maintenance
• maintenance of eroded turbine and maintenance scheduling
Hence, all technical, managerial and economical aspects are supposed to be considered in
this kind of study but only technical aspect has been studied extensively in present work.
It is believed that if a Francis turbine designer combines the hydraulic design and coating
of the critical parts, a significant reduction of erosion can be achieved, but detail analysis
of this aspect is out of the scope of this research work.

1.6 OUTLINE OF THE THESIS
This thesis is organized in eight chapters. Some chapters include their own conclusions.
An extensive literature review regarding the various form of wear and its mechanisms
caused by contact between a particle and a surface is presented in Chapter 2. The
sediment erosion types in hydraulic machinery components and the field observation at
Cahua hydropower plant is considered in Chapter 3. Chapter 4 is dedicated to the
laboratory studies of particle velocity measurement in highly swirl conditions similar to
turbine flow in curved path. It also includes a brief description of the developed test rig,
concept of critical diameter of particle inside a turbine, and experimental analysis. The
governing equations of fluid, particle equation of motion and two erosion models are
briefly presented in Chapter 5. A description of CFD computational model similar to
Cahua power plant Francis turbine design along with solution methods is included in
chapter 6. The simulations results and discussions are presented in Chapter 7. Finally, the
conclusions and recommendations for further work are given in Chapter 8. The auto
generated turbine rotor reports and three publications are included in the appendices.

Chapter 2
Literature Review
This chapter presents extensive literature review regarding the various forms of wear and
its mechanisms caused by contact between a particle and a surface. Furthermore, factors
responsible for sediment erosion with controlling parameters, and general erosion model
are also included.
2.1 INTRODUCTION
Sediment erosion is a phenomenon of mechanical wear of components. This is due to the
dynamic action of sediment flowing along with water impacting against a solid surface of
hydraulic components. Therefore, sediment flowing along with water passing through
the turbine is the root cause of sediment erosion in turbine components. The mechanical
wear in hydraulic machinery is mainly due to the suspended sediment in water, which is
subjected to kinetic energy, the force of gravity, viscosity, turbulence, centrifuge and
cavitation. Even if minor abrasion may take place in certain parts of hydro turbines,
erosion is the main caused of the damage.
Erosion occurs in a wide range of machinery (Thapa, 2004). The blades of gas turbines
are subjected to erosion from high velocity solid particle, and those of steam turbines
from liquid droplets (Tabakoff, 1995). The military aircrafts and missiles experience both
sand and rain erosion. The problem of hydro‐abrasive erosion of hydraulic machinery is
not limited to hydroelectric plants. Similar problems to those encountered in
hydroelectric work are also prevalent in the mining industry, dredging work, and waste
disposal. Even though erosion is normally harmful for the machine components, the
same mechanism has beneficial use in unconventional manufacturing techniques. Some
common examples of beneficial applications of erosion mechanism are water jet
machining, sand blasting, erosive drilling and rock cutting (Finnie, 1960).

2.2 MECHANISMS OF SOLID PARTICLE EROSION
In general, there are wide ranges of material degradation mechanisms. However, it can
be classified into three basic categories: mechanical, chemical and thermal actions, which
are considered the root causes of material separation as debris in erosion, but

Chapter 2 Literature Review
mechanisms for reaching to those actions are different (Thapa, 2004). Figure 2‐1 illustrates
the different action of solid particle erosion and its mechanisms. Stachowiak and
Batchelor, 1993, have discussed the different possible mechanisms for solid particle
erosion. From the subject point of view of sediment erosion damage caused by the
suspended solid particle in hydraulic machinery, only four solid particle erosion i.e.,
abrasive erosion, surface fatigue, brittle fracture, and ductile deformation are mainly
applicable.
Erosion by solid particle impingement
Cutting Fatigue Brittle Melting

fracture
Cyclic Non‐ cyclic Loss of fluid

failure failure state
Penetration Plastic
of deformation
cutting edge to
failure

Figure 2‐1 Mechanisms of solid particle erosion

2.3 ABRASIVE AND EROSIVE WEAR
Abrasive wear is the loss of material by the passage of hard particles over a surface. This
wear occurs whenever a solid object is loaded against particles of a material that have
equal or greater hardness. The abrasive wear involves processes such as micro cutting,
fatigue, grain detachment and brittle fracture. Erosive wear is caused by the impact of
solid and liquid particles on a surface. Erosive wear can resemble abrasive wear when
hard solid particles of microscopically visible size are eroding agent, the angle of
impingement is low and the impingement speed is of the order of 100 m/s. For all other
conditions, wear mechanisms not resembling abrasive wear become dominant. For
particle of microscopically visible size and an impingement speed of the order of 100 m/s,
wear at the high impingement angles proceeds by a combination of plastic deformation
and fatigue or by cracking for brittle materials.

Literature Review Chapter 2
2.3.1 Mechanism of abrasive wear

Generally, abrasive wear by grits or hard asperities closely resembled cutting by a series
of machine tools or a file. However, microscopic examination has revealed that the
cutting process is only approximated by the sharpest of grits and many other more
indirect mechanisms are involved. The particles or grits may remove material by micro
cutting, micro fracture, pullout of individual grains or accelerated fatigue by repeated
deformations as illustrated in Figure 2‐2.

(a) Cutting (b) Fracture

(c ) Fatigue by repeated ploughing (d) Grain pullout
Figure 2‐2 Mechanisms of abrasive wear (Stachowiak and Batchelor, 1993)

The first mechanism illustrated in Figure 2‐2 a, is cutting, which represents the classic
model where a sharp grit or hard asperity cuts the softer surface. The material that is cut
is removed as wear debris. When the abraded material is brittle, e.g., ceramic, fracture of
the worn surface may occur (Figure 2‐2 b). In this instance, wear debris is the result of
crack convergence. When a ductile material is abraded by a blunt grit, then cutting is
unlikely and the worn surface is repeatedly deformed (Figure 2‐2 c). In this case, wear
debris is the result of metal fatigue. The last mechanism illustrated (Figure 2‐2 d)
represents grain detachment or grain pullout. This mechanism applies mainly to
ceramics, where the boundary between grains is relatively weak. In this mechanism, the
entire grain is lost as wear debris.
Abrasive wear has the virtue of being possibly the easiest of the wear mechanism to
suppress completely if adequate materials are available. Rapid abrasive wear only occurs
when the ratio of particle hardness to materials hardness is greater than 1.2. When the
ratio of particle hardness to material hardness is less than 1.2, abrasive wear declines
rapidly but does not cease entirely until the material hardness is greater than the particle

hardness by a similar factor of 1.2. It is observed that when the particles have equal
hardness to the material, some scratching and wear of the material occurs although
damage to the particles by crushing and plastic deformation is extensive.

2.3.2 Mechanisms of erosive wear

Erosive wear involves several wear mechanisms, which are largely controlled by the
particle material, the angle of impingement, the impact velocity and the particle size. If
the particle is hard and solid then it is possible that a process similar to abrasive wear will
occur. Where liquid particles are the erodent, abrasion does not take place and the wear
mechanisms involved are the result of repetitive stresses on impact.
The term erosive wear refers to an unspecified number of wear mechanisms, which occur
when relatively small particles impact against mechanical components. This definition is
empirical by nature and relates more to practical considerations than to any fundamental
understanding of wear. The known mechanisms of erosive wear are illustrated in Figure
2‐3.

(a) Cutting erosion mechanism (b) Fatigue erosion mechanism

(c)Plastic deformation (d) Erosion by brittle fracture
Figure 2‐3 Mechanisms of erosive wear (Stachowiak and Batchelor, 1993)

2.3.2.1 Cutting erosion

As illustrated in Figure 2‐3 a, when particles strike the surface at low impact angle and
removal of material by cutting action, the erosion mechanism is called abrasive erosion.
The abrasive grits roll or slide when they strike on the surface and cause erosion by

abrasion or cutting mechanism. The material is removed by scouring or scrapping by
sharp edges of the particles forming short track‐length scars.

2.3.2.2 Surface fatigue

As illustrated in Figure 2‐3 b, surface fatigue erosion mechanism occurs, when the
particles strike the surface with large impact angle but at low speed. This mechanism of
erosion is similar to wear due to surface fatigue on rolling surfaces. The surface cannot be
plastically deformed. Instead, the surface becomes weak due to fatigue action and cracks
are initiated in surface after repeated hitting. The particles will be detached from the
surface after several strikes.

2.3.2.3 Plastic deformation

As illustrated in Figure 2‐3 c, plastic deformation of the surface takes place due to
formation of the flakes around the striking point when the particles strike the elastic
surface with medium speed and large impact angle. With repeated strike on the flakes,
the material will detach as debris.

2.3.2.4 Brittle fracture

As illustrated in Figure 2‐3 d, when particles strike the brittle surface with large impact
angle in medium velocity, brittle fracture erosion takes place. If the particles are sharp,
then brittle fragmentation is more likely to occur and the particles detach from the
material by subsurface cracking.

2.4 CONTROLLING PARAMETERS FOR EROSIVE WEAR
The mechanisms of erosive wear are not constant but are controlled by the angle of
impingement of a particle, its speed, its size and the phase of material that constitutes the
particles. The angle of impingement is the angle between the eroded surface and the
trajectory of the particle immediately before impact, as shown in Figure 2‐4. A low angle
of impingement favours wear processes similar to abrasion because the particles tend to
track across the worn surface after impact. A high angle of impingement causes wear
mechanisms, which are typical in erosion.
The speed of the erosive particle has a very strong effect on the wear process. If the speed
is very low, then stresses at impact are insufficient for plastic deformation to occur and
wear proceeds by surface fatigue. When the speed is increased, for example, to 20 m/s, it
is possible for the eroded material to deform plastically on particle impact. In this regime,
which is quite common for many engineering components, wear may occur by repetitive

plastic deformation. If the eroding particles are blunt or spherical then thin plates of worn
material form on the worn surface because of extreme plastic deformation. If the particles
are sharp, then cutting or brittle fragmentation is more likely. Brittle materials, on the
other hand, wear by subsurface cracking. At very high particle speeds, melting of the
impacted surface might even occur (C. S. Yust and R. S. Crouse, 1975). The angle of
impingement and the other listed parameters are illustrated schematically in Figure 2‐4.

Figure 2‐4 Physical and material parameters for controlling erosive wear (Batchelor et al.,
2002)

There are two modes of erosive wear, called the ductile mode of erosive wear (at
moderate particle speeds) and the brittle mode of erosive wear. The ductile mode of
erosive wear depends on tangential movement of the impingement particle across the
eroded surface for wear particle to be produced by searing or cutting. This tangential
movement is favoured by a very small angle of impingement as opposed to normal
impact. However, the angle of impingement is zero, there will be no kinetic energy of
impact between particle and worn surface to initiate indentation by the particle and chip
formation. A maximum in the ductile mode of erosive wear is usually found close to an
angle of 30 °, which represents a compromise between the requirements of tangential
particle movement and impact energy. For the brittle mode of erosive wear, the
maximum in erosive wear occur around 90 °, where the kinetic energy of impact is at a
maximum. The wear rates as a function of impingement angle for the brittle and ductile
modes of erosive wear are shown schematically in Figure 2‐5 .
Erosive wear by liquid particles is mostly limited to the brittle mode, where the rate of
wear is far more sensitive to particle velocity than when solid particles are involved. For
liquids, erosive wear is observed to be proportional to approximately the 5th power of
impact velocity while for solids wear, it is proportional to the square of impact velocity
approximately (Batchelor et al., 2002). Material with high toughness are usually the most
effective at resisting erosive wear, as hard materials are often too brittle to be wear
resistant. The size of the particle is critical to wear resistance as small particles, e.g. 8 μm
average sizes tend to erode by ductile shearing of the worn material. This characteristic is

observed because brittleness depends on the presence of at least one defect in the
deformed volume. Since the deform volume decreases in proportion to particle size,
erosive wear at small particle sizes is more likely to be ductile than erosion at large
particle sizes, where the apparent brittleness of the worn material is higher. High
hardness materials are therefore more likely to be effective in resisting erosive wear by
small particles, whereas tough materials are suitable for larger particles of 50 μm sizes
and larger.

Figure 2‐5 Contrast in dependence on impingement angle for brittle and ductile modes of
wear (Stachowiak and Batchelor, 2006)
2.5 SEDIMENT
River sediments flowing along with water passing through the turbine are the reasons
why the sediment erosion in hydraulic turbine components occurs. The river sediments
are in the form of clay, silt, sand and gravel with specific gravity approximately 2.6. In
the river hydraulics, sediment particles are classified in to bed load and suspended load
based on transport of sediment. All the particles, which move close to the bed by sliding,
rolling or jumping are called bed load. These particles have much lower velocity than
flowing water, whereas all those particles, which are carried away in suspension by
flowing water, are called suspended load and they have more or less same velocity as the
flowing water. Fraction of suspended load is settled down in the settling basins or
reservoirs and rest will pass through turbines causing erosion of components.
Sediments are made of fragmentation of rock due to chemical and mechanical
weathering. The sediments in river are mixtures of different particle sizes as shown in
Table 2‐1. This is a sand fraction of the sediment, which causes turbine erosion. The sand

fraction can be further classified in to fine (0.06 ‐ 0.2 mm), medium (0.2 ‐ 0.6 mm) and
coarse (0.6 ‐ 2 mm).
Table 2‐1 Classification of river sediment (Lysne et al., 2003)
Particle Clay Silt Sand Gravel Cobbles Boulders
Size (mm) <0.002 0.002 ‐ 0.06 0.06 ‐ 2 2 ‐ 60 60 ‐ 250 >250
2.6 FORCES ACTING ON THE PARTICLES
The fluid flow influence the particle velocity and direction, and several forces are also
acting on the particle for stabilizing and destabilizing its position on its path through the
turbine. The particle will be affected mainly two different force categories (Lysne et al.,
2003).
• Stabilizing forces: This is due to viscous and gravity forces that retard
movements of the particles and have stabilizing effects
• Destabilizing forces: This is due to centrifugal and coriolis forces that slide the
sediment along the bed with a low velocity, called bed load. The turbulence force
will cause irregular movement in the fluid and the particles as they move in the
same direction as the fluid surrounding them.
When particles move along curved path, the centrifugal force and, if the system is
rotating (including global rotations), the Coriolis force are pertinent destabilizing forces.
Exact interactions between all associated factors are complicated for numerical analysis.
Sediment transport analysis may give indication on location of attack on turbine, which
can be useful to focus on an optimal location of erosion resistant coating in turbine.

2.6.1 Drag and lift forces

The drag force works in the direction of the main flow, and the lift force in the transverse
direction. The forces are proportional to the square of the flow velocity.

1
FD = ⋅ C D ⋅ AP ⋅ ρ ⋅ U 2
2 Equation 2‐1

FL = C L ⋅ AP ⋅ ρ f ⋅ U 2
Equation 2‐2


2.6.2 The buoyancy force

The buoyancy force is an effect of the difference between the density of the fluid and
particles. The gravity effect influences the particle motion.

π
FB = ⋅ d P ⋅ (ρ P − ρ f ) ⋅ g
3
Equation 2‐3
6

2.6.3 Rotation force

The sum of coriolis and centripetal forces is the rotation force acting in a rotating domain.

FR = m ⋅ (− 2 ⋅ ω ⋅ U P − ω ⋅ ω ⋅ rP ) Equation 2‐4
2.6.4 Turbulence force

Eddies are formed in an irregular flow due to shearing action inside the flow. Eddies will
form and dissipate irregularly, and this disorderly flow is called turbulence effect.

2.7 FACTORS RESPONSIBLE FOR SEDIMENT EROSION
Since the sediment erosion damage is due to the dynamic action of sediment against a
solid surface, characteristics of the sediment, fluids (carrying sediments) and base
material are jointly responsible for sediment erosion (Thapa, 2004).

• Characteristics of the sediment (shape, size, hardness, concentration, material
etc.),
• Characteristics of fluids (velocity, acceleration, impingement angle, medium of
flow, temperature, turbulence etc.) and,
• Characteristics of the base material (chemistry, elastic property, hardness, surface
morphology etc.)


2.7.1 Characteristics of the sediment

Particle characteristics are important but relatively poorly researched aspect of the
erosion problem (Stachowiak and Batchelor, 2006). It is known that hard particles cause
higher wear rates than soft particles (Goodwin et al., 1969 ‐ 70). The sharpness of the
particle has also been recognized as accelerating erosive wear (Bahadur et al., 1990). Both
of these parameters have been included in numerical models of erosive wear (Stachowiak
et al., 1987). The ratio of particle hardness to substrate hardness seems to be a controlling
parameter (Stachowiak et al., 1987). The significance of particle hardness becomes
apparent when the hardness of some erosive, for e.g., alumina, are compared to that of
standard materials such as mild steel. In this instance, the ratio of particle to substrate
hardness is about 10. The effect of particle hardness on wear depends on the particular
mode of erosive wear taking place, e.g., ductile or brittle. In the brittle mode, the effect of
particle hardness is much more pronounced than in the ductile mode (Stachowiak et al.,
1987).
It is impossible to isolate hardness completely from other features of the particle such as
its shape. Even if the particle is hard but relatively blunt then it is unlikely to cause severe
erosive wear. A blunt particle has a mostly curved surface approximating to a spherical
shape while a sharp particle consists of flat areas joined by corners with small radii,
which are critical to the process of wear.
Variations in particle size in the range typical to engineering applications can cause
fundamental changes in the erosion mechanism. A series of erosion tests on glass, steel,
graphite and ceramics revealed that as particle size was increased from 8.75 μm to 127
μm in diameter, the mode of erosion changed from ductile to brittle. This caused the
erosive wear peak to move from 30 ° to 80 ° impingement angle and even more
significantly resulted in a dramatic increase in erosive wear rates as shown in Figure 2‐6
(Sheldon and Finnie, 1966). In both cases, silicon carbide impinging at a speed of 152 m/s
was used as the erosive agent.
It can also be seen from Figure 2‐6 that particle size not only affects the wear rate but
drastically alters the ranking of materials in terms of wear resistance. When the small
particles were used as the erosive agent, the materials ranked according to their wear
resistance are in the following order: high‐density alumina > annealed aluminium > plate
glass > high‐density magnesia > graphite and hardened steel. In this case, apart from the
annealed aluminium, erosive wear rate depends on the hardness of the material. Work
hardening of the aluminium could be significant in this instance. On the other hand,
when the large particles were used as the erosive agent, the order changes to annealed
aluminium > hardened steel > high‐density alumina > high‐density magnesia > plate glass
> graphite. Therefore, in this case toughness of the material is important. Materials, which
are neither tough nor hard, e.g., graphite, show inferior erosion resistance.



Figure 2‐6 Effect of particle size on mode and rates of erosive wear (Hojo et al., 1986)

The change in wear modes is believed to be a consequence of the average spacing of
defects, e.g., holes or cracks in a solid. If the impinging particles are very small, then only
a minority of the impingement sites will coincide with a defect. The impingement site is a
zone of highly stressed material directly beneath the particle on impact and similar in size
to the particle. Plastic deformation is encouraged by an absence of defects and is the
predominant mode of metal removal for small particles. Since repeated plastic
deformation is required to remove material, this form of wear is relatively slow. For
larger eroding particles, a defect is usually present in the impingement site and material
removal by brittle processes is therefore favoured. Since crack formation is rapid the
brittle mode of erosion can be a very destructive form of wear.
The incubation period of erosive wear refers to the period from the start of erosion to the
onset of measurable positive wear. During the incubation period, wear either may be
negligible or may appear to be negative. This latter characteristic is caused by eroding
particles becoming trapped in the worn material. The incubation period is generally
believed to relate to the accumulation of subsurface damage, e.g., cracks or strained
materials that are the precursors of wear particle release. Once the incubation period has
passed, wear usually proceeds at a constant rate.


2.7.1.1 Size and shape of the particle

The intensity of erosion is directly proportional to the size of the particles. Particle sizes
above 0.2 to 0.25 mm are extremely harmful. It has been found that large size sediment
particles (above 0.25 mm) even with hardness lesser than 5 on Moh’s scale cause wear.
Similarly fine silt even with size less than 0.05 to 0.1 mm, containing quartz wears out the
underwater parts. The fine sediment can also be dangerous if the turbine is operating
under high head. Sharp and angular particles cause more erosion in comparison to
rounded ones.
Based upon the information available from literature survey conducted on abrasive wear
in hydraulic machinery by Truscott, 1972, Beregoron, 1950, and Wiedenorth, 1970, state
that, in general, the absolute wear rate increases with grain size and sharpness.
Wellinger, 1958, and Worster, 1955, state that wear is directly proportional to size for
sliding or grazing abrasion, but is independent of size for direct impact. Goodwin’s tests,
1970, show that the erosion rate for impact abrasion becomes constant only above a
certain grain size (about 50 ‐ 100 μm depending on velocity). Stauffer, 1956, also states
that the relative wear of metals decreases with increasing size, but gives no results.
Bergeron, 1952, found from tests on Aluminum Bromide (Al‐Br), that wear was
proportional to (size)0.75, but he stated that for general application, wear was proportional
to size × function of coefficient of friction, densities, and size / surface curvature ratio.
Wellinger, 1958, shows the effects of particle shape on impact abrasion and states that
angular grains cause about twice the wear compared to rounded ones. Goodwin, 1970,
also discusses erosiveness of particles, and defines a shape‐factor; he states that hardness
and sharpness are interrelated. Wiedenroth’s wear tests, 1970, on a small dredge pump
impeller, using a lacquer removal technique, show differences in the blade wear pattern
depending on grain size (i.e. sand or gravel).
Particle size can be characterized mainly in two basic dimension, mass and length. For a
given velocity, the kinetic energy of particle is directly proportional to mass and mass of
spherical particle is proportional to (diameter)3.
Sheldon and Finnie, 1966, observed the change of ductile mode of erosion to brittle mode
when particle size is changed from small to larger. In the experiment with small and
larger size particles, maximum erosion rate shifted from impact angle 30 ° to 80 °. Small
size particles have more cutting effect while bigger particles deform material by elastic
deformation and fatigue. The erosion rate ranking depends on hardness in case of erosion
due to small particles, whereas, in case of large particles, it is dependent on toughness of
material. Most often, non‐cohesive natural sediments are characterized in term of particle
diameter in river hydraulics. This method can be acceptable for other particles that have
shape and density similar to natural sediments. Sediment particle sizes are defined in
terms of sedimentation diameter, standard fall diameter, and nominal diameter or sieve
diameter. Sedimentation diameter of a particle is the diameter of sphere that has the same
specific gravity of 2.65 and has the same terminal settling velocity as the given particle in
the same fluid under the same condition.

Generally, particle shapes are described qualitatively such as round, angular and
semi‐round based on visual observation. The basic shapes found in the nature are
generally uniform, but due to several reasons, actual shapes of particles are sharp and
complex, which cannot be described in simple mathematical terms.
The shape of the particle is a good indicator of erosion process, for instance, irregular
shape with sharp edge increases erosion rate, whereas, blunt particles with round edges
are less erosion rate in general. Most of the erosion models have incorporated the effect of
shape; hence, quantification of the shape parameter is essential for the estimation of
erosion by solid particles. Together with some of the approaches suggested by Bahadur
and Badruddin, 1990, roundness factor and other statistical parameters are also used to
describe the shape of the particles. Correlating the ductile erosion with cutting by single
point cutting tool, the shape of particles can also be defined in term of Rake angle, which
is the angle between the front face of particle and normal to the target surface. Winter
and Hutchings, 1975, also used this concept. H. Drolon et al., 2000, used multi‐scale
roughness descriptor technique, based on harmonic wavelet transform for sediment
particle analysis. Stachowiak, 2000, used Spike Parameter‐Linear Fit based on projected
particle boundary and Spike Parameter‐Quadratic Fit based on curve fitting for major
boundary to describe the angularity of the particles and studied their relation with
erosion rate. These parameters are based on representing the particle boundary by set of
triangles, which is directly related to particle erosivity. Chen and Li, 2003, simulated
erosion using computer model (Micro‐scale dynamic model, MSDM), and investigated
the difference in erosion rate by three basic shapes: triangle, square and circle. The
highest erosion loss in single particle impact is by triangular particle followed by circular
and square. This observation is in agreement with stresses induced by contact area. The
erosion loss changes when square particle is rotated at 45 ° and contact becomes the
smallest. In reality, several particles strike the surface and the ranking of erosion will be
different. In this case, erosion by square shape particle is bigger than circular, because of
plastic deformation after subsequent strikes is in a larger area. In general, the erosion rate
by triangular or square particles may be 1.5 times higher than the circular particles.
Shrestha and Sandeep, 2005, studied shape feature extraction and pattern recognition of
sand particles and their impact on hydraulic turbines. A systematic approach to the
problem was identified. Sand particles were collected from the erosion sensitive power
plants and its digital images were acquired. Software was developed on MATLAB 6.5
platform to extract the exact shape of sand particles collected. These shapes were further
been analyzed by artificial neural network. This network was first trained for the known
input and known output. After that, it was trained for unknown input and known
output. Finally these networks can recognize any shape given to it and gives the shape
which is nearest to the seven predefined shape. The software was trained for seven types
of shapes with shape number 1 to 7 in increasing number of sharp edges. The shape with
shape number seven is having large number of sharp edges, and is considered as most
erosive where as shape with shape number one is having round edges and considered as
the least erosive.


2.7.1.2 Hardness of the particle

The intensity of the erosion is also directly proportional to the hardness of particles
(irrespective of sizes). Those particles have a hardness value above 5 in Moh’s scale are
considered harmful. Incidentally, most of the Himalayan sediment contains more than 70
% Quartz particles on average with hardness value 7 in Moh’s scale. Hence, hardness of
the particle is a highly influencing factor in the erosion rate.
Truscott, 1972, mentioned that both Wehinger’s, 1958, and Stauffer’s, 1956, laboratory
tests show that, for metals in general, wear increases rapidly once the particle hardness
exceeds that of the metal, for both scouring and impact abrasion. Beyond this, the wear
rate may become constant, or even reduce, with increasing abrasive hardness. Stauffer
noted that the wear resistance of a 13 % Cr cast steel was only slightly better than that of
the unalloyed reference steel, whereas it is usually considerably better in practice; he
suggests this might have been due to the excessive hardness of the test abrasive. From
tests with various grades of very fine sand (< 200 μm) under dry conditions, Goodwin et
al., 1970, found that erosion varied as hardness, and depended on the amount of quartz
present. The stronger is an abrasive particle, the deeper it is pressed into the surface
before it is broken and the heavier is the resulting wear. The mechanical strength, σ , is
determined as
4⋅ N
σ = Equation 2‐5
π ⋅d2
Where, N is the load producing destruction of a single abrasive grain and d is its
diameter. The values of σ rises with reduction in d because of the scale effect
(Kragelsky, 1985).

2.7.1.3 Concentration
The sediment concentration is one of the dominating factor influencing erosive wear
rates. Concentration is the total mass (or volume) of imparting particles present in the
unit mass (or volume) of the fluid. Sometimes, it also called the particle flux rate, or the
mass of the imparting material per unit area and time. It can also be represented in terms
of percentage of particles in a given fluid mass (or volume). Especially for river
sedimentation, concentration is usually expressed in grams per litre (g/l). However, often
parts per million (ppm) by weight is used, which is equivalent to mg/l, with the
approximation of 1,000 ppm equal to 1 kg/ m3 of water being normal usage (1,000 ppm is
equivalent to 0. 1 %).
Erosive wear rate is proportional to the concentration up to a certain limiting value of the
wear. This limit has been observed in many studies and is believed to be the result of
interference between rebounding particles and arriving particles. It is also possible for
wear rates to decrease marginally when the limiting flux is exceeded.

There is surprisingly little quantitative information on the effect of solids concentration. It
is generally accepted that wear increases with concentration. Bak, 1966, and Antunes,
1966, consider this relationship to be direct. Bergeron, 1952, from tests on Al‐Br.,
suggested that this applies only to small amounts of solids, but for larger values, wear
increases more slowly; his theory states that wear times no. of grains per unit surface area
is dependent on concentration and flow pattern. Arnold and Hutchings, 1989, found that
the limiting particle flux rate is highly variable, ranging from as low as 100 kg/m2s for
elastomer to as high as 10,000 kg/m2s for erosion against metals by large particles
(Stachowiak and Batchelor, 2006). Mostly erosion rate is considered linearly proportional
to concentration. Bjordal, 1995, found relations for different metals and coatings as
follows. Erosion rate ∝ concentration0.25 to .27
However, for most of the materials, when tested for longer period, this value is close to
unity. Hence, considering erosion rate direct proportional to concentration with respect
to velocity is a satisfactory approximation.

2.7.2 Characteristics of fluids

The main characteristics of fluids include velocity and acceleration of water carrying
sediment, impingement angle, media of the flow, temperature, and turbulence. The effect
of each characterises on erosive wear is important.

2.7.2.1 Velocity of water carrying sediment

The intensity of erosion is normally proportional to the cube of the velocity of water,
carrying sediment particles in suspended condition. This seems true of Francis runners.
Any decrease in velocity therefore would substantially reduce the erosion damage. For
instance, 10 % decrease in water velocity would reduce erosion at least by 27 % (Naidu,
1999). Laboratory experiments conducted at corporate R and D, BHEL, Hyderabad, India,
by utilizing rotating disc apparatus with sediment concentration 75,000 ppm on an AISI
304 steel test piece proved that the sediment erosion rate increased in proportion to V
from 1.75 to 2.4. The same experiment also revealed that the reduction in sediment
erosion rate was 200 to 280 % when the particle sizes, were reduced by half. Sediment
erosion rate (W) can be generally governed by the formula,

W ∝ S1 ⋅ S 2 ⋅ S 3 ⋅ S 4 ⋅ M r ⋅ V n Equation 2‐6
Where, S1 = coefficient of sediment concentration, S2 = coefficient of sediment hardness, S3
= coefficient of sediment particle size, S4 = coefficient of sediment particle shape, Mr =
coefficient of wear resistance of base material, and Vn = relative velocity of water.

Based on experience of some hydropower stations in Latin America, following values of
exponent n have been suggested: n = 3 for Francis runner, n = 3 for Guide vanes and pivot
ring liners, n = 2.5 for Pelton nozzles, and n = 1.5 for Pelton runner buckets.
In actual practice, material damage due to plastic deformation and cutting occur
simultaneously and the ratio of these damage mechanisms depends on the velocity of
particle and the impingement angle together with other parameters. Up to certain
velocity, also referred as critical velocity or threshold velocity, the particle cannot skid in
the surface due to friction and cutting action does not take place. Yabuki et al., 1999,
found this critical velocity for silica sand and carbon steel pair as 2.5 m/s for 0.26 mm
particles. As the velocity increases higher than critical velocity, both cutting and plastic
deformation component increases, which amplify the erosion rate drastically. The modes
of erosion also vary depending on the velocity of the particles. At low velocity, the
particles do not have enough energy to erode the material by cutting action, but elastic
deformation or fatigue effect may be observed.
Most often quoted expression for relation between erosion and velocity of particle is
Erosion ∝ Velocityn, where the values of exponent n vary depending on material and
other operating conditions. Considering the impact of particles due to kinetic energy as
cause of material removal, theoretically, value of n is 3. However, the view and the
finding of different researchers on the value of this exponent n are not alike. Truscott,
1971, reported different values of the exponent, for instance n = 1.4 for steel St37 to n = 4.6
for rubber tested on sandblast apparatus. The value of the exponent will be lower in the
case of combined effect of corrosion and erosion with higher corrosion intensity.
However, sudden increase in this value can be observed in some velocity range when
corrosion scales are removed. In the combined case of solid particle erosion and
cavitation, Zhang et al., 1996, found the erosion rate proportional to peripheral speed of
disc with exponent value in between 3 ‐ 4.5 for non‐metallic coatings.
Similarly, Arnold and Hutchings, 1990, found the velocity exponent for unfilled
elastomer in between 2.9 and 5.1 for impingement angle 30 ° and 90 ° respectively. It may
be due to particle erosion at higher velocity is more significant than synergy due to
cavitation and erosion. Daun et al., 2002, observed different values for different type of
test rig for instance n = 2.5 ~ 3.0 for rotary type stand, n = 1.8 ~ 2.7 for disc stands, and n =
2.0 ~ 2.2 for water jet impact. There are big differences in values of the velocity exponent
and it is difficult to have uniformity in erosion rate prediction and simulation.
Most often, particle velocity is considered same as the fluid velocity, which is estimated
based on continuity equation (Wood, 1999). This is not true in actual practice because
generally particle velocity is less than the fluid velocity. Zahavi and Schmitt, 1981, found
sand velocity one third of air velocity in jet type of erosion test rig. The accuracy of
particle velocity is important in erosion models, but measurement of that is difficult in
practice. Chevallier and Vannes, 1995, have mentioned light speed photography, optic
gates and double rotating disc for measurement of particle velocity. Bjordal, 1995, used
the velocity based on rotational speed of specimen. In the experimental part of this

research, velocity of the particles will be computed by considering continuity equation of
the mixture.

2.7.2.2 Impingement angle

The impingement angle is defined as the angle between the eroded surface and the
trajectory of the particle just before impacting a solid surface. If the particles are moving
parallel to the surface, impingement angle is almost 0 ° and hence only minor erosion
may take place. When particles are moving normal to the surface the impingement angle
is 90 °.
There are differences in the ways of presenting the relation of erosion rate with respect to
impingement angle. Bhushan, 2002, showed that the maximum erosion rate for brittle
material is higher than that for ductile material. On the other hand, Matsumura and
Chen, 2002, have shown higher erosion rate for ductile material. Similarly, Bhushan, 2002,
has shown no erosion up to certain low impingement angles and on contrary, Stachowiak
and Batchelor, 1993, have shown erosion rate (about 10 % of maximum) even at zero
degree impingement. These differences in the presentation could be due to way of
defining the impingement angle of actual particles. Normally the jet angle is considered
as impingement angle of particles for practical purpose, but that is not the true impact
angle. The flow of particle in the straight pipe or parallel plates can be considered to have
impingement angle zero but even in such flow, erosion can be expected. In such cases, if
the flow is turbulent, the particles could be dancing or oscillating within the boundary
layer in the direction normal to the flow and hence effective impingement could be even
close to 90 °.

2.7.2.3 Effect of erosion media on erosive wear

Most erosive agents are conveyed by a medium, e.g., water or air. A mixture of erosive
particles and liquid medium is known as slurry. The characteristics of the medium have a
surprisingly strong effect on the final wear rate. Controlling factors relate to the bulk
properties of the medium, i.e., viscosity, density and turbulence, and to its microscopic
properties such as corrosivity and lubrication capacity. It has been shown that small
additions of lubricants to erosive slurries can significantly reduce wear (Levy et al., 1987).
The ability of the liquid medium to provide cooling during particle impingement is also
important (levy et al., 1987). In terms of bulk properties, the drag forces imposed by
viscous slurry on the erosive particles can affect wear by altering the impingement angle.
This is demonstrated schematically in Figure 2‐7(Hojo et al., 1986).


It can be seen that the increased particle drag forces, imposed by more viscous medium,
shift particle impingement to the sides of the eroding cylinder. The effect of the medium
is to alter the location and the form of wear since the impingement angle is reduced by
the shift to the cylinder sides. The medium‐induced reduction in impingement angle
causes an increase in abrasion‐type mechanisms of erosive wear. If an estimation of wear
rates in a real machine is required, then a comprehensive analysis of the particle
trajectories is essential. For example, an analysis performed for the inlet blades of a gas
turbine gave an excellent agreement between predicted and actual location of wear spots
(Tabakoff, 1987). An example of erosive particle trajectories between gas turbine blades is
shown in Figure 2‐8.

(a) Low viscosity media
(b) High viscosity media
Figure 2‐7 Effect of medium on impingement angle by erosive particles (Stachowiak
and Batchelor, 2006)



Figure 2‐8 Example of particle trajectory analysis to predict erosive wear (Stachowiak and
Batchelor, 2006)
The effect of a medium is assessed in terms of the ʹcollision efficiency’, which is the ratio
of particles that actually hits a wearing surface to the theoretical number of particle
impacts in the absence of any medium. It was found that the collision efficiency declines
from a limiting value of 1 for large particles, e.g., 750 μm size, to less than 0.1 for small
particles of 75 ‐ 90 μm size, at medium viscosities of 0.005 Pa (H. Mcl et al., 1991). The
reduction in collision efficiency is due to the viscous medium sweeping the particles past
the wearing surface. The erosive wear rate was found to closely follow the same trend as
collision efficiency, which indicates that the primary effect of a liquid medium is to divert
particles from the wearing surface. Increasing particle velocity reduces the influence of
medium, so that at high slurry velocities, only large particles are affected by the
mediumʹs viscosity (H. Mcl et al., 1991).

2.7.2.4 Effect of temperature on erosive wear

From the sediment erosion in hydraulic machinery point of view, temperature alone
cannot give any problem however; in general, this can influence the erosion rate and
mechanism of erosive wear. The primary effect of temperature is to soften the eroded
material and increase wear rates. The effects of temperature on erosion of stainless steel
are shown in Figure 2‐9 (Levy et al., 1986). The erosive agent is silicon carbide impinging
stainless steel at a speed of 30 m/s in a nitrogen atmosphere. It is not until temperatures
higher than 600 ~ are reached when the erosion rate showed significant increase. This
temperature coincides with the softening point of the steel. There is a strong correlation
between the mechanical properties of the material at the temperature of erosion and wear
rate as shown in Figure 2‐10 (Y. Shida et al., 1985).When high temperature erosion of
metals occurs in an oxidizing medium, corrosion can take place and it further accelerates
wear. Material is removed from the eroding surface as a relatively brittle oxide and this
process of wear can be far more rapid than the erosion of ductile metal. At sufficiently

high temperatures, however, the underlying metal does not come into contact with the
impinging particles because of the thick oxide layer present (Stephenson et al., 1986) and
then, oxidation rates, not mechanical properties, control the erosive wear.

Figure 2‐9 Effect of temperature on the erosive wear rate of stainless steel (Stachowiak
and Batchelor, 2006)

Figure 2‐10 Relationship between mechanical properties of materials and erosion rate at
elevated temperature (Y. Shida et al., 1985):
(1) Carbon steel, (2) 1.2Cr‐1Mo‐v steel, (3) 2.25Cr‐1Mo steel, (4) 12Cr‐1 Mo‐V steel, (5) 304
steel and (6) alloy 800


2.7.2.5 Effect of turbulence on erosive wear

Turbulence of the medium accelerates erosive wear, as particle impingement is more
likely to occur in turbulent flow than in laminar flow, where the medium tends to draw
the particles parallel to the surface (S. Dosanjh at el., 1985). The difference between
particle behaviour in laminar and turbulent flow of the medium is illustrated in Figure
2‐11.

Figure 2‐11 Effect of flow on erosive wear (Stachowiak and Batchelor, 2006)
2.7.3 Characteristics of the base material

The material used for the turbine components is equally important factor in the sediment
erosion damage. Hardness of the material, its chemical composition, microstructure and
its work hardening property influence the intensity of erosion. The choice of the material
for a particular component is to be made considering its ability to meet the functional
requirements like impact strength and ability to withstand cyclic loading in addition to
its wear resistance.
Experience has shown that it is sufficient for components, which are not very susceptible
to abrasion, such as spiral casings, nozzle pipes and draft tubes, to be made of plain

structural steel or castings of adequate strength, if their wetted surfaces are protected by
a tough, elastic coating. Various epoxy and polyurethane based plastics have proved very
suitable for this purpose. These coating materials are recommended for new components
and even for eroded parts after suitable pre‐treatment. Experience also indicates that the
erosion resistance of stainless steel is very good compared to other materials. Stainless
steel is stainless because of a thin layer of chromium oxide on their surfaces. The film of
chromium oxide is also responsible for the resistance of stainless steels to oxidation. It is
therefore necessary that stainless steels have a minimum of 12 % chromium. In addition
to chromium, stainless steels also have other alloying elements like carbon, Nickel,
Manganese, Molybdenum, Silicon and in some cases, Titanium and Columbium
(Niobium).

2.7.3.1 Erosive wear resistance of materials

Material characteristics exert a strong effect on erosive wear and have been extensively
studied. In a similar manner to abrasive wear, it is found that improvements in
mechanical properties do not always coincide with superior erosive wear resistance. For
example, erosive wear rates may increase when a material is deliberately hardened. The
difficulty with material optimization for wear reduction is that the characteristics of
erosive wear as well as the material characteristics control the wear rate. An illustration
of this rule is provided by the comparison of the relative erosion resistance of metals as a
function of impingement angle. When the impingement angle is shallow, a hardened
steel shows lower wear than a soft steel; the converse is true at high impingement angles.
This is illustrated in Figure 2‐12, where the erosive wear rate, at two different
impingement angles of 15 ° and 90 °, is shown as a function of material hardness for
various metals and grades of steel hardness (K‐H. Zum et al., 1987). The abrasive used
was silicon carbide of diameter about 1 mm impinging at a velocity of 30 m/s. At the
shallow impingement angle, it is evident that the hardness and work‐hardening ability of
materials suppress a quasi‐abrasive process of wear. In this case, materials can be rated
according to the hardness of the pure metal. It can be seen from Figure 2‐12 that, at an
impingement angle of 15 °, the most wear resistant metal is cobalt while the second worst
is copper. When the impingement angle is 90 °, the ranking of materials changes
significantly and copper has the second best while cobalt has the third worst wear
resistance. Heat treatment of steel to increase hardness improves erosive wear resistance
at low impact angles but lessens the erosive wear resistance at high impact angles. To
summarize, the effects of small differences in, for example, hardness or alloy content
between similar materials cannot be viewed in isolation from the overall system
characteristics of erosive wear. In order to define a materialʹs erosive wear resistance it is
only useful to consider broad classes of materials, e.g., polymers, ceramics and metals,
where distinctive differences are present and are not obscured by the effects of variables
such as velocity or impingement angle. There is no general recipe for a high level of
erosive wear resistance. Because of the two different erosive wear protection mechanisms
that can take place, high wear resistance can be achieved by more than one type of

material. In some cases, the material can be extremely hard and tough so that the
impacting particle is unable to make any impression on the surface. This is the approach
adapted when developing metallic or ceramic erosion resistant materials. Alternatively,
the material can be tough but with an extremely low elastic modulus so that the kinetic
energy of the particles is harmlessly dissipated. These contrasting wear protection
mechanisms are illustrated in Figure 2‐13.

Figure 2‐12 Effect of primary material characteristics and erosion parameters on erosive
wear rate (Zum et al., 1987)

Super hard material to resist erosion

Elastic Energy absorption
Figure 2‐13 Comparison of the high and low elastic modulus modes of erosive wear
protection (Stachowiak and Batchelor, 1993).

Rubber is generally believed to provide good erosion resistance by elastic absorption of
particle energy although this has not been demonstrated experimentally. It has been
shown that the first particle impact causes no visible damage to a rubber surface and that
wear depends on slow fatigue processes (Arnold and Hutchings, 1990). Unfilled rubber
shows good erosive wear resistance but surprisingly it is not resistant to abrasive wear
(Arnold and Hutchings, 1990). The choice of erosion resistant material may also be
compromised by other considerations such as operating temperature or material
transparency. Clearly, temperatures in excess of 200 °c preclude polymers from service,
but if a transparent material is required for a specific application then metals are not
particularly useful. For example, materials for aircraft windscreens, apart from being
transparent, are required to be resistant to high‐speed erosion by sand, dust and rain
(Rao and Buckley, 1986). The relative merits and demerits of metals, polymers and
ceramics as erosive wear resistant materials are summarized in Table 2‐2.
Table 2‐2 Relative qualities of erosive wear resistant materials (Batchelor et.al., 1993)
Material Relative qualities regarding erosive wear resistance
Metals Large range of toughness and hardness to suit any particle or impingement
angle. Prone to high temperature corrosion, and softening effects; corrosive
media also harmful.
Ceramics Very hard and increasingly tougher grades available. Resistant to high
temperatures and corrosive media. Poor erosive wear resistance when
brittle mode prevails.
Polymers Tough polymers and rubbers provide good erosion resistance even in
corrosive media. Usage is restricted however by a relatively low

temperature limit.
It was found that the erosive wear of steels shows the classical ductile erosion
characteristic, i.e., a maximum wear rate at a low impingement angle of 30 ° with
subsurface and surface cracking (Levy, 1987). This suggests that the erosive wear
resistance of steels is limited by a lack of ductility. For very soft erosive particles such as
coal, the inclusion of carbides promotes wear resistance slightly (Sargent and Saigal,
1986). Alloying of steel or cast iron to obtain a microstructure containing a significant
amount of retained austenite is an effective means of reducing erosive wear. Bahadur et
al., 1990, adding about 2.5 wt % of silicon to 0.7 wt % carbon steel or about 0.45 wt % of
silicon to 2.54 wt % cast iron results in good erosive wear resistance. The optimum heat
treatment of this steel or cast iron includes a relatively long austempering time where all
the martensite is removed and only retained austenite and bainitic ferrite are present.
Generally, ductility rather than hardness should be enhanced in steels for improved
erosive wear resistance.


2.8 WEAR THEORY
There are many different terms are used to define the wear. Wear can be classified and
characterized in several ways (Thapa, 2004). Several authors give different expressions of
wear and its rate, based upon their wear test results. The wear/erosion rate (mm/year) is
often expressed as functions of the velocity of the particles and proportional to Vn. The
velocity exponent n for pure erosion in between 3 to 4 has most often been reported.
Since wear rate is a function of velocity, this means that the relative importance of erosion
depends on the flow rate. Further, it depends on material hardness, grain size, solids
concentration and temperature. The one most often quoted expression of wear is
(Truscott, 1972),
Wear ∝ (Velocity )
n
Equation 2‐7
Where, the index n may vary depending on the material and other factors involved; the
most common value appears to be 3 (Truscott, 1972).
Some more detailed analyses consider wear as affected by the forces and velocities acting
on a particle in a liquid flow. Bovet, 1958, states that wear is directly proportional to
abrasive power (Pf) of a particle impinging on a surface (mentioned by Truscott, 1972),
and
μ ⋅ V ⋅ ( ρ s − ρ1 ) ⋅ c 3
Pf = Equation 2‐8
Rp
Where, μ = coefficient of friction between particle and surface, V = volume of particle ρs =
density of particle, ρ1 = density of liquid, c = velocity of particle, R = radius of curvature of
surface.
In a much more involved analysis, but starting with the same basic assumption,
Bergeron, 1952, develops a complicated expression based on the statement.
Wear ∝ ( accleration of main flow) ⋅ (coefficient of friction ) ⋅ (thickness of particle layer ) ⋅
( solid − liquid density difference ) ⋅ ( flow velocity )
He thus takes account of the difference between the solid and liquid velocities. He
attempts to predict wear rates in similar pumps handling solids with varying properties,
with simplified assumptions such as pure sliding of the particles over the surface, from
the initial expression:
U 3 ⋅ (P − ρ ) ⋅ d 3 ⋅ p ⋅ K
Wear ∝ Equation 2‐9
D
Where, U = characteristic velocity of liquid, P = density of particles, ρ =density of liquid, d
= diameter of particles (assumed spherical), D = characteristic dimension of machine, p =
no. of particles/unit surface area, and K = experimental coefficient depending on abrasive
nature of particles.

Bitter, 1963, in a fundamental study of erosion phenomena, but strictly for dry conditions,
gives expressions for cutting and deformation wear, also based on energy considerations
and the type of material eroded, i.e., whether brittle or ductile.
A few authors also develop expressions for pump service life. Both Bak, 1966, and
Bergeron, 1952 (Truscott, 1972), consider this in terms of pump total head for given
conditions. Vasiliev, 1970, gives a somewhat involved method, based on statistical
analysis of pump wear tests, to predict life based on a specified maximum permitted
wear.
It is perhaps debatable whether these more complex theories can be used to predict
absolute wear rates with any certainty; most involve empirical constants and other
parameters difficult to determine for an actual machine (Truscott, 1972). In fact,
Bergeronlo, 1952, admits that some of the assumptions made may be questionable.
However, such theories are of some value in predicting likely trends in wear rates when
only one or two of the relevant factors are altered (Truscott, 1972).
When considering a stationery plate as a unit of area with flow normally to its face
surface by a uniform steady liquid stream, it becomes possible to derive the mathematical
relationship representing the mass laws of hydro abrasive erosion. Daun et al., 2002, gave
relation of erosion and kinetic energy. He stated that without regard to the deterioration
pattern, the place erosion developed under the action on its surface of a single solid
particle Eʹ is proportional to the kinetic energy possessed by this moving particle, i.e.
m ⋅ c2
E '=α ⋅ Equation 2‐10
2
where, m is mass of the particle, c is average particle velocity of translation, and α is
coefficient defined by the flow conditions, the material of the particle and the plate, as
well as other factors. The number N of abrasive particles containing with the plate
surface for time interval t can be defined by the expression:
N = β ⋅ε ⋅v ⋅t Equation 2‐11
Where, β is coefficient dependent on the flow conditions around the plate and conveying
capabilities of the flow, v is mean velocity of the flow, ε is the particle concentration. The
plate erosion for time interval t is as follows:
m ⋅ c2
E = E '⋅N = α ⋅ β ⋅ ⋅ ε ⋅ν ⋅ t Equation 2‐12
2
It can be assumed that velocity c of the solid particles suspended in the flows is
proportional to flow speed v, i.e. c = γ ⋅ v , therefore, equation 2.12 will be,
m ⋅ ε ⋅ν 3
E = α ⋅ β ⋅γ 2 ⋅ ⋅t
2
or

E = k ⋅ m ⋅ ε ⋅ν 3 ⋅ t Equation 2‐13

Equation (2.13) shows that the abrasive erosion of stationery component passed by a
liquid flow with solid particles suspended varies in direct proportion to the mass of
particles, their volumetric concentration, the 3rd power of the flow velocity and duration
of the effect exerted by the flow. This is the theoretical evaluation of abrasion intensity. In
practice, a large number of additional factors complicate the erosion of hydraulic
machines. Moreover, there is no extra mathematical dependence, for the time being to
define them.
2.9 EROSION MODELS
Mathematical models of erosion are useful for designing of hydraulic turbine
components, sediment settling basin and optimization of hydropower plant operation in
Sediment‐laden River. Most often, individual particle dynamics are used for developing
erosion models. Empirical and statistical relations are also often developed from
experiments and field experiences. However, studies are heading toward numerical
modelling and simulations, the importance of analytical models are increasing day by
day (Thapa, 2004). Truscott, 1972, has found that the most often quoted expression for
erosion is proportional to (velocity)n.
Extensive research has been done to develop a wear model in terms of the material
properties involved but little attention has been given to clarify the influence of fluid
motion, especially in the turbulent flow regime (Humphry, 1990). Over the years, many
models have been proposed for many different situations. Unfortunately, most models
are co‐relational in nature and therefore system specific. They only work for the
particular material pair, contact geometry, operating condition and range, and the
particular environment and lubricant. The available equations are so confusing that few
designers can use any of them to predict product life with confidence. The complexity of
wear and the large number of parameters affecting the outcome are the primary reasons
for this situation.
However, several fundamental studies of erosion behaviours for different conditions are
found while reviewing the available literatures, only general erosion model and some
erosion models that have been developed for hydraulic machinery applications, is
discussed here.
2.9.1 General erosion model

The actual mechanism of the erosive wear is not yet clearly understood. Hence, simple,
reliable and generalized model for erosion is not yet developed for engineering purpose.
Most common expression for the erosive wear is based on experimental experience and
they are generally expressed as a function of the factors that associated with erosive wear.

The most often quoted expression is Wear ∝ (Velocity)n where index may vary upon the
factors (Thapa and Dahlhaug, 2003). The simplest way of writing equation for erosion is:
Erosion = f (operating condition, properties of particles, properties of base material)
Generally, this expression is given as a function of velocity, material hardness, particle
size, and concentration. Bardal, 1985, describes the most general formula for pure erosion
as,
W = K mat ⋅ K evn ⋅ c ⋅ V n ⋅ f (α ) Equation 2‐14
Here, W is erosion rate (material loss) in mm/year, Kmat is material constant and Kenv is
constant depending on environment, c is concentration of particles and f(a) is function of
impingement angle α. V is the velocity of particle and n is the exponent of velocity.
2.9.2 Erosion models for hydraulic machinery

Many erosion models are developed for specific purpose or condition (Thapa, 2004). For
example, Bitterʹs model is developed for dry condition, hence it is not clear whether this
equation realistically predict erosion rate for wet condition or not. Few researchers have
presented models specifically for hydraulic machinery. Truscott, 1972, presented the
equation of Bergeron, 1952, to predict the erosion rate of pump with simplified
assumptions such as pure sliding of spherical particles over the surface.
⋅ (ρ p − ρ ) ⋅ d 3 ⋅ p ⋅ K
V2
W∝ Equation 2‐15
D
Where, V is the characteristic velocity of liquid, D is the characteristic dimension of the
machine, ρp is density of particle, d is diameter of particle, p is number of particles per
unit surface area, ρ is density of liquid and K is experimental coefficient depending upon
nature of abrasive particles. This equation is depends on experimental coefficient, which
is dependent on abrasive nature of particles.
Daun et al., 2002, established the equation for surface erosion based on impact effect of
particles considering kinetic energy of single particle. They have anticipated deviation on
erosion estimated by equation due to uncertainties like non‐homogeneous particles,
variable concentration, continuous alteration and pulsation of velocities and pressure,
non‐uniform flow distribution and so on. On the contrary, to laboratory tests, Tsuguo,
1999, established the relationship of factors concerning erosion of turbines based on 8
years erosion data of 18 hydropower plants. The repair cycle of turbine is determined
according to calculation of turbine erosion from equation, which gives erosion rate in
term of loss of thickness per unit time (w).
W = λ ⋅ c x ⋅ a y ⋅ k1 ⋅ k 2 ⋅ k 3 ⋅ VChar
z
Equation 2‐16
Where ‘λ’ is turbine coefficient at eroded part; ‘c’ is the concentration of suspended
sediment, and V is the characteristic velocity. The term ‘a’ is average grain size coefficient
based on unit value for grain size 0.05 mm. The terms k1 and k2 are shape and hardness

coefficient of sand particles and k3 is abrasion resistant coefficient of material. The x, y
and z are exponent values for concentration, size coefficient and velocity respectively.
The value of x and y are close to the unity and any deviation of this linear proportionality
is determined from plot of wear versus parameter. The values of z are proposed for
different turbine components based on relation between relative velocity and erosion.
Minimum value of n is proposed as 1.5 for Pelton bucket and maximum value is 3 for
Francis turbine runner. Similarly, for Francis turbine guide vanes and Pelton turbine
needle, this value is proposed as 2.5.
IEC 2008, recommended the following theoretical model of abrasion rate in order to
demonstrate how different critical aspects impact the particle abrasion rate in the turbine.
dS/dt = f(particle velocity, particle concentration, particle physical properties, flow
pattern, turbine material properties, other factors)
However, this formula being of little practical use, several simplifications are introduced.
The first simplification is to consider the several variables as independent as follows:
dS/dt = f(particle velocity) . f(particle concentration) . f(particle physical properties,
turbine material properties) . f(particle physical properties) . f(flow pattern) . f(turbine
material properties) . f(other factors)
This simplification is not proven. Nevertheless, based on literature studies and
experience, it is felt that this simplification is justified for hydraulic machines. Finally,
IEC suggested the following expression with additional simplifications and new variable.
S = W 3 ⋅ PL ⋅ K m ⋅ K f
Equation 2‐17
Where, S is abrasive depth in mm, W is characteristic velocity, PL is particle load which is
obtained by particle concentration integrated over the time, Km is material factor and Kf is
flow factor.

2.10 CONCLUSION
The abrasive and erosive wear was initially thought to consist of one or two relatively
simple mechanisms, but it is now realized that many processes are involved and some of
them are not yet well understood. Most of the common expressions for erosive wear
models for hydraulic machinery are based upon the experimental experience only. The
sediment erosion damage of mechanical components is due to the dynamic action of
sediment flowing along with water, which involves different forces and many other
influencing factors that make the erosion mechanism further complex. More research and
development is needed to understand the actual mechanisms of particle movement and
erosion inside the hydraulic machines.

Chapter 3
Sediment Erosion in Hydraulic
Machinery
This chapter presents the sediment erosion types in hydraulic machinery components
based upon an extensive literature review and the field observation at Cahua
hydropower plant. It includes some recommended methods to minimize the effect of
sediment erosion in turbine components. An alternative design of a Francis turbine in
sediment‐laden water is also briefly discussed.
3.1 INTRODUCTION
Hydraulic turbines mainly divided into two groups: Impulse and Reaction. This
classification is based upon the principle of energy conversion. Pelton and Turgo are
examples of Impulse turbines. Francis, Kaplan and Bulb turbine are examples of reaction
turbines. The cross flow turbines are two stage impulse turbine used for smaller units.
In general, a number of factors influence the development of sediment erosion process of
hydraulic machinery. These factors include mean velocity of particles, mass of the
particle, concentration of the abrasive particles in a liquid flow, grain size and shape of
the particles and angle of attack at which the particles collide with the surface etc. In
practice, a large number of additional factors involved and further complicated the
erosion of hydraulic machinery. In addition, there is no exact mathematical dependence,
for the time being, to define them. Variable concentration and structural in homogeneity
of suspended particles, continuous alternation and pulsation of both velocities and
pressure during the motion of the flow, and variance in operation, and design features of
hydraulic machinery itself, caused the actual pattern of the erosion more complicated
and different (Duan et al.,2002).
Most often, it is difficult to distinguish the exact type of erosion on the hydraulic
machinery. Duan et al., 2002, have described the erosion of hydraulic turbines in six
categories as shown in Table 3‐1 based upon the visual appearance. This classification
could be use to evaluate the hydraulic patterns.

Chapter 3 Sediment Erosion in Hydraulic Machinery
Table 3‐1 Turbine erosion categories (Duan et al., 2002)
S.N. Type Description
1 Metallic luster A Shining surface with no traces of paint, scale or
rust
2 Fine‐scaly erosion A surface with rare, separately located and skin‐
deep minute scales
3 Scale erosion A surface entirely covered with skin‐deep fine scale
4 Large‐sized scaly erosion A surface entirely covered with deep and enlarged
scales
5 In‐depth erosion A surface covered with deep and long channels
6 Through hole or entire Out of the material
erosion
Similarly, B.S. Mann, 1999, collected data on wear patterns of different sediment affected
hydro turbine components at various hydropower stations in India in order to analyse
the wear patterns. This was compared with the data available with some foreign
hydropower stations from Switzerland, Pakistan, and China. It was observed that the
wear patterns have a resemblance and there is a significant relevance to the flow
characteristics of all the power stations. The wear pattern on a particular component is
found to be similar.
As mentioned by Thapa, 2004, Matsumura and Chen, 2002, classified the erosion
condition in Reaction turbines in three categories, as I, II and III, based upon difference in
flow velocity and impingement angle of particle. This classification is shown in Table 3‐2,
which was developed based on erosion test of specimens located at different turbine
components. From this classification, it is not possible to interpret the different type of
erosion on the same component of turbine, for example, the turbine blade, can have
different type of erosion at leading edge and trailing edge.
Table 3‐2 Classification of erosion (Matsumura and Chen, 2002)
Type of erosion Location Flow velocity Impingement angle
I Spiral casing Draft tube Low Small
II Runner blade Guide vane High Small
III Wearing ring High Large due to vortex
and turbulence
Brekke, 2002, classified the sediment erosion in hydraulic machinery into three different
categories, namely, micro erosion, secondary flow vertex erosion, and acceleration
erosion.
Micro erosion is found on the surface of turbine components where fine particles with
grains size less than 60 μm are moving at very high velocity. High shear stress in the
boundary layers gives high rotational motion to these particles causing several ripples in

Sediment Erosion in Hydraulic Machinery Chapter 3
the direction of flow. The patterns with such erosion are also compared as fish scale or
orange peel. Such type of erosion can appear in guide vane and runner blade in a Francis
turbine towards outlet and in the needle of a Pelton turbine.
Obstacles in the flow field or secondary flow in the corners of conduits causes secondary
flow vortex erosion. Any obstruction in the flow field causes secondary flow and
horseshoe vortex is generated around the cylindrical obstacles like guide vane leading
edge. Similarly the needle of the Pelton wheel have vortex behind the ribs supporting
needle and hence vortex erosion takes in the straight line downstairs the ribs. The
vortices in the corner of conduits like guide vanes‐facing plates and blades‐band also
cause this type of erosion. Such vortices and secondary flow are caused by combined
effect of boundary layer and change in flow acceleration. The design of hydraulic
machinery working in the range of high Reynolds numbers, (106 ‐ 108) will normally be
exposed all three types of erosion.
3.2 IMPULSE TURBINE: PELTON
Generally, Pelton turbines are designed for low speed number, range from 0.1 to 0.2, the
velocity in the jets will be higher than 100 m/s. The acceleration of the particles in the
buckets will normally be more than 50,000 m/s2, which depends on the size of the buckets
and head of the turbine. The high velocity and acceleration of particles at the buckets are
main reasons for the sediment erosion. Brekke, 2002, categorized the Pelton turbine
components into four groups in order to study the sediment erosion phenomenon. Those
are inlet system, nozzle system, turbine runner and the wheel pit.
3.2.1 Inlet system
Inlet system consists of manifold and valve. The velocity at inlet system is normally
maintained low. Brekke, 2002, provided the following velocity relations in order to
design the inlet system of the Pelton turbines.
At inlet manifold, C = k i ⋅ 2 ⋅ g ⋅ h [0.08 < ki > 0.1] Equation 3‐1
At valve, C = k v ⋅ 2 ⋅ g ⋅ h [0.095 < kv > 0.12] Equation 3‐2
At nozzle, C = k n ⋅ 2 ⋅ g ⋅ h [kn > 0.99] Equation 3‐3
Because of the low operating velocity at the inlet, the inlet system or pipes will have only
moderate effect of sediment. Hence, application of high erosion resistance rubber, and

epoxy based coating or paint may prevent erosion in such pipes. Slight erosion of this
system does not affect the performance of the turbine by any means, but severe erosion in
bifurcation and bends is likely to increase the leakage and in worst‐case rupture of pipe.
Regular inspection and maintenance should be carried out to prevent any catastrophic
effect.
3.2.2 Nozzle system

Nozzle system includes nozzle ring and needle. High head Pelton turbine, for example,
1,200 m head can have jet velocity up to 150 m/s. Such a high velocity can damage both
the nozzle and the needle. The flow is accelerated from the outlet of fins supporting
needles up to nozzle tip and all the pressure energy available in the water is converted
into kinetic energy. High velocity combined with the needle geometry creates strong
turbulence in the boundary layer close to needle tip. The fine particles bombarding due to
turbulence strikes the needle surface several times and severe erosion can be seen in short
time. Cavitation can follow in a short interval of time and severe damage of the needle
can take place. The nozzle tips are relatively sharp and less than 1 mm contact between
nozzle and needle are maintained to reduce cavitation damage and to obtain highest
possible efficiency. This makes the nozzle tip vulnerable to sand erosion. If nozzle
diameter increases by 5 % due to erosion at tip, the turbine will run at 10 % load even if
needle is at closed position. It may affect entire control system of the power plant.

The photographs presented in Figure3‐1 illustrate the extent of sediment erosion in
Pelton turbine nozzle and needle. These photographs were taken from the lecturer notes
provided by professor Ole Gunnar Dahlhaug, NTNU.

(a) Nozzle needle, Mel, Norway (b) Nozzle ring, Andhi Khola, Nepal
Figure 3‐1 Sediment erosion at Pelton turbine nozzle and needle

The protection of needle and nozzle surface by applying ceramic‐metallic coatings may
help to improve erosion resistance. Ceramic coating is not very effective in case of larger
size particles. There is very little scope to improve the erosion resistance of needle and
nozzle by hydraulic design, but maintainability can be improved by designing
replaceable nozzles tip.
3.2.3 Pelton turbine runner
Peltion turbine runner consists of splitter, bucket tip and bucket surface. Sediment
erosion can be found in all components however, the nature of erosion is different. In
high head Pelton turbines, the absolute acceleration range normal to the surface could be
50,000 ‐ 100,000 m/s2. Such a high acceleration is the main reason of sediment erosion in
turbine buckets, which has a strong effect on separation of particles from streamline. The
characteristics of damage due to fine and coarse sediment are different. With coarse
particles, most of the damages are in the area where the jet directly hits at the bucket
surface. Surface damage is observed due to the hammering action and not due to the
cutting action by sharp edge. Long scars are also seen in the flow direction in each side of
the bucket splitter but no damage is observed at the root of the bucket. Splitters and
entrance lips are most severely damaged portion of the buckets, because of direct hitting
of particles. The photographs presented in Figure 3‐2 illustrate the extent of sediment
erosion in Pelton turbine buckets and splitter.

(a) Bucket surface, Khimti, Nepal (b) Splitter, Rangjung, Bhutan
Figure 3‐2 Sediment erosion at Runner buckets
The acceleration of particles normal to the flow direction separates the particles from the
flow direction and such accelerating particle strikes the surface causing collision in the
water conduit surface. Large particles, for instance higher than 0.5 mm, cause severe
damage in the Pelton turbine bucket.

Fine particles may glide along with water inside the bucket and strike the surface toward
outlet edge, causing severe erosion around the outlet. Due to distortion of bucket profiles
near the outlet, but not at the edge where the acceleration is zero, the direction of flow
changes bending inward and strike backside of following bucket with braking effect. This
phenomenon is schematically explained in Figure 3‐3
Figure 3‐3 Illustration of separation of particle in a Pelton bucket (Thapa and Brekke,
2004)
Thapa and Brekke, 2004, have drawn some conclusions based upon the different
hydropower plants erosion patterns observations:
• If the particles are fine (silts), then there will be erosion on the needle but not
much erosion in the buckets
• If the particles are coarse (sand), then there will be erosion in the buckets and
there is less erosion of needles
• With medium size particles, both needle and bucket will be eroded

3.2.4 Criteria for Pelton turbine design
Sediment erosion problem cannot be solved completely by hydraulic design alone
however, this can be minimised to some extent. The three basic criteria for the design of
Pelton turbine to minimize the effect of sand erosion are as follows (Brekke, 2002).

• The radius of curvature should be as large as possible at the location where flow
direction changes
• The number of jets should be as low as possible
• The hydraulic radius of the bucket and nozzle size should be large. This brings
minimum sand particles in contact with the surface

3.3 REACTION TURBINE: FRANCIS
Generally, Francis turbines are designed for speed number, range from 0.2 to 1.5. High
head Francis turbines are mostly affected by sediment erosion. For the high head type of
Francis turbine, approximately 50 % of energy is converted in to kinetic energy at guide
vane and remaining 50 % is retained as pressure energy. The inlet velocity could reach up
to 85 ‐ 95 m/s on high head turbines. Hence, guide vane faces high absolute velocity. The
runner outlet has highest relative velocity. The velocity at the runner outlet is normally
selected around 40 m/s during design to ensure flow of water out of the turbine. Brekke,
2002, categorized the Francis turbine components in to four groups in order to study the
sediment erosion phenomenon. Those are inlet system, guide vanes system, runners and
labyrinth seals, draft tubes, and shaft seals.

3.3.1 Inlet system: Stay vane

Inlet system of Francis turbine consists of manifold, valves, bypass system, spiral casing
and stay vanes. Compared to inlet valve of Pelton turbine, its inlet valve will face 50 %
less pressure during closing due to pressure created by spinning runner. Hence inlet
valve of Francis turbine have rubber seal against movable steel seal, which has better
erosion resistance. It is important to make bypass system larger in order to create higher
possible pressure in the spiral casing before opening the valve seals because lower
pressure in the spiral casing during opening will increase the damage of the valve seals
and bypass valve seals. Hence, bypass system in the Francis turbine has to be stronger
than Pelton turbine to create higher pressure in spiral casing. The starting pressure in the
guide vane system gives indication of leakage due to sand erosion. The velocity at spiral
casing is higher than Pelton turbine manifold because of shorter distance between inlet
valve and guide vanes. The velocity at the inlet of the spiral casing is almost same as the
meridional velocity at the runner outlet and inlet of valve.

The stay vanes have the main purpose of keeping the spiral casing together. The
dimensions have to be given due to the stresses in the stay vanes. The vanes are designed
so that the flow is not disturbed, and they direct the flow into the turbine. Because of the
secondary flow in the spiral casing, an incorrect flow angle towards the top and the
bottom region of inlet of stay vanes in traditional design often cause secondary flow
erosion in high head turbines. Similar phenomenon has been observed at Cahua power

plant, where paint and material are removed due to the erosion at the stay vane inlet as
shown in Figure 3‐4 (a).

(a) Inlet of stay vane (b) Typical ring grooves near leading edge

Figure 3‐4 Erosion at stay vane at Cahua power plant
The corrosion followed by removal of paint accelerates the erosion rate. However,
modern parallel stay rings reduces incorrect flow and minimizes erosion at stay vane
inlet. Under normal condition, erosion resistant paints can be used in spiral casing and
stay vane, but for high head turbines, stainless steel stay vanes can be used to reduce the
effect of sediment erosion. The mid height of the leading edge part is exposed to less
erosion intensity. Near the upper and lower cover at the inlet, ring‐shaped erosion
grooves, as shown in Figure 3‐4 (b), were observed (Mette Eltvik, 2009).

3.3.2 Guide vane system

The guide vane system is highly affected by sediment erosion due to highest absolute
velocity and acceleration. For high head turbines, the relative velocity head (C2/2g)/Hn for
guide vane increases from 10 % at guide vane inlet to 50 % at runner inlet. At normal
speed, pressure drop across the guide vane will be approximately 40 % of net head at full
load and 50 % at small opening, which is one of the reason for cross flow and hence
erosion takes place at the junction of guide vanes and facing plates. The main reason for
cross flow is the pressure difference. The erosion of guide vane due to sand laden water
can be classified in following four categories (Brekke, 2002):
• Turbulence erosion at the outlet region and facing plate due to high velocity of
fine grain sand
• Secondary flow erosion in the corner between guide vane and facing plates due
to fine and medium size particles, which makes horseshoe grooves in the facing
plates following contours of guide vanes
• Leakage erosion at the clearance between guide vane and facing plate due to
local separation and turbulence increasing the horseshoe vortex in the suction

side. The leakage also causes local separation and turbulence at the pressure side
at inlet and suction side at outlet of guide vanes causing even a deep groove at
the bottom and top of the guide vanes
• Acceleration erosion is caused by separation of large particles from the
streamlines of main flow due to rotation of water in front of the runner. This
acceleration of particle is normal to the streamline and strikes guide vane surface
causing severe erosion. This acceleration also creates secondary flow causing
erosion at the corner between the guide vanes and the facing plates by fine
particles
In Cahua power plant, as shown in Figure 3‐5 (a), more erosion is observed near the
transitional zone to the lower cover, due to the high acceleration and absolute velocity in
the guide vane cascade. The acceleration erosion in the guide vane in reaction turbine can
be reduced by designing the flow with smoothest possible acceleration. The stay vane
outlet angle should be carefully chosen so that guide vane will be in neutral position in
normal operation condition. Similarly, reduction of clearance between guide vane and
facing plate avoid cross flow and secondary flow. Metal sealing are used to reduce gap
between guide vane and facing plates with the intention to improve efficiency, but this
could be more destructive once damage of such seal commence. The turbulence and
secondary flow create dangerous galling in facing plates. It destroys flow pattern and
reduces turbine efficiency. The covers will be exposed to erosion since the acceleration
normal to the streamline creates secondary flow, especially at the corners between the
facing plates and guide vane as clearly seen in Figure 3‐5 (b). This effect occurs because of
the horseshoe vortex and heavy erosion grooves are observed.

(a) Eroded guide vanes (b) Horseshoes vortex of facing plates
Figure 3‐5 Erosion at guide vane and facing plates at Cahua power plant

The guide vane clearance for new turbine is recommended roughly 0.1 ‐ 0.3 mm in
pressurized condition, which is dependent on deflection of the head cover. Low initial
dry clearance in the order of 0.05 to 0.1 mm may give low clearance in pressurized
condition, but such a low dry clearance may cause abrasion and adhesion between guide
vane and facing plates. The facing plates can be improved by cladding underneath the

guide vanes, but the difference in hardness between guide vanes and facing plates should
be maintained to avoid galling of the surface. The hardness of 16Cr5Ni Guide vanes with
350‐400 HB and facing plate of 17Cr1Ni with 300 HB is an example of appropriate
combination, which avoids galling and abrasion. The necessary tolerance and surface
finish at the mating part of guide vane and facing plate should also be maintained to get
rid of these problems. The maintainability of the facing plate is often improved by
designing replaceable layer to save maintenance time and expensive bulk material.

3.3.3 Runner
In the runner, the highest relative velocity occurs at outlet region while the highest
absolute velocity and accelerations may be found at the inlet of the blade. Hence, impact
due to kinetic energy is small compared to force exerted by large accelerating particle.
Contrary to this, the relative velocity is the highest at the outlet of the runner blade.
Hence, turbulence erosion due to fine sand is always susceptible at the trailing edge of
the blade. Also because of high relative velocity, most of the particles will move towards
outer diameter in the runner outlet and hence more effect of erosion is seen there. Inlet
region of the runner is sensitive to incorrect pressure distribution between pressure and
suction side and any separation caused by this may cause severe local erosion at the inlet
due to fine grain sand. Cross flow from the hub to the shroud caused by incorrect blade
leaning will also increase the so‐called horseshoe vortex in the blade roots. The erosion at
runner outlet of Cahua power plant is shown in Figure 3‐6.

(a)Erosion at runner outlet (b) Erosion at pressure side of blade
Figure 3‐6 Erosion at runner at Cahua power plant

The improvement of blade leaning and correct blade loading at inlet may improve
performance of Francis turbine against erosion. The splitter blades at the inlet of runner
help to reduce damage of flow around leading edge at off design operation, which
ultimately improves the resistance to sand erosion and cavitation. Incorrect blade leaning
may lead to cross flow between hub to band and such cross flow may intensify erosion
effect together with other loss associated with it.

3.3.4 Labyrinth seals

The velocity in labyrinth is in order of 45 m/s for high head turbine. Labyrinth seals
having small clearance and working with coarse sand may also have erosion as well as
abrasion effect. The efficiency of a labyrinth seal is inversely proportional to clearance
gap. Hence, gaps should be made the smallest possible to have minimum leakage, but it
has to be sufficiently large to avoid any direct contact between rotating and stationary
part. The clearance between the stationary and rotating parts is varied between 0.25 to 1.0
mm depending on the size of the turbine. The turbulence erosion due to fine sand is
always susceptible in the labyrinth seal because of high velocity in its surrounding. The
rotating labyrinth seals are made up of steels to achieve longer life, while stationary seals
are normally made of softer materials such as bronze. The replacement cost for the
stationary seal is cheaper. However, in the case of high sediment concentration, even the
stationary seals can be made up of steel, which has higher erosion resistance than softer
material.

3.3.5 Draft tube

The draft tube section closer to runner will be exposed to the highest velocity because of
high absolute velocity of water coming out of runner and some sediment erosion effect
can be anticipated. Apart from this portion, sediment erosion is normally no problem in
draft tube.

3.3.6 Shaft seal

Mechanical shaft seal will be damaged if the sand‐laden water is exposed to it. The
damage could be due to more abrasion than erosion. Shaft seals are normally made up of
carbon rings, bronze and steels with lower hardness compared to shaft material. Apart
from selection of erosion and abrasion resistance material for seals, insertion of
pressurized clean water in the shaft seal also avoids wear.
3.4 TURBINE DESIGN
The following IEC guidelines explain some recommended methods to minimize particle
abrasion and the effects thereof, by modifications to design for clean water. It should be
understood that every hydraulic power plant is a compromise between several
requirements. While it is possible to design a unit to be more resistant against particle
abrasion, this may adversely, affect other aspects of the turbine. Some examples are:


• Thicker runner blades may result in decreased efficiency and increased risk of
vibrations from von Karman vortices.
• Fewer runner blades (in order to improve the access to the blade surfaces for
thermal spray surface treatment) may result in reduced cavitation performance.
• Abrasion resistant coatings may initially result in increased surface roughness,
which may reduce the efficiency.
• Reduced runner blade overhang may result in reduced cavitation performance,
which in turn may reduce the output that can be achieved for a turbine upgrade.
• Many abrasion resistance design features will increase the total cost of the power
plant.
The optimum combination of abrasion resistant design features must be considered and
selected for each site based on its specific conditions.

3.4.1 Hydraulic design of turbine
3.4.1.1 Selection of type of machine

It is advantageous to select a type of machine that has low water velocity that can easily
be serviced and that can easily be coated with abrasion resistant coatings. Some general
guidelines are; in the choice between a vertical shaft Kaplan and a Bulb; the Kaplan will
normally have lower velocity. The serviceability and ease of coating is approximately
equal between the two. In the choice between a Kaplan and a Francis, the Francis will
normally have lower velocity. On the other hand, the Kaplan runner has better access for
applying abrasion resistant coatings. The serviceability is approximately equal between
the two. In the choice between a Francis and a Pelton, the Francis will normally have
lower maximum velocity. However, the parts in a Pelton turbine that are subject to the
maximum velocity (i.e. the needle tips and seat rings) are small and have better access for
applying abrasion resistant coatings. The Pelton turbine is also easier to service.

3.4.1.2 Specific speed

For the same plant, lower specific speed machines are normally bigger and have lower
water velocities in the runner outlet. However, the water velocities are not lower in the
guide vanes and in the runner inlet. For Kaplan, Bulb and low head Francis turbines,
most of the abrasion damage will be in the runner, so the specific speed is important. For
high head Francis turbines, much of the abrasion damage will be in the guide vane
apparatus, so the specific speed is not so important. For Pelton turbines, the water
velocity does not depend on the specific speed. However, a lower number of jets is
beneficial for a Pelton turbine since the buckets will be larger which in turn gives less
water acceleration in the buckets and thus less abrasion damage. A lower number of jets
will automatically result in a lower specific speed.

3.4.1.3 Variable speed

Even though variable speed machines are not frequent, they are less prone to cavitation,
even under a wide head range operation. Due to this characteristic, the variable speed
machine may better resist particle abrasion.

3.4.1.4 Turbine submergence

Cavitation and abrasion will mutually reinforce each other. For this reason, it is
recommended that the turbine submergence is higher for plants where abrasion is
expected.

3.4.1.5 Runner blade overhang

In case of Francis runner refurbishment, there is sometimes a need to increase the turbine
output significantly. One way to do that is to extend the runner band inside the draft tube
cone, in order to increase the blade area and to improve the cavitation performance.
However, this creates additional turbulences at the top of draft tube cone that will
increase metal removal if particles are present in water. A secondary effect of the
overhang blades is to create a lower pressure zone downstream of the runner band seal,
thus creating higher seal leakage and more particle abrasion at the band seal.

3.4.1.6 Thicker runner blades and guide vanes

Increased runner blade thickness, particularly at the outflow edge, gives some extra
margin before the removal of material on the runner blades becomes critical for the
structural integrity of the runner. A thicker blade design should be done with care. The
thicker blade may result in decreased efficiency and increased risk of vibrations from von
Karman vortices. In addition, the risk of cavitation damage on the runner band,
downstream of the blade, may increase.

3.4.2 Mechanical design of turbine

If abrasion is expected and the turbine type is defined, not only the hydraulic design but
also the mechanical design can take some precautions to reduce the abrasion rate and to
allow easy maintenance or replacement of the abraded parts. Generally, for the design,
the area exposed to the abrasive wear should be as small as possible. As well,
discontinuities and sharp transitions or direction change of the flow should be avoided.
The turbine shaft seal should have rubber rigs in place of carbon rings and its design
should be such as to allow dismantling and replacement in the shortest possible time. The
thickness of the runner blade in the area prone to erosion should be increased. These
areas are mainly at runner outlet. Increasing the wall thickness is a method to increase
the overhaul interval of a component due to abrasion. For structural components, which

do not influence the efficiency, the wall thickness can be increased in critical areas to
avoid early failure of the component due to higher stresses.

The material selection for components, which are subject to abrasive wear, is another
important criterion. Generally, weldable stainless steel materials are preferred. If both
corrosive and abrasive attack anticipated then stainless steel is preferred. Considering the
larger hardness, martensitic steel is preferred over austenitic steel. Furthermore, shaft seal
with clean sealing water is recommended. Normally, shaft seals in units, which are
operated with water and, which contains abrasive particles, have to be fed with clean
sealing water. It must be avoided that the contact surface or the wearing surface get
contact with the abrasive particles.

3.4.3 Operation of turbine

Operation of a turbine at part load and full load is marked by reduced efficiency,
increased flow turbulence and higher relative flow velocities at turbine runner outlet.
Presence of secondary flows and accompanying vortices lead to increased local velocities.
Since such flow conditions are conducive to increased sediment erosion, attempt should
be made to avoid such operations when water carries excessive sediment.
The following actions are recommended for consideration during operation of the units.

• Shut down units at higher particle concentration periods. This may avoid considerable
wear on the unit for a small amount of lost production. Especially for run of river
schemes, where large variation in particle concentration can happen very fast this
strategy can be useful.

• Minimize amount of debris passing through unit. Large solid items, for example logs,
gravel (larger than 2 mm), etc. may damage the hydraulic surfaces and any abrasion
resistant coatings. Damage to hydraulic surfaces may increase the turbulence of the
flow, which will increase the abrasion damage. This is especially important for high
head Francis and Pelton units, since the water velocities are very high and these units
rely on smooth hydraulic surfaces to keep the turbulence low.

• Do not operate the unit in case the abrasion damage jeopardizes the safety of
operation. As the abrasion damage progresses, the unit will eventually become unsafe
to operate. This could for example be due to seal leakage increasing so much that the
axial thrust exceeds allowable limits or that the remaining material thickness of some
component falls below acceptable minimum thickness. Regular inspections of critical
components should be made at least every year and inspection results must be
compared with predefined acceptance criteria.

• Avoid low load operation as much as possible. Low load operations are the worst
operating conditions with respect to abrasion for most components and turbine types.

• Close inlet valve at shutdown. With a turbine at standstill and the water shut off only
by the guide vanes, the water leaking past the guide vane clearances will have very
high velocity, close to the free spouting velocity. This will cause abrasion wear in the
guide vane apparatus. By closing the inlet valve, this abrasion is eliminated. To close
the inlet valve is especially important for high head units.

• Hard coatings are very sensitive to cavitation. Thus, in machines with such coatings
all operating conditions that lead to cavitation must be avoided and strictly stick to
the recommended operating range for the turbine.

3.5 ALTERNATIVE DESIGN OF FRANCIS TURBINE
A few instances have been noticed where under identical conditions of sediment, the
intensity of damages at different hydropower stations were not identical. While
components at a particular power station eroded very fast, damages to components at
other power stations were insignificant. This leads one to believe that equipment design
has a role to play in influencing the intensity of erosion (Naidu, 1999).

The author earlier designed a Francis turbine for sediment‐laden water that considering
the erosion of the Francis turbine occurs mainly at the outlet of the guide vanes and at the
outlet of the runner blades. In order to reduce the erosion rate of the turbine, the absolute
velocity at the inlet of the runner and the relative velocity at the outlet of the runner have
to be reduced. In this study, the flow and head were kept constant while the speed, inlet
peripheral velocity and outlet runner blade angle were changed according to Table 3‐3.

Table 3‐3 Variable input parameters
Speed rpm n 750 600 500 433 375 333 300 275
Inlet peripheral ‐ U 1 0.71 0.74 0.77 0.8 0.83 0.86 0.89 0.92
Velocity, reduced
Outlet blade angle degree β 2 17 19 21 23 25 27 29 31
The results show that the outlet diameter changes relatively little while the inlet diameter
changes drastically. The reduction of the erosion at the outlet is more than at the inlet.
This is shown in Figure 3‐7. The inlet angle of the turbine has changed so that the design
looks more like a pump‐turbine. This means that the turbine will be larger than the
traditional design.

The reduction of the erosion is linked to the reduction of the velocity and therefore the
size of the turbine increases. This result in a higher price of the turbine, but it will reduce
the maintenance costs during its lifetime. It has been shown from the calculation that the
design of the runner can decrease the sand erosion but it is not possible to avoid this
problem completely by design alone.

However, if a Francis turbine designer combines the hydraulic design and coating of the
critical parts, a significant reduction of erosion can be achieved.

Figure 3‐7 Variation of diameter for reduction of erosion

3.6 CONCLUSION
Sediment content of water can no more be overlooked in any phase of hydropower
project implementation. This includes all the phases like, investigation, design, operation
and maintenance, and refurbishment and upgrading. Due consideration of the problem
at every stage, would effect economies on one hand and long‐term solutions would
emerge on the other hand. However, one solution in order to decrease the sediment
erosion is to increase the size of the turbine, thereby increases the hydraulic radius of
curvature, and thus decreases the accelerations. Furthermore, while it is possible to
design a Francis turbine to be more resistant against sediment erosion, this may
adversely, affect other aspects of the turbine. It should be understood that every
hydraulic turbine is a compromise between several requirements.


Chapter 4
Particle Velocity Measurement in
Swirl Flow, Laboratory Studies
This chapter presents the laboratory studies of particle velocity measurement in highly
swirl conditions similar to turbine flow in curved path. It includes a brief description of
the developed test rig, concept of critical diameter of particle inside a Francis turbine and
experimental analysis.
4.1 BACKGROUND
4.1.1 General
Swirl flow can be used as a mechanism for separation of particles from the fluids creating
a centrifugal force as in many separation processes. Solid particles of different shapes and
sizes play a significant role in many separation processes. The separation of particles of
different shapes and sizes depends upon the variations in behaviour of the particles
when subjected to the action of moving fluid. A particle falling in an infinite fluid under
the influence of gravity will accelerate until the resistance force that includes buoyancy
and drag exactly balances the gravitational force. The constant velocity reached at that
stage is called the terminal velocity. The resistive drag force depends upon an
experimentally determined drag coefficient.

The separation of particles from streamline depends upon acceleration of particles, which
further depends on profile and curvatures of runner blade and bucket. Together with
several parameters affecting erosion rate (discussed in section 2.7), the modification in
design in term of turbine size, profile and curvatures also play a vital role for reducing
erosion rate. However, this aspect is not explored at large extent (Thapa, 2004).

There are very less literatures available for the separation of particles from streamline,
which can directly apply to hydraulic turbines applications. However, some studies have
been found in the field of particle transport and separation in erosion test rig together
with cyclone separator and conveying of particles in process industry.

Chapter 4 Laboratory Studies
Chevallier and Vannes, 1995, carried out numerical and experimental study of interaction
between particle and specimen. They studied the particle speed in non‐uniform flow by
generalizing Basset, Boussinesq and Oseen’s expression that gives equilibrium of the
particle. The added weight effect, effect of static and viscous pressure, Archimedes thrust
and gravity force are neglected. When the ratio ρ/ρp is low (<10‐3), and if the particle and
fluid acceleration are of the same order, those effects become negligible. After all these
simplifications, the equation of motion of particles becomes,

π ⋅d3 π
⋅ d p2 ⋅ ρ ⋅ C D ⋅ (V − C ) ⋅ V p − C
dV
⋅ρp ⋅ = Equation 4‐1
6 dt 8

Here, dp is the diameter of the particle, ρp is the density of the particle, ρ is density of the
fluid. The fluid velocity is C, and particle velocity is V. The coefficient of drag (CD) is
obtained from equation (Martin, 2008),

CD =
24
Re , p
(
⋅ 1 + 0.15 ⋅ Re, p
0.687
) Equation 4‐2

Similarly, Tabakoff et al. ,1991, used simplified governing equation of force of interaction
of particle motion in the turbo‐machinery flow with reference to cylindrical coordinates
relative to frame of reference, fixed with respect to the rotating blades as shown below,

1
3 ρ C D ⎡⎛ d (rp ⋅ θ p ) ⎞ ⎛ dz p ⎞ ⎤ 2
2 2 2
F= ⋅ ⋅ ⋅ ⎢⎜Vr −
4 ρ p d ⎢⎜⎝
drp ⎞ ⎛
⎟ + ⎜Vθ −
dt ⎟⎠ ⎜⎝
⎟⎟ + ⎜⎜V z − ⎟⎟ ⎥ ⋅ V − V p ( )
⎣
dt ⎠ ⎝ dt ⎠ ⎥⎦

Equation 4‐3

Here, rp, θp and zp define the particle location in cylindrical coordinates. Similarly, Vr, Vθ
and Vz represent relative gas velocities in the radial, the circumferential and the axial
directions respectively. This equation includes centrifugal force as well as Coriolis force.
The forces due to gravity and inter particle interaction are negligible in the case of turbo
machinery. The drag coefficient (CD) is a function of particle Reynolds number. Deng et
al., 2001, considered, CD = 0.44 for large Re,p (> 500), and provided equation 4‐4 for CD for
Re,p range, 0.2 – 500, which is almost same that of the values and the Reynolds number
rages provided by Martin, 2008.

⎛ −1 − ⎞
1
C D = 24 ⋅ ⎜ Re , p + 0.167 ⋅ Re , p 3 ⎟ Equation 4‐4
⎝ ⎠

Laboratory Studies Chapter 5
4.1.2 Characterization of non spherical particles

Particles in water may range in size from a few nanometres to few millimetres in
dimension. Natural particles also have various shapes, including rod, plate, and sphere,
with many variations in between, which make the treatment of the particle size difficult.
The discussion is vastly simplified if the particles are considered spherical. In this case
only one size parameter is needed (the diameter) and hydrodynamic properties are much
more easily treated. Of course, non‐spherical particles also occur in natural water, and
some way of characterizing them is essential. A common concept is that of the equivalent
sphere based on a chosen property of the particles (Gregory, 2006). For instance, an
irregular particle has a certain surface area and the equivalent sphere could be chosen
with the same surface area. The surface area of sphere with diameter d is π ⋅ d . So, if
2
the surface area of the non‐spherical particle is known, the equivalent spherical diameter
can easily be calculated. For an object of a given volume, the sphere has minimum surface
area and the volume (or mass) of a given particle must be equal to or less than that of the
sphere. Another common definition of equivalent spherical diameter is based on
sedimentation velocity. In this case, from the sedimentation velocity and density of
particle, the diameter of a sphere of the same material that would settle at the same rate
can be calculated. This is sometimes called the Stokes equivalent diameter.

4.2 OBJECTIVE OF EXPERIMENT
The objectives of development of experimental set‐up and conducting experiment are as
follows:
• To study the particle separation process in the flow in the curved path
• To study the forces on the particle on rotational motion or swirl
• To investigate the velocity and the drag coefficient relations based upon different
shape and size of the particle in swirl flow
• To establish an operating strategy for Francis turbine operating on sediment‐
laden water

4.2.1 Description of test rig and test procedure

A previously made test rig (Thapa, 2004), was reviewed and modified to create a strong
swirl flow in curved path. This flow was found similar to the flow between the guide
vane outlet and the runner inlet of a Francis turbine. The flow in the guide vane cascade
was simulated in order to verify the particle separation process and to investigate the
velocity and the drag coefficient relations based upon different shape and size of the
particle. There was a provision to introduce particles, with sizes ranging from 1 to 10 mm,

into the swirl, and to observe the motion of the particles from Plexiglas windows located
on the cover of the tank using a high‐speed digital camera.

(a) Outside view (b) Middle plate with vanes and Plexiglas

Figure 4‐1 Photographs of test rig

The experimental set up, as shown in Figure 4‐1, consisted of a main tank (1,100 mm
diameter and 700 mm height), 400 mm diameter inlet pipe and outlet cone with valve.
The main tank and other components of the test rig were designed and dimensioned for
50 m of head in order to carry out the experiment in high velocity. The main tank
consisted of two compartments with a 250 mm diameter opening at the centre of the
plate. This plate divides the main tank into two compartments. Thirty‐six curved vanes,
which resemble guide vanes of a Francis turbine, with a radius of 100 mm toward the
inlet and 90 mm straight section toward the outlet, were fixed at a pitch circle diameter of
900 mm to the middle plate. This is shown in Figure 4.2. This arrangement was made in
such a way that the inlet velocity direction should be almost in the radial direction and
the outlet would be 10 ° to the tangent. Each vane cascade has 30 mm x 5 mm opening
and altogether 36 numbers of vanes provide the total opening areas be about 5.4 x 10‐6 m2.
These vanes were located in between the upper part of the middle plate and the bottom
part of a 50 mm thick, 950 mm diameter transparent Plexiglas plate. Peripheral
component (Cu) of absolute velocity can be obtained from the orientation of blade angle,
which is fixed at 10 ° in this experimental set‐up. However, this angle varies at different
operating conditions in actual hydropower plant. The peripheral component is caused to
create a centrifugal force of particle whereas the radial component is caused to create a
drag force of particle. From this arrangement, the vanes caused swirl inside the tank,
which further emulate the swirl flow created by guide vanes in a Francis turbine. The
swirl flow in between the Plexiglas and the middle plate could be clearly observed from
the top of the Plexiglas windows, which satisfied a free vortex relation towards vanes
whereas; force vortex flow existed towards the centre of the tank. The edge of the
Plexiglas was made uniform to ensure uniform flow.


The top cover of the tank was fitted with five transparent windows of Plexiglas. Four of
those transparent windows of 160 mm diameter were located at 90 °, 180 °, 270 ° and 360°
respectively at the pitch circle diameter 700 mm to observe the motion of the particles.
The fifth window of 275 mm diameter was located at 315 ° at 700 mm pitch circle
diameter to measure the velocity of the flow (after inserting a Pitot tube as shown in
Figure 4‐3) and to observe the velocity of particle. The particle injection point was located
at 279 mm from the centre of the tank. The particle was released at the bottom of the
Plexiglas through a 15 mm diameter pipe and valves arrangement as shown in exploded
view in Figure 4‐2. Particles up to 10 mm in diameter were tested in this experiment.
Ø200
1 80
67°
°
Ø 10
90
0
0
Ø70
0
Ø16
Ø1
100
Ø275
Plan view
PlexiGlas Windows
Additional box
Pitot Tube
PlexiGlas
303
Vanes
Particle Injector
(<10 mm dia.)
Hole for Pitot Tube 250

350
400
Valve at Close position
Particle Filter
Sectional view

Figure 4‐2 Schematic diagram of experimental set up

To release particle, first lower valve was closed and particle was inserted from the upper
valve. Then upper valve was closed and lower valve was open. Once lower valve is
opened, the particle sinks and drops in between Plexiglas and steel plate. Depending
upon the velocity at the point of the injection, the particle either moves towards outer
radius or follow flow direction heading towards outlet at the centre of the tank and
ultimately sinks. A manometer was fitted into the injection pipe and valves arrangement
in order to measure the inlet pressure of the tank in the swirl flow field. There were two
additional manometers located in the system. One was connected to an air‐bleeding valve
located at 390 mm from the centre of the tank for measuring the inlet pressure, and
another was fitted at the outlet of cone for measuring the pressure at the outlet. The flow
rate was calculated with the help of an ultrasonic flow meter fitted in the inlet pipe.

Figure 4‐3 Pitot tube for measuring the velocity of flow

4.2.2 Measurement of particle velocity

The main purpose of this experiment was to determine the velocity of the particle that
would flow along with water in the given flow condition. To achieve this, firstly, the
operating condition was set at a certain velocity level by controlling the valve opening.
Then the particle was injected and the motion of the particle was observed through the
Plexiglas windows with the help of the high‐speed video camera.

Figure 4‐4 Photograph of middle plate with radial and angular markings


The particle images at different frame rate were saved and analysis was performed based
upon the actual time required to move the particle from one fixed location to another
fixed location. Location of the particle was identified based upon the radial and angular
position of particle inside the test rig. In order to calculate the correct location of the
particle inside the test rig, radial and angular markings were inscribed on the middle
plate at diameters of 219 mm, 239 mm, 259 mm, 279 mm, 299 mm, 309 mm and 22.5
degrees interval respectively. Using these markings, six different circles and sixteen
different angular lines could be clearly observed in the middle plate. Figure 4‐4
illustrated the photograph of middle plate inside the test rig with radial and angular
markings. A particle rotation radius and the angle traversed by the particle are needed to
calculate the radial distance travel by the particle. Then the velocity of a particle was
computed based upon the actual time obtained from the saved images to travel this
radial distance, which is the rate of change of position of particle.

4.2.3 Visualization of particle motion

In order to find out the size of the particle which will flow along with water in a given
flow condition, and to verify the validation of critical diameter expression that derived in
Equation 4.7, six different particles having different densities were considered. The
experiments were done based upon different shape and size of aluminium, brass,
ceramic, plastic, sand and steel particles. Most of these particles rotating at low velocity
can be clearly seen by naked eye. However, as the velocity of particle increase, it was not
possible to visualize the location of particles in flow field by naked eye but the location of
small particles at higher velocity was visualized by the high‐speed video camera. The
high‐speed digital camera was used to compare the rotation radius for different shape
and size of the particles at different operating conditions. The equilibrium condition was
clearly observed with adjustment of flow and sizes of the particles for most of the
samples except for plastic particles. Since the density of plastic particle is less than the
density of water, most of these particles were floating on water. At equilibrium condition,
the rotation radius of particle was observed very close to the injection radius of particle at
279 mm from the centre of the tank. The larger size of the particles that was assumed
larger than critical diameter of particle on that particular operating condition was clearly
observed by hitting the outer wall or suction side of the guide vane cascades in this
experimental set up and the smaller size of the particle was observed clearly by its flow
along with water and ultimately sinking.
4.2.4 Particle in swirl flow

When particles travel in swirling flow toward a turbine’s outlet, which is located in the
centre as shown in Figure 4‐5, the particles will be exposed to two main forces.
Centrifugal force (Fc ) moves the particles away from the centre, while the drag force
(FD ) pulls them toward the outlet, which is in the centre, or toward the runner in the

case of a Francis turbine (Thapa and Brekke, 2004). These two forces are given by the
following equations.
Relative velocity of particle

Orbit for particle rotation
Absolute velocity of water
Tangential velocity of particle

Inlet
vanes
outlet
Outlet
Figure 4‐5 Illustration of particle flow in spiral swirl
π ⋅d3 π ⋅ d 3 Cu
2
Fc = ρ p ⋅ ⋅ r ⋅ω 2 = ρ p ⋅ ⋅
6 6 r Equation 4‐5

1
FD = ⋅ C D ⋅ ρ ⋅ C m ⋅ A p
2
2 Equation 4‐6

Following three conditions prevail in such case:
• Particle will stay at the orbit of radius r, if Fc = FD
• Particle will strike outer wall, if Fc > FD
• Particle will flow along with water towards the centre of the tank, if Fc < FD

At equilibrium, these two forces balance each other and a particle of a given diameter will
stay at an orbit of radius (r ) until either the velocity component is changed or particles
become smaller by fracture due to impact. The diameter of a particle (d ) for the
equilibrium condition is given by Equation 4‐7. This is called the critical diameter.

⎛ ρ ⎞ ⎛ Cm
2
3 ⎞
dc = ⋅ CD ⋅⎜ ⎟⋅⎜ ⎟⎟ ⋅ r
⎜ρ ⎟ ⎜C
4 ⎝ p ⎠ ⎝ u ⎠ Equation 4‐7


The drag force is caused by the relative velocity of particles in radial direction (i.e.
towards the centre of the tank), and the centrifugal force is caused by the velocity of
particle in tangential direction (i.e., away from the centre of the tank).

The phenomenon of particle motion discussed here is similar to swirl flow in between
guide vane outlet and runner inlet of a reaction turbine. Once the larger particles enter in
to the swirl flow it will be rotating and continuously hitting the suction side of the guide
vanes. Direct application of this concept is on operating strategy for Francis turbines
operating in sediment‐laden water. Guide vane position can be manipulated to maintain
a velocity ratio in a way that the particle of given size should flow along with the water.

A damage caused by the larger particles has been observed in Tokke hydropower plant
in Norway. The damage on suction sides of guide vanes is clearly observed in Figure 4‐6.
The bigger particle sizes are settled down or removed before it enters the turbines. The
unsettled bigger size particles, especially in the monsoon season, pass thorough turbines.
The guide vanes of Francis turbine can have maximum angle as low as 12 ° at full load.
Whereas this angle could go up to 40 ° in some of the turbine design.

Figure 4‐6 Erosion damage of suction side of Tokke guide vane by large particles

The guide vane angles for some of Norwegian power plants are presented in Table 4‐1.
This inclination of guide vanes creates swirl flow in Francis turbine and the magnitude of
swirl depends upon angle of inclination of guide vanes. Even though irregular shape of
sand particles could have higher drag coefficient compared to spherical one, the
observation from the test rig can be utilized for determining the size of particle, which
will remain rotating in the swirl flow in between guide vane outlet and runner inlet. Any
particles higher than this size hit the suction side of guide vane continuously and damage
them severely.


Table 4‐1 The guide vane maximum angle at full load condition
S.N. Power plant Specific Speed Maximum Guide vane
(ns) angle at full load (a0)
1 Skjærka 66 12
2 Nedre Vinstra 69 12
3 Hol I 72 13
5 Røssåga 104 18
6 Grønsdal 113 23
7 Nore II 198 34
8 Dynjafoss 208 27.5
9 Oltesvik 264 38.5
10 Iverland 269 31.5
11 Fiskumfoss 308 40.5
12 Fiskumfoss 308 36.5
13 Gravfoss 346 37
14 Solbergfoss 365 38
Particles larger than critical diameter will remain rotating in the swirl flow and hitting
the guide vane wall but the particles smaller than critical size flows through turbines. The
relation between critical diameter and runner inlet diameter is shown in Figure 4‐7 a. This
relation was obtained based upon drag coefficient of 0.1. At higher Reynolds number (Re
> 106), the drag coefficient is around 0.1 to 0.2. Similarly, the relation between particle size
and drag coefficient for turbine of radius 1 m at inlet is shown in Figure 4‐7 b. This figure
indicates that the sand particles of diameter higher than 0.9 mm will stay rotating in the
swirl flow and damage guide vanes positioned around 10 °. This figure also shows that
smaller turbines are more prone to sand erosion because smaller turbines are having
small critical diameter, for example particles as small as 0.7 mm may remain rotating in
the turbines of inlet diameter 800 mm.
Critical diameter of particle Drag coefficient on Critical diameter

10 18
Critical diameter of particle [mm]
15
Diameter of particle [mm]
R1000 12
6 R800
Drag coefficient 0.1
R600 9
Drag coefficient 0.2
4 R400
R200 6
2
3
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Guide Vane angle [degree] Guide Vane angle [degree]

(a) Size of runner (b) Drag coefficient

Figure 4‐7 Critical diameter relation based upon size of runner and drag coefficient

This result can be utilized as guidelines for operating strategy for Francis turbine
operating in sediment‐laden water. If the particle size flowing with water is larger than
critical sizes of the particle, the turbine should not be operated at low guide vane
opening.
4.2.5 Drag coefficient for particles

Drag coefficients and terminal velocities are important design parameters in many
separation processes. Many equations have been developed and presented in literature
relating the drag coefficient (CD) to the Reynolds number (Re) for particles of spherical
shape falling at their terminal velocities. These correlations are of varying complexity and
contain many arbitrary constants. Many of these correlations are listed in Gabitto and
Tsouris, 2008. However, in the case of non‐spherical particles, less information is found in
literature. Gabitto and Tsouris, 2008, also mentioned that Heywood developed an
approximate method for calculating the terminal velocity of non‐spherical particle or for
calculating its size from its terminal velocity. The method was an adaptation of his
method for spheres. Several correlations have been proposed for C D over the entire
range, the one presented in Equation 4‐8 is that of Haider et al., 1989, which is claimed to
fit the data with a root mean square deviation of 0.024.

⎛ ⎞
⎜ ⎟
CD =
24
Re , p
(
⋅ 1 + 0.1806 ⋅ Re, p
0.6459
) ⎜ 0.4251 ⎟
+⎜
6880.95 ⎟
⎜⎜ 1 + ⎟
⎝ Re, p ⎟⎠
Equation 4‐8
Figure 4‐8 Standard drag curve for motion of particle in a fluid

At higher relative velocities, the inertia of the fluid begin to dominate (the fluid must
accelerate out of the way of the particle). Analytical solution of the Navier‐Stokes
equations is not possible under these conditions. However, experiments give the
relationship between the drag coefficient and the particle Reynolds number in the form of
the so‐called standard drag curve as shown in Figure 4‐8. Four different regions are
identified: the Stokes’ law region, the Newton’s law region in which drag coefficient is
independent of Reynolds number, an intermediate region between the Stokes and
Newton’s regions; and the boundary layer separation region. The Reynolds number
ranges and drag coefficient correlations for these regions are given in Table 4.2, (Martin,
2008)
Table 4‐2 Reynolds number ranges for single particle drag coefficient correlations
Region Stokes Intermediate Newton’s law
Re, p range < 0 .3 0.3 < Re, p < 500 500 < Re, p < 2 × 10 5

CD 24
Re , p
24
Re , p
(
⋅ 1 + 0.15 ⋅ Re, p
0.687
)
≈ 0.44

As the Reynolds number increases, there will be a separation of flow field with the
formation of vortex behind the body and further increase in Reynolds number causes
unstable and unsteady wake formation.

Figure 4‐9 Drag coefficient of the sphere (Spurk, 1997)

The drag coefficient depends upon the flow pattern behind the object. However, for
incompressible flow, the drag coefficient is only a function of particle Reynolds number

as shown in Figure 4‐9. The sudden drop of value of CD around Reynolds number 3 x105
is due to transition of laminar boundary layer on spherical surface to turbulent boundary
layer. The shear stress in the turbulent boundary layer is larger; hence, fluid at outer layer
is dragged close to the wall. This retards the flow separation and wake becomes
narrower. The non‐separated flow in the backside of the particle exerts force against the
flow direction and hence CD becomes smaller. This transition can occur in lower Reynolds
number when the surface is made rougher. Sand particle surface being rougher than the
regular spherical surface may have value of CD in this region. The observation of
Reynolds number at different flow condition and direction of flow components are
presented in results and discussions section. The values of CD from the experiment in the
swirl flow are smaller than that shown by Spurk, 1997, because of measurement
uncertainties and limitations of experimental study.
4.2.6 Uncertainties for the measurements

Due to the nature of physical measurements, it is impossible to measure a quantity
without error. A measurement system is often made up of series of components, which
are subjected to individual inaccuracy. Each of these individual uncertainties includes
random and systematic errors. There are many uncertainties in laboratory experimental
study, which combine to make this an imprecise prediction of drag coefficient for
particle, including:
• The shape and size range distribution of the particle,
• The effect of the shape of non‐spherical particles on their drag coefficient has
proved difficult to define. This is probably due to difficulty in describing particle
shape for irregular particles. The approached applied to define the irregular
shape of the particle with its spheroid may not be sufficient in some cases,
• When many particles flow in a fluid in close proximity to each other the motion
of each particle is influenced by the presence of the others. The simple analysis
for the fluid particle interaction for a single particle is no longer valid but can be
adapted to model the multiple particle system,
• While calculating the forces acting on the particle mainly centrifugal and drag
forces were considered, however, the friction force may acting on the particle
sliding on the surface of the middle plate then causes that particle slightly rotate.
The rotation condition of the particle depends on characteristics of the particle
and wall of the plate. If friction force is big enough the particle will rotate,
otherwise, the particle will slide.
However, the value of drag coefficient around 0.1 to 0.2 could be appropriate for
estimation of particle that will stay and rotating in the swirl flow at higher particle
Reynolds number, where separation of particles from streamlines takes place.

4.3 RESULTS AND DISCUSSIONS
Velocities of particles of different sizes and shapes are shown in Figures 4‐10 to 4‐12. The
velocity of the particle will increase if the velocity of water increases. Higher velocity of
the particle was obtained at higher operating head. The size of the particle is inversely
proportional to the velocity of the particle, and it depends upon the shape of the particle.
Different shapes of particles were tested for different operating head, and it was
with other shapes.

Velocity of particle in swirl flow Velocity of particle for different size & shape
15 13
12
Particle velocity [m/s]

11.5
Velocity [m/s]
9 5mm steel particle vel.

6mm ran. sand
average flow vel.
10 5mm sp.steel
6 maximum flow vel.
4.74mm cylin.alum.
3 8.5
0
10 15 20 25 30 35 40 45 50 7
20 25 30 35 40 45 50
Head [m]
Head [m]

(a) Head (b) Size and shape

Figure 4‐10 Particle velocity and head relation for different size and shape of particle

Velocity of particle & size Velocity of particle for same size but different shape
12 15
10
13
Velocity [m/s]
8 Spherical
11 triangularl
Steel particle at 30m head
cylindrical
6
9
7
20 25 30 35 40 45 50
2
2 4 6 8 10 12 Head [m]
Size of particle [mm]

(a) Different size (b) Same size but different shape

Figure 4‐11 Particle velocity for different sizes and same size with different shape

Many natural particles are usually in non‐spherical shape. These particles will tend to
have lower settling velocities because both decreases in spheroid and increases in
angularity tend to decrease velocities. Furthermore, larger cross‐sectional areas tend to be
directed perpendicular to the transport path. As a result, higher coefficient of drag,
higher rotational motion and more separation of flow likely to occur and hence more
erosion rate is anticipated.

Shape of the particle Velocity of particle for regualr & irregular shape
5 13
Velocity of Particles [m/s]

4
11.5
3 Smooth Cylindrical
3x3 mm cylin.regular
Smooth Triangular
10 3x6mm cylin. irregular shape
Smooth Spherical
2 2x4mm cylin. irregular
Rough Triangular
8.5
1
0 7
4 6 8 10 12 14 16 30 35 40 45 50
Flow Rate [LPS] Head [m]
(a) Different particle with different shape (b) Same particle with different shape

Figure 4‐12 Particle velocity relation for same/different particle with different shape

If centrifugal force and drag force on the particle is equal, the particle will rotate exactly
at the injecting radius, but if those forces do not balance each other, there will two
possibilities. The particle either moves toward the inner radius and ultimately sinks or
moves toward the outer radius and ultimately hits the sides of the vanes. The equilibrium
condition was observed for a given particle after manipulating the flow velocity,
verifying that the different forces were balanced in the test rig. This also revealed that the
particle with a given diameter would stay at the orbit of the injecting radius until either
the velocity components were changed or the particle became smaller by fracturing due
to impact with the outer wall. Different shapes of particles were tested with the same
operating conditions as shown in Figure 4‐13, and it was determined that triangularly
shaped particles were more likely to hit the suction side of the guide vane cascade. The
radius at which the particle is moving, called rotation radius. This rotation radius also
varied with different shapes of particles, which further supports the influence of shape
factor for particle velocity calculation.

Rotation radius for same size but different shape
0.4
Rotation Radius [m]
0.3
Sand particle 3-5mm size

Steel particle 5mm size
Trian. ceramic 6mm size
0.2
0.1
0 10 20 30 40
Flow Rate [LPS]

Figure 4‐13 Rotation radius for particles of the same size but different shape

Drag coefficient was also calculated and plotted for different shapes and sizes of particle
as shown in Figure 4‐14 and 4‐15.

Drag Co-efficient & Re,p no for same size Drag Co-efficient and Size
0.475 12
0.470 6
0.470 9
Drag Co-efficient
0.465
Drag Co-efficient
Size [mm]
drag co-efficient
0.460 4 drag co-efficient
size of particle 0.465 6
size of steel particle
0.455
0.460 3
0.450 2
2.E+04 3.E+04 3.E+04 4.E+04 4.E+04 5.E+04 5.E+04 6.E+04
0.455 0
Re,p 2.E+04 3.E+04 4.E+04 5.E+04 6.E+04 7.E+04
Re,p

(a) Same size (b) Different size

Figure 4‐14 Drag coefficient relation for same /different size of particle

Drag Co-efficient for regular & irregular shape Drag Co-efficient for different shape
0.473
0.472
Drag Co-efficient
0.470 0.471
Drag Co-efficient
0.468 3x3 mm cylin.alu. regular 0.470 cylin. shape (irregular)
3x6mm cylin.alu. irregular trian. shape (regular)
0.465 3x9mm cylin.alu. irregular 0.470 sp. shape (regular)
0.463 0.469
0.460 0.468
2.E+04 3.E+04 4.E+04 5.E+04 6.E+04 7.E+04 5.E+04 6.E+04 6.E+04 7.E+04 7.E+04 8.E+04 8.E+04 9.E+04
Re,p
Re,p

(a) Same particle different shape (b) Different particle different shape

Figure 4‐15 Drag coefficient relation for regular and irregular shape of same/different
particle

The drag coefficient C D is a non‐dimensional number that depends on the shape of the
particle, the fluid kinematic viscosity, and the grain size. It has been found that the effect
of the shape of non‐spherical particles on their drag coefficient can be defined in terms of
its spheroid. Moreover, shape affects drag coefficient far more in the intermediate and
Newton’s law regions than in the Stokes’ law region. However, the variation of drag
coefficient in the Newton’s law region is not so significant, but the influence of size and
shape of the particle has been clearly demonstrated. The result shows that the
triangularly shaped particles have a higher drag coefficient than other shapes because the
angular particles also tend to have lower settling velocity than the spherical ones.


Chapter 5
Computational Fluid Dynamics
Theory
In this chapter, the governing equations of fluid, the particle equation of motion, and the
two empirical erosion equations are briefly presented. The derivation of the governing
equations is not included in this chapter, however, it can be found in many CFD books
including ANSYS user guide, 2006.

5.1 INTRODUCTION
Computational Fluid Dynamics (CFD) is the analysis of systems, which involves fluid
flow, heat transfer and other related physical processes by means of computer based
simulation. It works by solving the equations of fluid flow over a region of interest, with
specified boundary conditions of that region. ANSYS CFX is general‐purpose CFD tool,
which is used for the numerical simulation analysis that combines an advanced solver
with powerful pre and post‐processing capabilities. This is capable of modelling: steady
state and transient flows, laminar and turbulent flows, subsonic, transonic and
supersonic flows, heat transfer and thermal radiation, buoyancy, non‐Newtonian flows,
transport of non‐reacting scalar components, multiphase flows, combustion, flows in
multiple frames of reference, particle tracking, etc. Furthermore, this includes, an
advanced coupled solver that is both reliable and robust, with full integration of problem
definition, analysis, and results presentation and an intuitive and interactive setup
process, using menus and advanced graphics features.

The CFX consists of five software modules that pass the information required to perform
a CFD analysis. These are the mesh generation software, the pre‐processor, the solver, the
solver manager, and the post‐processor.

Chapter 5 CFD Theory
5.2 GOVERNING EQUATIONS
The governing equations of fluid flow, which describe the processes of momentum, heat
and mass transfer, are known as the Navier‐Stokes equations. There are three different
streams of numerical solution techniques: finite difference, finite volume method and
spectral methods. The most common, and the CFX is based, is known as the finite volume
technique. In this technique, the region of interest is divided into small sub‐regions,
called control volumes. The equations are discretized, and solved iteratively for each
control volume. As a result, an approximation of the value of each variable at specific
points throughout the domain can be obtained. In this way, one derives a full picture of
the behaviour of the flow. The instantaneous equations of mass, momentum and energy
conservation can be written as follows in a stationary frame (ANSYS, 2006).
The continuity equation:
∂ρ
+ ∇ • (ρ ⋅ U ) = 0 Equation 5‐1
∂t
The momentum equations:

∂ (ρ ⋅ U )
+ ∇ • (ρ ⋅ U ⊗ U ) = −∇p + ∇ • τ + S M Equation 5‐2
∂t
⎛ ⎞
= μ ⋅ ⎜ ∇ ⋅ U + (∇ ⋅ U ) − ⋅ δ∇ • U ⎟
2
Where τ
r
Equation 5‐3
⎝ 3 ⎠
The total energy equation:
∂ (ρ ⋅ htot ) ∂ρ
− + ∇ • (λ ⋅ ∇T ) + ∇ • (U • τ ) + U • S M + S E Equation 5‐4
∂t ∂t
Where ht ot is the total enthalpy, related to the static enthalpy h(T , p ) by,
1 2
htot = h + ⋅U Equation 5‐5
2

The term ∇ • (U • τ ) represents the work due to viscous stress and is called the viscous
work term. The term U • S M represents the work due to external momentum sources
and S E is the energy source.

CFD Theory Chapter 5
5.2.1 Two equation turbulence closure models

Two‐equation turbulence models are widely used to provide a ‘closure’ to the time
averaged Navier‐Stokes equations. Two principle closure models exist commercially, are
the k − ε (k‐epsilon) model and the shear stress transport (SST) model. The two‐
equation models are much more sophisticated than the zero equation models. In these
equations both the velocity and the length scale are solved using separate transport
equations. The detail of these equations is not considered in this section. This can be
found in ANSYS user guides, 2006.
The k − ε model has proven to be stable and numerically robust and has a well‐
established regime of predictive capability. For general‐purpose simulations, the k − ε
model offers a good compromise in terms of accuracy and robustness; however, it can
lack prediction accuracy for complex flow. Such complexities include rapid variations in
flow area, flows with boundary layer separation, flows with sudden changes in the mean
strain rate, flows in rotating fluids, flows over curved surfaces etc.
A Reynolds Stress model may be more appropriate for flows with sudden changes in
strain rate or rotating flows, while the SST model may be more appropriate for separated
flows.
5.3 PARTICLE EQUATION OF MOTION
The most widely applied method to determine the motion of the dispersed phase is by
tracking a large number of individual particles in the flow field. This method is called
Lagrangian tracking. In this method, each particle represents a sample of particles that
follow an identical path, which describe the average behaviour of the dispersed phase.
The CFD code contains a Lagrangian particle‐tracking algorithm that numerically
predicts trajectories of solid particles through the flow field. These calculations use
information generated by the flow field simulation. The code also has the capability to
couple the particle equation of motion, with the flow solution. Coupling the governing
equations for the fluid and the particle allows effects, such as fluid displacement by
particles and particles‐induced turbulence, to be investigated. In many cases of
engineering interest, especially solid /liquid flows, this coupling allows the investigation
of particle concentration effects. However, at low particle concentrations, the particles do
not affect the flow and this coupling is not necessary. While setting up the Lagrangian
way particle model, the following assumptions are implied:
• Particle‐particle interactions are neglected due to the low particle concentrations
experienced.
• The particles are spherical. The physical properties of each phase are constants.
• The mean flow is steady. The turbulent flow is locally isotropic.
• The geometry modification caused by the removal of wall by the sand particles,
are neglected

To derive a particle motion of equations, consider a discrete particle travelling in a
continuous fluid medium. The forces acting on the particle, which affect the particle
acceleration, are due to the difference in velocity between the particle and fluid, as well
as to the displacement of the fluid by the particle. Basset, Boussinesq and Oseen derived
the equation of motion for such a particle for a rotating reference frame (ANSYS, 2006):

dU P
mP ⋅ = FD + FB + FR + FV M + FP + FBA Equation 5‐6
dt

Where, FD is drag force acting on the particle, FB is Buoyancy force due to gravity, FR is
forces due to domain rotation (centripetal and coriolis forces), FVM is virtual (or added)
mass force. This is the force to accelerate the virtual mass of the fluid in the volume
occupied by the particle. This term is important when the displaced fluid mass exceeds
the particle mass, such as in the motion of bubbles. FP is pressure gradient force. This is
the force applied on the particle due to the pressure gradient in the fluid surrounding the
particle caused by fluid acceleration. It is only significant when the fluid density is
comparable to or greater than the particle density. FBA is Basset force or history term,
which accounts for the deviation in flow pattern from a steady state. This term is not
implemented in ANSYS CFX. The particle and fluid mass values are given by,

π
mP = ⋅ d p ⋅ ρ p
3
Equation 5‐7
6
π
mF = ⋅ d p ⋅ ρF
3
Equation 5‐8
6
5.3.1 Drag force

The aerodynamic drag force on a particle is proportional to the slip velocity, U S between
the particle and the fluid velocity. This is given by,

⋅ C D ⋅ ρ F ⋅ AF ⋅ U S ⋅ U S = ⋅ C D ⋅ ρ F ⋅ AF ⋅ U F − U P ⋅ (U F − U P )
1 1
FD =
2 2
Equation 5‐9
Where, C D is the drag coefficient and AF is the effective particle cross section. The drag

coefficient C D is introduced to account for experimental results on the viscous drag of a

solid sphere. In the simplest form of equation of motion, the main term on the right hand
side is the drag force that includes the consideration of both skin and form effects. The
total drag force is most conveniently expressed in terms of the non‐dimensional drag
coefficient, as defined by the Schiller‐Naumann correlation, which modifies this to ensure
the correct limiting behaviour in the inertial regime by taking,

⎛ 24
(
C D = max⎜⎜ ⋅ 1 + 0.15 ⋅ Re
0.687
) ⎞
, 0.44 ⎟⎟ Equation 5‐10
⎝ Re ⎠
However, this correlation has been developed based upon spherical shape of the particle;
the trajectories of the non‐spherical particles can be modelled through the application of
shape factors, allowing non‐uniform drag distribution and bounce characteristics to be
classified.

5.3.2 Buoyancy force

The buoyancy force is the force on a particle immersed in a fluid. The buoyant force is
equal to the weight of the displaced fluid and is given by,

⎛ ρ ⎞ π
FB = (m p − m F ) ⋅ g = m p ⋅ ⎜1 − F ⎟ ⋅ g = ⋅ d p3 ⋅ (ρ p − ρ F ) ⋅ g Equation 5‐11
⎜ ρ p ⎟⎠ 6
⎝
5.3.3 Rotation force

In a rotating frame of reference, the rotation term is an intrinsic part of the acceleration
and is the sum of Coriolis and centripetal forces:
FR = m p ⋅ (− 2 ⋅ Ω ⋅ U P − Ω ⋅ Ω ⋅ rp ) Equation 5‐12
The implemented rotation term also contains contributions from the pressure gradient
and the virtual mass force due to the domain rotation, which leads to the following final
contribution of the rotation term.

5.3.4 Virtual or added mass force

This force is caused by the fact that the particle has to accelerate some of the surrounding
fluid, leading to an additional drag, which has the following form:


CV M ⎛ dU F dU P ⎞
FV M = ⋅ mF ⋅ ⎜ − ⎟ Equation 5‐13
2 ⎝ dt dt ⎠

If the virtual mass force is included, the coefficient CV M is normally set to one. However,
when the virtual mass force is not included, then CV M effectively has the value zero, and
RV M is equal to one.
5.3.5 Pressure gradient force

The pressure gradient force results from the local fluid pressure gradient around the
particle and is defined as,

mF
FP = − ⋅ ∇P Equation 5‐14
ρF
This force is only important if large fluids pressure gradients exist and if the particle
density is smaller than or similar to the fluid density. Neglecting diffusive and source
terms in the steady state momentum equation, the pressure gradient can be replaced by
the velocity gradient. Assuming a constant fluids density, the pressure gradient force can
be written as,

ρF
Fp = m F ⋅ (U F ⋅ ∇U F − RF ) = m p ⋅ ⋅ (U F ⋅ ∇U F − RF ) Equation 5‐15
ρp
5.4 RESTITUTION COEFFICIENT FOR PARTICLES
The parallel and the perpendicular restitution coefficients describe the action of particles
when they hit a wall. The reflected velocity of the particle is lower than the incoming
velocity due to energy transfer. Energy is dissipated as heat, noise and target material
deformation. This effect is described by the momentum‐based restitution coefficient. The
coefficient values of 1 describes an elastic collision, while values less than 1 describes an
in‐elastic collision. The parallel coefficient is usually equal to 1. The perpendicular
coefficient will depend on the particle material. Particle that bounces off the walls will
have a perpendicular coefficient close to 1, while particles that stick to walls will have a
perpendicular coefficient of 0. In this simulation, the perpendicular and parallel
coefficients of restitutions are chosen equal to 0.9 and 1 respectively.

5.5 BASIC EROSION MODEL
There are two erosion models available in ANSYS CFX, namely, Finnie and Tabakoff. The
choice of one model over another is largely simulation‐dependent. In general, the
Tabakoff model provides more scope for customization with its larger number of input
parameters. The erosion model can be set on a per‐boundary or per‐domain basis. When
enabled for the domain, the domain settings will apply for all boundaries that do not
explicitly have erosion model settings applied to them.

5.5.1 Model of Finnie

The wear of a wall due to the erosive effect of particle impacts is a complex function of
particle impact, particle and wall properties. For nearly all metals, erosion is found to
vary with impact angle and velocity according to the relationship (ANSYS, 2006)

E = k ⋅ V pn ⋅ f (γ ) Equation 5‐16
Where, E is a dimensionless mass, V p is the particle impact velocity and f (γ ) is a
dimensionless function of the impact angle. The impact angle is the angle in radians
between the approaching particle track and the wall. The value of the exponent, n , is
generally in the range 2.3 to 2.5 for metals.
Finnieʹs model of erosive wear relates the rate of wear to the rate of kinetic energy of
impact of particles on the surface, using n = 2
E = k ⋅ V p ⋅ f (γ )
3
Equation 5‐17
Where,
f (γ ) =
1 1
⋅ Cos 2 γ if tan γ > , and Equation 5‐18
3 3
f (γ ) = Sin(2γ ) − 3 ⋅ Sin 2 γ
1
if tan γ ≤ Equation 5‐19
3

5.5.2 Model of Tabakoff and Grant

The Tabakoff erosion model requires the specification of five parameters (ANSYS, 2006).
The k12 constant, 3 reference velocities and the angle of maximum erosion γ 0 must all be
specified. Table 5‐1 provides some values for the Tabakoff model coefficients for Quartz‐
Aluminium, Quartz‐Steel and Coal‐Steel.

Table 5‐1 Coefficients for some materials using the Tabakoff erosion model
Variable Coefficient Value Material
k12 k12 5.85 x 10
‐1
Quartz‐ Aluminium
Ref Velocity 1 V1 159.11 m/s

Angle of Maximum
γ0 25 degree
Erosion
k12 k12 2.93328 x 10‐1

Quartz ‐ Steel
Ref Velocity 2 V3
352.99 m/s
Angle of Maximum
γ0 30 degree
Erosion
k12 k12 ‐1.321448 x 10‐1

Coal ‐ Steel
Ref Velocity 2 V3
87.57 m/s
Angle of Maximum
γ0 25 degree
Erosion

In the erosion model of Tabakoff and Grant, the erosion rate E is determined from the
following relation (ANSYS 2006):

[ 2
]
E = k1 ⋅ f (γ ) ⋅ V p ⋅ Cos 2 γ ⋅ 1 − RT + f (VPN ) .
2
Equation 5‐20

Where,

2
⎡ ⎛ π / 2 ⎞⎤
f (γ ) = ⎢1 + k 2 ⋅ k12 ⋅ Sin⎜⎜ γ ⋅ ⎟⎟⎥
⎣ ⎝ γ 0 ⎠⎦ , Equation 5‐21
RT = 1 − k 4 ⋅ V P ⋅ Sinγ , Equation 5‐22
f (VP N ) = k 3 ⋅ (VP ⋅ Sinγ )

4
, Equation 5‐23
⎧1.0 if γ ≤ 2 ⋅ γ 0
k2 = ⎨ Equation 5‐24
⎩0.0 if γ > 2 ⋅ γ 0

Here, E is the dimensionless mass (mass of eroded wall material divided by the mass of
particle). V p is the particle impact velocity. γ is the impact angle in radians between the
approaching particle track and the wall, and γ 0 is the angle of maximum erosion. k1 to
k 4 , k12 and γ 0 are model constants and depend on the particle and wall material
combination.


Chapter 6
CFD Simulation Model
In this chapter a description of computational model, similar to Cahua power plant
Francis turbine design is presented, in which the governing equations will be solved and
solution is obtained.

6.1 CAHUA HYDROPOWER PLANT (HPP)
Cahua hydropower plant is a Run‐of‐River hydropower plant built in Peru. It is located
60 km upstream the mouth of Pativilca River, which is about 200 km north of Lima. The
construction of power plant was finished in 1967. SN‐Power Norway purchased this
power plant in 2003.

Figure 6‐1 Cahua hydropower plant

Chapter 6 CFD Simulation Model
There are two Francis turbines installed, which were originally delivered by Riva. Both
Francis turbines composed of 20 stay vanes, 20 guide vanes and 17 runner blades. The
rated power output is 42 MW, the gross head is 215 m, and the mass flow is 22 m3/s.

The river carries large amount of sediments and has caused severe erosion damage on
turbine components, mainly, turbine runner, guide vanes, and covers. The runner, guide
vanes, and covers of the turbines are taken out every year and refurbished. The original
turbines are operated at maximum 3 g/l (or 3,000 ppm) sediment concentration and the
sediment that passes the turbine has about 35 % of quartz and 30 % of feldspar content on
average. This means that the turbine has a maximum sediment load of 33 kg/s and 65 %
of minerals are harder than the base material of turbine. This is the main reason of
excessive sediment erosion in the hydraulic machinery operating in this power plant.
Furthermore, the sediment study conducted at this power plant indicated that, the
sediment concentration exceeds 120,000 tons of sediment, i.e. 33.06 kg/s only after six
weeks of operation during intense monsoon period. Table 6‐1 presents the sediment load
of Cahua power plant for four monsoon periods. According to the engineers at the power
plant, the turbine can take maximum 120,000 – 140,000 tons of sediment loads. If the load
is even higher, the turbine parts gets eroded so much that welding and grinding is
impossible.

Table 6‐1 Sediment load of Cahua power plant (Ole et al., 2009)
Turbine Turbine put in to service Turbine taken out of service Sediment load[tons]
1 16.04.2001 27.03.2002 165,6354
1 28.03.2002 02.05.2003 131,799
1 03.05.2003 29.05.2005 136,327
1 30.05.2005 01.07.2006 115,538
2 11.06.2001 28.07.2002 220,032
2 29.07.2002 13.07.2003 82,583
2 14.07.2003 19.04.2005 122,891
2 22.04.2005 01.07.2006 144,865

6.2 DESCRIPTION OF COMPUTATIONAL MODEL
The erosion prediction procedure mainly consists of three different models. These are,
flow model, particle transport model, and erosion model. CFX code contains the ability to
couple the equations governing fluid motion and the particle equation of motion.

The flow simulation contains the information necessary to perform all subsequent
calculations. Velocity components, turbulence quantities (turbulence kinetic energy and
dissipation rate), and carrier fluid properties (density and viscosity), are all contained
within the flow field simulation. Once a simulated flow field is obtained, the solution is
introduced with a large number of particles at the inlet of stay vanes. A large numbers of

CFD Simulation Model Chapter 6
particles are normally required in order to obtain a reasonable distribution and to reduce
scatter in the erosion predictions. Each particle is tracked separately through the flow
field. For each particle impingement, a set of empirical equations is applied. These
erosion equations account for the impingement speed and angle, as well as the particle
shape and the mechanical properties of the wall material.

In order to visualize erosion predictions in a convenient manner, predicted erosion data
is transferred to a post‐processor. This post‐processor is used to generate contour plots of
predicted erosion quantities. This allows not only the simultaneous examination of the
flow solution, particle trajectories, and erosion predictions, but also provides the ability
to identify areas of high erosion.

6.2.1 Flow model

The flow simulation was obtained through use of a commercially available
computational fluids dynamics (CFD) code, namely ANSYS CFX. The code utilizes a
finite‐volume, multi‐block approach to solve the governing equations of fluid motion
numerically on a user‐defined computational grid.
The flow solution procedure first generates the computational grid. A pre‐processor is
available in the software to perform this task. Second, the solution option such as inlet
and boundary conditions, turbulence model, and discretization scheme, are specified.
The final step is running the flow solver to generate the actual flow field simulation.
CFX contains several models for turbulence behaviour. Isotropic and non‐isotropic
turbulence models are available. In addition, multitudes of discretization schemes are
available to obtain the most accurate flow solution possible. For this work, a differential
Reynolds stress turbulence model, and Forward Euler implicit scheme, second order time
in high‐resolution advection scheme, with convergence criteria at 10‐4 were used.

6.2.2 Mesh generation

A computational domain is necessary for the simulation of a Francis turbine. The
computational domain was made based upon a Cahua Francis turbine design, which
composed of stay vanes, guide vanes, and runner vanes, as shown in Figures 6‐2. This
domain corresponds to the water passages and can fully consider non‐uniform flow field
incoming from the stay vanes. Two computational domains, as shown in Figure 6‐2 (a)
and (b); were considered during simulation analysis, one based upon best efficiency
point and another based upon full load conditions.

For each independent component, a hexahedral structure mesh was generated using
Turbo Grid 11.0. The Turbo Grid was selected for meshing because this is a highly
automated hexahedral mesh generator, specially designed for turbo‐machinery
applications.

The Grid generation relied on three different factors; namely, number of cells, Reynolds
number, and turbulence model desired. Two equation turbulence models was chosen
initially for grid generation but the Shear stress transport (SST) model gave the most
realistic results, and it was decided to continue the simulation with this model only.
Figure 6‐2 shows single blade cascade computational model for three different vanes.
(a) BEP (b) Full load

Figure 6‐2 A single blade cascade computational model for three vanes

To achieve a result with second order accuracy and sufficient numerical stability, a high‐
resolution scheme was applied for discretization of the grid. The grids used in these
simulations have different mesh quality, as presented in Table 6‐2. Since runner blade is
more important than other components of turbine, higher numbers of nodes were
observed in runner blade than guide vane and stay vane. At areas where the flow
includes large gradients, the grid quality needs special attention.

Table 6‐2 Mesh statistics of numerical models
Computational Vane Number of nodes Number of elements
model (hexahedra)
Runner blade 132,125 119,952
Guide vane 110,345 100,800
Full load (FL) Stay vane 59,130 53,352
Runner blade 99,190 90,304
Guide vane 72,565 65,472
Best efficiency Stay vane 59,130 53,352
(BEP)

To get a smooth transition between the small and the large control volumes, an
expansion factor of approximation 1.25 was used. The GTM files (mesh files) were
directly imported into a Turbo machinery mode in pre‐processor for defining simulation.

Figure 6‐3 A single blade cascade mesh model for three vanes

6.2.3 Properties of sand

The density of sand particles used in the simulation needs to be defined in general mode.
To calculate the effect of particles on continuous fluids, between 100 and 1,000 particles
are usually required (ANSYS, 2006). However, if accurate information about the particle
volume fraction or local forces on the wall boundaries is required, then a much larger
number of particles are needed to be modelled. After creating a domain, full coupling or
one‐way coupling between the particle and continuous phase has to be chosen. Full
coupling is needed to predict the effect of particles on the continuous phase flow field but
has the higher computational cost than one‐way coupling. One‐way coupling simply
predicts the particle paths during post processing based on the flow field, but without
affecting the flow field. To optimize CPU usage, the two sets of identical particles were
created. The first set is fully coupled and between 100 and 1,000 particles are used. This
allows the particles to influence the flow field. The second set uses one‐way coupling but

a much higher number of particles are used. This provides a more accurate calculation of
the particle volume fraction and local forces on the wall. In inflow or in opening
boundary, the particle velocity, injection position, diameter distribution, and mass flow
rate are needed to be specified. Sand particles are uniformly injected at the stay vane inlet
with the same conditions as the fluid. The particles will follow through the domains and
exit at the outlet. The sand particles are defined as solid particles, and the size
distribution is uniform in diameter. The turbulence dissipation force is activated, and the
Schiller Naumann model is chosen to calculate the drag force acting on the particle. The
two empirical erosion models, namely Finnie, and Tabakoff, were used for predicting
erosion.

6.2.4 Boundary conditions

In a numerical simulation, it is impossible and unnecessary to simulate the whole system.
Generally, the interesting regions with certain boundary conditions are sufficient to be
chosen for conducting proper simulations. Numerical simulations also have to consider
the physical processes in the boundary region. Different boundary conditions may cause
quite different simulation results. Improper sets of boundary conditions may introduce
nonphysical influences on the simulation system, while a proper set of boundary
conditions can avoid that. So arranging the boundary conditions for different problems
becomes very important. While at the same time, different variables in the environment
may have different boundary conditions according to certain physical problems. If all
residual are below their targeted criteria then convergences of equations will prevail. To
improve the convergences of simulation and to obtain a convergence of solution with less
numbers of iterations in relatively short interval of time, only one set of blades was
considered in this simulation analysis. Generally, the meshed domains are imported into
the pre‐processor, and connected to each other by domain interfaces. The runner domain
is a rotating domain, which is coupled to the stationary domain of guide vane and stay
vane parts by so‐called frozen rotor domain interfaces. In addition to these domain
interfaces, inlet and outlet boundary conditions are defined. Fluid properties (density
and viscosity) and particle properties (diameter distribution, density, and concentration)
are also defined in the pre‐processor. In the post‐processor, there is a provision to
assemble all components of a turbine. The two most interesting operating points were
considered, at best efficiency point and at full load with guide vane angle of 16 ° and 22 °
respectively.

6.2.4.1 Inlet and outlet

When setting boundary conditions, ANSYS CFX best practices guide for Turbo
machinery recommends specifying a total inlet pressure and mass flow outlet. This will
give more appropriate calculation of the flow field than with the mass flow inlet
condition for machines drawing fluid directly from a static reservoir. This may be valid
for hydraulic pump simulation but in the case of radial turbine, the mass flow inlet and a

total pressure outlet were found more appropriate and robust than the total inlet
pressure and mass flow outlet. In addition, this provides better performance of turbine,
when looked at pressure distribution, y+ values, power output, head, and flow rate.
Thus, the inlet condition was set as mass flow rate per passage with velocity components.
The outlet pressure in this case is arbitrary and is usually set at zero or close to zero to
reduce round off error. In this simulation, this was chosen equal to 1 atm. The detail
boundary conditions with appropriate value are presented in Table 6‐3.

Table 6‐3 Boundary conditions
Variable Value
Water density 997 kg/m3
Quartz particle density 2.65 gm/cm3
Diameter of the particle 15 μm to 2 mm
Inlet mass flow rate of water at BEP 540 kg/s
Inlet mass flow rate of water at full load 685 kg/s
Total pressure outlet 1 [atm.]
Inlet mass flow rate of particle 1 to 50 kg/s
Inlet flow direction at Stay vanes (α , r , θ ) 0 , ‐0.4 , ‐0.9165
6.2.4.2 Wall
The covers, hub, shroud and vanes were defined as smooth walls with no‐slip condition.
The wall friction then decreases fluid velocities near the wall. The runner has an angular
velocity 600 rpm, while the stay vane and guide vane are stationary domains. A periodic
boundary condition was set to couple two adjacent blades. This simplified the
computational model. When a particle reaches a symmetry plane boundary condition, it
is reflected. If simulation includes any solid domains, then it needs to set particle options
at the fluid‐solid interface. To do this, first it needs to create the required fluid‐solid
domain interfaces, and then edit the automatically created interface boundary condition
for the fluid side of the interface. The options on the fluid side of fluid‐solid interfaces are
specified in the same way as for wall boundaries.

Chapter 7
Simulation Result and Discussions
This chapter presents contour plots of erosion rate density on different components of the
Francis turbine based upon the simulation results. The relationship between particle
movement and erosion inside the turbine based upon the shape, size, and concentration
of particle at two operating conditions namely, best efficiency point (BEP) and full load
(FL) are compared and the results are discussed.

7.1 HYDRAULIC PERFORMANCE OF TURBINE
Turbines are always designed for specific conditions of head, speed and output.
However, a turbine may actually be required to operate under conditions widely
different from those it has been designed. It is, therefore, important to determine and
have complete information about the performance of the turbine over a wide range of
operating conditions. The hydraulic performance of a simulated turbine is presented in
Figure 7‐1. This is found in good agreement with the actual data obtained from the Cahua
turbine. The auto generated hydraulic turbine reports of simulation analysis are
presented in Appendix A and B.
Hydraulic performances of turbine Hydraulic performance of turbine at constant head
15 300 95 26
flow head
14
Power output [MW]
Volume flow rate
90 24
Efficiency
13
[m3/s]
Head
[m]
200
12
85 22
efficiency power
11
10 100
80 20
14 16 18 20 22 24 14 16 18 20 22 24
Guide vane angle [degree] Guide vane angle[degree]
(a) Volume flow rate and guide vane angle (b) Efficiency and guide vane angle

Figure 7‐1 Hydraulic performance of turbine

Chapter 7 Simulation Results and Discussions
7.2 SEDIMENT EROSION ON TURBINE COMPONENTS
An important consideration in these simulations is erosion to the different components of
a turbine due to different shape and size of the sand particles. A good indication of
erosion is given by the erosion rate density (kg/s/m2) parameter, which corresponds to
pressure and shear stress due to the flow. In order to visualize erosion predictions in a
convenient manner, predicted erosion data is transferred to a post‐processor. This post‐
processor is used to generate contour plots of predicted erosion quantities. This allows
not only the simultaneous examination of the flow solution, particle trajectories, and
erosion predictions, but also provides the ability to identify areas of high erosion. This
can be seen as coloured spots. A Red colour indicates the highest value of erosion
intensity, whereas a blue colour denotes the lowest intensity of erosion. The predicted
erosion on different components can be used to better understand how parameters such
as inlet conditions, fluids properties, flow rate, particle size & shape, concentration, and
geometry, affect erosion behaviour.

7.2.1 Stay vane

Figure 7‐2 (a‐b) shows predicted erosion at stay vane and its surface. The mid height of
the leading edge of stay vane is exposed to less erosion intensity. Near the upper and
lower cover at the inlet, ring‐shaped erosion grooves are observed, as seen in red colour
in Figure 7‐2 (a). On the blade surface, scattered erosion spots are observed at suction
side. There is no erosion intensity observed at pressure side of vanes. This may be due to
the boundary layer effect, which is caused by the initial velocity values at inlet of stay
vanes or incorrect mesh properties. The boundary layer effect is one of the main
limitations in many CFD turbulence models. These simulation results are found in
agreement with Cahua turbine stay vanes as presented in Figure 3‐2.
(a) Stay vanes with covers (b) Suction side of stay vane

Figure 7‐2 Predicted erosion pattern on stay vanes

Simulation Results and Discussions Chapter 7
7.2.2 Guide vane

Figure 7‐3 (a – b) shows the predictions of erosion rate density along the guide vane
surface. The highest erosion is predicted on outlet regions and both covers. This is due to
the turbulence erosion caused by high velocity with fine grain sand, which is in good
agreement with the field test at Cahua, where significant surface deterioration was
detected on the surface near the transitional zone to lower and upper cover. The particle
velocity inside a turbine at both designed and maximum load conditions are presented in
Figure 7‐4 (a‐b). The highest absolute velocity was observed at guide vane cascade and
the highest relative velocity was observed at runner outlet surface. On the suction side of
the vane, the simulations indicated relatively low erosion rate density on the surface
towards the trailing edge whereas there is no erosion tendency found on the pressure
side of the blades. This may be due to poor boundary conditions.

(a) Outlet and covers (b) Suction side

Figure 7‐3 Predicted erosion pattern on guide vanes
(a) Design load (b) Full load

Figure 7‐4 Velocity of sand particle inside a turbine

Similar erosion tendency was observed at both design and off design conditions except at
the leading edge of the vane, where surprisingly the highest erosion is predicted. This
may be due to the combination of strong acceleration with bigger sizes of particle, and
high velocity and strong turbulences with smaller sizes of particle. This creates secondary
flow and strong vortex in the corners, and leads more erosion damage.

7.2.3 Runner blade

In the runner, the highest relative velocities occurred at the outlet regions while the
highest accelerations may be found at or close to the blade inlet. Figures 7‐5 (a‐ d) show
the prediction of erosion on runner blade surface. The highest erosion is predicted at
outlet blade surface than other component of turbine because the relative velocity is
highest at the outlet of the blades. Hence, turbulence erosion due to fine sand particles is
always susceptible at the trailing edge of the blades. In addition, because of high relative
velocity, most of the particles will move towards outer diameter in the runner outlet and
hence more effect of erosion is seen there. These simulations revealed that the outlet area
of pressure side of the runner is highly exposed to erosion, which was found at the
middle of the blade and towards the outlet area. This is due to the cross flow forming
horseshoe vortices, which is caused by incorrect blade leaning. The incorrect blade
leaning leads to cross flow between hub and shroud and such cross flow may intensify
erosion effect. Minor erosion spots were also observed at the leading edge and at lower
part of the outlet area on the suction side of the blades. The leading edge erosion is found
due to the local accelerations.

The predicted erosion rate density on pressure side is more than on the suction side of the
blade. This is even more in larger guide vane opening. The erosion rate density is more
towards the outlet as compare to the inlet of the blade. This is due to the blade profile‐tail
vortex flow, which leads to higher erosion rate density in the blade outlet.

(a) Pressure side (b) Suction side


(c) Leading edge and trailing edge (d) Blade with hub and shroud

Figure 7‐5 Predicted erosion pattern on turbine blades
7.3 EFFECT OF SEDIMENT CONCENTRATION ON EROSION
Based upon the simulation results obtained from Tabakoff erosion model, effects of
sediment concentrations on erosion rate density on turbine blade are presented in Figure
7‐6(a ‐ d). In order to investigate the effect of sediment concentration on erosion, different
sediment concentration rate were considered, while other parameters, i.e., sediment size,
shape, operating condition of a turbine and impingement angle were kept constant. It can
be observed from the figures that with the increase in concentration rate, the erosion rate
density increases, although it is not the same at different guide vane openings.

(a) 1 kg/s (b) 5 kg/s

(c) 10 kg/s (d) 50 kg/s

Figure 7‐6 Effect of concentration rate on erosion rate density of turbine blade

Additional simulations were performed with different sediment concentration rate at
different guide vane opening i.e., 16 degree (BEP), 20 degree and 22 degree (full load)
based upon Finnie erosion model. The variations of relative erosion rate density with
concentration are presented in Figure 7 ‐7 (a ‐ d). It can be seen from the figures that the
relative erosion rate density increases almost linearly on increasing concentration rate.
This shows that the erosion rate is constant for a given concentration and guide vane
opening.

Erosion rate density & concentration at BEP Erosion rate density & concentration at 20 degree GV.
1.2 1.2
Relative Erosion rate density
Relative Erosion rate density
1 1
0.8 0.8
[kg/m2s]
[kg/m2s]
0.6 Finnie 16 0.6 Finnie 20
0.4 0.4
0.2 0.2
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Concentration [kg/s] Concentration [kg/s]
(a) Guide vane angle 16‐degree (b) Guide vane angle 20‐degree

Erosion rate density & concentration at Full load Erosion rate density & Concentration at Full load
1.2 1.2
Relative erosion rate density
1 1
0.8 0.8
[kg/m2s]
[kg/m2s]
0.6 Finnie 22 onew ay coupling 0.6 Finnie 22 bothw ay coupling
0.4 0.4
0.2 0.2
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Concentration [kg/s] Concentration [kg/s]
(c) Guide vane angle 22 degree (one‐way coupled) (d) 22 degree (both‐way coupled)

Figure 7‐7 Variation of relative erosion rate density with concentration

7.4 EFFECT OF SEDIMENT SIZE ON EROSION
Based upon the simulation results obtained from Finnie erosion model, effects of
sediment sizes on erosion rate density on turbine blade are presented in Figure 7 ‐8 (a ‐
d). In order to investigate the effect of sediment size on turbine blades, different sediment
sizes were considered while other parameters, i.e., sediment concentration, shape factor
of sediment, operating conditions of a turbine, and impingement angle were kept
constant. It can be observed from the figures that within the range of sediment size i.e.,
0.15 – 2 mm investigated, the erosion rate density increase with the increase in sediment
size. However, the rate of increase of erosion rate density is not same for all sizes range
and guide vane opening.
(a) 0. 015 mm (b) 0.15 mm
(c) 1 mm (d) 2 mm

Figure 7‐8 Effect of sediment size on erosion rate density of turbine blade

Hence, additional simulations were performed at two guide vane opening i.e., at best
efficiency point (16 degree) and at full load (22 degree). The variations of relative erosion
rate density with mean sediment size are presented in Figure 7 ‐9 (a – b). It can be seen
from the figures that the relative erosion rate density increases with sediment sizes.
However, the erosion rate is not constant for all sizes range. For the particle smaller than
0.15 mm, owing to their light nature, they respond strongly to turbulences and thus gain
the erosive energy, causing damage to the blade surface. In this range, the turbulence
structure in the boundary layer dominates the erosion process. For larger particles sized

above 0.5 mm, the inertia force dominates their behaviour. They virtually do not respond
to small turbulent structures. Their erosive power comes from the main flow where they
are entrained. Owing to the significant momentum of these particles, they usually cause
severe damage to the blade surface. For particles with size between these two extremes
cases, a combined mode may exist subject to flow structure encountered, or one mode is
more dominant than the other mode. They are likely to follow along the outer hydraulic
contour of the blade, and their tendency for damage found progressively less. This
further shows that the erosion rate is constant for a given size and guide vane opening.
Erosion rate density & particle diamenter at BEP Erosion rate density & particle diameter at full load
1.2 1.2

Relativeerosion rate density
1 1
0.8 0.8
[kg/m2s]
[kg/m2s]
0.6 Finnie 16 bothw ay coupling 0.6 Finnie 22 bothw ay coupling
0.4 0.4
0.2 0.2
0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
Particle diameter[mm] Particle diameter [mm]
(a) BEP (b) Full load

Figure 7‐9 Variation of relative erosion rate density with sediment size

7.5 EFFECT OF SEDIMENT SHAPE FACTOR ON EROSION
In order to investigate the effect of sediment shape on erosion, different shape factor of
the particles i.e., 1, 0.75, 0.5 and, 0.25 were considered. Simulations were carried out at
same concentration rates for both operating conditions while other influencing
parameters were kept constant. The trajectories of the non‐spherical particles can be
modelled through the application of shape factors, allowing non‐uniform drag
distribution and bounce characteristics to be classified.

(a) Shape factor 1 (b) Shape factor 0.5

Figure 7‐10 Effect of sediment shape factor on erosion rate density of turbine blade

A particle shape factor equals to one signifies that the shape of the particle is close to the
spherical shape. In other words, it can be more or less regular shape of the particle. The
lower value of shape factor gives the more irregular shape of the particle. Based upon
simulations results obtained from Tabakoff erosion model, effects of shape factor on
erosion rate density on turbine blade due to shape factor 1 and 0.5 are presented in Figure
7 ‐10 (a ‐ b).
The figures show the prediction of erosion rate density along the blade surface at 16‐
degree guide vane opening which is at the maximum efficiency point. It can be clearly
seen that maximum erosion occurs in the blade profile with shape factor of 0.5. The
particle shape factor assumed spherical by default during simulations has value equal to
1. This always calculates the diameter of the particle from the mass of the particle divided
by its density, assuming it is spherical. A cross sectional area factor can be included to
modify the assumed spherical cross section area and to allow for non‐spherical particles.
This affects the drag force. The surface area factor is analogous to the cross sectional
factor. For non‐spherical particle, it is the ratio of the surface area to the surface area of
spherical particle with the same equivalent diameter. This affects both mass transfer and
heat transfer correlation and hence the erosion prediction. Similar erosion pattern is
observed in each case. However, the blade profile area is highly affected and larger
intensity of erosion rate is clearly observed in the case of non‐spherical shape of the
particle.

7.6 EFFECT OF OPERATING CONDITION ON EROSION
In order to investigate the effect of operating condition of turbine on erosion, simulations
were carried out at mainly two operating points, i.e., at best efficiency and at full load
conditions, while other influencing factors maintained constant. Based upon the
simulation results obtained from Tabakoff erosion model, effects of operating conditions
on erosion rate density on turbine blade are presented, as shown in Figure 7 ‐11 (a ‐ d).

(a) BEP with shape factor 1 (b) BEP with shape factor 0.5


(c) Full load with shape factor 1 (d) Full load with shape factor 0.5

Figure 7‐11 Effect of operating conditions on erosion rate density of turbine blade

These simulations revealed that the predicted erosion rate density at full load operation
point is higher than at best efficiency operation point. This is due to increased flow
turbulence and higher relative flow velocities at turbine runner outlet. Presence of
secondary flows and accompanying vortices lead to increased local velocities, higher
rotational motion and caused more separation of flow. Such flow conditions are
conducive to increased erosion rate and hence more erosion rate density has been
observed.
7.7 VALIDATION OF NUMERICAL SIMULATIONS
The experimental results discussed in section 4.3, and field studies at Cahua power plant
discussed in section 3.3 are compared with CFD results to validate numerical simulations.

The velocity of particle was calculated in highly swirl conditions similar to turbine flow
in curved path during experimental study, and it was determined that the size of the
particle is inversely proportional to the velocity of the particle. Similarly, different sizes
of spherically shaped particles, i.e., 0.015 mm, 0.05 mm, 0.15 mm, and 1 mm, were
considered and velocities of those particles around a turbine blade were computed
numerically. Figure 7‐12 (a ‐ d) shows the velocity for different size of the particle. It was
found that higher velocity is obtained for smaller size of the particle and lower velocity is
obtained for larger size of the particle. These simulations were carried out at same
operating conditions keeping all other influencing parameters constant except the size of
the particle. This is in agreement with the experimental results.

The experimental results further revealed that the velocity of particle also depends upon
the shape of the particles. It was found that velocity of regular shape of the particle is
higher than the other shapes of the particle. In order to compare this results, numerical
simulations were performed on different shape and size of the particles, and it was

determined that spherically shaped particle had higher velocity than particle with other
shapes. This can be clearly seen from Figure 7‐13. These simulations were carried out at
same operating conditions keeping all other influencing parameters constant except the
shape of the particle.

(a) 0.015 mm (b) 0.05 mm

(c) 0.15 mm (d) 1 mm

Figure 7‐12 Particle velocity for different size around turbine blade
Generally, the erosion rate density increases when the velocity of particle increases.
However, this has not occurred in the simulation analysis because of shape of the
particle. Direct measurements of the erosion rate density from experimental analysis for
comparison were not available; however, the drag coefficient of different shape of the
particles, which were found from experimental analysis, can be utilized to verify the
numerical predictions. The experiment results show that the triangularly shaped particles
have higher drag coefficient than the other shapes.

This means that higher coefficient of drag caused higher drag force. The particle of given
size flow along with the water and damage the blade surface. In general, many natural

particles are usually in non‐spherical shape. These particles will tend to have lower
velocities because both decrease in spheroid and increase in angularity tend to decrease
velocities. The impact area of these particles on the eroded surface is relatively larger and
tends to be directed perpendicular to the transport path. As a result, higher coefficient of
drag, higher rotational motion and more separation of flow are likely to occur and hence
more erosion rate is predicted.

(a) Shape factor 1 and velocity (b) Shape factor 1 and erosion rate

(c) Shape factor 0.5 and velocity (d) Shape factor 0.5 and erosion rate

Figure 7‐13 Particle velocity, shape factor and erosion rate around turbine blade

Furthermore, the numerically obtained erosion patterns, and similar field observations at
Cahua power plant Francis turbine components discussed in section 7.2, are in good
qualitative agreement for further validation of the numerical simulations.

Chapter 8
Conclusions and Recommendations
for Further Work
8.1 CONCLUSIONS
Sediment erosion analysis of a Francis turbine gives an indication of relative erosion
intensity and critical zones of erosion damage on the turbine components. The most
realistic numerical prediction of erosion is found on a turbine runner blade. The highest
velocities and accelerations occurred at outlet of the runner blade and more erosion was
predicted especially at the pressure side of the blade outlet and at the shroud.
Furthermore, unexpected sediment erosion was found at the suction side of the guide
vane where concept of critical diameter can be utilized. It has been concluded that if the
particle size in the water increases beyond critical particle sizes, the turbine should not be
operated at low guide vane opening.

The numerically obtained erosion pattern and the field observation and inspection at
Cahua power plant Francis turbine components are in good qualitative agreement. The
encouraging agreement shows that, for this application, numerical simulation really can
be used in a predictive manner. This information may serve as an input in an early stage
of turbine design process to identify the regions where special surface treatment is
necessary in order to increase the lifetime of the components for new hydropower
projects involving risks of sediment erosion.

The velocity of a particle is inversely proportional to the size of the particle, and it was
with other shapes. However, non‐spherical shape of the particles will tend to have lower
settling velocities because both decrease in spheroid and increase in angularity tend to
decrease velocities. Moreover, larger cross‐sectional areas tend to be directed
perpendicular to the transport path. As a result, higher coefficient of drag, higher
rotational motion and more separation of flow are likely to occur and hence more erosion
rate is predicted. The roles played by shape of the particle significantly affect erosion rate
prediction inside the Francis turbine components.

Chapter 7 Conclusions and Recomandations for Further Work
Furthermore, it has been found that the erosion process is strongly dependent on the
particle size, shape, concentration, and operating conditions of the turbine. The reduction
of the erosion is not only linked to the reduction of particle velocity but also is linked to
the reduction of separation of flow, which further depends on the shape, size, and
concentration of the particle. A significant reduction of erosion rate can be achieved by
operating turbine at best efficiency point. The full load operations reduced efficiency,
increased turbulence, and increased relative velocity of flow at outlet of the blade.

8.2 RECOMANDATIONS FOR FURTHER WORK
Sediment erosion problem in hydraulic machinery in general, and Francis turbine in
particular, is one of the complex operation and maintenance problem that causes major
challenges in the planning, design, operation and maintenance of hydropower projects in
high sediment‐laden Rivers. As discussed in section 2.7, there are large numbers of
factors responsible for the sediment erosion damage inside the turbine. This research on
sediment erosion in Francis turbine is one of the few attempts to address the problem
area. The present knowledge and findings are not enough to deal with this problem
completely but can be utilised to achieve one major step forward in erosion prediction
and prevention. The ideas that were emerging during this study period are considered in
this section. The following specific recommendations are made from the experience
gained through this study, keeping a broader perspective on this specific problem.

Because of the simplicity of the empirical erosion model implemented in ANSYS CFX in
steady state conditions, the present study demonstrated the relative erosion prediction on
a qualitative basis. However, from the result provided, it is clear that the role of sediment
characteristics and erosion pattern inside a turbine are addressed based upon the steady
state conditions, which are quite different from the realistic operating conditions. Hence,
a transient simulation analysis is recommended for predicting erosion on the turbine
components, which was not possible during this study period due to the limitations in
software and time.

The individual particles can be tracked throughout the computational domain, given
knowledge of the flow field. However, in its basic form, CFX has no ability to produce
erosion or particle impact data. To overcome such limitations, it is recommended that
extensive FORTRAN routines have to be developed and embedded within CFX, allowing
the relevant particle impact data to be computed, manipulated and displayed. The
coefficient of restitution has to be modified to include the consideration of both impact
angle and material type. Furthermore, validation of CFD by simplified lab test is, though
challenging, also recommended.

It is recommended to carry out numerical simulation analysis including leakage flow in
the guide vanes of Francis turbine because the losses due to the secondary leakage flow
in the guide vanes represents a relatively large part of the total loss of the turbine

Conclusions and Recomandations for Further Work Chapter 7
efficiency. This leakage flow arises due to the head covers deflection, which increases the
gap between the guide vanes and the head covers, when it is pressurized. Moreover,
when generating the grids, a tip‐ clearance can be added on the guide vane.

During this study, the drag coefficient and velocity relations were investigated with
respect to different size and shape of the particles at fixed guide vanes position, i.e., at 16‐
degree guide vane opening. However, it is recommended to carry out more tests with
adjustable vanes for statistical confidence. Furthermore, a rotational motion of the
particle inside the flow field can also be included for assessing different forces acting on
the particle so that the realistic expression for critical diameter of a particle other than
spherical shape can be achieved.

It is recommended to carry out simulation analysis in CFD computational domain that
corresponds to the water passages and can fully consider the non‐uniform flow field
incoming from the spiral casing so that better simulation of the flow can be achieved.

Hydraulic design modification is also recommended in a turbine runner to make them
sediment friendly without much sacrificing the efficiency, size and cost parameters. This
needs to be confirmed through model tests and numerical analysis.


References
REFERENCES
A. V. Levy, “The solid particle erosion behaviour of steel as a function of microstructure”,
Wear Vol. 68, 1981, pp. 269‐287.
V. Levy, and G. Hickey, “liquid solid particle slurry erosion of steels”, Wear Vol. 117,
1987, pp. 129 ‐158.
Al‐Bukhaiti, M. A., S. M. Ahmed, F. M. F. Badran, and K. M. Emara, “Effect of
impingement angle on slurry erosion behaviour and mechanisms of 1017 steel and high‐
chromium white cast iron”, Wear Vol. 262, No. 9 ‐ 10, 2007.
Anderson, J. D., “Computational Fluid Dynamics (The basics with applications)”:
McGraw‐Hill Series in Mechanical Engineering, 1995.
ANSYS, Release 11.0 “Solver Theory Guide”. 2006.
ANSYS, Release 11.0 “Turbo Grid Users Guide”. 2006.
ANSYS, CFD online, “http://www.cfd‐online.com”/, [Accessed on 31 August 2009].
Arnold, J. C., and I. M. Hutchings, “The mechanisms of erosion of unfilled elastomers by
solid particle impact”, Wear Vol. 138, 1990, pp. 33 ‐ 46.
Avery, H. S., “Work hardening in relation to abrasion resistance”, Material for mining
industry symposium, editor: R. Q. Barr, Greenwich, conf. USA, Climax Molybdenum Co.,
Page 43 ‐ 37, 1974.
Bahadur, S., and R. Badruddin, “Erodent particle characterization and the effect of
particle size and shape on erosion”, Wear Vol. 138, 1990, pp. 189 ‐ 208.
Bajracharya, T. R., B. ,Acharya, C. B. Joshi, R. P. Saini, and O. G. Dahlhaug, “Sand erosion
of Pelton turbine nozzles and buckets: A case study of Chilime hydropower plant”, Wear
Vol. 264, No. 3 ‐ 4, 2008, pp. 177 ‐ 184.
Basnyat, S., “Monitoring of sediment load and its abrasive effect in Jhimruk power
plant”, 1999, Proceedings, Optimum use of Run‐of‐River hydropower schemes,
Trondheim, Norway.
Base, J. L., and K. G. Matthew, “Influence of structure on abrasive resistance of A1040
steel”, Wear Vol. 14, Page 199 ‐ 206, 1969.
Batchelor, A. W., L. N. Lam, and M. Chandrasekaran, “Materials degradation and its
control by surface engineering”: Imperial college press, 2002.
Bardal E., “Korrosjon og korrosjonsvern”, 1985, Tapir, Trondheim (Norwegian)
Behera, U., P. J. Paul, K. Dinesh, and S. Jacob, “Numerical investigations on flow
behaviour and energy separation in Ranque‐Hilsch vortex tube”, International journal of
heat and mass transfer Vol. 51, No. 25 ‐ 26, 2008, pp. 6077 ‐ 6089.
R-1 HPN PhD Thesis, NTNU

References
Bergeron P. and Dollfus J., “The influence of nature of the pump mixture and hydraulic
characteristics on design and installation of liquid/ solid mixture pump”, Proc. 5th
conference on hydraulic turbines, 1952.
Bergeron P., “Similarity conditions for erosion caused by liquids carrying solids in
suspension, application to centrifugal pump impellers”, La Houille Blanche, 5, 1950,
B.H.R.A. translation T 408.
Bergeron P., “Consideration of the factors influencing wear due to hydraulic transport of
solid materials”, 1952, Proc. 2nd conference on hydraulic transport and separation of solid
materials.
Bergeron, S. Y., T. C. Vu, and A. P. Vincent, “Silt erosion in hydraulic turbines: The need
for real‐ time numerical simulations”, 2002.
Bingley, M. S., and D. J. OʹFlynn, “Examination and comparison of various erosive wear
modelsʺ, Wear Vol. 258, No. 1 ‐ 4, 2005, pp. 511 ‐ 525.
Bitter J. G. A., “A study of erosion phenomena”, 1962, part 1 and part II, Wear, Vol. 6, pp.
5‐ 21, and 169 – 190.
Bishwakarma, M. B., “Online sediment monitoring in RoR hydropower plants, PhD
thesis”, Department of hydraulics and environmental engineering, Trondheim: NTNU,
2008.
Brekke, H., “A discussion of Pelton turbines versus Francis turbines for high head
plants”, ASCE‐ IAHR/AIHR ‐ ASME joint symposium on design and operation of fluid
machinery, Colorado State University, Fort Collins, Colorado USA, 1978.
Brekke, H., E. Bardal, and R. T., ʺNorwegian research work on erosion resistive coating
for water turbinesʺ, XVII IAHR symposium Beijing China, 1994.
Brekke, H., H. Wu, Y. L., and B. Y. Cai, “Design of hydraulic machinery working in sand
laden water”. Book series on Hydraulic machinery, Vol.2 edited by Duan C. G. and
Karelin V. Y., 2002.
Brown, G. J., ʺErosion prediction in slurry pipeline tee‐junctionsʺ, Applied mathematical
modelling Vol. 26, No. 2, 2002, pp. 155 ‐ 170.
Chevallier, P., and A. B. Vannes, ʺEffects on a sheet surface of an erosive particle jet upon
impactʺ, Wear Vol. 184, No. 1, 1995, pp. 87 ‐ 91.
Chen Q. and Li D. Y., “Computer simulation of solid particle erosion”, 2003, Wear, Vol.
254, pp. 203‐ 210.
Clark, H. M., ʺOn the impact rate and impact energy of particles in a slurry pot erosion
testerʺ, Wear Vol. 147, 1991, pp. 165 ‐ 183.
Daun, C. G., Karelin, and W. Y., ʺAbrasive erosion and corrosion of hydraulic machinery
published by Imperial College Pressʺ, 2002.
HPN PhD Thesis, NTNU R-2

References
De Pellegrin, D. V., and G. W. Stachowiak, ʺSimulation of three‐dimensional abrasive
particlesʺ, Wear Vol. 258, No. 1 ‐ 4, 2005, pp. 208 ‐ 216.
Deng T., Bradley M. S. A. and Bingley M.S., “An investigation of particle dynamics
within a centrifugal accelerator type erosion tester”, 2001, Wear Vol., 247, pp., 55‐65.
Desale, G. R., B. K. Gandhi, and S. C. Jain, ʺEffect of erodent properties on erosion wear of
ductile type materialsʺ, Wear Vol. 261, No. 7 ‐ 8, 2006, pp. 914 ‐ 921.
Doby, M. J., A. F. Nowakowski, E. Nowak, and T. Dyakowski, ʺNumerical and
experimental examination of swirl flow in a cylindrical container with rotating lidsʺ,
Minerals engineering Vol. 20, No. 4, 2007, pp. 361 ‐ 367.
Dosanjh, S., and J. A. C. Humphery, ʺThe influence of turbulence on erosion by a particle
laden fluid jetʺ, Wear Vol. 102, 1985, pp. 303 ‐ 330.
Edwards, J. K., and B. S. Mclaury, ʺModelling solid particle erosion in elbows and
plugged teesʺ, Journal of Energy Resources Technology Vol. 123, 2001.
Eltvik Mette, “Sediment Erosion in Francis Turbines”, 2009, Master thesis, NTNU
Escaler, X., E. Egusquiza, M. Farhat, F. Avellan, and M. Coussirat, ʺDetection of
cavitation in hydraulic turbinesʺ, Mechanical Systems and Signal Processing Vol. 20, No.
4, 2006, pp. 983 ‐1007.
Fathy Hussein, M., and W. Tabakoff, ʺComputation and plotting of solid particle flow in
rotating cascades”, Computers & Fluids Vol. 2, No. 1, 1974, pp. 1 ‐ 15.
Finnie, I., ʺSome observations on the erosion of ductile metalsʺ, Wear Vol. 19, No. 1, 1972,
pp. 81 ‐ 90.
Finnie, I., “Erosion of surfaces by solid particle”, 1960, wear Vol. 3, pp. 87‐102.
Forder, A., M. Thew, and D. Harrison, ʺA numerical investigation of solid particle
erosion experienced within oilfield control valvesʺ, Wear Vol. 216, No. 2, 1998, pp. 184 ‐
193.
Gabitto, J., and C. Tsouris, ʺDrag coefficient and settling velocity for particles of
cylindrical shapeʺ, Powder Technology Vol. 183, No. 2, 2008, pp. 314 ‐ 322.
Gahr, K.‐H. Z., ʺMicrostructure and Wear of Materials, Elsevier, Amsterdamʺ, 1981.
Goodwin, J. E., W. Sage, and G.P. Tilly, ʺStudy of erosion by solid particlesʺ, Proc. Inst.
Mech. Engrs., 1969 ‐ 1970, pp. 279 ‐ 289.
Gregory, J. (John), “Particles in water properties and processes”, University College
London England, 2006, CRC press, Taylor & Francis Group, p. 9‐10
H. Drolon, F. Druaux, A. Faure, “Particles shape analysis and classification using the
wavelet Transform”, Pattern Recognition Letters, Volume 21, Issues 6‐7, June 2000, Pages
473‐482.

References
Haider, A., and O. Levenspiel, ʺDrag coefficient and terminal velocity of spherical and
nonspherical particlesʺ, Powder Technology Vol. 58, No. 1, 1989, pp. 63 ‐ 70.
Head, W. J., and M. E. Harr, ʺThe development of a model to predict the erosion of
materials by natural contaminantsʺ, Wear Vol. 15, 1970, pp. 1 ‐ 46.
Hojo, H., K. Tsuda, and T. Yabu, ʺErosion damage of polymeric material by slurryʺ, Wear
Vol. 112, 1986, pp. 17 ‐ 28.
Honeycombe, R. W. K., ʺSteels, microstructure and properties, Edward Arnold
(Publishers) Ltd.ʺ 1981.
IEC, ʺHydraulic Machines ‐ Guide for dealing with abrasive erosion in water, 62364ʺ,
2008.
I.V. Kragelsky, A.I. Zolotar, A. O. Sheiwekhman, “Theory of material wear by solid
particle impact‐ a review”, Tribology, International Vol. 18, Issue 1, 1985 , pp. 3 ‐ 11
Jin, H., F. Zheng, S. Li, and C. Hang, ʺThe role of sand particles on the rapid destruction
of the cavitation zone of hydraulic turbinesʺ, Wear Vol. 112, No. 2, 1986, pp. 199 ‐ 205.
J. A. C. Humphrey, “Fundamental of fluid motion in erosion by solid particle impact”,
International Journal of heat and fluid flow, vol.11, 1990.
Klujszo, L. A. C., P. K. Songfack, M. Rafaelof, and R.K. Rajamani, ʺDesign of a stationary
guide vane swirl air cleanerʺ, Minerals Engineering Vol. 12, No. 11, 1999, pp. 1375 ‐ 1392.
Klujszo, L. A. C., P. K. Songfack, M. Rafaelof, and R.K. Rajamani, ʺDesign of a stationary
guide vane swirl air cleanerʺ, Minerals Engineering Vol. 12, No. 11, 1999, pp. 1375 ‐ 1392.
L. Xu, and N. F. Kennon, ʺA study of the abrasive wear of carbon steels, Wear vol.148,
Page 101 ‐ 112.ʺ Wear, 1991.
Levy, A. V., N. Jee, and P. Yau, ʺErosion of steels in coal solvent slurriesʺ, Wear Vol. 117,
1987, pp. 115 ‐ 127.
Levy, A. V., and Y.‐ F. Man, ʺSurface degradation of ductile materials in elevated
temperature gas particle streams ʺ, Wear Vol. 111, 1986, pp. 173 ‐ 186.
Li, J., T. Deng, M. S. Bingley, and M. S. A. Bradley, ʺPrediction of particle rotation in a
centrifugal accelerator erosion tester and the effect on erosion rateʺ, Wear Vol. 258, No. 1 ‐
4, 2005, pp. 497 ‐ 502.
Li, S., ʺCavitation enhancement of silt erosion‐An envisaged micro modelʺ, Wear Vol.
260, No. 9 ‐ 10, 2006, pp. 1145 ‐ 1150.
Liu, X. B., L. D. Zhang, Z. Liang, and L. J. Cheng, ʺNumerical prediction of silt abrasive
erosion in hydraulic turbineʺ, Fluids Engineering Division Conference Volume 1, ASME,
1996.

References
Lysne, D., B. Clover, H. Støle, and E. Tesaker, “Hydraulic Design”: Norwegian University
of Science and Technology, Department of hydraulic and environmental engineering,
2003.
Mack, R., P. Drtina, and E. Lang, ʺNumerical prediction of erosion on guide vanes and in
labyrinth seals in hydraulic turbinesʺ, Wear, 1999, pp. 685 ‐ 691.
Mann, B. S., ʺParticle erosion‐‐a new concept of flow visualization and boundary layer
investigations of rotating machines at high Reynolds numbersʺ, Wear Vol. 223, No. 1 ‐ 2,
1998, pp. 110 ‐ 118.
Mann, B. S., ʺHigh‐energy particle impact wear resistance of hard coatings and their
application in hydro turbinesʺ, Wear Vol. 237, No. 1, 2000, pp. 140 ‐ 146.
Mann, B. S., and V. Arya, ʺAbrasive and erosive wear characteristics of plasma nitriding
and HVOF coatings: their application in hydro turbinesʺ, Wear Vol. 249, No. 5 ‐ 6, 2001,
pp. 354 ‐360.
Mann, B. S., and V. Arya, ʺAn experimental study to correlate water jet impingement
erosion resistance and properties of metallic materials and coatingsʺ, Wear Vol. 253, No. 5
‐ 6, 2002, pp. 650 ‐ 661.
Matsumura M. and Chen B.E., “Erosion resistant materials”, In: Duan C. G. and Karelin
V.Y. Eds., “Abrasive erosion and corrosion of hydraulic machinery”, 2002, pp. 235‐ 314.,
Imperial college press, London.
Mazumder, Q. H., S. A. Shirazi, B.S. McLaury, J. R. Shadley, and E. F. Rybicki,
ʺDevelopment and validation of a mechanistic model to predict solid particle erosion in
multiphase flowʺ, Wear Vol. 259, No. 1 ‐ 6, 2005, pp. 203 ‐ 207.
Mazur, Z., R. Campos‐Amezcua, G. Urquiza‐Beltrán, and A. García‐Gutiérrez,
ʺNumerical 3D simulation of the erosion due to solid particle impact in the main stop
valve of a steam turbineʺ, Applied Thermal Engineering Vol. 24, No. 13, 2004, pp. 1877 ‐
1891.
Ministry of Water Resources, (MoWR) G. O. N., ʺCountry report presented on 1st
international summit on Sustainable use of water for energyʺ, Kyoto, Japan, 2003.
Moore, M. A., ʺThe relationship between the abrasive wear resistance, hardness and
microstructure of ferritic materialsʺ, Wear Vol. 28 1974, pp. 59 ‐ 68.
Moore, M. A., ʺAbrasive wear, ASM Material Science Seminar on Fundamentals of
Friction and Wear of Materialsʺ, Pittsburgh, Pennsylvania, 1981.
Mosonyi, E., “Water power development, Budapest”: Akademiai Kiado, 1965.
Naidu, D. B. K. S., ʺDeveloping Silt Consciousness in the minds of Hydro Power
Engineersʺ, Silting Problems in Hydro Power Plants, New Delhi, India, 1999.
NEA, ʺNepal Electricity Authority Annual reportʺ, 2009.

References
Neopane, Hari Prasad, Ole. G. Dahlhaug, and M. Eltvik, ʺNumerical prediction of
particle shape factor effect on sediment erosion in Francis turbine bladesʺ, 6th
International conference on hydropower, Tromsø Norway, 2010.
Neopane, Hari Prasad, Ole. G. Dahlhaug, and B. Thapa, ʺAlternative Design of a Francis
Turbine for sand‐laden waterʺ, International Conference on Small Hydropower ‐ Hydro
Sri Lanka, (22 ‐ 24 October) Sri Lanka, Kendy, 2007
Neopane, Hari Prasad, Ole. G. Dahlhaug, and B. Thapa, ʺExperimental examination of
the effect of particle size and shape in hydraulic turbinesʺ, Waterpower XVI, Spokane,
Washington, USA: HCI 2009.
Oka, Y. I., K. Okamura, and T. Yoshida, ʺPractical estimation of erosion damage caused
by solid particle impact: Part 1: Effects of impact parameters on a predictive equationʺ,
Wear Vol. 259, No. 1 ‐ 6, 2005, pp. 95 ‐ 101.
Oka, Y. I., and T. Yoshida, ʺPractical estimation of erosion damage caused by solid
particle impact: Part 2: Mechanical properties of materials directly associated with
erosion damageʺ, Wear Vol. 259, No. 1 ‐ 6, 2005, pp. 102 ‐ 109.
Ole, G. Dahlhaug., and B. Thapa, ʺSand erosion in Francis turbine‐ a case study from
Jhimruk power plant, Nepalʺ, 22nd IAHR Symposium on Hydraulic Machinery and
System, Stockholm, Sweden, 2004.
Ole Gunnar Dahlhaug, and P. E. Skåre, ʺDevelopment of a sediment resistive Francis
runnerʺ, Hydro 2009: Progress – Potential ‐ Plans, Lyon, France: The international journal
on hydropower & dam, 2009.
Padhy, M. K., and R. P. Saini, ʺA review on silt erosion in hydro turbinesʺ, Renewable
and Sustainable Energy Reviews Vol. 12, No. 7, 2008, pp. 1974 ‐ 1987.
Padhy, M. K., and R. P. Saini, ʺEffect of size and concentration of silt particles on erosion
of Pelton turbine bucketsʺ, Energy Vol. 34, No. 10, 2009, pp. 1477 ‐ 1483.
Rao, P. V., and D. H. Buckley, ʺAngular particle impingement studies of thermoplast
materials at normal incidenceʺ, ASLE transactions Vol. 29, 1986, pp. 283 ‐ 296.
Rhodes, M., “Introduction to Particle Technology”: Wiley, 2008.
Richardson, R. C. D., ʺThe Abrasive Wear of Metals and Alloysʺ, Inst. Mech. Engrs.
Volume 182, Pt. 3A, Page 410 ‐ 414, 1967, pp. 410 ‐ 414.
Richardson, R. C. D., ʺLaboratory simulation of abrasive wear such as that imposed by
soil,ʺ Inst. Mech. Engrs. Volume 182, Pt. 3A, Page 29 ‐ 31, 1967.
Sandip Kumar Suman, and Bim Prasad Shrestha, “Shape feature extraction and pattern
recognition of sand particles and their impact”, 2005, Proc. SPIE Vol. 5996
Sargent, G. A., and D. Saigal, ʺErosion of Low‐Carbon Steel by Coal Particlesʺ, ASLE
Transactions Vol. 29, 1986, pp. 256 ‐ 266.

References
Shah, S. M., S. Bahudur, and J. D. Verhoeven, ʺErosion behaviour of high silicon bainitic
structure II: high silicon steelsʺ, Wear Vol. 113, 1986, pp. 279 ‐ 290.
Shah, S. M., J. D. Verhoeven, and S. Bahadur, ʺErosion behaviour of high silicon bainitic
structures,I: Austempered ductile cast ironʺ, Wear Vol. 113, 1986, pp. 267 ‐ 278.
Sheldon, G. L., and I. Finnie,ʺOn the ductile behaviour of nominally brittle materials
during erosive cuttingʺ, ASME Vol. 88B, 1966, pp. 387 ‐ 392.
Sheldon, G.L., and A. Kanhere, ʺAn investigation of impingement erosion using single
particlesʺ, Wear Vol. 21, No. 1, 1972, pp. 195 ‐ 209.
Shida, Y., and H. Fujikawa, ʺParticle erosion behaviour of boiler tube materials at
elevated temperatureʺ, Wear Vol. 103, 1985, pp. 281 ‐ 296.
Solero, G., and A. Coghe, ʺExperimental fluid dynamic characterization of a cyclone
chamberʺ, Experimental Thermal and Fluid Science Vol. 27, No. 1, 2002, pp. 87 ‐ 96.
Spurk, J. H., Fluid Mechanics: Springer, 1997.
Stachowiak, G. W., and A. W. Batchelor, ʺDimensional analysis modelling Tribological
dataʺ, International conference, Melbourne, Australia: National conference publication,
1987.
Stachowiak, G. W., and A. W. Batchelor, ʺAbrasive, Erosive and Cavitation Wearʺ,
Engineering Tribology (3‐Edition), Burlington: Butterworth‐Heinemann, 2006, pp. 501 ‐
551.
Stachowiak, G. W., and A. W. Batchelor, Engineering Tribology, Elsevier, 1993
Stephension, D. J., J. R. Nicholls, and P. Hancock, ʺParticle surface interactions during the
erosion of a gas turbine material (MarM002) by pyrolytic carbon particlesʺ, Wear Vol.
111, 1986, pp. 15 ‐ 29.
Sugiyama, K., S. Nakahama, S. Hattori, and K. Nakano, ʺSlurry wear and cavitation
erosion of thermal‐ sprayed cermetsʺ, Wear Vol. 258, No. 5‐6, 2005, pp. 768 ‐ 775.
T. Bovet, “Contribution to study of the phenomenon of abrasive erosion in the realm of
hydraulic turbines”, 1958, Bull. Tech., Suisse Romande, Vol. 84, pp. 37‐49.
Tabakoff, W., ʺStudy of single stage axial flow compressor performance deteriorationʺ,
Wear Vol. 119, 1987, pp. 51 ‐ 61.
Tabakoff W. , Hamed A., and, Metwally M., “Effect of particle size distribution on
particle dynamics and blade erosion in axial flow turbines”, 1991, ASME, Jr., of Engg. Gas
and power, Vol. 113, pp. 607‐615.
Tabakoff, W., ʺHigh‐temperature erosion resistance of coatings for use in turbo
machineryʺ, Wear Vol. 186‐187, No. Part 1, 1995, pp. 224 ‐ 229.
Tabakoff, W., A. Hamed, and B. Beacher, ʺInvestigation of gas particle flow in an erosion
wind tunnelʺ, Wear Vol. 86, No. 1, 1983, pp. 73 ‐ 88.

References
Thapa, Bhola, ʺSand erosion in hydraulic machinery, PhD thesisʺ, Trondheim: Norwegian
University of Science and Technology, Faculty of Engineering Science and Technology,
Department of energy and process engineering, 2004.
Thapa, Bhola, and Brekke Hermod, ʺEffect of sand particle size and surface curvature in
erosion of hydraulic turbineʺ, IAHR symposium on hydraulic machinery and systems
Stockholm, 2004.
Truscott, G. F., ʺA literature survey on abrasive wear in hydraulic machineryʺ, Wear Vol.
20, No. 1, 1972, pp. 29 ‐ 50.
Tullis, J. P., “Hydraulics of pipelines: pumps, valves, cavitation, transients”, New York:
Wiley, 1989.
Tsuguo N., “Estimation of repair cycle of turbine due to abrasion caused by suspended
sand and determination of desilting basin capacity”, 1999, Proc., International seminar on
sediment handling technique, NHA, Kathmandu.
Veerapen, J. P., B. J. Lowry, and M. F. Couturier, ʺDesign methodology for the swirl
separatorʺ, Aquacultural Engineering Vol. 33, No. 1, 2005, pp. 21 ‐ 45.
Versteeg, H. K., and W. Malalasekera, “An Introduction to Computational Fluid
Dynamics (The Finite Volume Method)”: Longman Scientific and Technical, 1995.
Walley, S. M., and J. E. Field, ʺThe contribution of the Cavendish Laboratory to the
understanding of solid particle erosion mechanismsʺ, Wear Vol. 258, No. 1 ‐ 4, 2005, pp.
552 ‐ 566.
W. A. Stauffier, “The abrasion of hydraulic plant by sandy water”, 1956, Schweizer
Archiv. Angew. Wiss. Technik., 24, pp 3‐30. Also shorter version in Metal Prog., January
1956.
Yust, C. S., and R. S. Crouse, ʺMelting at particle impact sites during erosion of ceramicsʺ,
Wear Vol. 51, 1975, pp. 193 ‐ 196.
Zhang, Y., E. P. Reuterfors, B. S. McLaury, S. A. Shirazi, and E. F. Rybicki, ʺComparison
of computed and measured particle velocities and erosion in water and air flowsʺ, Wear
Vol. 263, No. 1 ‐ 6, 2007, pp. 330 ‐ 338.
Zhao, K., C. Gu, F. Shen, and B. Lou, ʺStudy on mechanism of combined action of
abrasion and cavitation erosion on some engineering steelsʺ, Wear Vol. 162 ‐ 164, No. Part
2, 1993, pp. 811 ‐ 819.
Zhong, Y., and K. Minemura, ʺMeasurement of erosion due to particle impingement and
numerical prediction of wear in pump casingʺ, Wear Vol. 199, No. 1, 1996, pp. 36 ‐ 44.

Appendix A
This appendix presents the auto generated hydraulic turbine report during simulation
analysis at best efficiency point. The performance results obtained for the turbine rotor
were used to plot the hydraulic performances of a turbine discussed in Chapter 7. This
also includes the different blade loading charts, blade geometry plots and contour plots.

Title
Hydraulic Turbine Rotor Report at Best Efficiency Point (BEP)
Author: Hari P. Neopane
Date: 2010/01/13 11:37:10

Contents
1. File Report
Table 1 File Information for SST without particle_001
2. Mesh Report
Table 2 Mesh Information for SST without particle_001
Table 3 Mesh Statistics for SST without particle_001
3. Physics Report
Table 4 Domain Physics for SST without particle_001
Table 5 Boundary Physics for SST without particle_001
4. Tabulated Results
Table 6 Performance Results
Table 7 Summary Data
A-1 HPN PhD Thesis, NTNU

Appendix A
5. Blade Loading Charts
Chart 1 Blade Loading at 20% Span
6. Streamwise Charts
Chart 4 Streamwise Plot of Pt and Ps
Chart 5 Streamwise Plot of C
Chart 6 Streamwise Plot of W
Chart 7 Streamwise Plot of Alpha and Beta
7. Spanwise Charts
Chart 8 Spanwise Plot of Alpha and Beta at LE
Chart 9 Spanwise Plot of Alpha and Beta at TE
8. Blade Geometry Plots
Figure 1 Isometric 3D View of the Blade, Hub and Shroud
Figure 2 Meridional View of the Blade, Hub and Shroud
9. Blade Mesh Plot
Figure 3 Mesh Elements at 50% Span
10. Blade to Blade Plots
Figure 4 Contour of Pt at 50% Span
Figure 5 Contour of Ptr at 50% Span
Figure 6 Contour of Ps at 50% Span
Figure 7 Contour of W at 50% Span
Figure 8 Velocity Vectors at 20% Span
11. Meridional Plots
Figure 11 Contour of Mass Averaged Pt on Meridional Surface
Figure 12 Contour of Mass Averaged Ptr on Meridional Surface
Figure 13 Contour of Mass Averaged W on Meridional Surface
Figure 14 Vector of Area Averaged Cm on Meridional Surface
12. Circumferential Plots
Figure 15 Contour of Pt at Blade LE
Figure 16 Contour of Ptr at Blade LE
Figure 17 Contour of W at Blade LE
Figure 18 Contour of Pt at Blade TE
Figure 19 Contour of Ptr at Blade TE
Figure 20 Contour of W at Blade TE
13. Streamline Plot
Figure 21 Velocity Streamlines at Blade TE
Chart 10
HPN PhD Thesis, NTNU A-2

Appendix A
1. File Report
Table 1: File Information for SST without particle_001
Case SST without particle_001
C:/Documents and Settings/haripra.old/Desktop/16 final/Without
File Path
particle/SST without particle Rotating Frame/SST without particle_001.res
File Date 13 January 2010
File Time 09:18:30 AM
File Type CFX5
File
11.0
Version
Fluids Water
Solids None
Particles None
2. Mesh Report
Table 2: Mesh Information for SST without particle_001
Domain Nodes Elements
R1 95988 87320
Table 3: Mesh Statistics for SST without particle_001
Domain Maximum Edge Length Ratio
R1 9684.12
3. Physics Report
Table 4: Domain Physics for SST without particle_001
Name Location Type Materials Models
Turbulence Model = SST
Turbulent Wall Functions = Automatic
R1 Passage 3 Fluid Water
Buoyancy Model = Non Buoyant
Domain Motion = Rotating
S1 Passage Fluid Water
Domain Motion = Stationary

Appendix A
S2 Passage 2 Fluid Water
Table 5: Boundary Physics for SST without particle_001
Domain Name Location Type Settings
Mass And Momentum =
R1 to R1
Conservative Interface Flux
R1 Periodic 1 Side PER1 Passage 3 Interface
Turbulence = Conservative
1
Interface Flux
Mass And Momentum =
R1 to R1
2
Interface Flux
Mass And Momentum =
R1 to R1 Conservative Interface Flux
R1 PER1 2 OUTBlock Interface
Periodic Side 1 Turbulence = Conservative
Interface Flux
Mass And Momentum =
Interface Flux
Mass And Momentum =
R1 R1 to S2 Side 1 INFLOW Passage 3 Interface
Interface Flux
Flow Regime = Subsonic
Mass And Momentum =
Average Static Pressure
R1 R1 Outlet OUTBlock OUTFLOW Outlet
Relative Pressure = 1 [atm]
Pressure Averaging = Average
Over Whole Outlet
Wall Influence On Flow = No
R1 R1 Blade BLADE 3 Wall
Slip
OUTBlock HUB, Wall Influence On Flow = No
R1 R1 Hub Wall
Passage HUB Slip
OUTBlock SHROUD, Wall Influence On Flow = No
R1 R1 Shroud Wall
Passage SHROUD Slip

Appendix A
Flow Direction = Cylindrical
Components
Unit Vector Axial Component
= 0
Unit Vector Theta Component
= ‐0.9165
Unit Vector r Component = ‐
S1 S1 Inlet INFLOW Inlet 0.4
Mass Flow Rate = 540 [kg s^‐1]
Mass And Momentum = Mass
Flow Rate
Turbulence = Medium
Intensity and Eddy Viscosity
Ratio
Mass And Momentum =
S1 to S1
S1 Periodic 1 Side PER1 Passage Interface
1
Interface Flux
Mass And Momentum =
S1 to S1
2
Interface Flux
Mass And Momentum =
S1 S2 to S1 Side 2 OUTFLOW Passage Interface
Interface Flux
S1 S1 Blade BLADE Wall
Slip
S1 S1 Hub HUB Wall
Slip
S1 S1 Shroud SHROUD Wall
Slip
Mass And Momentum =
S2 R1 to S2 Side 2 OUTFLOW Passage 2 Interface
Interface Flux
Mass And Momentum =
S2 S2 to S1 Side 1 INFLOW Passage 2 Interface
Interface Flux

Appendix A
Mass And Momentum =
S2 to S2
S2 Periodic 1 Side PER1 Passage 2 Interface
1
Interface Flux
Mass And Momentum =
S2 to S2
2
Interface Flux
S2 S2 Blade BLADE 2 Wall
Slip
S2 S2 Hub HUB 2 Wall
Slip
S2 S2 Shroud SHROUD 2 Wall
Slip
The first table below gives a summary of the performance results for the turbine rotor.
The second table lists the mass or area averaged solution variables and derived quantities
computed at the inlet, leading edge (LE Cut), trailing edge (TE Cut) and outlet locations.
The flow angles Alpha and Beta are relative to the meridional plane; a positive angle
implies that the tangential velocity is the same direction as the machine rotation.

Table 6: Performance Results
Rotation Speed ‐62.8318 [radian s^‐1]
Reference Diameter 0.9850 [m]
Volume Flow Rate 11.0270 [m^3 s^‐1]
Head (LE‐TE) 214.8660 [m]
Head (IN‐OUT) 216.8290 [m]
Flow Coefficient 0.1836
Head Coefficient (IN‐OUT) 0.5551
Shaft Power 21894500.0000 [W]
Power Coefficient 0.0955
Total Efficiency (IN‐OUT) % 93.6583


Appendix A
Table 7: Summary Data
Quantity Inlet LE Cut TE Cut Outlet TE/LE TE‐LE Units
[kg m^‐
Density 997.0000 997.0000 997.0000 997.0000 1.0000 0.0000
3]
‐
Pstatic 1559540.0000 1353300.0000 208617.0000 203652.0000 0.1542 [Pa]
1144680.0000
‐
Ptotal 2397140.0000 2395050.0000 294253.0000 277153.0000 0.1229 [Pa]
2100790.0000
Ptotal
387806.0000 386660.0000 351069.0000 253295.0000 0.9080 ‐35591.6000 [Pa]
(rot)
U 51.2049 46.7384 30.9449 25.5322 0.6621 ‐15.7934 [m s^‐1]
Cm 10.4949 11.5194 12.2979 9.7932 1.0676 0.7784 [m s^‐1]
Cu ‐39.5827 ‐43.6257 0.4033 ‐1.7434 ‐0.0092 44.0290 [m s^‐1]
C 41.0781 45.6422 13.7646 11.2940 0.3016 ‐31.8777 [m s^‐1]
Distortion
1.0030 1.0444 1.0835 1.1949 1.0375 0.0391
Parameter
Flow
Angle: 75.1921 52.6567 ‐12.1631 11.4885 ‐0.2310 ‐64.8197 [degree]
Alpha
Wu 11.6222 3.1143 31.3542 23.7888 10.0679 28.2399 [m s^‐1]
W 15.7712 13.1874 33.7696 26.1468 2.5608 20.5822 [m s^‐1]
Flow
Angle: ‐48.9092 ‐33.7478 ‐69.5526 ‐66.8545 2.0610 ‐35.8048 [degree]
Beta


Appendix A

5. Blade Loading Charts

Chart 1: Blade Loading at 20% Span


Appendix A


6. Streamwise Charts

Chart 4: Streamwise Plot of Pt and Ps

Appendix A

Chart 5: Streamwise Plot of C

Chart 6: Streamwise Plot of W


Appendix A

Chart 7: Streamwise Plot of Alpha and Beta

Appendix A
7. Spanwise Charts

Chart 8: Spanwise Plot of Alpha and Beta at LE

Chart 9: Spanwise Plot of Alpha and Beta at TE

Appendix A
8. Blade Geometry Plots

Figure 1: Isometric 3D View of the Blade, Hub and Shroud

Figure 2: Meridional View of the Blade, Hub and Shroud

Appendix A
9. Blade Mesh Plot

Figure 3: Mesh Elements at 50% Span
10. Blade to Blade Plots

Figure 4: Contour of Pt at 50% Span

Appendix A

Figure 5: Contour of Ptr at 50% Span

Figure 6: Contour of Ps at 50% Span

Appendix A

Figure 7: Contour of W at 50% Span

Figure 8: Velocity Vectors at 20% Span

Appendix A



Appendix A
11. Meridional Plots

Figure 11: Contour of Mass Averaged Pt on Meridional Surface

Figure 12: Contour of Mass Averaged Ptr on Meridional Surface

Appendix A

Figure 13: Contour of Mass Averaged W on Meridional Surface

Figure 14: Vector of Area Averaged Cm on Meridional Surface

Appendix A
12. Circumferential Plots

Figure 15: Contour of Pt at Blade LE

Figure 16: Contour of Ptr at Blade LE

Appendix A

Figure 17: Contour of W at Blade LE

Figure 18: Contour of Pt at Blade TE

Appendix A

Figure 19: Contour of Ptr at Blade TE

Figure 20: Contour of W at Blade TE

Appendix A
13. Streamline Plot

Figure 21: Velocity Streamlines at Blade TE

Chart 10

Appendix B
This appendix presents the part of auto generated hydraulic turbine rotor report during
simulation analysis at full load operating condition. The performance results obtained for
the turbine rotor were used to plot the hydraulic performances of a turbine discussed in
Chapter 7.

Title: Hydraulic Turbine Rotor Report at Full Load
Author: Hari P. Neopane
Date: 2010/01/13 11:51:47

Contents
1. File Report
Table 1 File Information for 685 SST without particle after corecting_001
2. Mesh Report
Table 2 Mesh Information for 685 SST without particle after corecting_001
Table 3 Mesh Statistics for 685 SST without particle after corecting_001
3. Physics Report
Table 4 Domain Physics for 685 SST without particle after corecting_001
Table 5 Boundary Physics for 685 SST without particle after corecting_001
Table 6 Performance Results
Table 7 Summary Data

B-1 HPN PhD Thesis, NTNU

Appendix B
1. File Report

Table 1: File Information for 685 SST without particle after corecting_001
Case 685 SST without particle after corecting_001
C:/Documents and Settings/haripra.old/Desktop/22 final/685 SST without
File Path
particle after correcting/685 SST without particle after corecting_001.res
File Date 13 January 2010
File Time 11:40:03 AM
File Type CFX5
File
11.0
Version
Fluids Water
Solids None
Particles None
2. Mesh Report

Table 2: Mesh Information for 685 SST without particle after corecting_001
Domain Nodes Elements
R1 131425 119952
Table 3: Mesh Statistics for 685 SST without particle after corecting_001
Domain Maximum Edge Length Ratio
R1 1703.81
HPN PhD Thesis, NTNU B-2

Appendix B
3. Physics Report

Table 4: Domain Physics for 685 SST without particle after corecting_001
Name Location Type Materials Models
R1 Passage 3 Fluid Water
Domain Motion = Rotating
S1 Passage Fluid Water
S2 Passage 2 Fluid Water
Table 5: Boundary Physics for 685 SST without particle after corecting_001
Domain Name Location Type Settings
Mass And Momentum =
R1 to R1
1
Interface Flux
Mass And Momentum =
R1 to R1
2
Interface Flux
Mass And Momentum =
Interface Flux
Mass And Momentum =
Interface Flux
Mass And Momentum =
R1 R1 to S2 Side 1 INFLOW Passage 3 Interface Conservative Interface Flux

Appendix B
Interface Flux
Mass And Momentum =
Average Static Pressure
R1 R1 Outlet OUTBlock OUTFLOW Outlet
Relative Pressure = 1 [atm]
Pressure Averaging = Average
Over Whole Outlet
R1 R1 Blade BLADE 3 Wall
Slip
OUTBlock HUB, Wall Influence On Flow = No
R1 R1 Hub Wall
Passage HUB Slip
OUTBlock SHROUD, Wall Influence On Flow = No
R1 R1 Shroud Wall
Passage SHROUD Slip
Flow Direction = Cylindrical
Components
Unit Vector Axial Component
= 0
Unit Vector Theta Component
= ‐0.9165
Unit Vector r Component = ‐
S1 S1 Inlet INFLOW Inlet 0.4
Mass Flow Rate = 685 [kg s^‐1]
Mass And Momentum = Mass
Flow Rate
Turbulence = Medium
Intensity and Eddy Viscosity
Ratio
Mass And Momentum =
S1 to S1
1
Interface Flux
Mass And Momentum =
S1 to S1
2
Interface Flux
Mass And Momentum =
S1 S2 to S1 Side 2 OUTFLOW Passage Interface
Interface Flux

Appendix B
S1 S1 Blade BLADE Wall
Slip
S1 S1 Hub HUB Wall
Slip
S1 S1 Shroud SHROUD Wall
Slip
Mass And Momentum =
S2 R1 to S2 Side 2 OUTFLOW Passage 2 Interface
Interface Flux
Mass And Momentum =
S2 S2 to S1 Side 1 INFLOW Passage 2 Interface
Interface Flux
Mass And Momentum =
S2 to S2
1
Interface Flux
Mass And Momentum =
S2 to S2
2
Interface Flux
S2 S2 Blade BLADE 2 Wall
Slip
S2 S2 Hub HUB 2 Wall
Slip
S2 S2 Shroud SHROUD 2 Wall
Slip


Appendix B
The first table below gives a summary of the performance results for the turbine rotor.
The second table lists the mass or area averaged solution variables and derived quantities
computed at the inlet, leading edge (LE Cut), trailing edge (TE Cut) and outlet locations.
The flow angles Alpha and Beta are relative to the meridional plane; a positive angle
implies that the tangential velocity is the same direction as the machine rotation.
Table 6: Performance Results
Rotation Speed ‐62.8318 [radian s^‐1]
Reference Diameter 0.9835 [m]
Volume Flow Rate 13.9743 [m^3 s^‐1]
Head (LE‐TE) 205.9170 [m]
Head (IN‐OUT) 216.2730 [m]
Flow Coefficient 0.2338
Head Coefficient (IN‐OUT) 0.5554
Shaft Power 25261500.0000 [W]
Power Coefficient 0.1110
Total Efficiency (IN‐OUT) % 85.4894
Table 7: Summary Data
Quantity Inlet LE Cut TE Cut Outlet TE/LE TE‐LE Units
[kg m^‐
Density 997.0000 997.0000 997.0000 997.0000 1.0000 0.0000
3]
‐
Pstatic 1872070.0000 1758220.0000 277735.0000 198927.0000 0.1580 [Pa]
1480490.0000
‐
Ptotal 2490610.0000 2485050.0000 471754.0000 376068.0000 0.1898 [Pa]
2013290.0000
Ptotal
877233.0000 872237.0000 821108.0000 550511.0000 0.9414 ‐51128.8000 [Pa]
(rot)
U 51.2055 46.7381 30.8987 25.3838 0.6611 ‐15.8394 [m s^‐1]
Cm 13.2556 14.3394 15.5960 12.5494 1.0876 1.2566 [m s^‐1]
‐
Cu ‐32.2993 ‐34.6462 9.6645 7.5696 44.3107 [m s^‐1]
0.2789
C 35.2618 37.7048 19.7635 16.2150 0.5242 ‐17.9412 [m s^‐1]
Distortion 1.0042 1.0434 1.0745 1.2007 1.0297 0.0310

Appendix B
Parameter
Flow
‐
Angle: 65.0771 66.6155 ‐39.2267 ‐30.0992 ‐105.8420 [degree]
0.5889
Alpha
Wu 18.9062 12.0921 40.5654 32.9534 3.3547 28.4733 [m s^‐1]
W 23.2236 19.2514 43.5515 35.3219 2.2622 24.3000 [m s^‐1]
Flow
Angle: ‐56.2440 ‐39.9337 ‐70.7207 ‐70.9315 1.7710 ‐30.7870 [degree]
Beta

Hydro Sri Lanka 07
Appendix C

ALTERNATIVE DESIGN OF A FRANCIS TURBINE FOR SAND LADEN WATER

Hari Prasad Neopane1, Ole Gunnar Dahlhaug2, and Bhola Thapa3

This paper is presented at the Hydro Sri Lanka 07, International Conference on small
Hydropower, held in Kandy, Sri Lanka, 22 ‐ 24 October 2007. The paper has also
published in the proceedings for this symposium.

ABSTRACT

The main objective of this paper is to propose an alternative design of a Francis turbine for
minimizing the sand erosion effect in sand laden river. To achieve this objective, one erosion model
for hydraulic machinery has been selected and all technical and managerial aspects have been
considered in this study. Hydraulic turbine components operating in sand‐laden water subjected to
erosive wear. Erosion reduces efficiency and life of turbine. It also causes problem in operation &
maintenance. Himalayan rivers carry large amount of hard abrasive particles. Withdrawal of the
clean water from the river for power production is expensive due to design, construction and
operation of sediment settling basins. Even with the settling basins, 100 % removal of fine
sediments is impossible and uneconomical. The design of the Francis turbine can be done in two
main stages. The first step is the initial design, based on empirical data and the Euler’s turbine
equation. The second step is a three dimensional CFD analysis for fine tuning of the design,
normally based on the assumption of uniform flow field from the guide vanes cascade towards the
runner blade inlets. The hydraulic design strategy for the Francis turbine has played the major role
in the design. The main objectives in the design of the medium head Francis turbines are to reduce
the pressure pulsation, avoid cavitations, to reduce sand erosion and to gain maximum efficiency.
The ideal goal is to design a runner that has the widest possible operating range for head variations
beyond the normal design head, and that would require the minimal maintenance. The initial step
is the most important. The basic hydraulic design of the runner is done during this stage, and the
balancing of the pressure distribution on the blades by controlled parameters gives the designer an
important feeling, which assists in evaluating the result of the CFD analysis. This feeling is very
useful in helping to make the right changes in the geometry and to improve the performance of the
runner. The CFD analysis is also important in the study of the model turbine behaviour and forms
the basis for final tuning of the prototype Francis runners.

C-1 HPN PhD Thesis, NTNU

Hydro Sri Lanka 07
1 CONVENTIONAL DESIGN OF FRANCIS TURBINE

1.1 THE IMPORTANCE OF THE RUNNER

The hydraulic energy of the water is converted into mechanical shaft energy in the
turbine runner at the expense of the interaction between flow and the runner blades
constituting a rotating cascade of profiles. Runner is one of the most important
components of the turbines. Francis turbine is a reaction type of turbine where the
specific energy in front of the runner consists of partly pressure energy and partly kinetic
energy. For the high head type of the Francis turbine approximately 50 % of the energy is
converted into kinetic energy in front of the runner and there is a pressure drop through
the runner of approximately 50 % of the total energy drop through the turbine. So, the
design of the runner is very important.

1.2 INPUT PARAMETER

The main input parameter of the design is the net head (He), which is calculated from the
available head after deducting the friction losses on the penstock, flow at the full load (Q)
and available submersion. The available submersion is the distance between the
centerline of the turbine and the tailrace. For the purpose of this research work, the input
parameters have been selected based upon the 12 MW Jhimruk Hydro Power Plant,
[Flow at the full load (Q) = 7.05 m3/s, net head (Hn) = 201.5 m, and available submersion =
0 m], which is located in western part of Nepal.

1.3 CALCULATIONS OF THE DIMENSIONS OF THE RUNNER

The development of hydraulic turbines is aimed towards two goals: to achieve the
maximum efficiency and to avoid cavitation damage with the highest possible
circumferential speed and meridional velocity in the runner for the given submergence of
the turbine.

Figure 1.1 Main dimension of the runner
HPN PhD Thesis, NTNU C-2

Hydro Sri Lanka 07
For high head runners the cavitation problems are most likely to occur at the outlet
section of the blades. In order to study the cavitation a careful study of the blade loading
must be done. For such study, the curvature of shroud and crown as well as curvature of
the blades, the blade angles and the blade leaning are the important parameters. The
empirical relation for the outlet angle and peripheral velocity is given by [8]:

130 < β2 < 220 lowest values for the highest head
35 < U2 < 43 highest values for the highest head

For maximum efficiency, maximum peripheral speed at the outlet of the runner, U2 = 40
m/s at the ring for moderate setting assumed. In some case, the noise and vibrations
problems are likely to increase with the speed. To avoid the noise and vibrations
problem, higher values of peripheral velocity couldn’t be used. U2 >43 m/s is not
recommended. If the higher values of U 2 is chosen the blade outlet angle should be
reduced. Assume, β 2 = 17 0. The lower values of β 2 is also not recommended due to
welding constraints. From the maximum efficiency point of view, no swirl at the outlet
i.e. C2u= 0 m/s has been selected. The meridional velocity at the outlet of the runner at best
efficiency with no rotation in draft tube will get a normal value in this case. (See the
outlet velocity vector diagram). Then the following meridional velocity will be obtained:
C m 2 = U 2 ⋅ tan β 2 = 12 .2 m/s (1.1)

U 2 = ω 2 ⋅ R2
β2

C2=Cm2

V r2

Figure 1.2 Outlet velocity diagram

The ratio between the full load flow and the best efficiency flow

κ = Q = 1 .1 − 1 .3 [ ‐ ] (1.2)
Q*

We choose κ = 1.2

So, Q* = Q = 5.875 m3/s, (* denotes the best efficiency point)
1.2

The outlet diameter may now be found by assuming C m 2 to be constant across the outlet
area across the runner outlet diameter:

Hydro Sri Lanka 07

4⋅Q *
D2 = = 0.808 m (1.3)
(π ⋅ C m 2 )

The speed of the turbine will then be according to the chosen values of U 2 and β 2

60 ⋅ U 2 60 ⋅ 40
n= = = 946 rpm (1.4)
π ⋅ D2 π ⋅ 0.808

This will not normally be a synchronous speed and an adjustment of the speed has to be
made by adjusting the circumferential speed U2 and the diameter. We want to keep the
velocity triangle the same at the outlet therefore the outlet blade angle β2 is constant.

In order not to increase the necessary submergence of the turbine, the speed should be
reduced to nearest synchronous speed.

The synchronous speed is found by checking the formula

3000
n= [rpm] (1.5)
Z

Where Z = number of the pairs of the poles in the generator for an electric grid with a
frequency of 50 Hz.

If Z= 3, then n = 1000 rpm and ω = 104.71 rad/s
If Z = 4, then n = 750 rpm and ω = 78.53 rad/s

Using an empirical equation for calculation of the Net Positive Suction Head [4],

( a ⋅ C 2 m 2 + b ⋅ U 2 2 )
NPSH = [m] (1.6)
2⋅ g
Where a and b are constant and depends upon the speed number (Ω )

Ω = ω ⋅ Q *
*
[ ‐ ] (1.7)

If Ω < 0.55 then a= 1.12, b = 0.055
If Ω > 0.55 then a = 1.12, b = 0.1 ⋅ Ω *

Ω = 0.51 for n = 1000 rpm
Ω = 0.38 for n = 750 rpm

Hydro Sri Lanka 07

In both case Ω < 0.55 and thus a = 1.12 and b = 0.055

If the outlet angle β 2 shall be unchanged as well as the best efficiency flow Q* the
velocity vector diagram will be homogeneous and the following equation will be valid.

π ⋅ D2
2
⋅ C m 2 = Q * [m3/s] (1.8)
4

n ⋅ π ⋅ D2
U2 = [m/s] (1.9)
60

Then from simplification of above (1.8) and (1.9) equations we get

240 ⋅ Q *
D2 = 3 [m] (1.10)
π ⋅ n ⋅ tan β 2
2

For determining the required submergence of the runner the following equation has been
used [5],

hs = hb − hva − NPSH [m] (1.11)

NPSH is depends on turbine parameters. It is also a requirement for the turbine design
and has been denoted as a required net positive suction head for turbine. The NPSH of
the power plant has been included the necessary submergence of the runner (‐hs) the
barometric pressure (hb) and the vapor pressure (hva) which ultimately depends upon the
water temperature.

The practical experience shows that for not exceed the cavitation limits, the difference of
the barometric pressure and the vapor pressure in the above equation is around ‐10 m.
then the above equation becomes,

hs = 10 − NPSH [m] (1.12)

Before going to choose the best synchronous speed of the runner we have to calculates
the outlet diameter, outlet peripheral velocity of the runner and required submergence
by changing the outlet blade angles as shown in the following tabular form,


Hydro Sri Lanka 07

n = 1000 rpm

β2 D2 U2 hs

15 0.81 42.41 ‐2.42

16 0.79 41.5 ‐2.89

17 0.77 40.31 ‐3.21

n = 750 rpm

β2 D2 U2 hs

15 0.89 34.95 1.58

16 0.87 34.16 1.26

17 0.85 33.5 0.94

From the above result we can choose the best parameter of the outlet of the runner. The
selection of the parameters depends upon the fulfillment of the main objective of the
Francis runner, which have been defined in the beginning. So the best outlet parameters
and appropriate speed of the runner are as follows:

β2 = 17 0 D2 = 0.85 m U2 = 33.5 m/s Cm2 = 12.2 m/s n = 750 rpm.

The inlet dimensions may now be found by means of the Euler turbine equation. The
energy converted by the runner will be

E1 − E 2 = U 1 ⋅ C u1 − U 2 ⋅ C u 2 = g ⋅ H ⋅ η h [m2/s2] (1.13)

The hydraulic efficiency η h is the ratio of the available energy which is transfer to and
converted to mechanical energy by the runner and the net energy drop from the
upstream to down stream side of the turbine. (Mechanical losses, friction losses and
leakage losses are not taken into consideration in the hydraulic efficiency.) That is:


Hydro Sri Lanka 07
U1 ⋅ C u1 − U 2 ⋅ C u 2 [ ‐ ] (1.14)

ηh =
g⋅H

Principally every turbine is designed according to the available discharge Q, net head Hn
and a chosen optimal rotational speed n. These parameters however, differ over wide
ranges from one site to the other. For this variability it is very useful to have similarity
relations at hand for comparison means. In the following it is therefore, introduced some
ratio parameters which are designated as reduced quantities.
c
c= as reduced absolute velocity [ ‐ ] (1.15)
− 2 ⋅ g ⋅ Hn
u
u = as reduced peripheral velocity [ ‐ ] (1.16)
− 2 ⋅ g ⋅ Hn
v
v= as reduced relative velocity [ ‐ ] (1.17)
− 2⋅ g ⋅ H n
Velocity diagrams based on dimensional values of the velocities are valid for only one
single value of the net head Hn. If reduced velocities however, present the corresponding
velocity diagrams, these diagrams keep a similar shape. The velocity diagrams based on
reduced velocities are therefore beneficial because these diagrams are valid for any value
of Hn.
Additional useful reduced quantities are:

h is the reduced head
h= [ ‐ ] (1.18)
− Hn
Q
Q= is the reduced discharge [ ‐ ] (1.19)
− 2 ⋅ g ⋅ Hn
ω
ω= is the reduced angular velocity [ ‐ ] (1.20)
− 2 ⋅ g ⋅ Hn
By using dimensionless reduced parameters the above equation yields

(
η h = 2 ⋅ u1 ⋅ c u1 − u 2 ⋅c u 2 ) [ ‐ ] (1.21)

Assume that the hydraulic efficiency of 96 % and the approximately 50 % of the energy in
front of the runner is converted to kinetic energy, i.e.


Hydro Sri Lanka 07
2 2
C1 C1
= 0.5 ⋅ H n ⇒ = 0.5 ⇒ C1 ≅ 0.7 [ ‐ ]
2⋅g 2 ⋅g ⋅ Hn

By drawing the inlet velocity vector diagram, will try to obtain the stagnation point on
the blade inlet tip and the inlet angle so the relative velocity does not lead to separation
and possible inlet cavitations. (Especially for low head turbines)

Cu1 U1
α1 β1
r1
Cm1
V
C
1

Figure 1.3.The inlet velocity diagram

Besides the assumption that C1 ≅ 0.7 .Now will use Eq. (2.21) when assume the
hydraulic efficiency of the runner to be 96 % and no swirl conditions at the outlet of the
runner. The following equation yields:

u1 ⋅cu1 = 0.48 [ ‐ ] (1.22)

The inlet velocity diagram clearly illustrates that the smallest variation of the inlet flow
angle with variation in the guide vane angle is obtained if the angle between the absolute
and relative velocity in close to 90 0 at best efficiency point of operation.

Using the following empirical relation for calculation of the reduced dimensionless
circumferential speed of the blade inlet [8]

0.7 < u1 < 0.75 [ ‐ ] (1.23)

The reduced dimensionless circumferential speed of the blade inlet then from experience
may be chosen to 0.71, for a low specific speed and then

0.48
cu1 = = 0.68 [ ‐ ] (1.24)
0.71

The inlet diameter of the runner can now be found by the absolute value of U1

U 1 = 0.71 ⋅ 2 ⋅ g ⋅ H = 44.6 [m/s] (1.25)


Hydro Sri Lanka 07
The inlet diameter of the runner will then be

U ⋅ 60
D1 = 1 = 1.137 [m/s] (1.26)
n ⋅π

The meridional velocity at the inlet may from experience be chosen approximately 10 %
[8] lower than at the outlet of the runner in order to obtain a slight acceleration of the
meridional flow. (However, this choice will be different for different manufacturers due
to the philosophy of blade shape etc,) Thus,

C = 0.9 ⋅ 10.3 = 9.3
m1
[m/s] (1.27)

Then the height of the blade at the inlet =B1, can now be found by means of equation of
continuity as follows,

Q* [m] (1.28)
B1 = = 0.176
π ⋅ D1 ⋅ Cm1

The inlet blade angle = β1 can be found from the following relation (see the above inlet
velocity triangle)

⎛ C m1 ⎞
tan β 1 = ⎜⎜ ⎟⎟ ⇒ β1 = 77 o [degrees] (1.29)
⎝ U 1 − C 1u ⎠

Note: In this preliminary calculation the displacement of the blade thickness has been
neglected. By taken the blade thickness into consideration the blade angles must be
corrected due to increased relative velocity or meridional velocity.

2 EROSION MODELS AND ALTERNATE DESIGN OF RUNNER

The mathematical models of erosion are useful for design of turbine components,
sediment settling basin and optimization of hydropower plant operation in Sand‐laden
River. Most often, individual particle dynamics are used for developing erosion models.
Empirical and statistical relations are also often developed from experiments and field
experiences. As erosion studies are heading toward numerical modeling and simulations,
the importance of analytical models are increasing day by day. Truscott [6] has found
that the most often quoted expression for erosion is

Erosion ∝ velocity n
(2.1)


Hydro Sri Lanka 07
2.1 EROSION MODELS FOR HYDRAULIC MACHINERY

The erosion models are basically developed for specific purpose or condition. For
example, Bitterʹs model is developed for dry condition, hence it is not clear whether this
equation realistically predict erosion rate for wet condition or not. Few researchers have
presented models specifically for hydraulic machinery. Truscott [9] presented the
equation of Bergeron (1952) to predict the erosion rate of pump with simplified
assumptions such as pure sliding of spherical particles over the surface. He presented
equation for erosion as:

Erosion ∝ char ⋅ (ρ p − ρ )⋅ D 3p ⋅ p ⋅K
V3 (2.2)
D
Where Vchar is the characteristic velocity of liquid, D is the characteristic dimension of the
machine, ρp is density of particle, Dp is diameter of particle, p is number of particles per
unit surface area, ρ is density of liquid and K is experimental coefficient depending upon
nature of abrasive particles. This equation is proportional to experimental coefficient,
which is dependent on abrasive nature of particles. Karelin et al. [10] established the
equation for surface erosion based on impact effect of particles considering kinetic energy
of single particle.

They have anticipated deviation on erosion estimated by equation due to uncertainties
like non‐homogeneous particles, variable concentration, continuous alteration and
pulsation of velocities and pressure, non‐uniform flow distribution and so on. On the
contrary to laboratory tests, Tsuguo [11] established the relationship of factors concerning
erosion of turbines based on 8 years erosion data of 18 hydropower plants. The repair
cycle of turbine is determined according to calculation of turbine erosion from equation,
which gives erosion rate in term of loss of thickness per unit time.

W = λ ⋅ c x ⋅ f y ⋅ k 1 ⋅ k 2 ⋅ k 3 ⋅ Vchar
z
(2.3)
Where λ is turbine coefficient at eroded part; c is the concentration of suspended
sediment, V is relative velocity. The term f is average grain size coefficient on the basis of
unit value for grain size 0.05 mm. The terms k1 and k2 are shape and hardness coefficient
of sand particles and k3 is abrasion resistant coefficient of material. The x, y and n are
exponent values for concentration, size coefficient and velocity respectively. The value of
x and y are close to the unity and any deviation of this linear proportionality is
determined from plot of wear versus parameter. The values of z are proposed for
different turbine components based on relation between relative velocity and erosion.
Minimum value of z is proposed as 1.5 for Pelton bucket and maximum value is 3 for
Francis turbine runner. Equation 2.3 has been chosen for calculation of outlet dimensions
of modify turbine runner.


Hydro Sri Lanka 07
As, W ∝ V r and also assume all the coefficients in equation 2.3 has a constant values,
3
and

W = k1 ⋅ Vr
3
(2.4)

Where, k1 = 0.3 ‐0.5. For the discussion, the constant, k1 is chosen to be 0.3.

3 RESULTS AND DISCUSSION

The main erosion of the Francis turbine occurs at the outlet of the guide vanes and at the
outlet of the runner. In order to reduce the erosion rate of the turbine, the absolute
velocity at the inlet of the runner and the relative velocity at the outlet of the runner have
to be reduced. In this study, the flow and head are kept constant while the speed, inlet
peripheral velocity and outlet runner blade angle has been changed according to Table
3.1

Table 3.1 Variable input parameters
Speed rpm n 750 600 500 433 375 333 300 275
Inlet peripheral ‐ U 1 0.71 0.74 0.77 0.8 0.83 0.86 0.89 0.92
Velocity, reduced
Outlet blade angle degree β 2 17 19 21 23 25 27 29 31

Figure 3.1 Results from the new design of the Francis runner

Hydro Sri Lanka 07
The results show that the outlet diameter changes relatively little while the inlet diameter
changes drastically. The reduction of the erosion at the outlet is more than at the inlet.
This is shown in Figure 3.1. The inlet angle of the turbine has changed so that the design
looks more like a pump‐turbine. This means that the turbine will be larger than the
traditional design. The reduction of the erosion is linked to the reduction of the velocity
and therefore the size of the turbine increases. This result in a higher price of the turbine,
but it will reduce the maintenance costs during its lifetime.

It has been shown from the above calculation that the design of the runner can decrease
the sand erosion. If a Francis turbine designer combines the hydraulic design and coating
of the critical parts, a significant reduction of erosion can be achieved.

4 CONCLUSION

The reduction of the erosion is linked to the reduction of the velocity and therefore the
size of the turbine increases. This result in higher price of the turbine, but it will reduce
the maintenance costs during its lifetime.

It has been shown from the above calculation that the design of the runner can decrease
the sand erosion. However, if a Francis turbine designer combines the hydraulic design
and coating of the critical parts, a significant reduction of erosion can be achieved.

5 FURTHER WORK

This paper is a part of the PhD study of the author. Only few aspects of sand erosion
investigation are presented in this paper. CFD analysis has not done yet. The
modification in the conventional design of the runner has played significant role for
increasing manufacturing cost, reducing repair and maintenances cost, and increase the
life and efficiency of the runner. Hence, detail technical, managerial and economical
consideration is needed along with experimental and computational fluids dynamics
analysis. The author is expected to explore more about it and will be presented in PhD
thesis.


Hydro Sri Lanka 07
REFERENCES

[1] T. R. Bajracharya, C. B. Joshi, R. P. Saini and O. G. Dahlhaug, Efficiency improvement
of hydro turbines through erosion resistant design approach, ICPS Conference
Proceedings (2004), IOE/TU/IITB/IIIE
[2] T. R. Bajracharya, C. B. Joshi, R. P. Saini, O. G. Dahlhaug, Sand erosion of Pelton
nozzles and buckets: A case study of Chilime Hydropower Plant, Wear (2007) doi:
10.1016/j.wear.2007.02.021

[3] B. S. Mann, High‐energy particle impact water resistance of hard coating and their
application in hydro turbines, Wear (2000), 140‐146
[4] B. Thapa, Ole G. Dahlhaug, Sand erosion in hydraulic turbines and wear rate
measurement of turbine materials, CD ROM proceedings of international
Conference‐Hydro Africa 2003
[5] T. R. Bajracharya T. R., D. Sapkota, R. Thapa, S. Poudel, C. B. Joshi, R. P. Saini, O. G.
Dahlhaug (2006), Correlation Study on Sand Led Erosion of Buckets and Efficiency
Losses in High Head Power Plants, Proceedings of First National Conference on
Renewable Energy Technology for Rural Development 12‐14th October Kathmandu,
Nepal.
[6] B. Acharya, B. Karki, and L. Lohia, (2005), Study on the Sand Erosion Led Damages of
the Pelton Turbine Component and their effects (A Case Study of Chilime
Hydroelectric Project), BE Thesis, Department of Mechanical Engineering, Pulchowk
Campus, Institute of Engineering, Tribhuvan University
[7] B. Thapa, R. Shrestha, P. Dhakal, (2004) Sediment in Nepalese hydropower projects,
Proc, Int. Conf. on the great Himalayas: climate, health, ecology, management and
conservation, Kathmandu
[8] H. Brekke: Hydraulic Turbines, Design, Erection and Operation, NTNU, Trondheim,
2000.
[9] G. F. Truscott, A literature survey on abrasive wear in hydraulic machinery, Wear
(20), Elsevier (1972) pp 29‐49
[10] V. Y. Karelin, Fundamentals of hydro‐abrasive erosion theory, Imperial College press
(2002) pp 1‐52
[11] N. Tsuguo, Estimation of repair cycle of turbine due to abrasion caused by
suspended sand and determination of desiting basin capacity, Proceedings of
International seminar on sediment handling techniques, NHA, Kathmandu (1999)


Hydro Sri Lanka 07
LIST OF SYMBOLS AND ABBREVIATIONS
α Guide vane blade angle degrees k3 Abrasion resistance coefficient ‐
a Constant (used in equation 1.6) ‐ K Experimental coefficient ‐
A Cross sectional area of pipe m2 λ Turbine coefficient ‐
b Constant (used in equation 1.6) ‐ k Capacity ratio ‐
B Height of the runner m n Speed rpm
β Blade angle degrees NPSH Net Positive Suction Head m
c Concentration of the sediments ppm ω Angular velocity rad/s

C Absolute velocity at the inlet m/s ω Angular velocity, reduced value ‐
C Absolute velocity, reduced ‐ Ω Speed number ‐
value
D Diameter m p Number of particles 1/m2
E Energy m2/s2 ρ Density kg/m3
ηh hydraulic efficiency ‐ Q Flow rate m3/s
f Grain size coefficient ‐ Q Flow rate, reduced value ‐
g Gravity m/s2 U Peripheral velocity m/s
H Head m U Peripheral velocity, reduced ‐
value
h Head m V Relative velocity m/s
h Head, reduced value ‐ V Relative velocity, reduced value ‐
Hn Net head m Vchar Characteristic velocity m/s
k1 Shape constant ‐ Z Number of generator pole pairs ‐
k2 Hardness constant ‐ W Erosion rate ‐

SUB‐SYMBOLS
b Refers to the atmospheric pressure s Refers to the suction head
h Hydraulic x Concentration exponent
m Refers to the meridional direction y Size exponent
n Refers to the net value z Characteristic velocity exponent
p Refers to a particle 0 Refers to the centerline of the stay vane shaft
u Refers to the peripheral direction 1 Refers to the inlet of the turbine runner
va Refers to the vapor pressure 2 Refers to the outlet of the turbine runner
r Relative value * Refers to the best efficiency point of the
turbine
o Refers to the full load of the turbine


Waterpower XVI, Spokane, USA, 09
Appendix D
AN INVESTIGATION OF THE EFFECT OF PARTICLE SHAPE AND SIZE IN
HYDRAULIC TURBINES

Hari Prasad Neopane , Ole Gunnar Dahlhaug2, and Bhola Thapa3
1

This paper is presented at the Waterpower XVI 09, International Conference, held in
Spokane, USA, 27 ‐ 30 July 2009. The paper has also published in the proceedings for this
symposium.

ABSTRACT

Hydraulic turbine components operating in sand‐laden water are subject to erosive wear.
This wear is not only reduces efficiency and life of turbine but also causes problems in
operation and maintenance. A test rig designed to insert different shapes and sizes (1 to
10 mm) of particles was built. The flow in the guide vane cascade was simulated in
order to find the drag force of a particle in swirl flow. When a particle is flowing in swirl
flow, drag force and centrifugal force are two major forces influencing the particle
equilibrium. The equilibrium of these two forces provides the critical diameter of the
particle. A particle larger than the critical diameter moves away from the centre and hits
the wall, but a smaller particle flows along with the water. The critical diameter of the
particle is such that the particle continues to rotate in the turbine. Different shapes and
sizes of particles were tested with the same operating conditions and found that
triangularly‐shaped particles were more likely to hit the suction side of the guide vane
cascade. Furthermore, it supports the concept of separation of particles from streamlines
inside the test rig, and lead to the development of an operating strategy for a Francis
turbine processing sand‐laden water. A high‐speed digital camera was used to capture
images of particles. This study also permitted experimental verification of a given size
and shape of a particle as it orbits in the turbine until either the velocity components are
changed or the particle becomes smaller by fracturing due to impact against the outer
wall.

D-1 HPN PhD Thesis, NTNU

1. INTRODUCTION

Sand erosion is caused by impacts of sand particles against the solid surface. These
particles are contained in liquid flow medium and possess kinetic energy that is sufficient
to damage even metallic surface. The process of material removal from a solid surface by
a stream of impacting solid particles has been studied by a number of researchers, and
considerable information has been documented on the mechanism of material removal
processes. It has been well established that the erosion wear per impacting particle W is a
function of the impacting velocity ‘v’ as well as the particle diameter ‘d’ as given by the
equation W = K ⋅ v ⋅ d (Sheldon et al., 1972). Similarly, several other authors have
a b
provided simple expressions, based on wear test results, for wear rate as a function of
velocity, material hardness, and grain size or solids concentration (Truscott, 1972). The
one most often quoted is: Wear ∝ (Vel.) Where, the index n may vary depending on the
n
material and other factors involved; the most common value appears to be 3 (Truscott,
1972). Some more detailed analyses consider wear as affected by the forces and velocities
acting on a particle in a liquid flow. Truscott, 1972, stated that wear is directly
proportional to “abrasive power” of a particle impinging on a surface. Many equations
have been developed and presented in literature relating wear rate based upon velocity
of particles, but less information is found relating wear rate based upon the shape and
size of the particles.

Furthermore, solid particles of different shapes and sizes play a significant role in many
separation processes. Swirl flow is a primary mechanism for separation of particles that
creates a centrifugal force in many separation processes including in a gap between the
guide vane outlet and the runner inlet in a Francis turbine. The separation of particles of
different shapes and sizes depends upon the variations in behaviour of the particles
when subjected to the action of moving fluid. A particle falling in an infinitely large fluid
under the influence of gravity will accelerate until the gravitational force is exactly
balanced by the resistive force that includes buoyancy and drag. The constant velocity
reached at that stage is called the terminal velocity. The resistive drag force depends
upon an experimentally determined drag co‐efficient. The drag co‐efficient and terminal
velocities are important design parameters for many separation processes (Gabitto et al.,
2007).

Particles in water may range in size from a few manometers up to millimetres
dimensions. Natural particles also have various shapes, including rods, plate, and
spheres, with many variations in between, which make a treatment of particle size
difficult. The discussion is vastly simplified if the particles are considered spherical. In
this case only one size parameter is needed (the diameter) and hydrodynamic properties
are much more easily treated. Of course, non‐spherical particles also occur in natural
water, and some way of characterizing them is essential. A common concept is that of the
‘equivalent sphere’ based on a chosen property of the particles (Gregory, 2006). For
instance, an irregular particle has a certain surface area and the equivalent sphere could
HPN PhD Thesis, NTNU D-2

be chosen as that having the same surface area. The surface area of sphere, with diameter
d, is just π ⋅ d .
2
So, if the surface area of the non‐spherical particle is known, the equivalent spherical
diameter can easily be calculated. For an object of a given volume, the sphere has
minimum surface area and the volume (or mass) of a given particle must be equal to or
less than that of the sphere. Another common definition of equivalent spherical diameter
is based on sedimentation velocity. In this case, from the sedimentation velocity and
density of particle, the diameter of a sphere of the same material that would settle at the
same rate can be calculated. This is sometimes called the ‘Stokes equivalent diameter’.

In this paper, the results of an experimental study of the effect of various shapes and
sizes of different materials in swirl flow are reported. The main focus of the paper is to
discuss the critical diameter of particles, and also support the concept of separation of
particles from streamlines inside the test rig and to establish the operating strategy for a
Francis turbine processing sand‐laden water.

1.2 MOTION OF PARTICLE THROUGH FLUIDS

There are three forces acting on a particle moving through a fluid: 1) The external force,
gravitational or centrifugal; 2) The buoyant force, which acts parallel with the external
force but in the opposite direction; 3) The drag force, which appears whenever there is
relative motion between the particle and the fluid.
1.1.2 EQUATIONS FOR ONE‐DIMENSIONAL MOTION OF PARTICLE THROUGH
FLUID
Consider a particle of mass m moving through a fluid under the action of an external
force Fe . Let the velocity of the particle relative to the fluid be u , let the buoyant force on
the particle be FB and let the drag be FD , then,
du
m⋅ = Fe − FB − FD [N] (1)
dt
The external force can be expressed as a product of the mass and the acceleration a e of
the particle from this force,
Fe = m ⋅ ae [N] (2)
The buoyant force is, be Archimedes’ law, the product of the mass of the fluid displaced
by the particle and the acceleration from the external force. The volume of the particle

m ⎛ m ⎞
is , the mass of fluid displaced is ⎜ ⎟ ⋅ ρ , where ρ is the density of the fluid. The
ρp ⎜ρ ⎟
⎝ p ⎠
buoyant force is then,
m ⋅ ρ ⋅ ae
FB = [N] (3)
ρp
The drag force is
C d ⋅ U 2 ⋅ ρ ⋅ Ap
FD = [N] (4)
2
Where C d is the drag coefficient, A p is the projected area of the particle in the plane
perpendicular to the flow direction.
By substituting the forces into Eq. (1), we have
ae ⋅ ρ C d ⋅ U ⋅ Ap ⋅ ρ ⎛ ρ p − ρ ⎞ C d ⋅ U 2 ⋅ Ap ⋅ ρ
2
du
= ae − − = ae ⋅ ⎜ ⎟−
dt ρp 2⋅m ⎜ ρ ⎟ 2⋅m
⎝ p ⎠
[m/s2] (5)
Motion from gravitational force:
In this case, a e = g [m/s2]
du ⎛ ρ p − ρ ⎞ C d ⋅ U 2 ⋅ Ap ⋅ ρ
= g ⋅⎜ ⎟− [m/s2] (6)
dt ⎜ ρ ⎟ 2 ⋅ m
⎝ p ⎠
Motion in a centrifugal field:
ae = r ⋅ ω 2 [m/s2]
du 2 ⎜
⎛ ρ p − ρ ⎞ C d ⋅ U 2 ⋅ Ap ⋅ ρ
= r ⋅ω ⋅ ⎟− [m/s2] (7)
dt ⎜ ρ ⎟ 2⋅m
⎝ p ⎠

In this equation, u is the velocity of the particle relative to the fluid and is directed
outwardly along a radius.
1.2 TERMINAL VELOCITY
In gravitational settling, g is constant. Also, the drag always increases with velocity. The
acceleration decreases with time and approaches zero. The particle quickly reaches a
constant velocity which is the maximum attainable under the circumstances. This
maximum settling velocity is called terminal velocity.
du ⎛ ρ p − ρ ⎞ C d ⋅ U 2 ⋅ Ap ⋅ ρ
= g ⋅⎜ ⎟− = 0 [m/s2] (8)
dt ⎜ ρ ⎟ 2⋅m
⎝ p ⎠
2 ⋅ g ⋅ (ρ p − ρ )⋅ m
ut = [m/s] (9)
Ap ⋅ ρ p ⋅ C d ⋅ ρ
In motion from a centrifugal force, the velocity depends on the radius and the
acceleration is not constant if the particle is in motion with respect to the fluid. In many
du du
practical use of centrifugal force, is small. If is neglected, then
dt dt
2 ⋅ r ⋅ (ρ p − ρ ) ⋅ m
ut = ω ⋅ [m/s] (10)
Ap ⋅ ρ p ⋅ C d ⋅ ρ
Motion of spherical particles:
π ⋅ Dp3 ⋅ ρ p π ⋅ Dp 2
If the particles are spheres of diameter D p , then, m = and, A p =
6 4
Substitution of m and A p into the equation for u t gives the equation for gravity settling
of spheres:
4 ⋅ g ⋅ (ρ p − ρ ) ⋅ D p
ut = [m/s] (11)
3 ⋅ Cd ⋅ ρ

1.3 DRAG COEFFICIENT
Drag coefficient is a function of Reynolds number. The drag curve applies only under
restricted conditions:
• The particle must be a solid sphere;
• The particle must be far from other particles and the vessel wall so that the flow
pattern around the particle is not distorted;
• It must be moving at its terminal velocity with respect to the fluid.
u ⋅ Dp ⋅ ρ
Particle Reynolds number: Re , p = [ ‐ ] (12)
μ
Where, u : relative velocity of particle, D p : diameter of the particle, ρ : density of
fluid, μ : viscosity of fluid.
Stokes’ law applies for particle Reynolds number less than 1.0
24
Cd = [ ‐ ] (13)
Re , p
From Eq. (4)
Cd ⋅ ut ⋅ Ap ⋅ ρ Dp ⋅ π 1 Dp ⋅ π
2 2 2
24 24 ⋅ μ
FD = = ⋅ ut ⋅ ρ ⋅ ⋅ = ⋅ ut ⋅ ρ ⋅ = 3 ⋅ π ⋅ μ ⋅ ut ⋅ D p
2 2
2 Re, p 4 2 ut ⋅ D p ⋅ ρ 8
FD = 3 ⋅ π ⋅ μ ⋅ u t ⋅ D p [N] (14)
From Eq. (11)
4 ⋅ g ⋅ (ρ − ρ p ) ⋅ D p 4 ⋅ g ⋅ (ρ − ρ p ) ⋅ D p g ⋅ (ρ p − ρ ) ⋅ D p ⋅ u t
2
ut = = =
3 ⋅ Cd ⋅ ρ 24 ⋅ μ 18 ⋅ μ
3⋅ ⋅ρ
ut ⋅ D p ⋅ ρ
g ⋅ D p ⋅ (ρ p − ρ )
2
Ut = [m/s] (15)
18 ⋅ μ

In the case of centrifugal, g is replaced by r ⋅ ω
2
For 1000 < Re , p < 200,000 , use Newton’s law
C d = 0.44 [ ‐ ] (16)
FD = 0.055 ⋅ π ⋅ D p ⋅ u t ⋅ ρ
2 2
[N] (17)
g ⋅ D p ⋅ (ρ p − ρ )
ut = 1.75 ⋅ [m/s] (18)
ρ

At higher relative velocities, the inertia of the fluid begin to dominate (the fluid must
accelerate out of the way of the particle). Analytical solution of the Navier‐Stokes
equations is not possible under these conditions. However, experiments give the
relationship between the drag co‐efficient and particle Reynolds number in the form of
the so‐ called standard drag curve. Four different regions are identified: the Stokes’ law
region, the Newton’s law region in which drag coefficient is independent of Reynolds
number, an intermediate region between the Stokes and Newton’s regions; and the
boundary layer separation region. The Reynolds number ranges and drag co‐efficient
correlations for these regions are given in Table 1, (Martin, 2008).
Table 1
Region Stokes Intermediate Newton’s law

Re , p range < 0 .3 0.3 < Re , p < 500 500 < Re, p < 2 × 10 5
Cd 24
Re , p

24
Re , p
(
⋅ 1 + 0.15 ⋅ Re, p
0.687
)
≈ 0.44
Several correlations have been proposed for C d over the entire range, the one presented
in equation (19) is that of Haider et al. (1989), which is claimed to fit the data with a root
mean square deviation of 0.024.
⎛ ⎞
⎜ ⎟
Cd =
24
Re , p
(
⋅ 1 + 0.1806 ⋅ Re , p
0.6459
) ⎜ 0.4251 ⎟
+⎜
6880.95 ⎟
[ ‐ ] (19)
⎜⎜ 1 + ⎟
⎝ Re , p ⎟⎠

1.4 PARTICLE IN SWIRL FLOW
When particles travel in swirling flow

Relative velocity of particle
Orbit for particle rotation
toward a turbine’s outlet, which is
Absolute velocity of water
located in the centre as shown in Figure
1.1, the particles will be exposed to two Inlet
Tangential velocity of particle
main forces. Centrifugal force (Fc )

vanes
moves the particles away from the outlet
centre, while the drag force (FD ) pulls Outlet
them toward the outlet, which is in the

centre, or toward the runner in the case
of a Francis turbine (Thapa et al., 2004). Figure1.1 Illustration of particle flow in
These two forces are given by the spiral swirl
following equations.
π ⋅d3 π ⋅ d 3 Ct
2
Fc = ρ p ⋅ ⋅ r ⋅ω = ρ p ⋅
2
⋅ [N] (20)
6 6 r
1
FD = ⋅ C d ⋅ ρ ⋅ Vr ⋅ A p
2
[N] (21)
2
Following three conditions prevail in such case:
• Particle will stay at the orbit of radius r, if Fc = FD
• Particle will strike outer wall, if Fc > FD
• Particle will flow along with water towards the centre of the tank, if Fc < FD

At equilibrium, these two forces balance each other and a particle of a given diameter
will stay at an orbit of radius (r ) until either the velocity component is changed or
particles become smaller by fracture due to impact. The diameter of a particle (d ) for the
equilibrium condition is given by equation (22). This is called the critical diameter.

⎛ ρ ⎞ ⎛ Vr
2
3 ⎞
d c = ⋅ Cd ⋅⎜ ⎟⋅⎜ ⎟⎟ ⋅ r [m] (22)
⎜ρ ⎟ ⎜C
4 ⎝ p ⎠ ⎝ t ⎠

The drag force is caused by the relative velocity of particles in radial direction (i.e.
towards the centre of the tank), and centrifugal force is caused by the velocity of particle
in tangential direction (i.e. away from the centre of the tank).

2. DESCRIPTION OF TEST RIG AND TEST PROCEDURE

A test rig was developed and designed at the waterpower laboratory, Norwegian
University of Science and Technology, Norway, to create strong swirl flow, similar to the
flow between the guide vane outlet and the runner inlet of a turbine. There was a
provision to introduce particles, with sizes ranging from 1 to 10 mm, into the swirl, and
to observe the motion of the particles from Plexiglas windows located on the cover of the
tank using a high‐speed digital camera.

The experimental set up, as shown in
Figure2.1, consisted of a main tank (1100
mm diameter and 700 mm height), 400
mm diameter inlet pipe and outlet cone
Ø9 0
Ø1
with valve. The main tank and other
Ø9
10
00
5
0
components of the test rig were 0
Ø70
Ø
designed and dimensioned for 50 m of
40
50
0
Ø2
head in order to carry out the
experiment in high velocity. The main
tank consisted of two compartments Ø1
60
with a 250 mm diameter opening at the
centre of the plate. This plate divides the Plan view
main tank into two compartments. Inlet pipe
Thirty‐six curved vanes, which resemble PlexiGlas Windows

Particle Injector
guide vanes of a Francis turbine, with a (<10 mm dia.)
Additional box
radius of 100 mm toward the inlet and a
Pitot Tube
90mm straight section toward the outlet, PlexiGlas
303
were fixed at a pitch circle diameter of Vanes
900mm to the middle plate in such a

Hole for Pitot Tube 250
way that the inlet velocity direction
350
should be almost in the radial direction 400
and the outlet would be 10 0 to the

tangent. These vanes were located in
Valve at Close position
between the upper part of the middle
plate and the bottom part of a 50mm
thick, 950 mm diameter transparent Particle Filter
Plexiglas plate. From this arrangement,

the swirl flow in between the Plexiglas Sectional view

and the middle plate could be clearly
Figure2.1 The experimental set up
observed. The edge of the Plexiglas was
made uniform to ensure uniform flow.

The top cover of the tank was fitted with five transparent windows of Plexiglas. Four of
those transparent windows of 160 mm diameter were located at 90 0, 180 0, 270 0 and 360 0

respectively at the pitch circle diameter 700 mm to observe the motion of the particles,
and the fifth window of 275 mm diameter was located at 315 0 at 700 mm pitch circle
diameter to measure the velocity of the flow (after inserting a Pitot tube) and to observe
the velocity of particle. The particle injection point was located at 279 mm from the centre
of the tank. The particle was released at the bottom of the Plexiglas through a 15 mm
diameter pipe and valves arrangement. Particles up to 10 mm in diameter were tested in
this experiment.

A manometer was fitted into the injection pipe and valves arrangement in order to
measure the inlet pressure of the tank in the swirl flow field. There were two additional
manometers located in the system. One was connected to an air‐bleeding valve located at
390mm from the centre of the tank for measuring the inlet pressure, and another was
fitted at the outlet of cone for measuring the pressure at the outlet. The flow rate was
calculated with the help of an ultrasonic flow meter fitted in the inlet pipe.

The main purpose of this experiment was to determine the velocity of the particle that
would flow along with water in the given flow condition. To achieve this, firstly, the
operating condition was set at a certain velocity level by controlling the valve opening.
Then the particle was injected and the motion of the particle was observed through the
Plexiglas windows with the help of the high‐speed camera. The particle image was saved
and analysis was performed based upon the time required to move the particle from one
fixed location to another fixed location. Location of the particle was identified based
upon the radial and angular position of particle inside the test rig. In order to calculate
the correct location of the particle inside the test rig, radial and angular markings were
inscribed on the middle plate at diameters of 219 mm, 239 mm, 259 mm, 279 mm, 299
mm, 309 mm and 22.5 degrees interval respectively. Using these markings, six different
circles and eight different angular lines could be clearly observed in the middle plate.

3. RESULTS AND DISCUSSION

Velocities of particles of different sizes
and shapes are shown in Figures 3.1 and Velocity of particle for different size & shape
3.2. The velocity of the particle will 13
increase if the operating head increases.
11.5
The size of the particle is inversely 6mm ran. sand
proportional to the velocity of the 10 5mm sp.steel

4.74mm cylin.alum.
particle, and it also depends upon the 8.5
shape of the particle. Different shapes of
7
particles were tested, and it was 20 25 30 35 40 45 50
Head [m]
determined that spherically‐shaped
particles had higher settling velocities Figure 3.1 Velocity of particle for
than particles with other shapes. different sizes and shapes

Many natural particles are not usually in
Velocity of particle for same size but different shape
spherical shape. These particles will
15
tend to have lower settling velocities

because both decreases in sphericity and 13
increases in angularity tend to decrease 11

Spherical
triangularl
cylindrical
settling velocities. Furthermore, larger 9
cross‐sectional areas tend to be directed
7
perpendicular to the transport path. 20 25 30 35 40 45 50
Head [m]
Separation of flow is more likely to
occur for non‐spherical particles, and
Figure 3.2 Velocity of particle for same
less spherical particles may rotate,
sizes but different shapes
follow wobbling paths, etc.

If centrifugal force and drag force on the and it was determined that triangularly‐
particle are equal, the particle will rotate shaped particles were more likely to hit
exactly at the injecting radius, but if the suction side of the guide vane
those forces do not balance each other, cascade. The radius at which the particle
there will two possibilities. The particle is moving, called rotation radius. This
either moves toward the inner radius rotation radius also varied with
and ultimately sinks or moves toward different shapes of particles, which
the outer radius and ultimately hits the further supports the influence of shape
sides of the vanes. The equilibrium factor for particle velocity calculation.
condition was observed for a given
particle after manipulating the flow
velocity, providing verification that the Rotation radius for same size but different shape
different forces were balanced in the test 0.4
rig. This also revealed that the particle

Rotation Radius [m]
with a given diameter would stay at the 0.3
Sand particle 3-5mm size
orbit of the injecting radius until either Steel particle 5mm size
Trian. ceramic 6mm size
the velocity components were changed 0.2
or the particle became smaller by

fracturing due to impact with the outer 0.1
0 10 20 30 40
wall. Different shapes of particles were Flow Rate [LPS]

tested with the same operating Figure 3.3 Rotation radius for particles
conditions as shown in Figure 3.3, of the same size but different shapes

Drag co‐efficient was also calculated and plotted versus different shapes, sizes and
particle Reynolds number as shown in Figures 3.4 and 3.5.

Drag Co-efficient and Size

0.475 12 Drag Co-efficient for different shape
0.472
0.470 9 0.471
Drag Co-efficient
Drag Co-efficient
Size [mm]
0.470 cylin. shape (irregular)
drag co-efficient
0.465 6 trian. shape (regular)
size of steel particle
0.470 sp. shape (regular)
0.469
0.460 3
0.468
5.E+04 6.E+04 6.E+04 7.E+04 7.E+04 8.E+04 8.E+04 9.E+04
0.455 0
2.E+04 3.E+04 4.E+04 5.E+04 6.E+04 7.E+04 Re,p
Re,p

Figure 3.4 Drag co‐ efficient and size Figure 3.5 Drag co‐ efficient & shapes
The drag co‐efficient, C d is a non‐dimensional number that depends on the shape of the
particle, the fluid kinematic viscosity and grain size. It has been found that the effect of
the shape of non‐spherical particles on their drag co‐efficient can be defined in terms of
its sphericity. Moreover, shape affects drag co‐efficient far more in the Intermediate and
Newton’s law regions than in the Stokes’ law region. However, the variation of drag co‐
efficient in the Newton’s law region is not so significant, but the influence of size and
shape of the particle has been clearly demonstrated. The result shows that the
triangularly‐shaped particles have a higher drag co‐efficient than other shapes because
the angular particles also tend to have lower settling velocity than the spherical ones.

4. CONCLUSIONS

The equilibrium between centrifugal force and drag force holds true in the case of particle
movement in swirl flow. The concept of particle separation due to high velocity and
acceleration in swirl flow is not only correlated for guide vane outlets and runner vane
inlets, but also correlated for other turbine components like flow through stay vanes,
guide vane cascades, runner vanes of reaction turbines and Pelton buckets. Direct
application of this concept leads to an operating strategy for Francis turbines that process
sand‐laden water. For example, guide vane position could be manipulated to maintain
the velocity ratio in such a way that particles of a given size should flow along with the
water without striking the wall. This means that the guide vane opening should be
adjusted to maintain a velocity ratio, which would be appropriate to the characteristic
diameter of the particles. This will keep the particles in the orbit without striking the
wall and, thus, the effect of sand erosion can be minimized.

Furthermore, the particle velocity and drag co‐efficient not only depend on particle size
but are also sensitive to the shape of the particle along with other variables. The
reduction of the erosion is also linked to the reduction of particle velocity.


REFERENCES

1. Brekke H. (2002) “Design of hydraulic machinery working in sand laden water,”
In: Duan C.G. and Karelin V.Y. (eds), “Abrasive erosion and corrosion of
hydraulic machinery,” pp 155‐181, Imperial college press, London.
2. Chevallier, P. and A. B. Vannes (1995). ʺEffects on a sheet surface of an erosive
particle jet upon impact.ʺ Wear 184(1): 87‐91.
3. Doby, M. J., A. F. Nowakowski, et al. (2007). ʺNumerical and experimental
examination of swirl flow in a cylindrical container with rotating lids.ʺ Minerals
Engineering 20(4): 361‐367.
4. Gabitto, J. and C. Tsouris (2008). ʺDrag co‐efficient and settling velocity for
particles of cylindrical shape.ʺ Powder Technology 183(2): 314‐322.
5. Gregory, J. (John), Particles in water properties and processes, University college
London England, 2006, CRC press, Taylor & Francis Group, p. 9‐10
6. Haider, A. and O. Levenspiel (1989). ʺDrag coefficient and terminal velocity of
spherical and nonspherical particles.ʺ Powder Technology 58(1): 63‐70.
7. Jie Zhang , Particle Technology ‐ Study Notes.
8. K. Harris January 21, 2003, “Sediment transport processes in coastal
environments,” Lecturer notes.
9. Martin Rhodes, 2008 “Introduction to Particle Technology,” second edition, p. 29‐
42.
10. Neopane, Hari Prasad, Ole G. Dahlhaug and Bhola Thapa, 22‐24 Oct.2007,
“Alternative Design of a Francis Turbine for Sand Laden Water.” Proc.
International Conference on Small Hydropower ‐ Hydro Sri Lanka.
11. Sheldon, G. L. and A. Kanhere (1972). ʺAn investigation of impingement erosion
using single particles.ʺ Wear 21(1): 195‐209.
12. Truscott, G. F. (1972). ʺA literature survey on abrasive wear in hydraulic
machinery.ʺ Wear 20(1): 29‐50.
13. Tabakoff, W., A. Hamed, et al. (1983). ʺInvestigation of gas particle flow in an
erosion wind tunnel.ʺ Wear 86(1): 73‐88.
14. Tabakoff, W. (1995). ʺHigh‐temperature erosion resistance of coatings for use in
turbomachinery.ʺ Wear 186‐187(Part 1): 224‐229.
15. Thapa, B. (2004). “Sand erosion in hydraulic machinery.” Trondheim, Norwegian
University of Science and Technology, Faculty of Engineering Science and
Technology, Department of Energy and Process Engineering. 2004:105.
16. Thapa B. and Brekke H. (2004) “Effect of sand particle size and surface curvature
in erosion of hydraulic turbine,” IAHR symposium on hydraulic machinery and
systems, Stockholm.
17. Zhao, K., C. Gu, et al. (1993). ʺStudy on mechanism of combined action of
abrasion and cavitation erosion on some engineering steels.” Wear 162‐164 (Part
2): 811‐819.


LIST OF SYMBOLS AND ABBREVIATIONS
FD Drag force N Ap Projected area m2
a Constant ‐ CD Drag co‐efficient ‐
A Cross sectional area of pipe m2 r radius m
b Constant ‐ Ut Terminal velocity m/s
m Maas of the particle kg Re,p Particle Reynolds number ‐
Fe External force N μ Viscosity of fluid Ns/m2
FB Buoyant force N ω Angular velocity rad/s
Ct Absolute velocity at the inlet m/s Fc Centrifugal force N
dc Critical dia. of particle m Cd Drag co‐efficeint ‐
Dp Diameter of particle m ρ p Density of particle kg/m3
g Gravity m/s2 ρ Density of fluid kg/m3
H Head m Q Flow rate m3/s
h Head m U Peripheral velocity m/s
W Erosion rate ‐ u Relative velocity of particle m/s
Vr Relative velocity m/s a e acceleration m/s2
d Diameter of particle m π
pi Const.
n Refer to exponential factor ‐


Hydropower 010, Trømso, Norway
Appendix E
NUMERICAL PREDICTION OF PARTICLE SHAPE FACTOR EFFECT ON
SEDIMENT EROSION IN FRANCIS TURBINE BLADES

Hari Prasad Neopane1, Ole Gunnar Dahlhaug2, and Mette Eltvik3

This paper is presented at the Hydropower’10, the 6th International Conference on
Hydropower, held in Trømso, Norway, 1 ‐ 3 February 2010. The paper has also published
in the proceedings for this symposium.

ABSTRACT

Hydraulic turbine components operating in sand‐laden water are subject to abrasive and
erosive wear. This wear not only reduces the efficiency and life of a turbine but also
causes problems in operation and maintenance. The prediction of the erosion on a Francis
turbine blade will be described based upon the two different shapes of particles at two
operating conditions, namely, best efficiency (BEP) and full load, for which the results of
experimental tests are available for discussion, comparison and validation. A numerical
simulation is based on the Lagrangian particle tracking multiphase model that is
available in ANSYS CFX, which is capable of modelling dispersed phases, which are
discretely distributed in a continuous phase. The modelling involves the separate
calculation of each phase with source terms generated to amount for the effects of the
particles on the continuous phase. It is found that erosion on a turbine blade is strongly
dependent on the shape of the particle. The predicted erosion rate density is in good
agreement with the particle velocity obtained from experiments.

E-1 HPN PhD Thesis, NTNU

1. INTRODUCTION

Hydraulic turbine components operating in sand‐laden water are subject to sediment
erosion. This type of erosion is not only reduces efficiency and the life of a turbine but
also causes problems in operation and maintenance (Brekke, 2002; Neopane, 2009).
Suspended sediments in water are subjected to kinetic energy, the force of gravity,
viscosity, turbulence, centrifuge and cavitation. The sediment erosion is a result of the
mechanical wear of components. This is due to the dynamic action of sediment flowing
along with water impacting against the solid surface. Therefore, sediment flowing along
with water passing through the turbine is the root caused of sediment erosion in turbine
components (Naidu, 1999). The sediment content in rivers may cause abrasion on
different components of hydraulic turbines. The erosion intensity depends on the
sediment type, its characteristics (particle shape, size and concentration etc.), the
operating conditions of the machine (flow rate, head and rotation speed etc.), and the
hydraulic design itself, as well as the material used for the turbine component (Finnie,
1972; Truscott, 1972; Thapa and Brekke, 2004). Various researchers have conducted
experiments to study the effect of these parameters on erosive wear, but most of these
experiments are on small‐size samples in different types of test rigs to simulate the flow
conditions in the turbine, but actual flow conditions and the phenomenon of erosive wear
are too complex to simulate (Padhy and Saini, 2008).
Particle characteristics are an important but relatively poorly researched aspect of the
erosion problem (Stachowiak et al., 2006). It is impossible to isolate hardness completely
from other features of the particle such as its shape. Even if the particle is hard but
relatively blunt then it is unlikely to cause severe erosive wear. A blunt particle has a
mostly curved surface approximating a spherical shape, while a sharp particle consists of
flat areas joined by corners with small radii, which are critical to the process of wear
(Stachowiak et al., 2006). The sediment wear increases rapidly with the flow velocity, and
is often reported as being approximately ∝ (velocity)3, or ∝ (pump head)3/2, from both
theoretical considerations and test results (Truscott, 1972). The actual value of the index,
for any given conditions and the actual mechanism of the erosive wear is not yet clearly
understood. Hence, a simple, reliable and generalized model for erosion is not yet
developed for engineering purposes. Many erosion models are developed for a specific
purpose or condition (Thapa, 2004). Most of the common expressions for erosive wear are
based upon the experimental experience so that a multidimensional study approach is
needed to investigate the relationship between sediment movement and sediment
erosion mechanism inside the hydraulic machines (Naidu, 1999). There is a pressing need
to study sediment erosion and develop more realistic empirical relations based upon the
combined experimental, numerical and field experiences with respect to the influencing
factors is increasing day by day.
Traditionally, manufacturers of hydraulic turbines have only relied on empirical data and
experience gathered from site to investigate the component damage due to silt abrasion
between outages. This information is very helpful, but for new hydropower projects
involving the risk of sand abrasion, it is desirable to numerically study the abrasive
HPN PhD Thesis, NTNU E-2

mechanism during the design process in order to predict and quantify, the zones, where
there is a high risk of erosion damage (Vu and Alain, 2000).
The main objective of this paper is to investigate the relationship between the different
shapes of particle movement and the tendency of erosion inside the Francis turbine
blades. Then establish the operating strategy for a Francis turbine operating in sand‐
laden water. In order to achieve the main objective, numerical simulations are carried out
based upon the commercial Computational Fluid Dynamics (CFD) code, namely CFX on
a Francis turbine of Cahua Hydropower plant. The prediction of the erosion on turbine
blade will be described based upon the two different shapes of particles at two operating
conditions, namely, best efficiency (BEP) and maximum load. Here the results of the
experimental tests are available for discussion, comparison and validation.

1.1 CAHUA HYDROPOWER PLANT (HPP)

Cahua HPP is a Run‐of‐River hydropower plant in Peru. It is located 200 km north of
Lima, by the Pativilca River. The power plant has two low head vertical Francis turbines
and total installed capacity of 43 MW with a gross head of 215 m and a mass flow of 22
m3/s. The average annual output is 280 GWh. Both Francis turbines composed of 20 stay
vanes, 20 guide vanes and 17 runner blades. The power plant was completed in 1967 and
purchased by SN Power Norway in 2003.
The sediment study conducted in this power plant indicated that, the sediment
concentration exceed 120,000 tons of sediment only after six weeks of operation. The
average content of quartz particles in the sediment is found to about 35 % and feldspar is
found to about 30 %, which are harder than the turbine material, is main reason of
excessive sediment erosion in the hydraulic machinery operating in this power plant.

2. CFD METHODOLOGIES
The numerical unsteady simulations are carried out with the commercial software
ANSYS CFX. There are a number of different solution methods which are used in CFD
codes. The most common, and the one on which ANSYS CFX is based, known as the
finite volume technique. In this technique, the region of interest is divided into small sub‐
regions, called control volumes. The equations are discretized and solved the
incompressible Unsteady Reynolds Averaged Navier‐Stokes (URANS) equations in their
conservative form iteratively. As a result, an approximation of the value of each variable
at specific points throughout the domain can be obtained. In this way, one derives a full
picture of the behaviour of the flow. To provide closure to the time averaged Navier‐
Stokes equations, it is necessary to utilize a turbulence closure model. The two equations
turbulence models, k − ε and SST were tested.

The k − ε model has proven to be stable and numerically robust and has a well‐
established regime of predictive capability. For general‐purpose simulations, the k − ε
model offers a good compromise in terms of accuracy and robustness; however, it can
lack prediction accuracy for complex flow. Such complexities include rapid variations in
flow area, flows with boundary layer separation, flows with sudden changes in the mean
strain rate, flows in rotating fluids, flows over curved surfaces etc.
A Reynolds Stress model may be more appropriate for flows with sudden changes in
strain rate or rotating flows, while the SST model may be more appropriate for separated
flows. Hence, the Shear stress turbulence (SST) model was utilized as default closure
model. Furthermore, the discretization of the equations was made with the Forward
Euler implicit scheme, second order time, and in high resolution advection scheme. The
assumed convergence criterion for residuals was RMS of 10‐4.

2.1 PARTICLE TRANSPORT THEORY
Particle transport modelling is one type of multiphase model, where particulates are
tracked through the flow in a Lagrangian way, rather than being modelled as an extra
Eulerian phase. The full particulate phase is modelled by just a sample of individual
particles. The tracking is carried out by forming a set of ordinary differential equations in
time for each particle, consisting of equations for position, velocity, temperature, and
masses of species. These equations are then integrated using a simple integration method
to calculate the behaviour of the particles as they traverse the flow domain. While setting
up the Lagrangian way particle model, the following assumptions are implied:
• Particle‐particle interactions are neglected due to the low particle concentrations
experienced.
• The particles are spherical. The physical properties of each phase are constants.
• The mean flow is steady. The turbulent flow is locally isotropic.
• The geometry modification caused by the removal of wall by the sand particles,
are neglected.

2.2 PARTICLE TRAJECTORIES COMPUTATION
To evaluate the particle trajectories, its velocity and finally erosion due to the particle
movement, it is necessary to develop equation of motion for that particle. Consider a
discrete particle travelling in a continuous fluid medium. The forces acting on the
particle, which affect the particle acceleration, are due to the difference in velocity
between the particle and fluid, as well as to the displacement of the fluid by the particle.
The equation of motion for such a particle was derived by Basset, Boussinesq and Oseen
(BBO) for a rotating reference frame.

dU P
mP = FD + FB + FR + FV M + FP + FBA (2.1)
dt
Where, FD is drag force acting on the particle, FB is buoyancy force due to gravity, FR is

forces due to domain rotation (Centripetal and Coriolis forces) , FVM is virtual (or added)
mass force. This is the force to accelerate the virtual mass of the fluid in the volume
occupied by the particle. This term is important when the displaced fluid mass exceeds
the particle mass, such as in the motion of bubbles, FP is pressure gradient force. This is
the force applied on the particle due to the pressure gradient in the fluid surrounding the
particle caused by fluid acceleration. It is only significant when the fluid density is
comparable to or greater than the particle density and, FBA is basset force or history term
which accounts for the deviation in flow pattern from a steady state. This term is not
implemented in ANSYS CFX.
A moving fluid exerts an aerodynamic force which is sum of the drag force (acts in
opposite direction to the fluid motion) and lift force (acts normal to the fluid motion).
CFX gives consideration to pressure gradient forces of the surrounding fluids and
gravitational effects. The lift force effect is simply ignored for the computational ease.
Furthermore, virtual mass force, basset force and other forces are normally neglected in
view of the large differential density between phases. The equation for the rate of change
of velocity within CFX is derived directly from Newton’s second law. The velocities are
obtained by simply integrating the force balance in a Lagrangian frame of reference.

dU P 1 π m
mP ⋅ = ⋅ C D ⋅ ρ F ⋅ AF ⋅ U F − U P ⋅ (U F − U P ) + ⋅ d p ⋅ (ρ P − ρ F ) ⋅ g − F ⋅ ∇P
3
dt 2 6 ρF

(2.2)
In the simplest form of equation of motion, the main term on the right hand side is the
drag force that includes the consideration of both skin and form effects. The total drag
force is most conveniently expressed in terms of the non dimensional drag coefficient, as
defined by the Schiller‐ Naumann correlation. ANSYS CFX modifies this to ensure the
correct limiting behaviour in the inertial regime by taking,

⎛ 24
(
C D = max⎜⎜ ⋅ 1 + 0.15 ⋅ Re
0.687
) ⎞
, 0.44 ⎟⎟ (2.3)
⎝ Re ⎠

However, this correlation has been developed based upon spherical shape of the particle;
the trajectories of the non‐spherical particles can be modelled through the application of
shape factors, allowing non‐uniform drag distribution and bounce characteristics to be
classified.


2.3 RESTITUTION COEFFICIENT FOR PARTICLES

The parallel and perpendicular restitution co‐efficient describe the action of particles
when they hit a wall. The reflected velocity of the particle is lower than the incoming
velocity due to energy transfer. Energy is dissipated as heat, noise and target material
deformation. This effect is described by the momentum based restitution coefficient. The
coefficient values of one described an elastic collision, while values less than one describe
an inelastic collision. The parallel coefficient is usually equal to one. The perpendicular
coefficient will depend on the particle material. Particle that bounces off walls will have a
perpendicular coefficient close to one, while particles that stick to walls will have a
perpendicular coefficient of zero. In this simulation, the perpendicular and parallel
coefficients of restitutions are chosen equal to 0.9 and 1 respectively.

2.4 TABAKOFF EROSION MODEL
In general, the Tabakoff model provides more scope for customization with its larger
number of input parameters. The choice of one model over another is largely simulation‐
dependent. The erosion model can be set on a per‐boundary or per‐domain basis. When
enabled for the domain, the domain settings will apply for all boundaries that do not
explicitly have erosion model settings applied to them. The Tabakoff model requires the
specification of five parameters. The k12 constant, 3 reference velocities and the angle of
maximum erosion γ 0 must all be specified. The erosion rate E is determined from the
following relations:
[ 2
]
E = k1 ⋅ f (γ ) ⋅ V p Cos 2 γ ⋅ 1 − RT + f (VP N )
2
(2.4)
2
⎡ ⎛ π / 2 ⎞⎤
f (γ ) = ⎢1 + k 2 ⋅ k12 ⋅ Sin⎜⎜ γ ⋅ ⎟⎥ (2.5)
⎣ ⎝ γ 0 ⎟⎠⎦
RT = 1 − k 4 ⋅ VP ⋅ Sinγ (2.6)
f (VP N ) = k 3 ⋅ (VP ⋅ Sinγ )

4
(2.7)
⎧1.0 if γ ≤ 2 ⋅ γ 0
k2 = ⎨ (2.8)
⎩0.0 if γ > 2 ⋅ γ 0
Here E is the dimensionless mass (mass of eroded wall material divided by the mass of
particle). V P is the particle impact velocity. γ is the impact angle in radians between the

approaching particle track and the wall, γ 0 being the angle of maximum erosion. k1 to
k 4 , k12 and γ 0 are model constants and depend on the particle/wall material
combination. The overall erosion rate of a wall due to a particle is computed from the
following relation:
•
Erosion Rate = E ⋅ N ⋅ m p (2.9)
•
Here m p is the mass of the particle and N is its number rate. The overall erosion of the
wall is then the sum over all particles. This gives an erosion rate in kg/s, and an erosion
rate density variable in the res file and post‐processor in kg/s/m2. Furthermore, this
erosion rate is only a qualitative guide to erosion, unless precise values for the model
constants are known.

2.5 NUMERICAL MODEL

To achieve a result with second order accuracy and sufficient numerical stability, the
high‐resolution scheme was applied for discretization of the grid. Turbo Grid 11.0 was
selected for meshing because this is a highly automated hexahedral mesh generator,
specially designed for turbo machinery. Grids generation is relied on three different
factors; namely, number of cells, Reynolds number and turbulence model desired. Two
equations turbulence model were chosen initially for grids generation but the SST model
gave the most realistic results, and it was decided to continue the simulation with this
model only. For setting up a new simulation in ANSYS CFX, the mesh files were
imported and merged together to form computational domains as shown in Figures 2.1
and 2.2.

Figure 2.1 Computational Model at BEP Figure 2.2 Computational Model at Full Load


The key steps of the numerical modelling process mainly include:

(a) Component definition: The component definition panel is used to import meshes and
to select the rotation speed of each component and set a tip clearance.
(b) Physics definition: All the physical settings, including fluid type, simulation type,
inlet and outlet boundary conditions, interface types and solver parameters are set in this
panel. The selection of boundary template, which provides quick setup of the most
common turbo machinery boundary conditions, has to be specified in this stage. For this
simulation mass flow inlet, pressure static outlet is selected per passage basis.
(c) Interface definition: The frozen rotor interface was selected as the interface default
type and set the convergence control to physical timescale with a value of 1/ ω [rev/min].
Generally the interfaces are automatically created using the region information from the
component definition panel.
(d) Boundary definition: Boundary conditions are also automatically created using the
region information from the component definition panel and information from the
physics definition panel.
(e) Final Operation: The final operation panel is used to enter in to the general mode of
operation for applying other CFX features including sand properties definition and
particle tracking etc.
(f) Defining the properties of Sand: The material properties of sand particles used in the
simulation needs to be defined in general mode. Heat transfer and radiation modelling
are not used in this simulation, so the only property that needs to be defined is the
density of sand.

To calculate the effect of particles on continues fluids, between 100 and 1 000 particles are
usually required. However, if accurate information about the particle volume fraction or
local forces on the wall boundaries is required, then a much larger number of particles
need to be modelled. After creating a domain, full coupling or one‐way coupling between
the particle and continues phase has to be chosen. Full coupling is needed to predict the
effect of particles on the continuous phase flow field but has the higher CPU cost than
one‐way coupling. One‐way coupling simply predicts the particle paths during post
processing based on the flow field, but without affecting the flow field. To optimize CPU
usage, the two sets of identical particles were created. The first set will be fully coupled
and between 100 and 1,000 particles will be used. This allows the particles to influence
the flow field. The second set will use one‐way coupling but a much higher number of
particles will be used. This provides a more accurate calculation of the particle volume
fraction and local forces on the wall.

Sand particles are uniformly injected at the stay vane inlet with the same conditions as
the fluid. The particles will follow through the domains and exit at the outlet. The sand
particles are defined as solid particles and the size distribution is uniform in diameter.
The turbulence dissipation force is activated, and the Schiller Naumann model is chosen
to calculate the drag force acting on the particle.

2.6 BOUNDARY CONDITIONS

If all residual are below their targeted criteria then convergences of equations will
prevail. To improve the convergences of simulation and to obtain a conversion of
solution with less numbers of iterations in relatively short interval of time, only one set of
blades was considered in this simulation analysis. In the post processing, there is a
provision to assemble all components of turbine. The two most interesting operating
points are at the best point efficiency and full load, with guide vane angle of 16 and 22
respectively. To study the effect of different shape of the particle for predicting erosion
rate density, the simulations were carried out at same concentration rates for both
operating conditions.

2.6.1 INLET AND OUTLET

When setting boundary conditions, ANSYS CFX best practices guide for Turbo
machinery recommended specifying a total inlet pressure and mass flow outlet. This will
give a more appropriate calculation of the flow field than with the mass flow inlet
condition for those machines drawing fluid directly from a static reservoir. This may be
valid for hydraulic pump simulation but in the case of radial turbine, the mass flow inlet
and a total pressure outlet was found more appropriate and robust than the total inlet
pressure and mass flow outlet. In addition, this provides better performance of turbine,
when looked at pressure distribution, y+ values, power output, head and flow rate. Thus,
the inlet condition was set as mass flow rate per passage with velocity components. The
outlet pressure in this case is arbitrary and is usually set at or close to zero to reduce
round off error. In this simulation, this was chosen equal to 1 atm. The detail boundary
condition with appropriate value is presented in Table 2.1.

Table 2.1: Boundary conditions
Variable Value
Water density 997 kg/m3
Quartz particle density 2.65 gm/cm3
Diameter of the particle 0.1 to 0.2 mm
Inlet mass flow rate of water at BEP 540 kg/s
Inlet mass flow rate of water at Full Load 685 kg/s
Total pressure outlet 1 [atm.]
Inlet mass flow rate of particle 0.05 kg/s per passage
Inlet flow direction at Stay vanes ( α , r , θ ) 0 , ‐0.4 , ‐0.9165

2.6.2 WALL

The covers, hub, shroud and vanes were defined as smooth walls with no‐slip condition.
Fluid velocities near the wall will then be decreased by the wall friction. The runner has
an angular velocity 600 rpm, while the stay vane and guide vane are stationary domains.

A periodic boundary condition was set to couple two adjacent blades. This simplified the
computational model.

3. RESULTS AND DISCUSSIONS

Figure 3.1 and 3.2 shows the prediction of erosion rate density along the blade surface at
16 0 guide vane angle which is at maximum efficiency point. It can be clearly seen from
the figures that maximum erosion occurs in the blade profile with shape factor of 0.5. The
particles shape factor assumed to be spherical by default in ANSYS and its value is equal
to 1. If the particle is deviated from spherical shape then shape factor value is less than 1.
ANSYS CFX always calculates the diameter of the particle from the mass of the particle
divided by its density, assuming it is spherical. A cross sectional area factor can be
included to modify the assumed spherical cross section area to allow for non‐spherical
particles. This affects the drag force calculated by ANSYS CFX. The surface area factor is
analogous to the cross sectional factor.
For non‐spherical particle, it is the ratio of the surface area to the surface area of spherical
particle with the same equivalent diameter. This affects both mass transfer and heat
transfer correlation.

Figure 3.1 Erosion rate at BEP @ SF 1 Figure 3.2 Erosion rate at BEP @ SF 0.5

Figure 3.3 Erosion rate at Full Load @ SF 1 Figure 3.4 Erosion rate at Full Load @ SF 0.5

Similarly, Figure 3.3 and 3.4 shows the prediction of erosion rate density along the blade
surface at 22 0 guide vane angle which is at full load condition. It can also be clearly seen
from the figures that maximum erosion occurs in the blade profile with shape factor 0.5.
A similar erosion pattern is observed in each case. However, the blade profile area is
highly affected and larger intensity of erosion rate is clearly observed in the case of non‐
spherical shape of the particle. The erosion rate density of pressure side is more than of
the suction side of the blade. This is even more in larger the guide vane opening. The
erosion rate density is more towards the outlet as compare to the inlet of the blade. This
may be due to the blade profile tail vortex flow, which leads to increasing the erosion rate
density in the blade outlet.
To validate simulation result, a test rig, was developed and designed at the Waterpower
Laboratory, Norwegian University of Science and Technology, Norway. The velocity of
particle was calculated in highly swirl conditions similar to turbine flow in curved path,
and found that the size of the particle is inversely proportional to the velocity of the
particle, and it also depends upon the shape of the particle. Different shapes of particle
were tested, and it was determined that spherically shaped particle had higher velocity
than particle with other shapes. Generally the erosion rate density will increase when the
velocity of particle increase. However, this has not occurred in this simulation analysis
because of shape factor of the particle.
In general, many natural particles are usually in non‐spherical shape. These particles will
tend to have lower velocities because both decreases in spheroid and increases in
angularity tend to decrease velocities. The impact area of these particles on the eroded
surface is relatively larger and tends to be directed perpendicular to the transport path.
As a result, higher co‐efficient of drag, higher rotational motion and more separation of
flow likely to occur and hence more erosion rate is found.

4. CONCLUSIONS

The operation of turbine at full load not only caused poor efficiency but also increased
turbulence and higher relative velocity of flow at outlet of the blade. This also creates
strong swirl flow in turbine. The role of operating conditions on erosion rate density
clearly demonstrated in simulation analysis. It has been shown that predicted erosion
rate density is in good agreement with the experiments. It has been found that erosion
process is strongly depended on the shape of the particle. Furthermore, the significant
reduction of erosion rate can be achieved by operating turbine in best efficiency point as
far as possible.

REFERENCES

1. ANSYS: CFD online, http://www.cfd‐online.com/[Accessed on 31 August 2009]
2. ANSYS Turbo Grid Release 11.0. Users guide, 2006.

3. ANSYS CFX Release 11.0 Solver Theory Guide, 2006.
4. Brekke Hermod, 2002. Design of hydraulic machinery working in sand laden
water. Abrasive erosion and corrosion of hydraulic machinery, London.
5. Bergeron S. Y., T. C. Vu, and A. P. Vincent, 2002. Silt erosion in hydraulic
turbines: The need for real‐time numerical simulations.
6. Doby M. J., A. F. Nowakowski, E. Nowak, and T. Dyakowski, 2007. Numerical
and experimental examination of swirl flow in a cylindrical container with
rotating lids. Minerals Engineering 20(4): 361 ‐ 367.
7. Forder A., M. Thew, and D. Harrison, 1998. A numerical investigation of solid
particle erosion experienced within oilfield control valves. Wear 216(2): 184 ‐ 193.
8. Finnie I., 1972. Some observations on the erosion of ductile metals. Wear 19: 81 ‐
90.
9. Gabitto J. and C. Tsouris, 2008. Drag coefficient and settling velocity for particles
of cylindrical shape. Powder Technology 183(2): 314 ‐ 322.
10. Liu X.B., L.D. Zhang, Z. Liang, and L.J. Cheng, 1996. Numerical prediction of silt
abrasive erosion in hydraulic turbine. Fluids Engineering Division Conference
Volume 1, ASME.
11. Mack R., P. Drtina, and E. Lang, 1999. Numerical prediction of erosion on guide
vanes and in labyrinth seals in hydraulic turbines. Wear. 233 ‐ 235: 685 ‐ 691.
12. Neopane Hari P., Ole G. Dahlhaug, and Thapa Bhola, 2009. Experimental
examination of the effect of particle size and shape in hydraulic turbines.
Waterpower XVI, Spokane, Washington, USA.
13. Naidu D. B. K. S., 1999. Developing silt consciousness in the minds of hydro
power engineers. Silting problems in hydro power plants, New Delhi, India.
14. Ole G. Dahlhaug and Thapa Bhola, 2004. Sand erosion in Francis turbine: a case
study from Jhimruk power plant, Nepal. 22nd IAHR Symposium on Hydraulic
Machinery and System, Stockholm, Sweden.
15. Padhy M. K. and R. P. Saini, 2008. A review on silt erosion in hydro turbines.
Renewable and Sustainable Energy Reviews 12(7): 1974 ‐ 1987.
16. Stachowiak G. W. and A. W. Batchelor, 2006. Abrasive, erosive and cavitation
Wear. Engineering Tribology (3 Edition). Burlington, Butterworth‐Heinemann:
501 ‐ 551.
17. Thapa Bhola and Brekke Hermod, 2004. Effect of sand particle size and surface
curvature in erosion of hydraulic turbine. IAHR symposium on hydraulic
machinery and systems, Stockholm.
18. Thapa Bhola, 2004. Sand erosion in hydraulic machinery, PhD thesis, Norwegian
University of Science and Technology, Faculty of Engineering Science and
Technology, Department of Energy and Process Engineering, 2004: 105.
19. Truscott G. F., 1972. A literature survey on abrasive wear in hydraulic machinery.
Wear 20(1): 29 ‐ 50.
20. Zhang Y., E. P. Reuterfors, B. S. McLaury, S. A. Shirazi, and E.F. Rybicki, 2007.
Comparison of computed and measured particle velocities and erosion in water
and air flows. Wear 263(1 ‐ 6): 330 ‐ 338.

Analysis of Surge
Macmillan Civil Engineering Hydraulics
General Editor: E. M. Wilson, PH.D., M.sc., F.I.C.E.
Reader in Civil Engineering, University of Salford
FLOW IN CHANNELS: R. H. J. Sellin

ENGINEERING HYDROLOGY: E. M. Wilson
COASTAL HYDRAULICS: A.M. Muir Wood
Analysis of Surge
John Pickford, M.Sc.(Eng.), C.Eng., M.I.C.E.

Department of Civil Engineering
Loughborough University of Technology
Macmillan
© John Pickford 1969
Softcover reprint of the hardcover 1st edition 1969
First published 1969
Published by
MACMILLAN AND CO LTD
Little Essex Street London wc2
and also at Bombay Calcutta and Madras
Macmillan South Africa (Publishers) Pty Ltd Johannesburg
The Macmillan Company of Australia Pty Ltd Melbourne
The Macmillan Company of Canada Ltd Toronto
Gill and Macmillan Ltd Dublin
ISBN 978-1-349-00162-0 ISBN 978-1-349-00160-6 (eBook)

DOI 10.1007/978-1-349-00160-6
Preface
Many engineers claim that they "do not understand surge", but for many
others there is a fascination in the varying surge pressures resulting from
changes of flow. It is hoped that this book will satisfy both groups by
providing clear explanations of surge phenomena, and by investigating a
variety of methods of dealing with the subject.
The material is presented in such a way that it should be a valuable
textbook for undergraduates dealing with surge problems for the first time.
Postgraduate students and practising engineers should find it a useful
reference for detailed study and a guide for design analysis.
Waterhammer is investigated from first principles along well-trodden
paths. Particular reference is given to graphical analysis.
In the chapters devoted to surge tanks a number of methods of cal-
culating surge heights are considered, and special attention is given to the
'logarithmic method', and to finite difference methods which are particu-
larly suitable for computer analysis.
The author has carried out an intensive investigation of mass oscillations
in surge tanks: over 20 000 measurements of surge heights were taken in a
laboratory model, and several thousand surge heights were obtained by
analogue and digital computers. The methods described in this book have
been well-tried, and have been included only when they have proved
satisfactory.
Many worked examples are given and S.I. units are used throughout,
although traditional British units are mentioned when it is thought that
familiar quantities may be helpful.
The material has been selected so as to deal with the practical aspects of
surge which are most useful for civil engineers. Reference is made
throughout the book to the literature of the subject so that readers wishing
to know more about particular aspects can readily find further information.
John Pickford
Loughborough
v
Contents
Preface v
1 Waterhammer, surge tanks and open

channel surge 1
2 Introduction to waterhammer 5
2.1 Incompressible water column theory
2.2 Valve movement
2.3 Limiting pressures
2.4 Equivalent length of pipe
3 Elastic theory for a simple pipe 18

3.1 Sudden closure
3.2 Pressure increase in a rigid pipe
3.3 Strain energy waterhammer theory
3.4 Fundamental differential equations of water-
hammer
3.5 Velocity of propagation
3.6 Direct application of elastic waterhammer theory
4 End conditions 32
4.1 Reflection at a reservoir
4.2 Reflection at a dead end
vii
viii Contents
4.3 Effects of changes of pipe
4.4 Effect of junction
4.5 Effect of restriction
4.6 Oscillation of the pressure wave
5 Non-instantaneous closure and opening 42

5.1 Rapid closure of downstream valve
5.2 Slow closure of downstream valve
5.3 Other types of valve movement
6 Graphical analysis 50
6.1 Pressure-velocity diagrams at fixed points
6.2 Pressure-discharge diagrams at fixed points
6.3 Schnyder-Bergeron diagrams
6.4 Surge diagrams for uniform pipe with hypothetical
conditions
6.5 Comparison of surge diagram with analytical
method
6.6 Intermediate times and positions
6. 7 Series pipes
6.8 Branched pipes
6.9 Effect of hydraulic losses
6.10 Other conditions
7 Waterhammer in pumping mains 70

7.1 Sudden stopping of pumping
7.2 Separation of water columns
7.3 Pump with inertia
7.4 Reverse flow
Surge control 78
7.5 Air vessels
7.6 Pump by-passes and feeder tank
7.7 Parallel pumps
7.8 Waterhammer in delivery pipe from reciprocating
pump
Contents ix
8 W aterhammer in hydro-electric supply
ptpes 87
8.1 End conditions at a surge tank
8.2 Simple tank near valve
8.3 Throttled tank near valve
8.4 Tank between pipeline and penstock
9 Introduction to surge tanks 98

9.1 Purpose of surge tanks
9.2 Oscillation phenomena
9.3 Design considerations
9.4 Effect of friction
9.5 Surge tank types
10 Theory of mass oscillation 108

10.1 Fundamental equations
10.2 Simple undamped oscillation
10.3 Logarithmic equations
10.4 Logarithmic curves
10.5 Stability
11 Finite difference methods I 129

11.1 Outline of the methods
11.2 Fundamental finite difference methods
11.3 Successive trials (Pressel's method)
11.4 Simple arithmetic method
11.5 Jakobsen's method
11.6 Escande's method
12 Finite difference methods II 141

12.1 M odificati ons
12.2 Flow to turbines during oscillation: sudden valve
movements
X Contents
12.3 Linear change of flow
12.4 Constant valve opening
12.5 Constant power output
12.6 Flow from several sources
12.7 Variable area tank
12.8 Spilling tank
12.9 Differential tank
12.10 Branching tank
12.11 Tail race tank
12.12 Multiple tanks
13 Approximate methods 162

13.1 Relative values--simple tank
13.2 Other approximate equations for simple tanks
13.3 Classified diagrams for throttled tanks
14 Graphical methods 175

14.1 Schoklitsch method
14.2 Calame and Gaden radial method
14.3 Surge height trace by analogue computer
15 Open channel surge 184

15.1 Unsteady flow in open channels
15.2 Backwater surges
List of Notations 193
Index 197
A list of references and Suggestions for further reading is given

at the end of each chapter.
1 Waterhammer, Surge Tanks and
Open Channel Surge
I saw him beat the surges under him,
And ride upon their backs.
SHAKESPEARE: The Tempest
In the following pages we deal with some aspects of hydraulic transients-

the variations of flow and pressure which occur when one steady state
changes to another steady state.
For example, when an hydro-electric turbine comes 'on load' the initial
steady state is zero load at static head and the final steady state is the flow
and related head needed to provide the power demanded by the generator.
During the time required for the water in the conduits leading to the turbine
to settle down to final steady conditions there are rapidly-moving pressure
waves in the water. These waves move at the speed of sound and the
phenomenon is known as waterhammer.
If the water is supplied through a tunnel or pipeline which has a surge
tank, much slower mass oscillation or surges also occur, whereas in an
open headrace channel a drawdown surge occurs.
There are similar hydraulic transients when the discharge from a pump
alters, and the waterhammer pressure variations following pump stoppage
are especially important. 'Waterhammer' can often be heard when a tap
is suddenly closed.
The term 'surge' is used to cover the three main types of hydraulic
transient-waterhammer, mass oscillation and open channel surge.
The high pressure sometimes associated with surge in closed conduits
can become dangerous, occasionally leading to fracture of a pipe or pump
casingl-4, The high pressure may be due to resonance, 5 caused by the
I
2 Analysis of Surge
relationship, under certain conditions, of the natural frequency of the
pipeline and the hydraulic transient.
Failure of a pipeline can have very serious consequences, as in the
Lac Blanc-Lac Noir accident in 1934, when the force of the rupture was
so great that a 12 ton piece of concrete surrounding the pipe was thrown
50 metres. 2 •6
In recent years computers have been increasingly used to analyse surge
conditions. Numerical methods form the basis for most computer studies
of surge tanks and have been dealt with at some length in Chapters 11 and
12. Few descriptions of computer techniques for general surge tank
problems have been published. On the other hand there is a wealth of
literature for those interested in the application of computers to water-
hammer,7-20 and the emphasis in the first part of this book is on the use
of surge diagrams for waterhammer analysis.
Joukowski 21 first established the basic waterhammer equations in the
closing years of last century and Allievi 22 in the early twentieth century
set down his 'classical' theory. Graphical surge analysis was developed by
Schnyder23,24 and Bergeron25 in the 1930s and was introduced to English
readers by Angus. 26 Lupton27 in 1953 wrote an extremely lucid and
readable account of the graphical method.
The most valuable work on surge tanks was published by Calame and
Gaden in 1926; their Theorie des chambres d' equilibre 28 introduced non-
dimensional 'relative values' and a system of notation which is widely used
on the Continent. Earlier, Johnson29,so had introduced the differential
tank which became popular in the American continent but has not been
generally accepted elsewhere. Since the last war, Escande* has published
extensively on various aspects of surge tanks and Jaeger* has investigated
a variety of surge problems.
There is fascination for small boys and engineers alike in the movement
of water. When the flow itself is changing, the study of the movement
becomes doubly interesting. It is hoped that your study of surges will be
full of interest, and that the complexities of the subject are made reasonably
clear.
The symbols suggested by the American Society of Mechanical
Engineers form the basis for the notation in the chapters dealing with
waterhammer. For surge tanks there are wide variations in usage, and
when referring to other literature readers are warned to be careful! An
innovation is the use of the subscript 'f' to indicate final steady conditions.
This has been of especial value when writing computer programmes
which may be equally applied to partial or complete opening or closure.
* See index.
Waterhammer, Surge Tanks, Open Channel Surge 3
REFERENCES
1. FERRAND, G.: 'A propos d'un accident survenu a une conduit forcee apres
quarante ans d'exploitation,' Houille Blanche, (1953), p. 20.
2. JAEGER, C.: 'Water-hammer effects in power conduits,' Civ. Engng pub!.
Wks Rev., (1948), 43, pp. 74-76, 138-140, 192-194, 244-246.
3. JAEGER, C.: Water hammer, Chapter XX of 'Hydro-electric engineering
practice,' Ed. J. Guthrie Brown (Blackie, 2nd Edn, 1964).
4. WooDWARD, A. C.: 'Some practical aspects of pressure surge control,'
Allen Engng Rev., (1964), No. 55, pp. 16-19.
5. JAEGER, C.: 'The theory of resonance in hydropower systems,' J. bas. Engng,
Trans. Am. Soc. mech. Engrs, (1963), 85, 631-640.
6. JAEGER, C.: 'Water-hammer effects in power conduits: accidents due to
water-hammer', J. Instn civ. Engrs, (1948), 29, pp. 334-338 (Abstract).
7. BIRO, C.: 'Etude du coup de belier dans des conduits a caracteristiques
multiples,' Cong. int. Ass. hydraul. Res., (1961), pp. 673-679.
8. CHAPPY, A.: 'Calcul des coups de belier et des phenomenons transitoires
per calculateurs electronique,' Houille Blanche, (1967), pp. 193-194.
9. Fox, J. A.: 'The use of the digital computer in the solution of water-hammer
problems,' Proc. Instn civ. Engrs, (1968), 39, pp. 127-131.
10. HARDING, D. A.: 'A method of programming graphical surge analysis for
medium-speed computers,' Symp. Surges Pipelines, Proc. Instn mech.
Engrs, (1966), 180, 3E, pp. 83-97.
11. HAYASHI, T.: 'Water-hammer in bifurcating pipe-lines,' Cong. int. Ass.
hydraul. Res., (1961), pp. 740-748.
12. MARCHAL, M., FLESCH, G., and SuTER, P.: 'The calculation of water-
hammer problems by means of the digital computer,' Intern. Symp.
Waterhammer pumped storage Projects, ASME, (1965), pp. 168-188.
13. PAYNTER, H. M.: 'Methods and results from M.I.T. studies in unsteady
flow,' J. Boston Soc. civ. Engrs, (1952), 39, pp. 120-165.
14. STREETER, V. L.: 'Waterhammer analysis of pipelines,' J. Hydraul. Div.,
Proc. Am. Soc. civ. Engrs, (1964), 90, HY4, pp. 151-172.
15. STREETER, V. L.: 'Computer solution of surge problems,' Symp. Surges
Pipelines, Proc. Instn mech. Engrs, (1966), 180, 3E, pp. 62-82.
16. STREETER, V. L.: 'Water-hammer analysis of distribution systems,' J. Hydraul.
Div., Proc. Am. Soc. civ. Engrs, (1967), 93, HY5, pp. 185-201.
17. STREETER, V. L. and WYLIE, E. B.: Hydraulic transients (McGraw-Hill, 1967).
18. THOMASSON, P. G.: 'The development of a method of using analogue
computers in surge analysis,' Br. Hydromech. Res. Ass., (1967), 9th
Members Conf, SP921.
19. VEGGEBERG, J. M.: 'Digital computer solution for complex tunnel surges,'
Pure Water, (1959), 11, pp. 54-68.
20. WooD, D. J., DORSCH, R. G., and LIGHTNER, C.: 'Wave-plan analysis of
unsteady flow in closed circuits,' J. Hydraul Div., Proc. Am. Soc. mech.
Engrs, (1966), 92, HY2, pp. 83-110.
21. JOUKOWSKI, N.: 'Waterhammer' (Translated by Miss 0. Simin), Proc. Am.
WatWks Ass., (1904), 24, pp. 341-424.
22. ALLIEVI, L.: The theory of water hammer (Translated by E. E. Halmos),
Am. Soc. mech. Engrs, (1929).
4 Analysis of Surge
23. ScHNYDER, 0.: 'Druckstosse in Pumpensteigletrungen,' Schweiz. Bauztg.,
(1929), 94, pp. 271-273, 283-286.
24. ScHNYDER, 0.: 'Druckstosse in Rohrleitungen,' WassKraft WassWirt,
(1932), 27, pp. 49-54, 64-70.
25. BERGERON, L.: Waterhammer in hydraulics and wave surges in electricity
(Wiley, 1961).
26. ANGUS, R. W.: 'Waterhammer in pipes, including those supplied by
centrifugal pumps: graphical treatment,' Proc. lnstn mech. Engrs, (1937),
136, pp. 245-291.
27. LUPTON, H. R.: 'Graphical analysis of pressure-surge in pumping systems,'
J. Instn Wat. Engrs, (1953), 7, pp. 87-125.
28. CALAME, J., and GADEN, D.: 'Theorie des chambres d'equilibre' (Gautier-
Villars, 1926).
29. JOHNSON R. D.: 'The surge tank in water power plants,' Trans. Am. Soc.
mech. Engrs, (1908), 30, pp. 443-474.
30. JOHNSON, R. D.: 'The differential surge tank,' Trans. Am. Soc. civ. Engrs,
(1915), 78, pp. 760-784.
FURTHER READING
DUBIN, C.: Protection of pipelines against water hammer (Int. Wat. Supp. Ass.,
1958), pp. 74-135.
GIBSON, A. H.: Hydraulics and its applications (Constable, 5th Edn., 1952),
pp. 222-245.
JAEGER, C.: Engineering fluid mechanics (Blackie, 1956).
KENNEDY, J. F.: 'A short bibliography of references on water hammer', J. New
Engl. WatWks Ass., (1964), 78, 270-272.
PARMAKIAN, J.: 'Waterhammer analysis' (Dover, 1955).
2 Introduction to Waterhammer
Any alteration of the rate of flow of a fluid passing through a pipe causes
a change of pressure. Sudden alteration of the flow rate can give rise to
surges of pressure which move up and down the pipe causing it to 'knock'.
This effect is sometimes noticed when a bath-tap is shut quickly and the
knocking noise can occasionally be heard all over the house. 'Water-
hammer' is the term applied to pressure surges set up in this way whether
actual 'hammering' or knocking occurs or not.
Even if the velocity head is negligible, the closure of a valve at the down-
stream end of a pipe through which fluid is passing results in an immediate
rise of pressure, as Fig. 2.1 (a). Opening of a downstream valve results in
an immediate drop of pressure, Fig. 2.1 (b).
(a)
(b)
FIG. 2.1
Conversely, when a valve shuts at the upstream end of a pipe there is an

immediate drop of pressure, and opening of an upstream valve causes a
rise of pressure. The same effect occurs when a pump delivering water to
a rising main suddenly stops or starts.
5
6 Analysis of Surge
When the fluid is a gas (i.e., the fluid is 'compressible') the change of
pressure causes change of density. Liquids are almost incompressible,
but we will see in Chapter 3 that the compressibility, or elasticity, of a
liquid has an influence on the pressure produced by a change of flow.
Although the phenomenon with which we are dealing is called 'water-
hammer' it applies equally to other liquids, and although 'water' is referred
to continually in the following pages, it may be assumed that the same
considerations apply to other liquids. 1
2.1 Incompressible water column theory

If water were completely incompressible and if it were flowing in a
completely rigid pipe, there would be no 'waterhammer', but there would
still be a change of pressure on altering the rate of flow, due to the inertia
of the flowing liquid. The assumption of complete incompressibility and
rigidity may be made when the valve movement, which causes the change
of flow, is very slow or the pipe is short. 2 The valve movement must also be
smooth, so that there is no shock at the beginning and end of the closure.
Consider a rigid pipe between two reservoirs A and C, Fig. 2.2, in
which there is a valve at B. If the velocity head and secondary losses are
neglected, the hydraulic gradient abc in Fig. 2.2 (a) gives the pressure
for all points along the pipe when there is a steady flow Qo with the valve
fully open. The slope of the hydraulic gradient is a function of the flow.
For example if the friction loss is proportional to the square of the velocity,
i = (4f/d)vo 2/2g. In terms of flow i ~ (f/3d5)Qo 2 when dis measured in
metres and Qo in m3 js, or i ""(Jj10d 5 )Qo 2 when dis measured in feet and
Qo in cusecs.
If the valve is closed gradually there is a change of pressure throughout
the pipeline as soon as the valve starts to close, even before there is a
sensible change of flow. Immediately upstream of the valve the pressure
rises by an amount h1 so that the hydraulic gradient in the length AB is
along the line ab1, Fig. 2.2 (b). Immediately downstream of the valve the
pressure falls by an amount h2 so that the hydraulic gradient in the
length BC is along the line b2c.
The change of pressure head h1 and h2 is caused by inertia. The water
approaching the valve from A finds its path impeded by the valve; it is
unable to move with its previous velocity vo (vo= Qo/A), and since the
water is incompressible the whole of the column AB is retarded. In other
words it suffers from a negative acceleration.
By- Newton's second law the mass of water in the pipe multiplied by the
acceleration is equal to the force applied. The force is associated with the
inertia head h1 acting over the cross-section of the pipe at B.
Introduction to Waterhammer 7
al Initial steady conditions valve open I bl Immediate! y offer valve movement begins
f = +sf where sf is very small
I
I
I
I
I
IP
I cl Same cond1tions as for I b) I d) During valve movement
-+--==·-·-·-·-.II£ I
I
I
I
I
I
I
f
I
I
I
I!I"t-·-·-·-
I
L__v=_o_:._o_=_o_ _--.+.-----__.l
I e) Immediately before valve movement ends

f =T -df where df is very small
FIG. 2.2
8 Analysis of Surge
The mass of water in column AB = pAL~,
. . dv
The acce1eratwn IS dt
Therefore (2.1)
L1 dv h dv. .
and at B h 1 =- - -d w ere -d 1s negative.
g t t
At other points along the pipe the inertia head depends on the length of
column retarded. For example, at point P, Fig. 2.2 (c), distance l from A,
the column AP is retarded and the inertia head is given by
l dv
hz = - - -
g dt
In the same way the column of water downstream of the valve is
retarded by the force associated with the inertia head h2 acting over the
cross-section of the pipe at B, and
h2 = _ L2 dv
g dt
Ldv
Therefore in general, a h=- - - (2.2)
g dt
For simplicity, the effect of valve closure may be considered as a secon-
dary loss acting at the valve. Closure results in a reduction of flow and
the head loss due to friction is reduced. Put in other words, the secondary
loss at the valve leaves less head available to overcome friction and the
flow is reduced correspondingly.
At time t after the start of closure the valve is partly closed: the velocity
of the water is v and the flow is Q. The friction head loss in the sections
AB and BC, assuming the 'v 2-law' applies, is given by
, 4fL2 v2
hj2 =---
d 2g
The hydraulic gradient and the total pressure are as shown in Fig. 2.2 (d);
h1' and h2' are the inertia heads at time t.
Immediately before the valve movement stops, the water in the pipe
has been brought to rest and therefore v = 0 and h1 = 0. The pressure
head upstream of the valve is above the static head by hi" and the pressure
head downstream of the valve is below the static head of the lower reservoir
by h2"·
If the valve is closed in time Tin such a way that the retardation is at
a constant rate, then the acceleration in AB is -vofT where v0 is the
initial steady velocity = Qo/A. In BC the initial steady velocity may be
considered as being -vo since it is away from the point B, and the change
of velocity in time Tis 0 -( -vo) = +vo.
Hence h1 = h1' = h1" = (LI/g)vo/T
h2 = h2' = h2" = (L2jg)vo/T
h1 = (1/g)vo/T
or in general terms, h=!:'!!!. (2.3)

gT
Example 2.1
Water flows from a reservoir, along a rigid horizontal pipeline 20m below the
water surface of the reservoir. The pipe is 0·15 m diameter and 1500 m long and
the water discharges to atmosphere through a valve at the end of the pipe. If the
valve closes in 4 seconds to give uniform retardation, calculate the total pressure
in the pipe just upstream of the valve and at a point P 500 m.from the valve when
the water is flowing with half its original velocity. Neglect the velocity head and
secondary losses and assume that the water is incompressible. f = 0·005
SOLUTION
_ 4fL vo 2 • _ J(ht X d X 2g)
ht - d 2g .. vo - 4fL
= J( 20 X 0·15 X 2 X 9·81)
4 x 0·005 x 1500 = 1'40 m/s
For uniform retardation the inertia head just upstream of the valve
L vo 1500 1·40
h = gT = 9·81 X 4 = 53·6 m.
2
When the flow is reduced, the friction head h/ = ( : 0 ) X ht
i.e. at half flow h/ = (t)2 x 20 = 5 m

:. Pressure head at valve = Hs - h/ + h = 20 - 5 + 53·6 = 68·6 m
At point P, hP = (1/L)h = (1000/1500) x 53·6 = 35·7 m
(ht)p' = (1/L)h/ = (1000/1500) X 5 = 3-3m.
:. Pressure head at point P = Hs - (h/)P + hP = 20 - 3·3 + 35·7 = 52·4 m
The variation of the pressure head with time can be shown graphically.
Figure 2.3 (a) shows this variation for the length AB when the retardation
is constant. 'h' is a constant pressure above the steady head corresponding
10 Analysis of Surge
to the instantaneous velocity at any time. The variation of velocity or
flow during closure, when plotted against time for constant retardation, is
a straight line, as shown in Fig. 2.3 (b).
~,-,---,~ I h{ // .
. ~/ I
I / i
I / I
_L·
i~
/
/ .I
--~/ I
I
-----T -----~~ Time
(b) Closure Closure

starts ends
vo, Oo j
_L_j_·-·-"'~- -·- ---:::..~--
FIG. 2.3
Partial changes of flow. Since the pressure changes which we have been
considering depend upon the rate of change of velocity, they apply, with
appropriate changes of sign, for opening as well as closure. They also
apply whether the opening or closure is complete or partial.
For example, if a valve is partly open at first and is adjusted to give a
larger or smaller flow, there will be a corresponding change of pressure.
For constant acceleration/retardation, equation (2.3) becomes
h = ~(vo- Vf) ( 24)
g T .
v0 is the initial steady velocity and Vf is the final steady velocity of the
water in the pipeline.
In Chapter 3 we will deal with elastic waves which are generated in a
pipeline when a valve moves. The incompressible theory which we have
been considering in this section gives a true account of the pressure varia-
tions only when the wave caused by the valve movement returns to the
valve to find conditions there substantially the same as when it left
the valve. We shall find in Chapter 3 that for slow valve movements the
pressure calculated by the elastic theory is twice that obtained by the
incompressible theory.
2.2 Valve movement

In Section 2.1 two separate cases were dealt with together. Length AB
has a downstream control, and reduction of flow results in positive inertia
head. Length BC has an upstream control, and reduction of flow results
in negative inertia head. The case of pump failure results in a negative
inertia head in the rising main in the same way as in pipe BC.
If the valve were to open instead of closing, the inertia heads would be
reversed in sign. A pipeline with downstream control has negative inertia
pressure; a pipeline with upstream control has a positive inertia pressure.
(a)
-----~-=-=--·==-= ._--=-=.: ::-::-:.·_-:: ·_--::= ~-1- .

I
I H
1L ___ _
(b)
1·0
'1.-6/-- vr /
0·8
~~
v (y :Le-1'V/
0
<,_"
0·6 I I I
...!::. I I I
Vo
I /
OA
I V'_
0·2 I I /
J /
0
(£, v
0 0·2 OA 0 6 0 8 1·0
A0 /A
FIG. 2.4
We considered especially the case of uniform acceleration/retardation.
In fact it is impracticable to make a valve which will give uniform rate of
change of flow. 4 Often a gate valve is used to control the flow, and the
movement of the valve is frequently uniform, especially if it is motorized,
but this does not give uniform acceleration.
A partly shut valve results in a head-loss hL where
Ao is the area of the passage through the partly-shut valve and Cc is the
coefficient of contraction. AoCc is the flow area just downstream of the
valve.
For a uniform pipe with constant difference of head between upstream
and downstream ends, the total head-loss is equal to the available head H,
Fig. 2.4 (a). If the friction loss in the pipe of length L is given by the
Darcy equation with constant J, then
H= {(~
AoCc
-t)2 + 4JL}
d 2g
v2 (2.5)
The variation of v with AoCc may be shown by curves, which may be of

non-dimensional form.
Example 2.2
Draw curves to show the variation of velocity against variation of flow area for
pipelines in which 4fL/d = 0·2, 2·0 and 20 when there is a constant head loss in
the pipeline:
SOLUTION {( AoCc
~ _ 1)2 + 4fL}
d 2g
v2 = H
= 2gH
= - - . H- vo 2 4fL
when A/AoCc = 1, v2 vo2 .. - 2g X d
4fL/d
For other values of AoCc,
v2= 2H/{(~ _1)2+

g AoCc
4fL}
d
= vo2xd 4fL/{(~
AoCc
_1)2 + 4fL}
d
Therefore
i; =
J[
v 4fL/d
{(A/AoCc)-1}2 + 4fL/d
J
4JL A
If -=F --=a
d AoCc
then
~ = J((a - ~2 + F)
TABLE 2.1 (as shown in Fig. 2.4 {b))
F= 0·2 F= 2·0 F= 20
(a- 1)2 I (a- 1)2 I (a- 1) 2

a a-1 (a- 1) 2 +F v vo +F v vo vfvo
a +F
0·9 1·111 0·111 0·0123 0·2123 0·97 2·0123 0·99 20·01 1·00
0·8 1·25 0·25 0·0625 0·2625 0·87 2·0625 0·985 20·06 1·00
0·7 1·429 0-429 0·184 0·384 0·72 2·184 0·96 20·18 1·00
0·6 1·667 0·667 0·444 0·644 0·56 2·444 0·90 20·44 0·99
0·5 2·0 1·0 1·0 1·2 0·41 3·0 0·82 21·0 0·98
0·4 2·5 1·5 2·25 2·45 0·29 4·25 0·69 22·25 0·95
0·3 3·33 2·33 5·446 5·646 0·19 7·446 0·52 25·45 0·89
0·2 5·0 4·0 16·0 16·2 0·11 18·0 0·33 36·0 0·75
0·1 10·0 9·0 81·0 81·2 0·05 83·0 0·15 101·0 0·45
With a gate valve controlling flow in a circular pipe, the open area Ao
is not proportional to the movement of the valve spindle, as shown in
Fig. 2.5, where x is the distance which the valve has moved from the fully-
open position and dis the diameter or depth of the valve.
10
~ "'-..
'' ~ /C~rcul6r valve
"' "'\
0·8 '
''
'
' ',,
06
Rectangular valve ''
"
Ao '
""'
A ''
04 '
' 'J\
'~
0·2 '
'----
'\
0
0 0 2 04 0·6 0·8
1'\
1·0
xld
FIG. 2.5
A rotary type of valve, Fig. 2.6 (a), is sometimes used because the rate
of area reduction becomes less in the final part of closure. For a constant
rate of rotation the open area varies in the way shown in Fig. 2.6 (b).
Several investigators have considered the effect of rate of closure for other
types of valve. 5 •6 •7
(b) 10
['\,
0·8 ~'
\
r\' ',
0·6
\ ''
Ao
A
'\ ' ' '
\ ''
0-4 '
~ ''
'
0·2
""""'-..' ' ''
'
0
''
""'-.. ~'---'
0 0·2 0·4 0·6 0·8 1·0
!IT
FIG. 2.6
2.3 Limiting pressures

Maximum permissible bead. The maximum pressure in a pipe caused by
the combination of static head and waterhammer pressure is limited by
the strength of the pipe. To prevent dangerous conditions, with the
possibility of the pipe bursting, the stresses produced by high positive
pressure must be limited to those with which the pipe walls are able to
cope.s,9
Minimum permissible bead. When water hammer causes a drop of pressure

(e.g. opening of a downstream valve) the total head must not be allowed
to fall so low that the water vaporizes. This will be discussed again in
Section 7.2.
The vapour pressure of water varies with the temperature (see Table 2.2)
and may be taken as equivalent to 0·3 m (1 ft) water if temperature con-
ditions are unknown. Atmospheric pressure varies with altitude and
TABLE 2.2. Some properties of water
p w Vapour
Mass Weight Bulk modulus K pressure head
Temperature density density hvp
oc kg/m 3 lb/ft 3
109 N/m. 2 105 lbfin.2 m ft
0 1000 62·4 2·03 2·9 0·062 0·20
10 1000 62·4 2·14 3·0 0·125 0·41
20 998 62·3 2.23 3·2 0·238 0·78
30 996 62·2 2·30 3-3 0·431 1-41
40 992 62·0 2·33 3·4 0·75 2·46
50 988 61·7 2·34 3·4 1·26 4·13
60 983 61·4 2·32 3·4 2·03 6·65
70 978 61·0 2·28 3-3 3·17 10·4
80 972 60·6 2·22 3·2 4·82 15·8
90 965 60·2 2·15 3·1 7·15 23-4
100 958 59·8 2·06 3·0 10·30 33-8
weather, but is about 10·3 m (34ft) water. The maximum negative head
for water is therefore about 10m (33ft) measured relative to atmospheric
conditions. A 'gauge pressure' head of -10 m (- 33 ft) is the minimum
permissible. The rate of opening of a downstream valve may be limited
by the need to keep within this requirement.
When the liquid contains air or gas bubbles in solution-for example
when sewage is being pumped-the pressure of the 'vacuous space'
formed at low pressures may be much higher than the vapour pressure of
air-free liquid.
2.4 Equivalent length of pipe

The inertia pressure in a rigid pipeline consisting of two or more lengths
of different diameters may be calculated from equation (2.2) if the
'equivalent length', L', is substituted for L.
Considering a pipeline of two sections having lengths L1 and L2 and
diameters d1 and d2 (Fig. 2.7), the dynamic equations for the water
columns are similar to equation (2.1), i.e.
dv1
for section 1 : pA1L1 dt = pg(h1 - h2)A1
for section 2:
v1 and v2 are the velocities of the water in sections 1 and 2. By continuity,
AlVl = A2V2, so
h1 = - -1 ( L1 A1) dv1
+ L2- -
g A2 dt
The equivalent length L' is therefore L1 + L2(A1/A2).
f
-sect<on 2 - 1 - Section 1 - 1
Ar
1----L2- - - • + • - - -
FIG. 2.7
For a pipeline consisting of n sections, the equivalent length is given by
LnAl
.+-
An
where L1 is the section nearest to the valve.

The equivalent length of a tapered pipe of length L and end diameter d1
and d2 is given by
REFERENCES
1. KERR, S. L.: 'Surges in pipe lines-oil and water,' Trans. Am. Soc. mech.
Engrs, (1950), 72, pp. 667-677.
2. VALLENTINE, H. R.: 'Rigid water column theory for uniform gate closure',
J. Hydraul. Div., Am. Soc. civ. Engrs, (1965), 91, HY4, pp. 27-33.
3. ANGUS, R. W.: 'Waterhammer in pipes, including those supplied by centri-
fugal pumps: graphical treatment,' Proc. Instn mech. Engrs, (1937), 136,
p. 249.
4. KERR, S. L.: 'Effect of valve action on waterhammer', J. Am. Wat. Wks.
Ass., (1960), 52, pp. 65-72.
5. LUPTON, H. R.: 'Rate of closure of a single flap reflux valve,' J. Instn. Wat.
Engrs, (1960), 14, pp. 209-214.
6. LIVINGSTONE, A. C., and WILSON, J. N.: 'Effects of valve operation,' Symp.
Surges Pipelines, Proc. lnstn mech. Engrs, (1966), 180, 3E, pp. 41-51.
7. PAGDIN, B. C., and SARGINSON, E. J.: 'The analysis and elimination of
waterhammer in small bore mains,' Civ. Engng pub/. Wks. Rev., (1967), 62,
pp. 179-183, 343-346.
8. KERR, s. L., MARSTON, F. A., WOLFE, T. F., CATES, w. H., MAAHS, A. J.,
and KENNISON, H. F.: 'Standard allowances for waterhammer,' J. Am.
Wat. Wks. Ass., (1952), 44, pp. 972-995.
9. PEABODY, R. M.: 'Pump discharge valves on The Colorado River Aqueduct,'
Trans. Am. Soc. mech. Engrs, (1940), 62, pp. 555-563.
3 Elastic Theory for a Simple Pipe
3.1 Sudden closure

In the theory in Section 2.2 we assumed that water is completely incom-
pressible and that a pipe may be completely rigid. The pressure changes
obtained in this way are reasonably accurate if the change of flow is slow
and smooth. If, however, the change of flow is sudden, i.e. if T = 0,
where T is the time of the valve movement, then dvfdt would be infinite
and the pressure increase or decrease would be infinite. Experiments show
that the change of pressure is finite.
An incompressible water column brought to rest would act like a rigid
bar striking a plate-the stresses resulting from the collision would be
transmitted instantaneously through the whole column. In fact, because
of the elastic nature of the water and the pipe, the column acts rather like
a goods train with loose couplings suddenly stopping, 1 Fig. 3.1.
l,c;auo~
~DO~
Mauoa
fiuXJo~"
FIG. 3.1
The same sort of thing happens to water passing along a pipe when an
outlet valve is suddenly closed. Before closure, the water is moving with
velocity v0 , Fig. 3.2 (a). At first we will neglect the friction head as it is
18
Elastic Theory for a Simple Pipe 19
very small compared with waterhammer pressures, but the effect of
friction and other losses will be considered later.
If the valve is now suddenly shut, a pressure wave moves upstream
from the valve with velocity a. Behind the wave front the water is com-
pressed and the pipe walls are stretched, and at the same time water still
~1~---vo---~
L-----
~_LT--i
(bT : 1 -~
1!-------=-T-
v0 I
II
:-- - _-_(1)=
------------j
~ ------------~-- ID=
I
~1 / L':.v / A
clr
FIG. 3.2
enters the pipe from the upstream end with the original velocity vo, Fig.
3.2 (b).
The wave front continues upstream until it reaches the upstream end,
Fig. 3.2 (c), taking time Lja to reach there. The time 2Lja is known as
the period of the pipe, f.l, and we will see later that it is an important
factor in waterhammer analysis. Obviously
L
- = lf.l (3.1)
a
If the total quantity of water which enters the pipe during this time
!p is Ll V, then because it is moving with velocity vo,
Ll V = lvoAp (3.2)
3.2 Pressure increase in a rigid pipe

If the pipe walls are absolutely rigid, the whole of the extra volume ~ V
is added to the original amount of water in the pipe, Fig. 3.2 (d). The
extra water and the original water now occupy the same space (A XL)
as was filled by the water before the valve closed.
The resulting increase of pressure p is given by the approximation,
~v ~v
p=K-=K- (3.3)
V AL
where K is the coefficient of compressibility, or the 'bulk modulus', of the
water.
Substituting from equations (3.1) and (3.2),
voAfl vo
p = K 2AL = K; (3.4)
The wave front meets the moving water with a relative velocity
vo- (-a) = vo +a. Since vo is very small compared with a (whose
value we investigate in Section 3.5), the relative velocity is virtually a.
The mass flow entering the wave front is pvoA and the rate of change of
momentum is therefore approximately pvoA x a since the water is brought
to rest downstream of the wavefront.
The rate of change of momentum is equal to the applied force due to
the waterhammer pressure p, so
pvoAa = pA a= pfpvo
Substituting this value of a in equation (3.4),
p = Kvo -pvo
p
p = voy(pK) h __ vgo J(K-p) (3.5)
We can see from equation (3.5) that the waterhammer pressure is inde-
pendent of the pipe dimensions for a given velocity.
If equations (3.4) and (3.5) are combined,
K~ = voy(pK) a= J(f;) J(X:)= (3.6)
Values of the bulk modulus K are given in Tables 2.2 and 3.2. For water
at normal pressure K = 2·1 x 109 Nfm2 or 300 000 lb/in.2
p = 103 kgjm3 or w = 62·4lbjft3
Therefore a = y(2·1 X 109 X 10-3) mjs = 1450 mjs
a = y(300 000 x 144 x 32·2/62·4) ft/s = 4700 ft/s
TABLE 3.1. Elasticity of materials used for pipe walls

(Figures in italics are values suitable for approximate calculations)
Young's modulus E Poisson's ratio

Material v
109 Nfm 2 10 6 lb/in 2
Aluminium alloy 68-73 0·33

70 10
Asbestos-cement 24 3·5
Brass 78-110 0·36
100 15
Concrete 14-30 0·1-0·3
20 3·0
Copper 107-131 0·34-0·37
120 18
Glass 46-73
70 10
Iron (cast) 80-170 0·21-0·30
100 15
Lead 4·8-17 0-43
10 1·5
Plastics: perspex 6 1·0 0·33
polyethylene 0·8 0·1 0·46
polystyrene 5 0·7 0·40
Steel (mild) 189-226 0·28
210 30
Rocks: Granite 50 7
Limestone 55 8
Quartzite 40 6
Sandstone 3 0·4
Schist 14 2
TABLE 3.2. Some Properties of Liquids

Mass Weight Bulk Modulus K Wave Vapour Pressure
density density Velocity
Liquid p w
Temp Head
kgfm,3 lb/ft3 T~~P J09N/m.2 I o5 Ib/in. 2 T~::';'P m/s oc m
lb/in.2
Benzene 880 55 15 1·05 1·5 20 1320 15 1·2 1·45

Castor oil 950 60 18·6 1500
Carbon tetrachloride 1600 100 20 1·12 1·6 20 940 15 0·8 1·74
Ethyl alcohol 790 49 0 1·32 1·9 20 1160 20 0·8 0·7
Glycerine 1260 79 15 4·43 6·3 30 1920 15 1·1 X 10-6 2 X 10-6
Mercury 13 570 847 20 26·2 38 20 1450 20 1·2 X 10-5 2·3 X J0-5
Oil 900 54 15 1·5 2·2 10 1625 15 4·6 5·7
Petrol 730 45 16·5 1·46 2·1 IS 919 10·0
Sea water 1025 64 15 1510 15 0·2 0·25
This is the velocity of the propagation of sound in water, and we can see
that it is large compared with vo. The value vo is unlikely to exceed 10 mfs.
It is therefore reasonable to neglect vo when considering the change of
momentum.
Combining equations (3.5) and (3.6) we can get
Waterhammer pressure p = vopa
p voa
Waterhammer head h =-=-=Avo (3.7)
pg g
where A = afg.
Equation (3.7) is a form of Joukowski's Law, 2 sometimes called the
Allievi formula, which in general terms is
a
h = - - tlv or h = - Allv (3.8)
g
where h is the increase of pressure associated with the increase of velocity
tlv, and A is the ratio (a/g).
3.3 Strain energy waterhammer theory

When, as is usually the case, the pipe is elastic rather than rigid, the value
of the waterhammer pressure may be obtained by considerations of strain
energy. The kinetic energy of the water before the valve closes is equated
to the strain energy of the water and the pipe wall after closure.
The volume of water in the pipe = AL = !TTd 2L, Fig. 3.3 (a). The
Kinetic Energy of the water in the pipe is therefore given by
K.E. = lpALvo 2
When the pressure wave has travelled the whole length of the pipe and
all the water is at rest, the whole of this kinetic energy has been converted
to strain energy if friction is neglected. The water is compressed and the
pipe walls are stretched.
The strain energy is l the stress x the strain.
Strain energy of the water. The stress is equal to the waterhammer
pressure p, and the strain is equal to the change of volume. From equation
(3.3) the change of volume in a rigid pipe, tlV, is given by pAL/K. In a
stiff elastic pipe (e.g. one made of steel) the increase in diameter resulting
from the increase in pressure is so small that it may be neglected as far
as the water is concerned.
Hence the strain energy of the water (SE)w = !ALp2fK.
Strain energy of the pipe walls. The stress depends upon the method of
anchoring the pipe and the material from which the pipes are made.
If /L is the longitudinal stress,
fc is the circumferential stress,
E is Young's Modulus for the wall material,
and v is Poisson's ratio for the wall material,
the longitudinal strain = (JL/E) - v (fc/E)
the circumferential strain = (fc/E) - v (JL/E)
The strain energy per unit volume of pipe wall = ~t (stress and strain).
The volume of the pipe wall in length L is approximately TTdeL where e
is the thickness of the pipe wall, Fig. 3.3 (b)
A
(~ r---------------------------~>/
--------------L------------
(b)
FIG. 3.3
The strain energy of the whole length of the pipe wall is (SE)p where
2
(SE)p = 1 { /L (/LE - v fc)
E + fc (fcE - /L)} TTdLe
vE
= {(/L 2 + /c 2 - 2vfcfL)TTdLe}f2E
(a) For a thin walled pipe without expansion joints which is fixed at
the upper end and is free to move in a longitudinal direction throughout
its length,
Therefore
But bd2 = A, and so
(SE)p = ALp2 ~ (~ -
2E e 4
v)
(b) For a thin-walled pipe without expansion joints which is anchored
against longitudinal movement throughout its length,
dp pd
/L = Y - and fc = -
2e 2e
ALp 2 d
and (SE)p = 2E ; (1- v2)
(c) For a thin-walled pipe which has expansion joints throughout its
length,
pd
/L = 0 and fc = -
2e
ALp2d
and (SE)p = 2E ~ (1 - lv)
In general (SE)p = ALp2 ~ c

2E e
where c is a constant depending upon the method of anchoring.
Energy balance. The energy before closure is equal to the energy after
closure, i.e. KE = (SE)w + (SE)p
~pALvo2
2
= ALp2
2
{2.K + eEde}
(3.9)
1 1 de
or p = voVpK' where-=-+- (3.10)
K' K eE
and 3 c = (i - v) for a pipe free to move longitudinally
c = 1 - y2 for a pipe anchored against longitudinal movement
c = 1 - tv for a pipe with expansion points
If longitudinal stress can be neglected so that c = 1, then,
(3.11)
In a rigid pipe the term dfeE is negligible compared with 1/K and
equation (3.11) becomes identical with equation (3.5).
Equations (3.9) and (3.11) may also be obtained by considering the
quantity of water contained in the pipeline when the waterhammer wave
has reached the upstream end of the pipe. This may be done by obtaining
partial differential equations applying to a column of water contained
within a short length of pipe.
Example 3.1
A 200 mm diameter cast iron pipe with wall thickness of 15 mm is carrying water
when the outlet is suddenly closed. Calculate the maximum discharge if the pressure
rise is not to exceed 180 kN/m 2 ,
(a) assuming the pipe is rigid;
(b) allowing for hoop stress but not longitudinal stress in the pipe wall;
(c) assuming the pipe to have expansion joints throughout its length.
SOLUTION (a) p = vy(pK) . -

· · v- vCpK)
p
For water p = 10 3 kg/m 3 180 X 103

K= 2·1 GNfm2 v = y(103 x 2·1 x 109) = 1"24 m/s
= Av = !7T x (0·20) 2 x 1·24 = 0·0314 x 1·24 m 3/s = 39·0 litresfsecond
J(i: ,~) v~p J(i: e~)

Q
(b)
p~v
" ~ J(z ~ O·~.:;
180 x 10, 1 IO' +
02
x ; :}. x IO')
= 1·41 m/s
Q = 0·0314 x 1·41 m 3/s = 44·1litresfsecond.
(c)
c = 1 - tv; v = 0·25 :. c = 0·875
V = 180 X 103 J( 2·1 ! 109 + 0·0~~2~ ~~· 8: 510 9)

103
= 1-39 m/s
Q = 0·0314 X 1·39 m3 /s = 43·5 litresfsecond.
3.4 Fundamental differential equations of waterhammer

In this section we obtain fundamental equations for a pipeline with
downstream control while the flow in the pipe is varying-i.e. the valve is
being opened or closed. For simplicity we take a horizontal pipe, Fig. 3.4,
but the same equations are obtained if the pipe is sloping since the gravita-
tional force on the column of water in the pipe balances the force caused
by difference of static head.
Static level
·-·-·-·-·-·-·-·-·
FIG. 3.4
The convention used by Allievi is adopted. The positive direction of x

is upstream from the valve, and the positive direction of v is towards the
valve.
If the pressure varies along the element of length ox, the unbalanced
force acting along the axis of the pipe on the element is A(opfox)ox. The
mass of the element of water is pAox and the acceleration is dvfdt, so from
Newton's second law,
Since v is a function of x and t,

dv ov ov dx ov ov
-=----=--V-
dt ot ox dt ot ox
and op = p (ov- v ov) (3.12)

ox ot ox
However, ovfox is very small compared with ovfot and may be neglected.
Equation (3.12) becomes approximately
op ov oh 1 ov ov oh
ox = P ot or ox = got or ot = g ox (3.13)
By continuity, the difference between inflow to and outflow from the

element in time otis A(ovfox) . ox. ot, and this is equal to the additional
volume due to the elasticity of the pipe walls plus the additional volume
which may be contained in the element due to the compressibility of the
water.
The elasticity of the walls gives an additional volume in time l5t
approximately equal to
~ de op l5t . bx
E e ot
and the volume stored because of compressibility is
A bx op bt
K ot ·
If differentials higher than the first order are neglected,
ov = (2.. + de) op
ox K eE ot
If we introduce the symbol a where
P (2..K + eE
de) = _!_
a2
and express the pressure in terms of head, then
ov g oh oh a2 ov
- = - - or - = - - (3.14)
ox a2 ot ot g ox
The general solutions of equations (3.13) and (3.14) are 4
h=H-Ho=F(t-~)+f(t+~) (3.15)
flv = v- vo = - ~ {F(t- ~) -f(t + ~)} (3.16)
Equations (3.15) and (3.16) are the fundamental equations of water-

hammer. The symbols 'F' and 'f' may be used as abbreviations to repre-
sent F(t - xfa) and f(t + xfa), and the fundamental equations become
h = H- Ho = (F +f); flv = v- vo = -?. (F- f) (3.17)

a
The significance of these equations is that two systems of co-existent
variable pressures ('waves' or 'surges') are propagated along the pipe in
opposite directions with constant velocity a. The F-wave moves in the
positive x direction (i.e. upstream from the valve) and the f-wave moves
in the negative x direction (i.e. towards the valve). The waves may be
of any shape, and move as a whole without deformation along a length of
uniform pipe. Both F and/ depend upon the boundary conditions, which
in the pipe we are dealing with are the reservoir at the upstream end and a
valve at the downstream end. The waves are such that they do not
mutually interfere; when an F-wave passes an f-wave, neither wave is
changed in any way.
3.5 Velocity of propagation

We saw in Section 3.2 that in a rigid pipe the waterhammer waves move
with the velocity of sound in water, a, which for a rigid pipe is about
1440 mfs (4700 ft/s). For a pipe with elastic walls the value of the bulk
modulus used in equation (3.6) is modified by equation (3.10) so that
a= y(K'fp) = J{gfw(~+ :~)} (3.18)
For other types of pipe and tunnel the 'modified' bulk modulus is as
follows:
Thick-walled pipe ..!._ = _!_ 2c (D2 + d2)

K' K+ E D2- d2
Rock tunnel 4 ·5
de
K + (ec
R ·-r d . 1 1
etworce concrete ptpe K' = As)
;:;; +-:;- Es
1 1 'lfJ
Generally -=-+-
K' K E
where ec is the thickness of the concrete

m is the ratio of the modulus steel/concrete
As is the cross-sectional area of each longitudinal reinforcing bar
s is the spacing of the reinforcing bars
Es is Young's Modulus for steel in a reinforced concrete pipe
'lfJ is the rigidity factor.6,7
The value of a depends upon the elasticity and inertia of the liquid
column in its containing pipe. Equation (3.18) may be written as
stiffness
)
density
If we think again of the train mentioned in Section 3.1, the speed of the
'shock wave' that brings the wagons to rest depends on the ratio of the
stiffness of the buffer springs to the weight of the wagons.
Looking at equation (3.18) we see that the velocity of the waterhammer
wave depends on three factors. A comparison of the wave-velocity in
standard commercial pipes conforming to British Standard Class C makes
this clear. s
Pipe size. Small pipes are stiffer than large pipes and the wave-velocity
is therefore greater. The velocity in a 3 in. diameter steel pipe is 4320 ftfs,
in a 12 in. diameter steel pipe it is 3620 ft/s, and in a 72 in. diameter pipe
it is 3450 ft/s.
Pipe material. Young's Modulus varies for different materials and
weaker materials need a greater thickness of pipe wall. If we compare
6 in. diameter pipes, each designed for the same pressure, we find the
wave-velocity in a steel pipe is 4100 ft/s, in a cast-iron pipe it is 3905 ft/s,
and in an asbestos cement pipe it is 3730 ft/s.
Liquid. The density and the elasticity of the liquid both effect the wave-
speed. In a 6 in. diameter Class C steel pipe the wave in glycerine moves
at 4560 ft/s, in sea-water at 4220 ft/s, in fresh water at 4100 ftfs, and in
petrol at 3630 ft/s.
The value of a is affected by temperature;9,10 in water it increases by
about 1% for every soc at normal atmospheric temperature, and reaches
a maximum at about 75°C. High pressures also increase K and hence a. 9
Free gas bubbles have a considerable effect, which may be important
if sewage is being pumped. For example, at a pressure head of about
10m (30ft) the presence of one part of air in a hundred parts of water
reduces the velocity to about 10% of that in air-free water.ll
A rough average figure for a is 1200 m/s, or 4000 ft/s.l2
Example 3.2
Calculate the velocity of propagation of waterhammer waves in a 500 mm
diameter pipe under the following conditions. Neglect longitudinal stress, i.e. c = 1.
(a) Rigid pipe containing fresh water;
(b) rigid pipe containing glycerine;
(c) elastic steel pipe 10 mm thick containing fresh water;
(d) cast iron pipe 100 mm thick (thick walls) containing castor oil;
(e) reinforced concrete pipe 50 mm thick with 20 mm diameter longitudinal
steel reinforcing bars 100 mm apart, containing fresh water.
SOLUTION
(a) a = J( pK) = J(2·1 X 109)

103 = 1450 mfs.
(b) a = J( !\
pJ = J(4·43 X 109)
1260 = 1840 m/s
(c) a =
j( (K 1
1
d
+ eE)
)
= J( 103 (2·1
1
1
50
x 109 + 2·1 x 1011)
)
J ts
· p
= e·1 106) = 1183 m/s
(d) a = J( (-" p K
2(~' +
+ E(D2 - d2)
d')J)
J( ( J) = 1132 m/s
J
= 1 1 2(0·36 + <>25)
960 1·48 X 109 + 1·0 X 1011(0·36- 0·25)
(e) a~ (P (~+ (~+;) £.))

= J( HXMl (21 ! 10' + (~ + ~)
1
2·1 X ton)
) = 1100m/s
3.6 Direct application of elastic waterhammer theory

The equations for the waterhammer pressure which we have found apply
to 'sudden' closure.
We saw that h = vo a (3.5)

g
=; J(;)
=; Jc(~ ~ fH))
for a rigid pipe (3.5)
h
and for a thin walled elastic pipe h (3.9)
Subsequent variation of the pressure will be considered in Chapter 4

and we will see in Chapter 5 how the pressure varies when the closure is
neither slow, as we considered in Chapter 2, nor sudden. We will also see
that as far as pressure increase near the valve is concerned the closure may
be considered as 'sudden' when the time taken for the valve closure, T,
is less than the time it takes for the waterhammer wave to pass along the
pipe to the open end and return to the valve. We noted in equation (3.1)
that the time taken by the wave to traverse the pipe is !p. The returning
wave, which is anf-wave, travels at the same speed and therefore also takes
a time !t-t to traverse the pipe. Therefore the limiting time for valve
closure such that the maximum pressure at the valve is that given by
equations (3.7) and (3.9) is given by
L
T ~ p where p = 2 -
a
REFERENCES
1. LUPTON, H. R.: 'Surge control in pipelines,' J. Instn Wat. Engrs, (1965),
19, pp. 81-85.
2. JouKOWSKI, N.: 'Waterhammer' (Translated by Miss 0. SIMIN), Proc. Am.
Wat.Wks Ass., (1904), 24, pp. 341-424.
3. PARMAKIAN, J.: Waterhammer analysis (Dover, 1955).
4. A.LLIEVI, L.: Teorie del colpo d'ariete (Atti Collegio Ing. Ach, 1913). Trans-
lation by Halmos, E. E.: The theory of waterhammer (Am. Soc. mech.
Engrs, 1929).
5. JAEGER, C.: 'Present trends in the design of pressure tunnels and shafts
for underground hydro-electric power stations,' Proc. Instn civ. Engrs,
(1955), 4, pp. 116-174.
6. SWAMINATHAN, K. V.: 'Velocity of waterhammer waves in embedded steel
penstocks,' Civ. Engng pub!. Wks Rev., (1964), 59, pp. 1409-1413.
7. SWAMINATHAN, K. V.: 'Waterhammer wave velocities in concrete tunnels,'
Wat. Pwr, (1965), 17, pp. 117-121.
8. LINTON, P., 'A simple guide to waterhammer and some notes on pressure
surges in pump delivery lines,' Br. Hydromech. Res. Ass., (1961), TN411.
9. GIBSON, A. H.: The mechanical properties o,(.fluids (Blackie, 1923) p. 210.
10. GOLDMAN, G. 0.: Water hammer: its causes, magnitude, prevention (Colum-
bia Graphs Inc., 1953).
11. PEARSALL, I. S.: 'The velocity of waterhammer waves,' Symp. Surges Pipe-
lines, Proc. Instn mech. Engrs, (1966), 180, 3E, pp. 12-20.
12. KERR, S. L.: 'Practical aspects of waterhammer,' J. Instn Wat. Engrs,
(1949), 3, pp. 67-74.
FURTHER READING
ENEVER, K. J.: 'An introduction to pressure surges in gas-liquid mixtures,'
Br. Hydromech. Res. Ass., (1967), 9th Members Conf, SP 920.
HALLIWELL, A. R.: 'Velocity ofwaterhammer wave in an elastic pipe,' J. Hydraul.
Div., Proc. Am. Soc. civ. Engrs, (1963), 89, pp. 1-21.
KENNISON, H. F.: 'Surge-wave velocity-concrete pressure pipe,' Trans. Am.
Soc. mech. Engrs, (1956), 78, pp. 1323-1327.
4 End Conditions
In Chapter 3 we found that sudden closure of a valve causes a wave of

pressure intensity given by equation (3.7) or (3.8) to move upstream with
velocity a. We also found that two counter-moving waves, the F-wave
and the f-wave, co-exist in a pipe when there is a change of flow.
These waves continue along a uniform pipe without change. At the
end of the uniform pipe the waves are reflected or transmitted in a modi-
fied form, depending upon the conditions at the end. The effects of different
types of end-condition are considered in this chapter.
4.1 Reftection at a reservoir

A reservoir is considered to be such that the pressure is always constant
there. It may therefore be an 'open' reservoir or tank so large that varia-
tion of flow or pressure in the pipe does not effect the water level. It may
also be a closed conduit which acts in the same way-for example a large
water main in which the water is scarcely affected by waterhammer in a
smaller branch-or it may be the 'vacuum' between two columns of
separated water such as may occur in pumping mains (Chapter 7).
Consider the simple case of a reservoir feeding a uniform pipeline at the
end of which is a valve, Fig. 4.1 (a). On sudden closure of the valve an
F-wave is propagated towards the reservoir with velocity a, taking time
!11- from the instant of closure to traverse the pipe, Fig. 4.1 (b). When the
F-wave reaches the reservoir, a reflected f-wave is generated in a down-
stream (negative x) direction. At the reservoir the head is constant and
therefore H = Ho. Since at the reservoir x = L, equation (3.15) gives
32
End Conditions 33
and when this is substituted in equation (3.16) we obtain the velocity,
v = vo - 2: F ( t - ~)
The meaning of these expressions is that at a reservoir the reflected
wave is equal to the incident wave but of opposite sign, and the change of
velocity at the reservoir caused by the wave is twice what it is in the
uniform pipe, Fig. 4.1 (c).
-----------L------------
TL---~--t-____,Ho'--+
!:::. v=-h!A. =-vo
(b)
h_--n--.
v=IQ + D.v =0
F
FIG. 4.1
In general this reflection applies to a reservoir whether the incident wave

is an F-wave or an f-wave and whether the velocity before and behind
the wave is positive, negative or zero.
4.2 Reflection at a dead end

In the case of a uniform pipe fed by a reservoir, the f-wave which is
reflected from the reservoir arrives at the valve after a further time i,u
(i.e. t = ,u) and finds that the valve is completely shut-it is a 'dead-end'
or 'barrier'. Obviously at such a barrier the velocity both before and after
the arrival of the wave is zero. That is, as far as the barrier is concerned
v = vo = 0, and from equation (3.16)
which can be substituted in equation (3.14) to give
h = 2F (t- ~)
The meaning of this expression is that at a dead end the reflected wave
is equal to the incident wave and of the same sign, as shown in Fig. 4.2.
Before
+h 4.3 Effect of changes of pipe
6v [ 0
If an F-wave travelling along a pipe
reaches a point where the pipe changes,
there is both reflection and transmission
of waves. Such a 'change of pipe'
occurs where there is a change of
After
+h diameter-i.e. an enlargement or a
<
contraction-or if the material or
H0 +2h I:
thickness of the pipe-wall alters so
6v 0 that K' changes. Whatever change the
<
pipe undergoes in these ways, the ratio
FIG. 4.2 A/a is different on either side of the
point.
To make the notation clearer, Fig. 4.3 has been drawn with the F-wave
moving from left to right: the positive x-direction is also towards the right
and the positive v-direction is towards the left.
Fi is the incident wave, Ft is the transmitted wave and fr is the reflected
wave. Applying equations (3.16)
hr = Fi + fr; V1 - Vol = - g (Fi - fr)

Gl
and by continuity, A1vo1 = A2vo2; A1v1 = A2v2.

If velocity head and the head loss at change of area are neglected, the
solution of these equations is
(4.1)
End Conditions 35
and Ct- Cr = 1, where
A1la1 - A2/a2
Cr is the reflection factor
A1/a1 + A2la2
. . . 2A1/a1
I
Ct IS the transmissiOn factor =
A1 a1 + 2Ia2
A
h,
Before
H0 +h, i Ho p Ho
vo,+ L':.v, b Vol voz

F,
Pipe I 2
Area
Wove velocity a,
A, Az
oz
L':.v, =±.h!At
"i -(v01 tL':.~) =-t/:>.. 1
hr ht
v2 -v02 =h,!:X. 2
After
Ho+ h,. 1 H' pH' j Ho
VOlt L':,vl .J v, vz b voz

f, Fi
P may be an enlargement, a constnction,
or a change of pipe material or thickness.
FIG. 4.3
You can see that at an enlargement the transmitted wave has the same
sign and smaller amplitude and the reflected wave has the same sign and
smaller amplitude as the incident wave.
At a constriction the transmitted wave has the same sign and greater
amplitude and the reflected wave has the same sign and smaller amplitude
as the incident wave.
4.4 Effect of junction

The way in which a wave is reflected and transmitted at a junction is
obtained in the same manner as in Section 4.3, see Fig. 4.4.
~Vl = h/Al
VI-(Vo + ~Vl) = -hr/Al
V2 - Vo2 = ht/A2
va- Voa = ht/Aa
fr = CrFi; F2 = Fa = CtFi; Ct - Cr = 1
Before
P1pe
Area
Wave velocity
After
FIG. 4.4
4.5 Effect of restriction
The most common restriction or 'throttle' in waterhammer is a partially-
open (or partially-shut) valve, but any type of restriction gives rise to
reflected and transmitted waves which differ from the incident wave in
amplitude.l However, both the reflected wave and the transmitted wave
have the same sign as the incident wave.
If the head loss across the restriction is KLvn, the diagram showing the
waves is as shown in Fig. 4.5.
Before
+h
H0 th Ho Ho- kL vo"
!
v0 t /':;v vo vo
F;
After
hr ht
H 0+ h Hr H, H0 -kL v;
.:.
v0 + /':;v v' v' vo
~
"
F,
FIG. 4.5
End Conditions 37
The equations relating the pressure and velocity changes are
hr = Hr- (Ho +h)= -.A.{v'- (vo + Av)}
ht = Ht - (Ho - KLvn) = .A.(v' - vo)
and hr - KL( v')n = ht
When a valve discharges to atmosphere, the flow derived from
Bernoulli's equation is
Q = Av = CcAoy'(2gH)
where Ao is the 'open area' of the valve as in Section 2.2. If now we
designate the term (CcAo)y'(2g)/A as B, then v =By'H.
For full opening vo = Bov'Ho and v may be expressed in non-dimen-
sional terms as
V = Vo !!._
Bo
J(H)Ho
= TZVo (4.3)
where T = B/Bo is the proportion of the full valve opening and

z = y'(H/ Ho) the square root of the ratio of the total head to the initial
head.
By combining equation (4.3) with Joukowski's Law (equation 3.8) and
introducing the further non-dimensional group (avo)/2gH0 = ~. which is
called the 'pipe constant', we obtain Allievi's classical equations.
Z12 -1 = 2~(ToZo - TIZl) (a)

Z1 2 + Z2 2 - 2 = n(TIZl- T2Z2) (b) (4,4)
z2 2 + za 2 - 2 = 2~(T2Z2 - Taza) (c)
The suffixes o, 1, 2 • . • refer to t = 0, t = p,, t = 2p, . .
Allievi systematically analysed these equations for a variety of valve

movements, and presented the results on charts.
4.6 Oscillation of the pressure wave

Because we know the effect of end conditions, we are now able to follow
the variation of pressure in a simple pipeline, such as that we considered
in Section 3.1, Fig. 3.2 (a), where a uniform frictionless pipe leads from a
reservoir to a valve which is suddenly closed.
When the F-wave (a positive pressure wave) reaches the reservoir at
time it-t the whole of the kinetic energy originally in the moving water has
r-- --·- -·-

-1~:
(a) -
been converted to strain energy. The
energy is 'stored' in compression of the
n ---------- water and stretching of the pipe walls,
Fig. 4.6 (a).
- [: Just as a spring rebounds from its
lowest position after a weight has been
placed on it, the pressure wave is
reflected, as we have seen in Section
4.1. When the F-wave reaches the
reservoir the strained condition of the
water and the pipe walls is unstable and
the stored energy forces water out of
the pipe at the upstream end into the
reservoir. Then ...
During time itt to tt. Fig. 4.6 (b)

An f-wave moves downstream at
velocity a with the original pressure
behind it (since the wave is equal to the
F-wave but of opposite sign).
Behind the wave-front the water
decompresses.
Behind the wave-front the pipe resumes
its original diameter.
Water behind the wave-front moves
upstream with negative velocity -vo.
The f-wave takes time itt to reach the
valve, and therefore arrives there at time
t = tt· The conditions in the pipe are
now ...
At time tt. Fig. 4.6 (c)

All the water is at its original pressure.
All the water is at its original density.
All the pipe is its original diameter.
All the water in the pipe is moving
with velocity -vo.
The water and pipe have returned to
FIG. 4.6 exactly the same conditions as at the
End Conditions 39
instant when the valve closed, except that the water is now moving with
negative velocity.
The f-wave arrives at the valve to find it shut. The end of the pipe is a
'dead-end' and therefore the reflection described in Section 4.2 occurs.
During time p, to Itp,, Fig. 4.6 (d)

A negative F-wave moves upstream at velocity a, causing a reduction of
the pressure below the static level.
This F-wave reaches the reservoir tp, after it leaves the valve-i.e. when
t = ltp,, and then ...
At time Itp,, Fig. 4.6 (e)

All the water is at pressure h below the static pressure.
All the water is at reduced density.
All the pipe is at reduced diameter.
All the water is at rest.
However, the condition is unstable. The 'stored' negative energy draws
water in at the upstream end. . . .
During time 1tp, to 2p,, Fig. 4.6 (f)

A positive f-wave moves downstream at velocity causing the pressure
to rise to the original static level.
At time 2p,, Fig. 4.6 (g)
All the water is at its original pressure.
All the water is at its original density.
All the pipe is its original diameter.
All the water is moving with velocity +vo.
The waterhammer wave has now passed through a complete cycle

of operation oscillation, and the conditions are as they were when the
valve closed.
Valve
closes
FIG. 4.7
In the absence of friction the cycle would be repeated ad infinitum. The
variation of pressure with time can be shown graphically; near the valve
the change of pressure head is 2h every half-cycle, as shown in Fig. 4.7.
The oscillation at a point distance nL from the valve is shown in Fig.
4.8.
FIG. 4.8
Stereograms which show the variation of pressure with time at any

point in a pipe-line, by a three-dimensional diagram, have been developed
by Professor du Juhasz.2 (See Fig. 4.9).
Stereogram of oscillation following sudden complete

closure of downstream valve.
Valve
closes
FIG. 4.9
End Conditions 41
Effect of friction. Because of friction in the pipeline the pressure-time
pattern is modified, and the oscillation gradually dies out. Near the valve
the pressure variation with time may be of the form shown in Fig. 4.10. 3 • 4
FIG. 4.10
REFERENCES
1. CoNTRACTOR, D. N.: 'The reflection of waterhammer pressure wave from
minor losses,' J. bas. Engng, Trans. Am. Soc. mech. Engrs, (1965), 87,
pp. 445-452.
2. AomsoN, H.: A treatise of applied hydraulics (Chapman and Hall, 5th Ed.,
1964).
3. BINNIE, A. M.: 'The effect offriction on surges in long pipelines,' Q, J1 Mech.
appl. Math., (1951), 4, pp. 330-343.
4. GIBSON, A. H.: The Mechanical properties ojjfuids (Blackie, 1923).
FURTHER READING
STEPHENSON, D.: 'Waterhammer charts including fluid friction,' J. Hydraul
Div., Proc. Am. Soc. civ. Engrs, (1966), 92, HY5, pp. 71-94.
5 Non-Instantaneous Closure
and Opening
5.1 Rapid closure of downstream valve T < fl

In practice, valve closure is rarely instantaneous. However, as was men-
tioned in Section 3.6, the maximum pressure rise near the valve is the same
as for sudden closure when the time of closure Tis equal to or less than
the critical time fl· The increase of pressure builds up during the time T.
The retardation resulting from the closure of a downstream valve may
be considered as consisting of n steps each lasting a time Tjn. If the retard-
ation is constant the change of velocity at each step is -vofn and by
Joukowski's Law (equation 3.8) the corresponding pressure rise for each
step, !J..h, is given by
!J..h =it Vo
n
Since there are n steps and each gives a pressure rise of !J..h, the total
pressure rise h = n!J..h = il.vo, which is the same value as for sudden closure.
If n is infinitely large, the steps become infinitely small, and the pressure
rise near the valve can be represented by a straight line, as in Fig. 5.1 (a).
Considering a finite number of steps again, at each time-interval Tjn a
positive F-wave moves up the pipe with velocity a. Each wave in turn
reaches the reservoir at time !t-t after it left the valve. At the reservoir the
wave is reflected as a negative f-wave.
When the firstf-wave arrives at the valve (at time p after closure started)
it finds the valve completely closed since more than time T has elapsed.
The valve therefore represents a dead-end, and a negative F-wave is
reflected. The effect at a point near the valve is that in a further period of
time Tfn the pressure drops by 2!J..h = 2hfn. This rate of pressure change
continues until the last of the n positive F-waves returns to the valve as a
42
Non-instantaneous Closing and Opening 43
negative f-wave to be reflected there as a negative F-wave! This occurs at
time (p, + T) after closure began. The variation of head with time can be
shown graphically by Fig. 5.1 (b).
(a)
I
I h = \ v0
--.---.,...-- - - _J_ __L - - -

I I
I I
I I
1-T~
1 T1me
(b)
T l __ _
I h
~~I_..___--' •
I = 0 T fi- fL+ T T1me
FIG. 5.1
If the retardation is not constant but the pipeline velocity is known at

intervals of Tfn during closure, the pressure variation following rapid
closure can still be found by the step-by-step method we have just
considered. The known variation of velocity in time Tfn is Llv and the
corresponding increment of pressure can be found by Joukowski's law.
Example 5.1
When the valve at the end ofa 3900 m long pipeline closes in4i seconds the velocity
is reduced as follows-
Time after start of

closure, seconds 0 0·5 1·0 1·5 2·0 2·5 3·0 3·5 4·0 4·5
Pipe velocity mfs 1·75 1·70 1·60 1·45 1·25 1·00 0·75 0·50 0·25 0
The modified wave velocity is 1300 mfs and the static head throughout the pipe
is 140m. Neglecting friction find the total pressure head near the valve during
closure.
2L 2 X 3900
SOLUTION p, = -a = 1300
= 6 seconds
Therefore T < p,; the closure is rapid.
a 1300
II. = g- = - -=
9·807
132·35 seconds.
t secondsO 0·5 1·0 1·5 2·0 2·5 3·0 3·5 4·0 4·5
~v m/s -0·05 -0·10 -0·15 -0·20 -0·25 -0·25 -0·25 -0·25 -0·25
~h = -ll.~v m 6·62 13·23 19·85 26·47 33·09 33·09 33·09 33·09 33·09
h m 6·6 19·8 39·7 66·2 99·3 132-4 165-4 198·5 231·6
H m 146·6 159·8 179·7 206·2 239·3 272-4 305-4 338·5 371-6
For discharge from a valve to atmosphere where the law of closure is

known, the pressure during closure can be found by applying equation
(4.3) or (4.4) (a). If we use the non-dimensional equation (4.4) (a), it is
clear that at the start of closure t = 0, To = 1 and zo = 1. The value of
z1 is found by solving the quadratic equation z1 2 - 1 = 2'(1 - nz1)
which gives
In the part of the pipe nearest the valve the maximum pressure for rapid
closure is the same as for sudden closure, but at a point near the reservoir
the maximum pressure is less. This is because returning.f-waves reach the
point before all the positive F-waves have passed. By Allievi's theory
(equation 3.15) the.f-waves are added to the F-waves.
Example 5.2
Obtain the total pressure head during the first 13 seconds after closure starts
at a point P, 2600 m from the valve in the pipeline in Example 5.1 for the same
conditions.
SOLUTION P is 2/3L from the valve.

The first F-wave reaches P at time 2/3Lfa after leaving the valve, i.e. t =
2600/1300 = 2 seconds.
The first fwave reaches P at time 4/3Lfa after leaving the valve, i.e. t =
4 seconds.
A negative F-wave reflected from the closed valve reaches P at p, + 2f3Lfa
(i.e. 8 seconds) after the valve starts to close and a positive fwave reflected from
the reservoir reaches P at p, + 4/3Lfa (i.e. 10 seconds) after the valve starts to
close.
!seconds 2·0 2·5 3·0 3·5 4·0 4·5 5·0 5·5
F1m 6·62 13·23 19·85 26·47 33·09 33·09 33·09
[1m -6·62 -13·23 -19·85
h m 6·6 19·8 39·7 66·2 92·6 112·5 125·7
Hm 146·6 159·8 179·7 206·2 232·6 252·5 265·7
t seconds 6·0 6·5 7·0 7·5 8·0 8·5 9·0 9·5

F1m 33·09 33·09 -6·62 -13·23 -19·85
[1m -26·47 -33·09 -33·09 -33·09 -33·09 -33·09
h m 132·4 132·4 99·3 66·2 33·1 -6·6 -19·8 -39·7
Hm 272·4 272-4 239·3 206·2 173·1 133-4 120·2 100·3
t seconds 10·0 10·5 11·0 11·5 12·0 12·5 13·0

F1m -26·47 -33·09 -33·09 -33·09 -33·09 -33·09
ji m 6·62 13·23 19·85 26·47 33·09 33·09
h m -66·2 -92·6 -112·5 -125·7 -132·4 -132-4 -99·3
Hm 73-8 47-4 27·5 14·3 7·6 7-6 40·7
Whether the retardation is uniform or not, the part of the pipe which
will be subjected to the same maximum and minimum pressures as at the
valve is that part where the whole of the first set ofF-waves has passed
before the first f-wave returns. If x is the distance from the valve of any
point, X, as shown in Fig. 5.2 (a), the last of the n F-waves which were
(a)
- - m-c~- -i;--x~~·~;x-~
-
~-------------L----------~~
(b)
-E
2 = L; / hrnax
II hmox II
L_i_ _ _ _j___ · __ _
I I I ,.
I
1-x '~ I
L-2Ta~
I II2 ra ------.-1
~----x--------
FIG. 5.2
generated by the closing of the valve reaches X at time T + xfa after

closure began. The first f-wave reaches X at time !Jt + (L- x)fa =
(2L - x)fa after closure began.
Let x' be the distance from the valve of the point which limits the part
of the pipe subject to the same maximum and minimum pressure as at the
valve. Thenl
T + x' fa = (2L - x')/a or x' = L - tTa
The part of the pipe at the reservoir end which is subject to reduced
maximum and minimum pressures has a length of !Ta.
If the retardation is uniform, the waterhammer pressure head at any
time t after the start of closure at the point X, where x is greater than
L - !Ta, is the sum of
due to the F-waves
2L-x
t----
a
and - hmax - - T - - due to the f-waves
where hmax is the pressure rise for instantaneous closure = AVo.

The maximum head at X is therefore
h -_ hmax 2(L- x)
aT
The variation ofthe maximum head along the pipe for uniform retardation
is shown in Fig. 5.2 (b).
To some extent the same reasoning can be applied to rapid opening of a
valve and to rapid partial movements. However, when the first f-wave
reaches the valve it does not come to a 'dead-end', and subsequent pressure
variation differs from that for rapid complete closure. In the case of
opening, the sign of the waves is reversed, so that the first F-wave is negative
and the first f-wave is positive.
5.2 Slow closure of downstream valve

When the time taken for the valve to close is greater than p, the maximum
waterhammer pressure does not attain the value hmax (=Avo) at any part
of the pipe. If we consider the valve movement as consisting of a number
of incremental movements each generating a corresponding F-wave we
can appreciate that the firstf-wave reflected from the reservoir reaches the
valve at time p, from the start of closure and this is before closure is
complete. 2
The first f-waves therefore reach the valve before the last of the initial
F-waves has been generated. If there are n incremental steps, it requires
a complete set of n F-waves to give a pressure of h. Obviously less than n
waves gives a smaller pressure.
When the first f-waves reach the valve they are partly transmitted and
partly reflected as a second set ofF-waves which are of the opposite sign
to the first F-waves. The pressure in the pipe therefore drops further
below h.
If the valve discharges to atmosphere the velocity may be obtained from
equation (4.3). The pressure can be calculated using Joukowski's Law
with equation (4.3) or by using equations (4.4).
As in the case of rapid closure, step-by-step methods can be used to
determine the variation of pressure. The duration of the steps need be
no less than the time which is taken for a wave to reach the reservoir and
the reflected wave to return. That is, at the valve the steps can be of
duration f-l and at other points distance x from the valve the duration need
to be no less than 2(L - x)fa. Shorter steps do not increase the accuracy
of the computation, but enable the pressure at intermediate times to be
determined.
If the retardation is constant, there is a linear rise of pressure during
time f-l from the start of closure, the maximum increase being J-lhmax/T.
Consider a pipeline in which a downstream valve closes in time T =
NJ-l so as to give uniform retardation. Tabulating, we obtain
F h
v vo- v A(vo- v) f A(vo- v) +f F+f
0 vo 0 0 0 0 0
ft (N- l)vo/N vo/N A.vo/N 0 A.vo/N A.vofN
2p, (N- 2)vo/N 2vo/N 2Avo/N -A.vo/N AvofN 0
3p, (N- 3)vo/N 3vo/N 3Avo/N -A.vo/N 2A.vo/N A.vo/N
4p, (N- 4)vo/N 4vo/N 4Avo/N -2A.vo/N 2A.vo/N 0
The pattern of the oscillation is therefore as shown in Fig. 5.3 (a).

In Chapter 2 (equation 2.3) we found that the pressure rise for slow
closure is
L vo
h=--
gT
when water is considered as incompressible. We now find that
h = Avo = (afg)vo
N Tff-l
Now p = 2Lfa, so by the elastic theory
( ~ vo) 2L
h= g a =2L!!!!
T g T
We therefore see that the elastic theory gives a pressure increase for
slow closure which is twice that given by the incompressible water column
theory. Charts obtained when waterhammer pressures have been
measured show how the two theories may be reconciled. Fig. 5.3 (b)
(a}
AV\1\ 1\
FIG. 5.3
shows the type of pressure-time variation which may occur when the
oscillations are damped by friction. It is drawn for the same conditions
as Fig. 2.3 (a). Elastic 'shock' waves, which are superimposed on the
inertia head increase, are due to the sudden application and cessation of
the retardation itself.
In the absence of friction, oscillations continue after the valve has
stopped moving except when the time of closure is a multiple of 2p (e.g.
T = 4p, T = 6p . . .). Typical pressure-time patterns are shown in Fig.
5.4, where the retardation is constant.
5.3 Other types of valve movement

The tabular method which we investigated in detail for complete closure
of a downstream valve can also be used for other types of valve move-
ment. In all cases care must be taken to use the correct sign for the
first F-waves, as follows:
(a) Opening of downstream valve: negative F-wave; ~vis positive.
(b) Closure of upstream valve: negative F-wave; ~vis negative.
(c) Opening of upstream valve: positive F-wave; ~v is positive.
Partial closure and opening can be treated in the same way.
4f1-
Noz.!.I (""' I I I I
I
I
I I I
: 4 :/I
I ~I I I
/I
1/1 I
I '\.
I
I
N I
I I ),----1
I
I I I '\. I I 1/ I
I
I
I I I I I I
I I I
I I I I I
I I I I I
I I I
I I I I I I
I I I
I I
No2~
FIG. 5.4
REFERENCES
1. AMERICAN WATER WORKS AssociATION: 'Waterhammer and surge,' J. Am.
Wat. Wks. Ass., (1961), 53, p. 1037.
2. GmsoN, N. R.: 'Pressures in penstocks caused by the gradual closing of
turbine gates.' Trans. Am. Soc. civ. Engrs, (1920), 83, pp. 707-775.
6 Graphical Analysis
6.1 Pressure-Velocity diagrams at fixed points

When pressure head is plotted against velocity for the passage of a water-
hammer wave past a fixed point in a pipe, a straight line is always obtained.
The slope of the line is ± A. since by Joukowski's law (equation 3.8)
h = -A.~v.
If we first deal with a simple pipeline in which water flowing from a
reservoir is suddenly stopped by the closure of a downstream valve, we
H (c) H
Closure
i
h
I
I
I
I
1!:--------JL..._
I
I
lv v
0 vo vo
ll;:::::====-~X
Negat1ve wave
X --..
Posit1ve wave
rr
(b) H (d)
11
H
Opening
Jl
I h
tafi 1l__ __
I I
I
1
Ho I
I
I
:H0 -h I v
v, 0 v,
FIG. 6.1
50
Graphical Analysis 51
know that the passage of the F-wave causes the pressure to increase from
Ho to (Ho +h). At the same time the velocity drops from v0 to 0. These
changes can be shown on an H-v diagram, as Fig. 6.1 (a). Similarly the
opening of a downstream valve gives rise to H-v changes as shown in
Fig. 6.1 (b). The H-v diagrams for closure and opening of an upstream
valve, as in a pumping main, are shown in Figs. 6.1 (c) and 6.1 (d).
In Chapter 4 we considered the effect of some end-conditions upon
incident waves. These effects may be shown on H-v diagrams, as shown
in Figs. 6.2 (a), (b) and (c). When there is a change of diameter it is
better to use H-Q diagrams, which are described in Section 6.2.
Cond1tion Posit1ve 1nc1dent wave
H
(a) Reflection
at a reservoir
H O,H0 +2h
(b) Reflection
at a dead end
D.v, H0 +h
H
( c ) Effect of
Reflected ---++--"-
restrict ion
Incident ----T-..A'"F
v,H
Transmitted Fi
/v,H-kL vn
_/
FIG. 6.2
We can now draw a complete H-v diagram for a uniform frictionless

pipe having two end-conditions. Fig. 6.3 (a) shows the diagram for a
pipeline fed by a reservoir and with a downstream valve which shuts
suddenly. Fig. 6.3 (b) is for a pumping main feeding a reservoir when the
pump suddenly stops in such a way that it becomes a 'dead end' and
water cannot pass through it in either direction. For a single wave, which
is produced by sudden opening or closure, the diagrams apply to any
point in the pipe, but the time at which the changes of pressure occur
varies according to the position of the fixed point. The diagrams may be
(a) H
fl X
b)
- va,Ha vo,Ho
F-wave
X
FIG. 6.3
rr
considered as showing the variations of H and v as seen by a stationary
observer.
6.2 Pressure-discharge diagrams at fixed points

When there is a sudden change of velocity under steady flow condition, as
at a change of diameter or a junction, H-Q diagrams are more useful
than H-v diagrams. They are drawn as shown in Fig. 6.4.
For a junction, Fig. 6.4 (c), the head Ho + ht is obtained by ensuring
that continuity is satisfied. That is, aa1 = aa3 + aa2.
Enlargement
(a)
Contraction
(b)
FIG. 6.4
Example 6.1
Water flows at a rate of 0·5 mSfs along a pipe (1) which has a cross-sectional
area of 0· 30 m 2 to a junction, from which it passes along two branches (2 and 3) of
cross-sectional area 0·10 m 2 and 0·05 m 2 • 0·3 mSfs goes along pipe 2 and 0·2 mSfs
along pipe 3. a is 1300, 1325 and 1250 mfs in pipes 1, 2 and 3 respectively, and
the head is 30 m at the junction.
If the flow is suddenly increased in pipe 1 to 0·6 mSfs find the head increase at
the junction.
1300
tan A-~-
1 - gA1 - 9·81 x 0·30 -442 sIm
SOLUTION 2
-
a2 1325
tan A2 = - = = 1351 s/m 2
gA2 9·81 X 0·10
as 1250
tan As = - = = 2549 s/m 2
gAs 9·81 X 0·05
Q1 = 0·5 ms/s; Q2 = 0·3 ms/s; Qs = 0·2 ms/s
LlQ = 0·6- 0·5 = 0·1 ms/s h =tan A1LlQ = 44·2 m
:. Ho +h= 30 + 44·2 = 74·2 m
Draw b1d and lines b2c2, baca, dc1'. By trial find the line aaaa2a1 so that
aa1 = aaa + aa2. Hence aaa = 0·223, aa2 = 0·344, aa1 = 0·567 m3 /s, and the
increase of pressure at the junction ilH = 58·9 m.
Other 'end-conditions' can be shown on an H-Q diagram. Charac-

teristic curves for a pump show its performance at various speeds, as Fig.
6.5 (a). The curve of the form shown in Fig. 6.5 (b) applies to an orifice
or other fixed throttle, e.g. a partly-open valve.
(a) Characterrsf1c (b) (c)
curves for
centnfugal pump
H
Q Q
FIG. 6.5
The friction loss in a pipe may be considered as being concentrated at

a single point (e.g. at the downstream end of the pipe) or at several points,
and is of the form shown in Fig. 6.5 (c).
6.3 Schnyder-Bergeron diagrams

The transient conditions for the most simple cases, such as a uniform
pipeline in which the change of flow is sudden, may be represented by the
H-v or H-Q diagrams for fixed points which we have noted in Sections
6.1 and 6.2. When the system is complicated this method becomes
excessively tedious. Each point in the system must be considered separately
and at any point the head may be due to a number of direct and reflected
waves. Often the end-conditions vary with time, as when a valve closes
or opens gradually or when a pump changes speed.
The difficulty is overcome by the method developed independently by
Schnyder1 •2 and Bergeron3,4 in the 1930's. This makes use of 'travelling
observers'. An 'observer' moving upstream along a pipe at the speed of
waterhammer waves will not be aware of other upstream-moving waves,
since all waves travel at the same speed, a; no wave can overtake or be
overtaken by another. The observer will of course observe waves moving
downstream; the pressure-head will change as he reaches them, and there
is a corresponding change in the velocity.
Schnyder-Bergeron diagrams, or surge diagrams show both observations
made at fixed points and observations made from waves by 'travelling
observers'. The axes of the diagrams are H and Q. Lines sloping at
± afg or ± A represent the passage of waves, as in Section 6.2. The
fixed point observations indicate the direction of the waves which pass
the point-waves moving in the positive v-direction have a positive slope
and those moving in the negative v-direction have a negative slope.
On the other hand an observer moving in the positive v-direction only
observes waves moving in the negative v-direction which have a negative
slope. Similarly observations made from a wave moving in the negative
v-direction are shown on the surge diagram by points on a line with positive
slope.
The notation used for the points on a surge diagram gives the position
in the pipeline to which the conditions apply and the time of the observa-
tion. The time is measured from the time of the start of the change from
initial steady conditions. A convenient time interval, T, is chosen, which
may be multiples of ft such as ft, lft or ift, or which may be a given
number of seconds. Thus point A2 shows the head and flow at point A
in the pipeline at time ft after the start of transient conditions if the time
interval T is tft.
6.4 Surge diagram for uniform pipe with hypothetical end conditions
Fig. 6.6 (a) shows a uniform pipe in which flow is initially steady at a rate
of Qo from A to B, the head being constant at Ho. At A and B are two
'devices' whose characteristics are known. These characteristics are H-Q
relationships, which can be shown by curves "PA and "PB· A 'device' might
be a valve, a pump, a turbine or a reservoir.
B starts to operate s x lft seconds after A starts, s being less than one.
Since the first change from the initial steady conditions occurs when A
starts, we call this timet = 0. We can show the initial conditions by point
P (co-ordinates Qo, Ho) on the surge diagram, Fig. 6.6 (b). P indicates the
head and discharge at A up to the time 0, where the device there starts to
operate, and is therefore called 'A0 '. If we take lft as the time interval T,
3
(a}
Oo
A 8
(b)
H
'fes
I
(c)
H
A o
(d) H I liz+<
I I
81 I
FIG. 6.6
the pressure wave from A, caused by the first operation of the device, will
not reach B until time 1 (i.e. t = T) and the device at B will have started to
operate at time s (i.e. t = s x T), so P also shows conditions at B up
to time s, and is therefore designed B •. The characteristics of the A
device at time 0, and the B device at time s (the initial characteristics) pass
through P.
An observer leaving A at time 0 travels downstream and meets negative
waves generated from B after time s, which are therefore of negative
slope -afgA =tan- A, shown by line PS on Fig. 6.6 (b). He reaches
B at time 1, i.e. when (t = T), when the characteristic of the device at B
has changed to 'l.fJ Bb and the conditions he finds at B are shown by point B1.
This observer now returns from B to A moving in the negative v-direction
and therefore only noticing positive waves shown by a surge line of
positive slope+ A. He reaches A at time 2, i.e. when (t = 2T), to find the
characteristics of the device there as 'l.fJ A2· The H-Q relationship on his
arrival at A are given by point Az, the intersection of the positive-slope
surge line and the characteristic 'l.fJ A2·
Moving up and down the pipe the observer then reaches B and A in
turn to find conditions as shown by points Ba, A4, B5, A6. . . . Each
'round trip' takes 2T.
In a similar way points on Fig. 6.6 (c) show the observations made by
a traveller leaving B at time s. At first he meets the positive wave from
A and reaches A to find conditions given by point A1+s on line PR at its
intersection with the characteristic '/.fJA(Hs)· Subsequent observations by
this traveller at the ends of the pipe are given by points Bz+s, Aa+s. B4+s,
A5+s ....
The observations of the two travellers (who left A at time 0 and B at
time s respectively) can be drawn on the same diagram, Fig. 6.6 (d).
You can see from this example that the points we have obtained repre-
sent conditions at intervals of time !11 apart. The nature of the change
of the characteristics of the devices at A and B does not effect the diagram.
Thus the characteristics may change continuously, or by a series of steps.
All we are concerned with is the characteristics at the times when the waves
were reflected. Thus point A 6 depends only on the initial conditions and
the characteristics represented by '/.fJB1, '/.fJA2, '/.fJB3, '/.fJA4, '/.fJB5 and '/.j)A6· All
these characteristics determine point A6, but characteristics at any other
time have no effect.
Also notice that each time a travelling observer comes to an end of the
pipe (at A or B) the surge line is reflected in a direction which is symmetrical
to the vertical, Q = constant. However, if the end condition is a 'dead-
end', as in Fig. 6.2 (b), the reflection is symmetrical about a horizontal
line.
6.5 Comparison of surge diagram with analytical method

Allievi's equations (4.4) can be used to obtain the relative waterhammer
head-relative to the steady head Ho-at times !p, apart. These are in
effect the same time intervals as those obtained by the surge diagrams.
Although Allievi adapted his method for a number of end conditions,
these conditions are severely limited. A 'device' such as that at B in Fig.
6.6 must have a characteristic which is a parabola with its vertex at the
origin. The characteristic of a pump or turbine must be approximated
to a parabola.
On the other hand, with the graphical method the true characteristic
of the device may be shown correctly.
A second advantage of using surge diagrams rather than Allievi's
equations is that the flow as well as the head is obtained for each position
in the pipe and at each time. Equations (4.4) yield only the head.
A third advantage is in speed of operation. A second order (quadratic)
equation must be solved to obtain each value of z from equations (4.4).
This obviously takes longer than drawing a straight line and a portion of
the characteristic curve on a surge diagram.
6.6 Intermediate times and positions

In Section 6.4 we obtained points on the surge diagram to represent
conditions at the ends of a uniform pipe at intervals of time equal to !p,
after the devices started to operate. The same method may be applied to
obtain points representing other times and for points along the pipe. To
illustrate the procedure we will take the same pipe as in Fig. 6.6 with the
same end-characteristics and assume that e = ! (i.e. the device B starts
when t = !p,).
Intermediate times. The observations of a traveller leaving A at some

time between t = -tp, and t = 0 is represented as before by point P
and line PS. The traveller arrives at B at time tp, after leaving A, and this
is between t = 0 and t = tp,. If he starts from A before t = tp,( -1 + e)
he arrives at B before the device there starts to operate and the condition
is shown by point P. If, however, he leaves at time -!(t = -lp,) he
arrives at B at time ! (i.e. t = ip,) when the characteristic of B is given by
"PBt, and point B1 shows the conditions at B when he arrives there, Fig.
6.7 (a). As the traveller reverses direction at each end of the pipe in turn
the conditions he finds are shown by points A11, B21, Aa1, B41..••
Similarly, for a traveller leaving B before time e, at some time between
-1 and e, the initial conditions are given by point P and the first journey
by line PR. For example, if he leaves B at time 0 he reaches A at time 1,
finding there the characteristic "P AI· The series of surge lines A 1, B2, A 3 ,
B4, As . . . may be drawn to show the further observations of the traveller
as he continues his forward and backward journey, Fig. 6.7 (b).
(a)
(b) 1j;B2
H I
/ A! -.\fAI
-----t:', A3
R ''
FIG. 6.7
A maze of lines can be drawn as close together as required, and the

variation of head and discharge at A and B at small intervals of time can
be found. However, these additional surge lines add nothing to the
accuracy of the original lines.
Intermediate position. If C is a point between A and B in the uniform

pipeline AB, Fig. 6.8 (a), a surge diagram can yield values of Hand Q
at C.
The waterhammer waves generated by any disturbance at A arrive at
C after a time /fa. Those generated at B arrive at C after a time (L - /)fa.
Therefore to find the conditions at C at any time we need to arrive with
a traveller who left A just time /fa before, and a traveller who left B just
time (L - /)fa before. The observations of these travellers may be
represented by surge lines, as Fig. 6.8 (b).
Turning again to the surge diagram for the uniform pipe with hypo-
thetical end-conditions, we can obtain points for C, where for example
I = iL. To find point C2 we draw a positive surge line from B11 and a
8 1 L-1
I--I L -I
--
0
A +C +D +E 8
L c,
-A
A~j,
(a) (b)
(c)
FIG. 6.8
negative line from A1!· C2t is at the intersection of a positive line from
B1i and a negative line from A2, and so on, as in Fig. 6.8 (c).
The maze of lines in Figs. 6.7 (b) and 6.8 (c) are drawn at time intervals
of i,u, and their intersections give points for the three-quarter positions
C, D and E without further construction.
6.7 Series pipes5,6

Consider a pipeline AB consisting of two pipes AC and CB which have
different diameters and wave velocities, as shown in Fig. 6.9 (a). At any
instant of time t a positive wave reaches C from A having started from A
(a)
al AI a2 A2
L, - I ..0.__ I
~-2P-i 02- 2 fl-2
ton 1 _E_Ar
g,
=AI ton- 1 i}. =Az
(b)
(c)
I
FIG. 6.9
at time (t - !,u1), when the characteristics of the device at A was tpA(t-!p,J·

At the same time t a negative wave arrives at C from B having started
from Bat time (t - !,u2), when the characteristic of the device at B was
"PB<t _ 1"'2). Observers travelling with the wave provide the surge lines
At-tp,-S and BH1, 2-R at slopes -A1 and +A2 respectively, as shown
in Fig. 6.9 (b).
At C, as at any position in a continuous fluid, the conditions of con-
tinuity must be satisfied and the pressure is the same in all directions.
This is only satisfied at the intersection of the line At-t" - S and
B1_ 1" 2 - R, i.e. at point Ct. Point Ct represents the conditions of the
pipes AC and CB at timet.
After time t one wave-born observer moves along CA at speed a1, and
another along CB at speed a2. The waves may either be considered as
(a)
A ==::::::;===~c==~2=~B
(b)
FIG. 6.10
'transmitted' through the junction with refraction of the slope of the

surge lines, or as 'reflected' at the junction with a surge line of the same
slope but opposite sign. Whichever way we look at it, one traveller
arrives at A, observing the conditions shown by a surge line of slope A 1
on the way. He reaches A at time (t + !.u1), so that the characteristic of
the device at A must then be 'II' A(t+Ml• and the conditions are given by
point AtHpl• Fig. 6.9 (c).
By working in this way we can obtain a complete surge diagram for
a pipeline consisting of two or more separate pipes of different diameter.
The most simple case is where the periods of two pipes are equal, i.e.
Lja1 = L2/a2. The lengths are not the same if the velocity of water-
hammer waves are different. The surge diagram is shown in Fig. 6.10,
where the devices at A and B both start to operate at time t = 0. The unit
of time is !.u, where .u is the period of one of the pipes (not the period of
the complete pipeline). A wave and its travelling observer therefore
takes time 2 x i!t = .u to travel from A to B or from B to A.
If the two sections have periods /tl and fl-2 which are not equal, but have
a common factor, then half the common factor is a suitable unit of time
for the surge diagram. For example, if ftl = 1!fJ-2, the time interval may
be taken as !(l/3ftl) = HifJ-2) = T. Then, referring to Fig. 6.11 (a), a
wave leaving A takes time 3T to reach C and a wave leaving B takes time
2T to reach C. Then, working as before, point C4 is obtained by drawing
(a)
A========~========~c~==~2==~B
2
...s.
..L fLt : a
1
(b)
FIG. 6.11
a surge line of slope - A1 from A1 and a line of slope + A2 from point

B2. The diagram is continued as Fig. 6.11 {b).
When the periods of pipes in series have no common factor it is neces-
sary to draw a series of surge lines back to the origin for every point on
the surge diagram. For example with the pipeline shown in Fig. 6.12 (a),
!,u1 may be taken as unit timeT. The devices at A and B begin to operate
simultaneously at t = 0.
The conditions at C at time t = l{tftl) are given by point P since no
wave reaches C until after time 1(t = iftl). We get the conditions at
time 3 by drawing a line of negative slope - A1 from point A2 and a line
of positive slope + A2 from Bl-43. Time 1·43 is used because a wave
leaves B then in order to reach C at time 3, taking time 1·57-r to travel
(a)
A====~====~c~~+~~2======~B
j_ fL = .!:-L j_ fL =..!:L
2 I a1 2 2 a2
FIG. 6.12
from B to C. To reach B at time 1·43 an observer would have to leave C
at time -0·14, when conditions would be the initial conditions shown by
point P.
The conditions at Cat time 5(t = 5 X !,u1) are obtained from a line of
negative slope - A1 leaving point A at time t = 4 and a line of positive
slope + A2 leaving B at time t = (5 - 1·57}r, i.e. from point Ba·43· Point
Ba·43 lies on the characteristic of device B at time 3·43 (i.e. on line 'lfB·a4a)
and on the negative line originating at Cat time !,u2 earlier, i.e. at point
C1·sa. CHa in turn is found from a positive line of slope +A2 from
point Bo·29 and a negative line of slope -A1 from point Ao.S6· Fig.
6.12 (b) shows the surge lines for obtaining points Ca, Cs and C7: the
lines for obtaining c7 are shown broken.
You can see that if we want to find point C21, for example, we have
to draw a large number of surge lines. The diagram may become too
confused to follow easily and it may be necessary to trace the continuation
onto a second sheet of paper to follow the lines.
The construction is easier to follow if the steps are set out in the tabular
form suggested by Bergeron, as shown in Fig. 6.12 (c). ·
Three or more pipes in series can be treated in the same way as two
pipes, although the surge diagram becomes more complex.
6.8 Branched pipes5,6

In order to determine the head and flow at a junction we must draw
a surge diagram for each pipe, and satisfy the two basic requirements of
common pressure at the junction and continuity of flow at the junction.
In Fig. 6.13 (b) we see the first stages of the surge diagrams for the
three pipes joining at junction C, Fig. 6.13 (a). In this case all three pipes
have the same period, but pipes having different periods can be examined
in the same way if the unit time T is a simple fraction of the period of each
pipe. The devices at A, B and C have been shown in general form-the
characteristics of all vary with time.
The initial conditions obtain at D until time 1 when the first travelling
observer from A brings a message that the device at A has started to
operate. These initial conditions at D are shown for the three pipes by
points PAD, PBn and Pen. The pressure of all is Ho and continuity is
satisfied because (QA n)o + (QBn)o = (Qcn)o where (QAn)o is the flow
to D from A at time 0, (QBn)o is the flow to D from B at time 0 and
(Qcn)o is the flow from D to Cat time 0.
An observer leaving D at time 0 along pipe AD arrives at A to find
conditions A1. Point A1 is at an intersection of a line of slope - A1
from DAo (i.e. PAD) representing the observation of a traveller moving
upstream and the characteristic of device A at time 1 (i.e. '!/'AI). If the
traveller now returns to D his observations are given by line A1-R at
slope A1. Point DA2 which gives the conditions at DA at time 2 is not
known yet; line A1-DA2 is 'undefined'.
Travellers reaching D along pipes BD and CD at time 2 make observa-
tions recorded by lines Dn2 from B1 (B1 being point Pnn) and Dc2 from
(a)
8
Pipe AD Q Pipe BD Q Pipe CD Q
FIG. 6.13
C1 (C1 being point Pen). Both lines are 'undefined' and points Dn2 and
Dc2 are at this stage unknown.
However, at time 2 (as at any other time) the basic requirements of
common pressure and continuity at D must be satisfied, and these require-
ments may be shown graphically on the surge diagram for any one of the
three pipes. Pipe AD has been chosen here but it might equally be either
of the other pipes. Now by continuity QAn = Qnc- Qnn. At time 2
the abscissa of point Dc2 (at present undefined) is (QAc)2 and the abscissa
of DB2 is (QAn)2. The abscissa of DA2, which is (QAn)2, is equal to the
difference between (Qnch and (Qnn)2, i.e. it is the abscissa of Dc2 less the
abscissa of D B2·
Line DA2 is drawn from PAn with a slope equal to (slope of line Dc2)-
(slope of line Dn2); for any common head the abscissa of any point on
line D A2 is equal to the abscissa of the corresponding point on line Dc2
less the abscissa of the corresponding point on line D B2· Continuity is
satisfied.
Line DA2, together with line At-R previously drawn, enables us to
define point DA2 which lies at their intersection. Points DB2 and Dc2
can now be defined on lines DB2 and Dc2 since they have the same ordinate
Hn2 as point DA2; line Hn2 is projected across the diagrams.
To find the conditions at D at time 3 the undefined lines A2S, DB3 and
Dc3 are drawn from fixed points. Since points B2 and C2 represent the
8 c
£
FIG. 6.14
same conditions as Bt and C1 (the devices at B and C do not start to

operate until time 2), lines DB3 and Dc3 are coincident with lines DB2 and
Dc2 and line DA3 is coincident with line DA2· Point DA3 is found at the
intersection of line A2S and line DA3 and points DB3 and Dc3 are on the
common ordinate HD3.
After time 3 the conditions at DB and De are affected by the change
from initial conditions at B and C after time 2. At time 4 travellers arriving
at D along pipes BD and CD have made observations corresponding to
lines DB 4 and Dc4. Line DA4 is therefore displaced horizontally from
line DA 3 by the displacement ofline DB4 from line DB3less the displace-
ment of line Dc4 from line Dc3. The procedure for finding, in turn,
points DA4, DB4 and Dc4 is the same as for similar points at time 3.
Junction of more than three pipes. The method we have been considering
may be applied to more than three pipes joining at a common point.
Surge diagrams are constructed for each pipe and lines similar to lines
DA2, DA3 and DA4 in Fig. 6.13 (b) are drawn on the diagram applying
to one pipe, the slope satisfying the conditions of continuity. For example,
for the pipes meeting at Fin Fig. 6.14 the slope of any line FA is equal to
(slope of line Fn +slope ofline FE) less (slope of line FB +
slope of line F c)
6.9 Effect of hydraulic losses

Velocity head, friction head and secondary losses (due to bends, etc.)
are often so small that they may be neglected. 7 However, in long pipelines
or in systems operating under a low head they may amount to a consider-
able proportion of the total head, and their effect should be included in
surge diagrams.
In most cases a reasonable approximation is obtained if we consider
the losses as being concentrated at a single throttle, which is often taken
as being close to a reservoir. For greater accuracy or when the pipeline
is very long the losses may be distributed between two or more throttles
along the pipeline.s,9
Fig. 6.15 shows a surge diagram for a pumping main, where the losses
are represented by a throttle near the reservoir shown by line L-L. Line
Pump hJ'
@r::'========::::;'L
L
FIG. 6.15
L-L gives the 'characteristic' at the end of the pipeline just upstream of
the throttle.
Obviously a major throttle, such as an orifice plate or a partly closed
valve, may be treated in the same way as the hypothetical throttle which
we have used to simulate minor hydraulic losses.
6.10 Other conditions

Surge diagrams can be used to determine the Q-H relationship under
other conditions. In the cases we have considered in this chapter the
characteristics of the 'devices' (i.e. the Q-H relationships at the devices)
have been known at intervals of time and these relationships have been
represented by 'If-curves.
With some other devices no such simple relationship exists. Q may
depend upon the rate of change of H with respect to time, or H may depend
on the rate of change of Q with respect to time. Devices which act in this
way include accumulators and air vessels, with which we will deal in
Chapter 7, and surge tanks which will be examined in Chapter 8.
REFERENCES
1. ScHNYDER, 0.: 'Druckstosse in Pumpensteigletrungen,' Schwiez. Bauztg.,
(1929), 94, pp. 271-273, 283-286.
2. ScHNYDER, 0.: 'DruckstOsse in Rohrleitungen,' WassKraft WassWeit,
(1932), 27, pp. 49-54, 64--70.
3. BERGERON, L.: 'Etude des coups de belier dans les conduits: nouvel expose
de la methode graphique,' Techq. mod., (1936), 28, pp. 33-38.
4. BERGERON, L.: Du coup de belier en hydraulique au coup de foundre en
electricite (Dunod, 1950): ASME translation: Waterhammer in hydraulics
and wave surges in electricity (Wiley, 1961).
5. ANGUS, R. W.: 'Waterhammer pressures in compound and branched pipes,'
Proc. Am. Soc. civ. Engrs, (1937), 59, pp. 661-668.
6. PEARSALL, I. S.: 'Waterhammer effects due to branched and stepped pipes,'
Symp. Surges Pipelines, Proc. Instn mech. Engrs, (1966), 180, 3E, pp. 28-40.
7. AMERICAN WATER WoRKS AssociATION: 'Waterhammer and surge,' J. Am.
Wat.Wks. Ass., (1961), 53, pp. 1034--1044.
8. ANGUS, R. W.: 'Waterhammer in pipes, including those supplied by centri-
fugal pumps: graphical treatment,' Proc. Instn mech. Engrs, (1937), 136,
pp. 245-291.
9. HAYASHI, T., and RANSFORD, G.: 'Sudden opening or closing of an outlet
valve on a pipeline,' Houille Blanche, (1960), 15, pp. 657-673.
FURTHER READING
BERGERON, L.: 'Method graphique generale de ca!cul des propagations d'ondes
planes,' Mem. C. r. Trav. Soc. Ing. civ. Fr., (1937), 90, pp. 407-497.
CoMBES, G., and ZAOUI, J.: 'Analyse des erreurs introduites par !'utilisation
pratique de Ia methode des characteristiques dans le calcul des coups de
belier,' Houille Blanche, (1967), pp. 195-200.
GRAY, C. A. M.: 'Analysis of water hammer by characteristics,' Proc. Am.
Soc. civ. Engrs, (1954), 119, pp. 1176-1189.
HARDING, D. A.: 'The simplification of graphical surge analysis by the use of
surge lines with a reduced slope,' Br. Hydromech. Res. Ass., (1962), SP671.
LUPTON, H. R.: 'Graphical analysis of pressure-surge in pumping systems,'
MARPLES, E. I. B.: 'The significance of surge diagrams,' Symp. Surges Pipelines,
Proc. Instn mech. Engrs, (1966), 180, 3E, pp. 3-11.
PARMAKIAN, J.: 'Water-hammer design criteria,' J. Pwr Div., Proc. Am. Soc.
civ. Engrs, (1957), 83, pp. 1216-1 to 1216--8.
STREETER, V. L.: 'Waterhammer analysis of pipelines,' J. Hydraul Div., Proc.
Am. Soc. civ. Engrs, (1964), 90, pp. 151-172.
7 Waterhammer in Pumping Mains
In Chapter 6 we dealt with pipes having 'devices' at the ends. In this

chapter we examine the practical case of a pump feeding a reservoir at the
opposite end of a pipeline.
7.1 Sudden stopping of pumping

The most simple case to deal with is when flow from a pump suddenly
stops in such a way that there can be no reverse flow. This could be caused
by failure of power to an inertia-less pump, or by the instantaneous closure
of a delivery valve. It is an unlikely occurrence, but illustrates the pro-
cedure which is followed in more complex situations. It is assumed that
separation does not occur at any point in the pipeline. The effect of
separation is dealt with in Section 7.2.
On the surge diagram Fig. 7.1 (b) the initial steady-flow values Qo and Ho
are found from the intersection at P of the two end-conditions-the pump
characteristic '!fJP and the pipe-line characteristic '!fJL given by adding
pipeline losses to the static lift. The discharge from A is Q 0 against
head Ho.
Immediately after the flow from A stops at time t = 0 the conditions
at A must be given by a point on the ordinate Q = 0, and by Joukowski's
Law the line PR is at a slope of afgA = tan A. At the reservoir inlet the
head and flow do not change until the negative wave from A teaches B
after time T = it-t· Point P therefore represents conditions at A before
closure and at B until time 1, which is equal toT = it-t·
At time 1 the negative wave from A (represented by line RS at slope
-A showing the observations of a stationary viewer) reaches Band point
S gives the conditions at B from time 1 onwards.
The wave is reflected from B as a positive wave and arrives at A at
time 2 to change the pressure at A to that given by point T. R shows
70
Waterhammer in Pumping Mains 71
conditions at A from time 0 to time 2, and T shows conditions at A from
time 2.
Therefore there is 'no-change' at A and B during the time fl during
which the wave does a round trip from one end to the other end of the pipe
and back again, and point S gives conditions at the reservoir B from time 1
(a)
(b)
I
I
I
I
I
I
I
I
I
I
I
I
-Q : -Q
FIG. 7.1
to time 3, point T gives the conditions at the pump A from time 2 to time 4,
and so on.
In Fig. 7.1 (b) the point R has a positive value of H. If Hwere negative
there is a danger of separation. A simple guide as to whether the pressure
will be positive or negative is the ratio tan A x (Qo/Ho). This is twice
the pipeline constant s, which we used in connection with Allievi's
equation in paragraph 4.5. In Fig. 7.1,
s
2 = tan A Qo = Ho - OR x Qo = Ho - OR
Ho Qo Ho Ho
~is here less than 0·5 and the maximum head at the. pump is positive.
If~ is greater than 0·5, the maximum head at the pump is negative, and
further checks must be carried out to ensure that there is no separation
of the water column.
We also notice from this example that when no separation occurs the
maximum pressure head does not exceed twice the normal head Ho.
7.2 Separation of water columnsl,2,3

If the absolute pressure in a pipe drops below hvp equal to the vapour
pressure of the water plus the partial pressure of released gases, a 'vacuous
space' is formed. This can occur at the pump or valve if~ > 0·5, or at a
high point in a rising main; we will
H consider these two cases in turn.
(a) Separation at pump or valve.
Fig. 7.2 is similar to Fig. 7.1 (b) except
that ~ is now greater than 0·5, so that
the surge line PR intersects the
abscissa corresponding to the absolute
pressure head hvp at a positive value
of Q.
What happens, of course, is that as
the pressure drops from Ho to hvp
the water near the pump or valve
'stretches' to allow the continued dis-
Q charge Qo downstream. When the
R head becomes hvp the water can
FIG. 7.2 stretch no more; at the reservoir end
of the pipeline the flow is still Qo
and a gap or 'vacuous space' is formed near the pump or valve. When the
gap closes a high pressure results. Waterhammer and separation can also
be caused by cavitation in a high-speed pump. 4
Example 7.1
Draw a surge diagram to obtain the maximum pressure head at the pump
described below when the flow is suddenly stopped.
The pump is placed in a dry well adjacent to the suction well and 4ft below the
water level in the suction well, which is considered as datum. Separation occurs
at a negative head of 30ft. The pipe normally discharges 8 cusecs along a 21 in.
diameter main which rises uniformly to a reservoir in which the water level is 60ft
above datum. The length of the main is 16 600ft and the friction head (assumed
proportional to velocity squared) is 19ft at normal flow. a = 3320 ft/sec.
2L
SOLUTION p, = -a = 10 seconds
A _ _!!..__ 3320 _ .
tan - gA - 32 .2 x 2_4053 - 42 87 ft/cusec
Let
The axes of the H- Q diagram are drawn as in Fig. 7.3. A horizontal line
at a head of 60ft represents the reservoir level and the friction curve 'PL is
1----------------------
IT
Max. head 267ft
I
I ft
I
120
I
100
p
80
Q
8 cusec
I
I
FIG. 7.3
drawn above the reservoir level for positive flows and below for negative flows.
Horizontal lines representing the suction level (H = 0) and the head at which
separation occurs (H = -30 ft) are also drawn.
The initial steady flow conditions are represented by point P. Immediately
the flow stops, at t = 0, a negative wave (represented by a surge line of slope A)
moves upstream. At point Ao-1 the head equals -30ft, and a 'gap' is formed.
At time t = f..! = ,. the wave returns to A, having been reflected from B at time
( = tT.
During the interval t = 2,. to t = 3,. the gap begins to reduce because the
flow is negative and it is finally closed at time 3·5,. found as follows-
Time interval
Q=
0 -h 5·45 cusec; volume of gap= 5·45T
1T-2T 1·10 cusec; volume of gap = 6·55T
2T-3T -3·08 cusec; volume of gap = 3·47T
3T-4T -6·93 cusec;
3·47T .
gap closes at 6 _93 after time 3T
i.e. at t = 3·50T,
When the gap closes the pressure rises as shown by line ST. This reaches the
axis Q = 0 at a head above the separation level equal to Q tan A = 6·9331 x
42·87 = 297·2 ft.
Therefore the maximum pressure at the pump = 297·2 - 30 + 4 = 271·2 ft
since the pump is 4 ft below the suction well level.
(b) Separation at high point. Even when the pressure near the pump or
valve remains above hvp there may still be separation in the pipeline if
there is an uphill slope to the pipe, as in Fig. 7.4. The 'solid' column at
1---
1 ---:----
--- ---f----
- - - - h.q.•
I ----- I
1
1
I
I
I
,.--- .
I
FIG. 7.4
the downstream end of the pipe must be treated independently, its end-
conditions being the vacuous space and the reservoir.
Whether separation occurs at pump or valve or at a high point in the
pipeline, a negative wave is reflected by the reservoir as a positive wave5
and the separated columns collide, the impact causing a large rise of
pressure. When there are a number of high points in a pipeline the
pressure resulting from there-coalescing portions at one gap is transmitted
to the next gap, increasing the closing velocity. The high surge pressure
is therefore increased and the effect may be cumulative.
7.3 Pump with inertia
In Section 7.1 we examined a pump without any inertia, which is impossible.
The pump itself, the driving motor, the couplings and the flywheel (if
fitted) all have inertia which increase the time during which the flow is
reduced to zero.
The equation relating the power available for a pumping set to the
speed change is
-1·62 X 106 P
tl.N = - tl.t (7.1a)
I N
where tl.N is the drop of rev/min in time tl.t seconds, I is the moment of
inertia in lb ft 2, and P and N are the mean horsepower and mean speed
during the time tl.t.
Expressed in SI units, equation (7.1) is
tl.N = -9000 !_ t1t (7.1b)
I N
where tl.N and N are measured in rev/min, I inN m2 , Pin watts and tl.t
in seconds.
--- ---
~====:::::
rotfLot2;
'tAO
lReservoir level
I
....... AI 1
'%1
I
I
-Q Q
0
FIG. 7.5
Failure of the power supply to a pump with inertia is the most common
cause of water hammer pressure in pump delivery mains.
Fig. 7.5 shows a surge diagram for a centrifugal pump feeding a reser-
voir through a long pipeline in which the losses are assumed concentrated
at a throttle at the reservoir entrance and in which there is no separation.
The construction is similar to that of Fig. 6.15 until the surge lines
reach Qo; if a reflux valve is fitted so that there is a 'dead-end' the maxi-
mum head (at point U) is about the same pressure above the reservoir
static head as point S is below. Obviously points R and S are important
when we are investigating the possibility of separation. Depending upon
the form of the pump characteristics (lines '1j)A4 • •• 'lj)Au) R may be at a
higher or lower pressure than S. However, S is found more easily-
approximate methods may be used-and gives a good indication of the
likelihood of separation.
The lines '1fJA are found from equation (7.1).
An approximate solution may be obtained by assuming that the values
of Nand P during the interval 11t are the same as those at the start of the
interval. The error so produced is small if 11t is small.
There may be danger of separation if the pipeline is long or the working
head is small. In this case the surge line through the initial point P may
reach zero head (or hvp) before the pressure is modified by the returning
wave at time p. after the power to the pump has been cut off.
This danger may be reduced by attaching a flywheel to the pump. This
increases the value of I and therefore the slowing-down of the pump takes
longer.
Flywheels have the disadvantage that more power is required to start
the pump, and this method of controlling waterhammer is therefore not
suitable for motors connected directly to the electricity supply. There are
other restrictions on the use of flywheels. The actual size of the flywheel
itself is a disadvantage, and the additional bearings which are necessary
increase the length of the set. Windage losses in the rotating flywheel
and additional losses in the extra bearing decrease the overall efficiency. 6
7.4 Reverse flow

In Section 7.1 we examined the waterhammer in a pipeline where no
reverse flow occurs. Under these conditions the maximum pressure is not
more than twice the normal operating pressure and pipe walls are usually
able to withstand stresses set up. In a system where the inertia of the set
is considerable and no cavitation occurs there is a similar limitation if
a non-return valve is installed, as shown in Fig. 7.5.
When a pump with inertia has no reflux valve and the power is cut off,
the water in the pipeline slows down and then flows back towards the
pump which is still rotating in the normal forward direction. The reverse
flow at first acts as a water brake, bringing the pump to rest, and then
the rotation of the pump is reversed. The pump now behaves as a
badly-designed turbine.
To study the H-Q relationship under these conditions we must know
the full pump characteristics. Fig. 7.6 (a) shows the H-Q characteristics
at constant speed,7 and Fig. 7.6 (b) the N-Q characteristics at constant
head, both diagrams covering negative as well as positive discharge. The
laws of similarity may be applied to obtain the characteristics at the pre-
vailing transient conditions. Alternatively a whole series of tests at
(a)
(b]
FIG. 7.6
R
I
I
I H
I
I
I I
I I
I I
I I
1 I
I I
I I
I I
I I
I /
I
FIG. 7,7
various speeds and heads may be carried out to obtain a full range of
characteristic curves.
Fig. 7.7 shows how the maximum positive pressure may be very great
if reverse flow can occur. For example, if a reflux valve operates auto-
matically at time 12 T = 6f1, then the surge pressure rises to point R which
is over three times the steady head Ho.
To overcome the danger of high positive pressure, a small hole is
sometimes drilled in a positive-acting reflux valve.
Surge control
Excessive surge must be avoided in a pipeline system. Negative surge
must be limited because of the possibility of separation, with subsequent
high 'impact' pressures when the separated columns unite. Positive surge
pressures may be due to reflected negative waves like those found in
Section 7.4, or to positive waves originating from the closure of a down-
stream valve. If the pressure is high there is a danger that the pipe or the
pump casing will burst.
We have already found two methods of controlling surge: a reflux
valve which prevents reverse flows and a flywheel which increases the time
during which the pump slows down. Other ways of limiting surge9 are:
Air vessels6,10,ll,12,13.
By-pass or feeder tank (Section 7.6).
Pressure relief valve or 'surge suppressor'14.
Air inlet valve 10 ·12.
7.5 Air vessels

A vessel containing compressed air may be used, as in Fig. 7.8 (a).
The vessel is connected to the pipeline by a pipe or orifice which is often
designed to give a high loss as water enters the vessel and a low loss as
water leaves. The air pressure is maintained by a compressor which is
actuated by a switch controlled by the water-level (e.g. a float switch).
The differential loss at inflow and outflow may be arranged by using a
bell-mouth orifice of the type shown in Fig. 7.8 (b), giving a loss against
flow diagram (or H-Q characteristic) as shown in Fig. 7.8 (c). Alterna-
tively the flow into the chamber may be through an orifice and the flow
out through a by-pass fitted with a non-return valve.
The volume of air in the vessel may be assumed to vary according to
the isothermal law p V = constant, where p is the absolute pressure and
Vis the volume of the air. Because of the rapidity of the volume changes,
the isothermal law is not strictly true. The adiabatic law is also not
accurate; it assumes very high temperatures at high pressures and anyway
the 'air' in the vessel contains a high proportion of water vapour and gases
which come out of solution. Consequently the expansion is often assumed
to follow the law
p v1.2 = constant (7.2)
A characteristic curve for the air chamber may be used to determine the
air volume.
(a)
Compressor
(b)
Air vessel
Pipe
(c l
~
9.
"0
"
<ll
I
Flow
FIG. 7.8
Design charts,I5,16,I7 by which the required size of an air vessel can be
obtained, are available. It is usually assumed that the column of water
below the air in the chamber has no effect. If the air chamber is intended
to protect a pumping main against excessive waterhammer following
power failure, it is also assumed that a check valve closes at the time of
failure so that all flow at the downstream end is into or out from the air
vessel.
A surge diagram can be drawn to include the effect of an air vessel,
making the assumptions made in the last paragraph. A step-by-step
trial-and-error process is used to obtain the absolute pressure head in the
air vessel at small intervals of time.
Air vessels are also used to restrict surge pressures in the following
situations:
(a) Starting of submersible pump supplying a long pipeline, Fig. 7.9 (a).
When the pump is not working the reflux valve is shut and so length AC
(a)
(c)
FIG. 7.9
contains air and length CE contains water at a static pressure corres-
ponding to the surface level in the reservoir. When the pump starts, air is
expelled from the valve at B, and water flows from A to C. When the
pressure on the upstream side of C exceeds that downstream, the reflux
valve opens. The moving water column in AC meets the stationary
column in CE and rapid deceleration results in a high rise of pressure if
there is no air vessel. With a large air-vessel at D this deceleration is
prevented because the high flow in AC goes into the air vessel when the
reflux valve opens. The pressure in D increases and the water in CE
accelerates and eventually equilibrium conditions are reached with the
flow in the whole pipe AE equal to the pump discharge.
(b) Sudden closure of a valve at the end of a long pipeline, Fig. 7.9 (b)
and (c). An air vessel may be used to protect a pipeline against the water-
hammer pressure produced when a downstream valve is suddenly closed.
The vessel must have sufficient capacity to ensure that no air enters the
pipeline when the pressure is a minimum. With hydro-power supply
pipes, Fig. 7.9 (c), it is common to use an open tank (a 'surge tank')
rather than an air vessel.
(c) Reciprocating pump. We deal with this separately in Section 7.8.
7.6 Pump by-passes and feeder tank

A simple method of preventing the pressure in a main from becoming
too high or too low is to connect the main to a tank at a point where
undesirable high or low pressure would otherwise occur. The tank must
be so large that water put into it, or taken from it, does not cause the level
to change very much. If the tank is required to protect the main from low
pressures, it is connected to the pipe by a non-return valve which only
opens when the mains pressure head falls below that in the tank.
In many cases the pump suction well can be used as a feeder tank to
prevent separation. A by-pass fitted with a non-return valve connects
the suction and delivery sides of the pump, Fig. 7.10 (a).
Example 7.2
Draw a surge diagram to obtain the maximum pressure head at the pump for
the system described in Example 7.1, if separation is prevented by a by-pass to the
pump. The by-pass is fitted with a non-return valve, which opens when the pressure
in the pipeline drops below the static level of the suction well, and the loss through
the by-pass, including the valve, is equal to 0· 15 Q2 , the loss being measured in
feet and Q in cusecs.
(a)
(b)
H 1~0
11/ls
-20 I Q
-2 -1 2 3 4 5 6 7 8 cusecs
FIG. 7.10
SOLUTION
The surge diagram, Fig. 7.10 (b) is constructed in the same way as Fig. 7.3,
but the negative surge at points Ao-1 and A1-2 is limited by the suction well
level, less the by-pass losses, i.e. by line 1fJ•· The maximum head, shown at
point Aa-4, is 116·9 ft compared with the maximum head of 267 ft when there
was no by-pass (Example 7.1).
The broken line shows the worst conditions which could occur, giving a
maximum head of 118·8 ft.
A pumping main which undulates may be protected by a number of
feeder tanks. These are located at points where the pressure head in the
main would otherwise drop below the vapour-pressure head, i.e. when
the negative head in the main reaches about 30ft. Fig. 7.11 (a) shows
such a main. The normal flow is such that in the absence of protective
devices separation would occur in the pump discharge if flow suddenly
stops.
A by-pass at the pump A limits the amplitude of the waterhammer
oscillation in AB to the line a1b1, which is parallel to the 'normal' hydraulic
gradient aoeo. At B, point b1 is 30 ft below the pipe; a feeder tank there
limits the downward oscillation of the waterhammer to the height of the
water surface in the tank, b2, and in pipe BC to the level given by line b2c1.
In the design of such a system it is necessary to know the total quantity
of water which passes from each feeder tank and the maximum rate of
flow from the tank. The capacity of the tank must be large enough to
supply the total quantity passing without significantly altering the water
(a)
Q
Normal flow
Rate
of
flow l:.Omax
Time
(b)
FIG. 7.11
level in the tank. The size of the valve between the tank and the main must
be sufficient to pass the maximum flow without too great friction loss.
Surge diagrams must be drawn for each length, AB, BC, CD, DE, and
also for combined lengths to show the waterhammer oscillations when
a valve is closed. For example when the valve from the feeder tank at B
is closed because the mains-head exceeds the water level in the feeder tank,
a diagram for AC is needed to determine the flow at C. From the surge
diagrams a flow-time diagram is obtained, Fig. 7.11 (b). The total quantity
of water taken from the tank is obviously represented by the hatched area
(= ~(Qout - Q;n)L1t) and the maximum rate of flow is shown by L1Qmax
7.7 Parallel pumpsls
If two or more pumps deliver to a simple main and the power supply to
one of the pumps fails, the waterhammer pressures may be obtained by
drawing surge diagrams in the way described in Section 6.8.
7.8 Waterhammer in delivery pipe from reciprocating pump

When there is no air-vessel, or if the vessel fails, the cyclic flow gives rise
to waterhammer pressures which can be represented by a surge diagram.
The discharge from a reciprocating pump is not affected by the operating
head, being only a function of time. Consequently the characteristic
curves are vertical lines.
£13
17
21
25
FIG. 7.12
FIG. 7.13
The form of the surge diagram depends on the relationship between the
time taken for one pump cycle, Tc, and the pipe period, f1·
If Tcf2f1 is an even integer we can obtain a diagram of the form shown
in Fig. 7.12, where Tcf2f1 = 12 and T = f1 = Tc/24; 1p-lines are drawn at
intervals of 2T.
For a double-acting pump the surge diagram is identical to that for a
similar single-acting pump for the first half-cycle, i.e. until time 12 in Fig.
7.12. After the first half-cycle the pressures and discharges progressively
increase; the oscillation is amplified until cavitation occurs.
If Tcf2fl is not an even integer there is no resonance and the oscillation
follows a pattern which is repeated every cycle. For example, if Tcf2fl = 5,
the surge diagram is as shown in Fig. 7.13.
REFERENCES
1. GAYED, Y. K., and KAMEL, M. Y. M.: 'Mechanics of secondary water-
hammer waves,' Proc. Instn mech. Engrs, (1959), 173, pp. 675-683.
2. RicHARDS, T. R.: 'Water-column separation in pump discharge lines,'
3. SHARP, B. B.: 'Cavity formation in a simple pipe due to rupture of the
water-column,' Nature, London, (1960), 185, pp. 302-303.
4. CARSTENS, M. R., and HAGLER, T. W.: 'Water hammer resulting from
cavitating pumps,' J. Hydraul. Div., Proc. Am. Soc. civ. Engrs, (1964),
90, HY6, pp. 161-184.
5. Due, J.: 'Negative pressure phenomena in pump pipelines,' Proc. Int. Symp.
Waterhammer pumped Storage Projects, ASME., (1965), pp. 154-157.
6. WooDWARD, A. C.: 'Some practical aspects of pressure surge control,'
Allen Engng Rev., (1964), vo 55, pp. 16-19.
7. SCHNYDER, 0.: 'Druckstosse in Pumpensteigletungen,' Schweiz. Bauztg.,
(1929), 94, p. 271.
8. ANGUS, R. W.: 'Protective air vessels for rising pipe lines,' Discussion,
Proc. Instn mech. Engrs, (1945), 153, pp. 20-22.
9. LIVINGSTONE, A. C., and WILSON, J. N.: 'Synopsis of surge control equip-
ment,' Wat. & Wat. Engng, (1963), 67, pp. 264-269.
10. ANGUS, R. W.: 'Air-chambers and valves in relation to water hammer,'
11. BINNIE, A. M.: 'Protective air vessels for rising pipe lines, Proc. Instn mech.
Engrs, (1945), 153, pp. 15-20.
12. BLAIR, J. S.: 'Controlling pipe line surges by means of air vessels,' Proc.
Instn mech. Engrs, (1945), 153, pp. 1-8.
13. GoLDMAN, 0. G.: Water hammer; its causes, magnitude, prevention (Colum-
bia Graphs Inc., 1953).
14. LuNDGREN, C. W.: 'Charts for determining size of surge suppressors for
pump-discharge lines,' J. Engng Pwr, Trans. Am. Soc. mech. Engrs, (1961),
83, pp. 43-45.
15. CoMBES, G., and BoRoT, G.: 'New chart for the calculation of air vessels,
allowing for friction losses,' Houille Blanche, (1952), pp. 723-729.
16. EvANS, W. E., and CRAWFORD, C. C.: 'Design charts for air chambers on
pump lines,' Trans. Am. Soc. civ. Engrs, (1954), 119, pp. 1025-1036.
17. TucKER, D. M., and YouNG, G. A. J.: 'Estimation of the size of air vessels,'
Br. Hydromech. Res. Ass., (1962), SP670.
18. LINTON, P.: 'Notes on pressure surge calculations by the graphical method:
No. 1, Pump stoppage after power failure,' Br. Hydromech. Res. Ass.,
(1954), TN447 (2nd Edn.), pp. 18-21.
FURTHER READING
BERGERON. P.: 'Complexite des phenomenes de coups de belier sur les installa-
tions de pompage et essai de classification des solutions generales pour y reme-
dier,' Houille Blanche, (1949), pp. 724-732.
BINNIE, A. M., and THACKRAH, D. G.: 'Water hammer in a pumping main and
its prevention,' Proc. Instn mech. Engrs, (1951), 165, pp. 43-52.
DoNSKY, B.: 'Complete pump characteristics and the effects of specific speeds
on hydraulic transients,' J. bas. Engng, Trans. Am. Soc. mech. Engrs, (1961),
83, pp. 685-696.
DoNSKY, B., and DEFAZIO, F. G.: 'Design analysis of waterhammer in the San
Louis pumping-generating plant.' Proc. Symp. Waterhammer pumped
Storage Projects, ASME, (1965), pp. 61-69.
GAULHET, A.: 'La protection des installations de pompage coutre les coups de
belier d'onde,' Houille Blanche, (1967), pp. 169-172.
JAEGER, C.: 'Water-hammer caused by pumps,' Wat. Pwr, (1959), 11, pp. 259-
266.
KITTREDGE, C. P.: 'Hydraulic transients in centrifugal pump systems,' Trans.
Am. Soc. mech. Engrs, (1956), 78, pp. 1307-1320.
KNAPP, R. T.: 'Complete characteristics of centrifugal pumps and their use in
prediction of transient behaviour,' Trans. Am. Soc. mech. Engrs, (1937), 59,
pp. 683-689.
LINTON, P.: 'Pressure surges on starting pumps with empty delivery pipes,'
Br. Hydromech. Res. Ass., (1950), TN402.
LINTON, P.: 'A simple guide to water hammer and some notes on pressure
surges in pump delivery lines,' Br. Hydromech. Res. Ass., (1961), TN411.
LINTON, P.: 'Notes on pressure surge calculations by the graphical method:
No.1, Pump stoppage after power failure,' Br. Hydromech. Res. Ass., (1954),
TN447.
LIVINGSTONE. A. C., and WILSON, J. N.: 'Effects of valve operation,' Symp.
Surges Pipelines, Proc. Instn mech. Engrs, (1966), 180, 3E, pp. 44-51.
LUPTON, H. R.: 'Graphical analysis of pressure-surge in pumping systems,'
MARPLES, E. B.: 'Introduction to the theory of surge,' Allen Engng Rev., (1958),
No. 39, pp. 14-17.
PARMAKIAN, J.: 'Pressure surges in pump installations,' Trans. Am. Soc. civ.
Engrs, (1957), 83, pp. 1216-1 to 1216-8.
PERLITER, S., and URBAIN, E. A.: 'Pipeline surges and their effect on pump
station design,' J. Am. Wat.Wks. Ass., (1964), 56, pp. 891-898.
ScHNYDER, 0.: 'Comparison between calculated and test results on water
hammer in pumping plants,' Trans. Am. Soc. mech. Engrs, (1937), 59,
pp. 695-700.
STEPANOFF, A. J.: 'Elements of graphical solution of water-hammer problems
in centrifugal pump systems,' Trans. Am. Soc. mech. Engrs, (1949), 71,
pp. 515-534.
STEPANOFF, A. J.: Centrifugal and axial flow pumps (Wiley, 2nd Edn., 1957).
STREETER, V. L.: 'Valve stroking to control water hammer,' J. Hydraul. Div.,
Proc. Am. Soc. civ. Engrs, (1963), 89, pp. 39-66.
TmRRIOT, C.: 'Methodes approchees de calcul des coups de belier dans des
conduites relativement longue,' Houille Blanche, (1967), pp. 179-192.
8 Waterhammer in Hydro-electric
Supply Pipes
A surge tank is often constructed when there is a long pipeline supplying

water to a hydro-electric power plant. We will consider the functions and
design of surge tanks later.
Short pipelines usually have no surge tank, and waterhammer pressures
in the pipeline can then be determined by using the methods discussed in
Chapters 4, 5 and 6. The 1p-lines showing the characteristics of the down-
stream 'device' depend upon the type of valve used. For impulse turbines
such as the Pelton wheel, the water supply is generally controlled by spear
valves. The characteristic curve is a parabola through the origin. For
reaction turbines, the characteristics of the moveable guide vanes (or
'gate') can be obtained by field or model tests on similar vanes.
In a system which includes a surge tank, the waterhammer oscillations
are superimposed on mass oscillations. In this chapter we examine the
use of surge diagrams to obtain water hammer pressures; mass oscillations
are examined in Chapter 9-14.
8.1 End conditions at a surge tank

As we shall see in later chapters, any movement of the valve results in
oscillation of the water surface in the surge tank. The flow into the surge
tank 'q' and the water level in the tank 'h' at any time t during the oscilla-
tion depend on the dimension of the pipeline and tank, and on the type
of valve movement. The valve movement may be opening or closure,
partial or complete, sudden or gradual.
We will consider a short interval of time !lt which begins at time t and
is of such duration that changes in the rate of flow into the surge tank may
be taken as linear. Changes of water-level during !lt are also linear.
87
Referring to Fig. 8.1, we see that for a tank of constant cross-sectional
area (A 8 = constant) the volume of water entering the tank in time f:.t
is given by
(8.1)
FIG. 8.1
B is the base of the surge tank BD, Fig. 8.2 (a), so using the notation
of Fig. 6.13 the conditions at B at the entrance to the surge tank are shown
(a)
c
(b)
s
/
r=., --1--qt--r
R~
FIG. 8.2
Waterhammer in Hydro-electric Supply Pipes 89
by point BDt for the timet. The co-ordinates ofBDt on the surge diagram,
Fig. 8.2 (b), are qe and he.
From equation (8.1) t1h =~=tanA' (8.2)

qe + q(t+3.t) 2As
Equation (8.2) can be shown on the surge diagram by the construction
drawn in broken lines, where R has the co-ordinates -qe and he, and tan A'
is made equal to tAt/As. Point S on the q = 0 axis can also be obtained
by drawing a line from point BDt at a slope of -A'. Line S-L is thus
defined; point Bn(t+ilt), which lies on line S-L, can be fixed only by a
knowledge of surge conditions in pipes 1 and 3. These may be drawn as
line J-K, and the intersection of lines S-L and J-K defines point Bn(t+llt)·
J-K depends on the position of the valve (whether or not the length of BC
is negligible) and the type of valve movement.
BDt-S and S-Bn<t+llt) are true surge lines. They represent the observa-
tions of a traveller moving with a surge wave. S represents conditions
found when the traveller reaches the water surface in the surge tank at D.
S may therefore be called Dn(t+!llt)·
8.2 Simple tank near valve (pipeline friction neglected)

(a) Sudden complete closure. The initial steady flow in the pipeline is Q0
and the initial steady head is Ho. Point P on the surge diagram, Fig.
8.3 (c), gives the initial conditions; P has the co-ordinates Qo, Ho. A is a
reservoir with so large a horizontal area that the level remains constant.
~:8~:--·-l~
L,
FIG. 8.3 (a). Before closure.
When the valve closes, the whole of the flow from the pipeline AB is
diverted to the surge tank BD. Subsequent variations of flow satisfy
continuity, so at all times after closure the upward flow in the surge tank,
qe, is equal to the flow down the pipeline, qpt. Consequently, conditions
at the downstream end of the pipeline are the same as at the base of the
surge tank and in the surge diagram we can use the notation B = BA = B n.
The time interval Tis made equal to t,u1 = L1/a1. and t1t for the surge
tank is made equal to 2T. Then, for the surge tank BD the slope of the
surge line A' = tan-1 (2T/2A 8) = tan-l(T/As).
Starting from the time when the valve closes, when t = 0, we draw
surge lines P-D1 and D1-L at slopes of -A' and A'. D1 corresponds to
S in Fig. 8.2 (b). These surge lines show the observations of a traveller
leaving B at time 0, going up the surge tank to reach the surface D at time
1 and then returning to B at time 2.
'"' """'~"" ~
!!----=---~0,,
FIG. 8.3 (b). After closure.
Now at time 2 the conditions at B depend on the waterhammer wave

which left the valve C at zero time. C is very close to B; the valve may
be taken as being at B. A traveller moving just ahead of the waterhammer
wave reaches A at time 1. He finds conditions at A still at their initial
value since he is just ahead of the wave. Therefore point A1 is the same
asP on Fig. 8.3 (c).
(c)
Ho
I
I
I
FIG. 8.3 (c)
On returning to B, which he reaches at time 2, the pipeline traveller

finds that the water movement in the surge tank has already affected the
flow and head at B, as shown by line D1-L. The surge line for the pipeline
traveller is P-K at a slope of -A1 = -tan-1 (a1/gA). Consequently point
B2 is at the intersection of D1-L and P-K.
Continuing the surge diagram we draw the observations of the pipeline
traveller as he goes back and forth along the pipeline. The lines are at a
slope of A1 and give points Aa, B4, As and so on. The surge tank traveller,
as he moves up and down the tank, finds conditions given by Da, B4, Ds
and so on. The diagram can be continued as Fig. 8.3 (d). Since there is
no friction, the upsurge y1 is equal to the downsurge y2 and at the end of a
cycle the diagram comes back to point P.
FIG. 8.3 (d)
(b) Slow closure. To prepare a surge diagram when the valve closes in
time T we first obtain the flow through the valve at intervals of time 2r
from the time of complete closure. For example, if the flow through the
valve varies as shown in Fig. 8.4 (a), the flow at timeT- 2r is Q<T-2,), and
at time T- 4r is Q(T-4,)· The 'remainder' of time may be termed Or,
and the corresponding flow Q<T-4,) = Q8. From this, Q(T-2,) is also
Q<e+2l· Of course 0 is always less than 2.
The first surge lines are drawn for the reduced time interval Or. An
observer leaving Bat time 0 and moving up the surge tank finds conditions
at D to be q = 0 and h = Ho because there is no change of flow or head
at the bottom of the surge tank at time 0. Hence, point 0 on Fig. 8.4 (b)
represents Band D. However, changing conditions will be noticed by the
observer as he returns from D to B, and these are shown by a surge line
of slope A 8 ' where tan Ae' = Or/2A 8 •
The pipeline traveller reaching Bat time eleft A at time (0 - 1), when
conditions had not changed from the initial conditions shown by point
P. The changes noted during the journey from A to Bare shown by a surge
line of slope A1 = tan a1jgA.
e
At time the difference between flow qe up the surge tank and qPe
down the pipeline is Q8 through the valve. Q8 is therefore the difference
between the abscissa of points Bne and BAe· Q 8 is shown by a horizontal
broken line on the surge diagram.
e• ...,.
(a)
(b)
(c)
FIG. 8.4
We continue by drawing surge lines to show the observations of

travellers in the pipeline and surge tank. Since the time interval is now
2T, the slope A' is TjA 8 •
At time ( () + 2) the difference of flow in pipeline and surge tank is
Q(o+2), which is the horizontal distance between points BD<o+2> and
BA(o+2)·
At time ( () + 4) = Tfr there is no flow through the valve and point
BD<e+4> coincides with point BA <e+4)· Thereafter the diagram continues in
the same way as Fig. 8.3 (c).
8.3 Throttled tank near valve

Many modern surge tanks include a restriction at the base. Chapter 9
explains this. We can allow for losses in a restriction and also losses due
to pipeline friction by plotting the losses against Q on the surge diagram.
The simplest is when closure takes place in time 2T. Instantaneous
closure requires two sets of surge lines-one for conditions immediately
before the passage of the wave and one after. Closure taking longer than
2r can be dealt with by the method shown in Fig. 8.4 (b).
Fig. 8.5 (b) shows the first few lines of a surge diagram to give the water-
hammer pressures in the hydraulic system of Fig. 8.5 (a) when the valve
closes in time 2T = /-ll·
(a)
\
(b)
FIG. 8.5
The head loss due to pipeline friction is shown by line L-L. This is
similar to line L-L on Fig. 6.15, except that the friction line is below the
reservoir level for positive flow; the losses are assumed to be concentrated
at a single point near the reservoir.
The initial conditions with steady flow Q0 along the pipeline and out
through the valve are given by point P on the diagram. P is on L-L at a
distance hFo below the static level Ho. The water surface in the reservoir
is initially at this level but for steady flow in the pipeline the flow into the
surge tank q = 0. Point R is given by the co-ordinates q = 0, H = Ho -
hFo, and shows conditions at D up to time 1 and at Bn at time 0.
Throttle losses are shown by curve T-T.
The construction of the first surge lines is similar to that of Fig. 8.4 (b).
From R a line is drawn at slope A' where tan A' = TjA 8 • From P a line
is drawn at slope A1 where tan A1 = a1jgA. By trial or graphical means
points Bnz and BAz are found such that the vertical distance between them
is equal to hTz obtained from curve T-T. This obviously shows the actual
conditions: when water enters the tank there is a head loss hTz such that
the pressure above the throttle is less than the pressure beneath the throttle
as the water moves up through the throttle.
From BD2 a line of slope -A' is drawn to meet the q axis at Da, which
gives the water level in the surge tank at time 3. From BAz a line of slope
A1 is drawn to meet L-L at Aa. The observers responsible for the surge
lines next return along surge tank and pipeline to find points BD4 and BA4
which have the same abscissa but are hT4 apart vertically; hT4 is obtained
from curve T-T.
The diagram can be continued for a full cycle. For the case considered
in Section 8.2, where there is no friction or throttle, the conditions at the
end of the cycle are the initial conditions at point P; the loci of Bn form
in ellipse. When allowance is made for pipeline and throttle losses, the
loci of Bn form a spiral. Downsurge is less than the preceding upsurge
and the flow in the pipeline at the end of the cycle is less than the initial
flow.
8.4 Tank between pipeline and penstock

The usual position for a surge tank is not close to the valve, although this
is advantageous because the amplitude of the waterhammer oscillations
then have the same range as the mass oscillation. In high head schemes
the power station is often in a valley and is fed by a pipeline or tunnel
from a reservoir in a neighbouring valley. The pipeline is fairly flat, to
limit static pressures. Often the tunnel or pipeline ends high up on the side
of the valley where the power station is built, and the final part of the
conduit falls steeply, often as several parallel penstocks. The most con-
venient position for the surge tank is near the junction of pipeline and
penstocks. It can then be excavated in the hillside.
Fig. 8.6 (a) shows a typical layout. Fig. 8.6 (b) shows the construction
of a surge diagram for an unrestricted tank; pipeline losses are neglected.
(a)
r=t,u.1=~,u.3
q3= Q
(b)
FIG. 8.6
The diagram can be modified for pipeline and throttle losses in the way
indicated in the last section. In Fig. 8.6 (b) /'1 = f'a, but the same values
of A1, Aa and A' have been taken.
The method of construction follows the lines of that for branched pipes
given in Section 6.8, but in this case we combine the diagrams for the
pipeline AB and the penstock BC, drawing the diagram for the surge
tank BD separately.
The initial conditions shown by P refer to the pipeline AB and the
penstock BC. The initial conditions shown by R refer to the surge tank-
i.e., D and Bn.
The characteristics of the 'valve' are shown by '!jJ-lines at intervals of
r = !,u1. For a Pelton wheel with a spear valve these lines are parabolas
through the origin Q = 0, H = 0. The guide vanes of reaction turbines
produce parabolas whose origin depends on the specific speed of the
turbine.
A travelling observer leaving B at time 0, when the valve starts to
operate, and going along the penstock BC reaches the valve at time 1,
when the discharge characteristic is given by line 'lj!e1. The surge line goes
from P (initial conditions at B at time 0) and the slope of the line is - Aa
where tan Aa = aafgAa. Point C1 is fixed at the intersection of the surge
line and '!jJ-line. The traveller then returns, reaching B at time 2. The
conditions he finds at Be lie on line Be2.
Meanwhile, a second observer does a round trip up the surge tank. He
finds initial conditions (point R) when he reaches the surface at D, and
returns to B n at time 2 to find conditions that must lie on line B n2 at
slope A' = r/As.
A third observer starting from B visits the reservoir to find initial
conditions (point P) there, and also returns to B at time 2. He finds
conditions that must lie on line BA2 at slope A1.
Now continuity must be satisfied at B, and the head at the three jointing
branches (i.e. BA, Be and Bn) must be the same. The corresponding
positions of points BA2, Be2 and Bn2 may be found by trial so that q1 =
q2 + qa. Alternatively, line B' A2 may be drawn from the point Y2 at a
slope equal to the sum of the slopes of lines Bc2 and Bn2. Point Y2lies at
the intersection of line Be2 with the projection of D1 (R). Point BA2 is
at the intersection of line B' A2 and line BA2· Point Be2 is fixed by drawing
a horizontal line, representing constant head at B, from point BA2 to cut
line Be2.
The diagram is continued by drawing surge lines and projections in the
directions shown on Fig. 8.6 (b).
When the cross section of the surge tank varies with elevation, the angle
A' must be varied to correspond with the area at the elevation of D.
REFERENCES
BERGERON, P., and KERR, S. L.: 'Waterhammer analysis on pumped storage
projects,' Proc. Int. Symp. Waterhammer pumped Storage Projects, (1965),
ASME, pp. 189-200.
BRATFISCH, A. E., and CARTWRIGHT, K. 0.: 'Water-hammer calculation and
test results-Owens Gorge Power Plant penstocks,' Trans. Am. Soc. mech.
Engrs, (1956), 78, pp. 1329-1336.
CAusoN, G. J.: 'Precautions against resonant water hammer,' Wat. Pwr, (1968),
20, pp. 20-22.
EscANDE, L.: 'Influence de la liaison entre chambre d'equilibre et canal
d'amenee d'une usine hydro-electrique sur le regime des surpressions-role
de l'etranglement,' Houille Blanche, (1949), 1, pp. 457-478.
EscANDE, L. : 'The transmission of water-hammer pressures through surge tanks:
communication,' Proc. Instn mech. Engrs, (1954), 168, p. 639.
JAEGER, C.: 'Present trends in surge tank design,' Proc. Instn mech. Engrs, (1954),
168, pp. 91-103.
JuPILLAT, R., and TRIVIDAC, A.: 'Sur I'evolution du coup de belier dans une
galerie d'amenee a l'amant de la cheminee d'equilibre,' Houille Blanche,
(1951), pp. 640-646.
PAYNTER, H. M.: 'Methods and results from M.I.T. studies in unsteady flow,'
J. Boston Soc. civ. Engrs, (1952), 39, pp. 120-165.
STROWGER, E. B., and KERR, S. L.: 'Speed changes of hydraulic turbines for
sudden changes of load,' Trans. Am. Soc. mech. Engrs, (1926), 48, pp. 209-
232.
WATER POWER: 'Water-hammer in pumped-storage projects,' Wat. Pwr, (1966),
18, pp. 119-123.
ZIENKIEWICZ, 0. C., and HAWKINS, P.: 'The transmission of water-hammer
pressures through surge tanks,' Proc. Instn mech. Engrs, (1954), 168, pp.
629-638.
9 Introduction to Surge Tanks
9.1 Purpose of surge tanks

If in a hydro-electric scheme the water is conveyed to the turbines through
a long conduit, a very large mass of water is contained in the conduit, and
considerable forces are necessary to retard or accelerate this mass when
the turbines' demand for water changes. For example, in an 8ft (2·4 m)
diameter pipeline 2 miles (3·2 km) long the mass of water is about 14 000
tons. Change of flow to the turbines also produces waterhammer
pressures, 1 which we dealt with in Chapter 8.
Such changes of flow occur whenever there is an alteration in the load
demand from the turbines, i.e. when the electrical power required to be
generated alters. Under normal operating conditions it may be possible
to regulate the turbines so that the change is gradual, but accidental rapid
changes must be allowed for. For example, failure of the power cable
from the hydro-electric station may result in an immediate rejection of
load; failure of another station connected to the same electricity network
may result in an immediate increase in demand.
Pelton wheels have spear valves to control the flow of water, but for
Francis and Kaplan turbines the flow is controlled by guide vanes. In
the following pages, by 'valve movements' we mean movements of the
mechanism controlling the flow to the turbines, however effected.
In the case of closure of the valve (load rejection) it is possible to
provide for a gradual change of flow by the provision of relief valves or
diverters. In some cases this may be the most satisfactory solution to
the problem, but such measures cannot assist when there is an increase
of demand and the valve is opened. In any case, relief valves and diverters
result in a waste of water, which may not be justified if valve movements
are frequent. The increase in pressure due to sudden change of flow may
produce dangerous conditions in the absence of an adequate surge tank.
Such conditions have occasionally resulted in the fracture of the conduit.2,3
98
Introduction to Surge Tanks 99
Fig. 9.1 shows the supply system for a hydro-electric plant with a single
surge tank at the upper end of the steeply-sloping penstock.
The inclusion in the hydraulic system of a surge tank near the power
station gives rise to the forces required to retard or accelerate the mass
of water in the pipeline. If the flow to the turbine is increased, the water
level in the surge tank drops, providing an accelerating head in the pipe-
line; if the flow to the turbines is reduced, the water level in the surge
tank rises providing a retarding head in the pipeline. The surge tank also
Tunnel w1th shaft surge tank
FIG. 9.1
provides a free surface near the turbine. The intensity of waterhammer

pressures in the pipeline is usually so reduced as to be practically negligible,
and the strength of the pipeline can therefore be reduced, with resulting
economy in construction costs.
A surge tank has been defined as a stand pipe or storage reservoir
placed at the down-stream end of a closed aqueduct to prevent undue rise
of pressure in case of a sudden diminution of draft and to furnish water
quickly when the gates are opened, without having to wait for the velocity
in the long feeder to pick up. 4
Summing up, upstream surge tanks fulfil the following purposes:
(a) They provide an accelerating head or retarding head to the mass of
water in the pipeline when the valve is opened or closed.
(b) They reduce waterhammer effects.
(c) They provide a reservoir from which water is drawn by the turbines
on increase of load before the water in the pipeline accelerates, and
to which water passes from the pipeline when the valve closes.
Where water leaves the turbines along a tail-race tunnel, as in some
underground plants, a downstream surge tank is sometimes provided. 5 ·6,7
9.2 Oscillation phenomena

When the valve is suddenly completely closed, the water in the penstock
stops almost at once because of its small inertia. The water in the pipe-
line, with large inertia, retards slowly. The difference in flows between
pipeline and penstock causes a rise in the water level in the surge tank.
The water rises above the static level of the reservoir water, producing a
counterpressure so that water in the pipeline flows towards the reservoir
and the level of water in the surge tank drops. The cycle is repeated, with
mass oscillation of the water in the pipeline and surge tank. In the
absence of damping, such oscillation would continue indefinitely with the
same amplitude. However, damping is provided by friction in the pipe-
line and may be increased by constructing a restriction (or 'throttle') at
the base of the surge tank. The pipeline and surge tank are considered
as rigid (i.e. they do not stretch under load), and the water is considered
as incompressible.
Similar mass oscillations occur when the closure is gradual and/or
partial, but the amplitude is reduced. For valve opening there are also
mass oscillations, but in this case the first movement of the water in the
surge tank is a downsurge. The mass oscillations are superimposed upon
oscillations due to waterhammer,I.s,9 The mass oscillations are slow;
waterhammer oscillations, which we have already dealt with in Chapter
8, are very rapid, and the two phenomena can be considered separately.
As the water level in the surge tank varies there is a corresponding
change in the head on the turbines. The turbine governors, in their
efforts to maintain constant power input to the turbines, tend to oscillate
in a similar manner to the mass oscillation of the water in the pipeline,
except in the case of complete closure. The oscillation of the governors
may, under certain conditions, maintain or even increase the mass
oscillation, giving rise to instability.
9.3 Design considerations

For any surge tank the following criteria10 must be complied with:
(a) The surge tank must be located so that waterhammer pressure
variations are kept within acceptable limits. The usual position is
at the junction of the pipeline and the penstock.
(b) The tank must be stable.
(c) The tank must contain the maximum possible upsurge unless a
spillway is provided.
(d) No air must be drawn into the pipe at lowest downsurge.
(e) The range of surges must not be so large as to cause undesirably
heavy governor movements or difficulty in picking up load.
Loading. Most surge tanks are designed so that item (c) is satisfied for
sudden complete valve closure after the flow in the pipeline has been
enough for full turbine load. This is known as 'full load rejection'. It
can occur if the electricity cables from the power station are broken.
The lowest downsurge, (d), is based upon the assumption that the
generators have to increase their output to full load from some lesser
load. Sometimes the increase in output is from no output at all (and
therefore no flow of water), and this is known as 'full load acceptance'.
More often it is assumed that before full load is demanded the turbines
are producing some power-for example 10% of fullload.7,11,12
The possibility of superimposed surges resulting from successive closure
and opening of the valve has been considered by several engineers and
digital computers have been used to investigate this problem.13 For the
Cameron Highlands hydro-electric scheme14 the surge tank had to be
sufficient for full-load rejection, for full-load acceptance, and also for an
increase of load from 50% to 100% followed by full-load rejection.

As in other aspects of fluid mechanics and hydraulic engineering, a friction-
less system provides a useful basis for theory. Of course, no actual
system is frictionless. Generally, friction is no disadvantage as it encour-
ages the mass oscillations to damp out. However, the first downward
surge following valve opening is increased with increase of pipeline
friction, and increased penstock friction makes for instability. 15
Pipeline friction. Whatever type of surge tank is used, the frictional resis-
tance of the pipeline is an important consideration. It is also difficult to
estimate.16 Because the first downsurge following load acceptance is
increased with increased pipeline friction whereas the first upsurge
following load rejection is reduced, different values for pipeline friction
are often assumed for the two cases. It is usual to assume that the friction
losses are proportional to the velocity squared, although in practice
the frictional index is likely to be rather less than 2. Losses at intakes,
screens, bends, etc., must also be considered and are usually taken as
proportional to the pipeline velocity squared.
Friction losses are again considered in Section 11.7 (page 138).
Throttle damping. The introduction of a throttle at the base of the surge

tank adds to the friction of the system. The amplitude of the first surge,
whether upsurge or downsurge, is reduced, and the mass oscillations
damp out more quickly.
Calame and Gaden17 give coefficients from which throttle losses can be
estimated, and the losses have been further considered by Levin, Is
Maionel9 and Zieman. 2o It is difficult to calculate the head losses in a
throttle because the flow is affected by changes of direction, contractions
and expansions. Only model studies21-25 of the whole system give the
true hydraulic conditions.
9.5 Surge tank types

The most common types of surge tank are shown in Fig. 9.2.
JL
Variable area tonk
Simple tonk
JL
Differential tonk
JL
Throttled lank Spilling tank
n
.Closed tonk
JUL
Multiple tanks
FIG. 9.2
(a) A Simple Tank is an unrestricted tank of constant cross-sectional

area in which the maximum variation of water-level is contained
within the tank.
(b) Variable-area Tanks provide additional reservoir capacity where it is
most required, i.e. at the top of the tank for load rejection and/or
at the bottom of the tank for load acceptance.
(i) Conical tanks have been built with the large diameter at the top
and at the bottom.26,27
(ii) Expansion chambers may be in the form of an increase of
diameter or galleries constructed off the main tank. 28 Many
recent surge tanks include expansion chambers.
(c) In a Throttled Tank a restricted orifice is provided at the base of the
tank to decrease the amplitude of the surge and to effect more rapid
reduction of the mass oscillation.
In recent years throttled tanks have become more popular
because they give an economy of up to 40% compared with simple
tanks, due to reduction of surge amplitude. 20 This reduction is
caused by losses between the pipeline and the surge tank in a
throttle. The losses are proportional to the square of the velocity
in the restriction, and the effect of the throttle is greatest for large
changes ofload. The simplest form of throttle is an orifice plate, and
a length of pipe of reduced diameter also gives a simple throttle.
When a throttled tank has to deal with large down-surges, such
as occur when the demand for water by the turbines increases from
no-flow to full-flow, there may be danger of the water column
breaking, if the loss is too severe, as water leaves the tank. In this
case the throttle can be faired on the upper side and square on the
f---------811. diam.--------1
FIG. 9.3
lower. Alternatively, if the tank is designed for complete stoppage

of flow and only partial acceptance, the throttle may be faired on the
lower side as in Fig. 9.3. 29 The same type of throttle is used when
the elevation of the tank is such that it is desirable to limit the
downsurge to prevent air from entering the pipeline.
(d) Differential Tanks are popular on the American continent.30,3I,a2
When flow to the turbines is stopped, water from the pipeline at
first rises in the narrow central part (the 'riser') of the differential
tank, giving a rapid increase of head to retard the flow in the pipe-
line. Further water from the pipeline then passes through the small
ports to the outer tank, which provides reservoir capacity.
Similarly, an increased demand by the turbines is met at first
from the central riser where the water level drops quickly. This
rapid change in the difference of head between upstream and down-
stream ends of the pipeline causes the water in it to accelerate.
While the pipeline water is accelerating the increased flow required
by the turbines comes from the outer tank through the ports.
(e) The volume of a Spilling Tank2B,33,34 is insufficient to contain the
whole surge volume under the most adverse conditions of closure,
and the excess passes to waste, often along a lined channel.
(f) In a Closed Tank the change of water level is reduced, due to a con-
fined air pocket above. 35-39 The effect is similar to that of an air
vessel on a pumping main,
(g) Multiple Tanks: 40 •41 In some modern plants more than one surge
tank is constructed, the lay-out often depending on the tunnelling
techniques used. In long pipelines intermediate surge tanks are
sometimes employed.
Many installations have a surge tank system which is a combination
of the types mentioned. 42 Fig. 9.4 gives diagrammatic sketches of a number
Rossens
(France)
S: d lin
7
Tongland
(Scotland) J:~
Mae
(Italy)
I nnert kirchen
( Sw1tzer land)
* *
Isere-Arc
(France)
* These galleries have
been turned through 90°
FIG. 9.4
of recent plants, all of which include throttles. The sketches are not to
the same scale and some have been distorted to bring different parts into
the same plane. 7 •43 ·44
The design of complex systems is greatly facilitated by the use of com-
puters.13,40,4S-49 Numerical methods, which we will consider in Chapter
11, may easily be adapted for digital computers, and the surge, or mass
oscillation, may be reproduced by an analogue computer.so
REFERENCES
1. WIDMANN, R.: 'The interaction between waterhammer and surge tank
oscillations,' Proc. Int. Symp. Waterhammer pumped Storage Projects,
1965, ASME, pp. 1-7.
2. JAEGER, C.: 'Water-hammer effects in power conduits,' Civ. Engng pub/.
Wks Rev., (19~8), 43, pp. 74-76, 138-140, 192-194, 244-246.
3. FERRAND, G.: 'A propos d'un accident survenu a une conduit forcee apres
quarante ans d'exploitation,' Houille Blanche, (1953), p. 20.
4. JoHNSON, R. D.: 'The surge tank in water power plants,' Trans. Am. Soc.
mech. Engrs, (1908), 30, pp. 443-474.
5. EscANDE, L.: 'Stabilite de deux chambres d'equilibre respectivement soli-
daires des canaux d'amenee et de fuite,' Houille Blanche, (1953), 8, p. 647-
654.
6. JAEGER, C.: Underground power stations, Chapter xxii of 'Hydro-electrical
engineering practice,' Ed. J. Guthine Brown (Blackie, 2nd Edn, 1964).
7. JAEGER, C.: 'Present trends in surge tank design,' Proc. lnstn mech. Engrs,
(1954), 168, pp. 91-103.
8. PEARSALL, I. S.: 'Comparative experiments on surge tank performance,'
9. ZIENKIEWICZ, 0. C., and HAWKINS, P.: 'The transmission of water-hammer
pressures through surge tanks,' Proc. Instn mech. Engrs, (1954), 168, pp.
629-638.
10. ELSDEN, 0.: Surge chambers, Chapter xviii of 'Hydro-electrical engineering
practice,' Ed. J. Guthrie Brown. (Blackie, 2nd Edn, 1964).
11. EvANGELISTI, G.: 'Present trends in surge tank design,' Communication,
Proc. Instn mech. Engrs, (1954), 168, p. 118.
12. ZieMAN, B.: 'Present trends in surge tank design,' Communication, Proc.
lnstn mech. Engrs, (1954), 168, p. 118.
13. MuRILLO, J.: 'Application d'un ordinateur a un problems de chambre
d'equilibre,' Cong. int. Ass. hydraul. Res., (1961), pp. 926-930.
14. DICKINSON, J. C., and GERRARD, R. T.: 'Cameron Highlands hydro-electric
scheme,' Proc. Instn civ. Engrs, (1963), 26, pp. 387-424.
15. JAEGER, C.: Engineering fluid mechanics (Blackie, 1957).
16. FRANKE, P. G.: 'Oscillations in a surge tank-calculation and measurement,'
Cong. int. Ass. hydraul. Res., (1963), pp. 33-36.
17. CALAME, J., and GADEN, D.: Theorie des chambres d'equilibre (Gautier-
Villars, 1926).
18. LEVIN, L.: 'De la determination des pertes de charge dans l'etranglement
des cheminees d'equilibre,' Houille Blanche, (1953), pp. 599-606.
19. MAioNE, U.: 'Perdite di carico vella strozzatura di un pozzo piezometrico,'
Energia elett., (1961), 38, 330-338.
20. ZieMAN, B.: 'Methodes nouvelles pour le calcul des cheminees d'equilibre,'
Houille Blanche, (1953), pp. 580-589.
21. DURAND, W. F.: 'Application of the law of kinematic similitude to the
surge tank problem,' Mech. Engng, (1921), 43, p. 643.
22. GIBSON, A. H.: 'The investigation of the surge-tank problem by model
experiment,' Proc. lnstn civ. Engrs, (1924), 219, pp. 161-173.
23. GmsoN, A. H., and CoWEN, W.: 'A comparison of the results of observ-
ations on surge tank installations and on their scale models,' Proc. lnstn
civ. Engrs, (1933), 235, pp. 327-351.
24. HAYISHI, T.: 'Dynamical similitude of surge tanks,' Trans. Japan Soc. civ.
Engrs, (1959), 61, extra paper 3-2.
25. PICKFORD, J. A.: 'Throttled surge tanks,' Wat. Pwr, (1965), 17, pp. 440-
445.
26. DURAND, W. F.: 'On the control of surges in water conduits,' Trans. Am.
Soc. mech. Engrs, (1912), 34, pp. 319-363.
27. JAKOBSEN, B. F.: 'Surge tanks,' Trans. Am. Soc. civ. Engrs, (1922), 85,
p. 1357.
28. HUDSON, W., and HuNTER, J. K.: 'The Galloway hydro-electric develop-
ment, with special reference to the construction work,' J. Instn civ.
Engrs, (1938), 8, pp. 323-375.
29. BLEIFUSS, D. J.: 'Diversion tunnel and power conduit of Nantahala hydro-
electric development,' Proc. Am. Soc. civ. Engrs, (1949), 75, pp. 1434-
1439.
30. GIBSON, W. L., and SHELSON, W.: 'An experimental and analytical investi-
gation of a differential surge-tank installation,' Trans. Am. Soc. mech.
Engrs, (1956), 78, pp. 925-938.
31. JoHNSON, R. D.: 'The differential surge tank,' Trans. A.m. Soc. civ. Engrs,
(1915), 78, pp. 760-784.
32. RICH, G. R.: Hydraulic transients (Dover, 1963).
33. HALCROW, W. T.: 'The Lochaber water-power scheme,' Proc. Instn civ.
Engrs, (1931), 231, pp. 31-63.
34. EscANDE, L.: 'Overflow-type surge tanks,' Wat. Pwr, (1953), 5, pp. 173-176.
35. BINNIE, A. M.: 'Oscillations in closed surge tanks,' Trans. Am. Soc. mech.
Engrs, (1943), 65, A-183.
36. FuLLARD, S. F.: 'Kiewa hydro-electric scheme,' Wat. Pwr, (1962), 14,
213-221.
37. JAGGER, B. K.: 'The Tamut I Project,' Wat. Pwr, (1960), 12, pp. 169-175.
38. SEDIJATMO, R. M.: 'Progress in the design of chamber surge tanks,' Trans.
4th Wid Pwr Conf, (1950), 4, p. 2410.
39. WATER PoWER: 'Mauvoisin-II,' Wat. Pwr, (1963), 15, pp. 295-302.
40. BATA, G. L., and MADICH, P. B.: 'Solution for multiple surge-tank systems
worked out on repetitive differential analyzer,' Cong. int. Ass. hydraul.
Res., (1961), pp. 1011-1023.
41. TORELL, P-A.: Surges in multiple tanks,' Wat. Pwr, (1966), 18, pp. 485-488,
498.
42. EmsoN, G.: 'Ampliaments del pozzo piezometrica dell 'impianto di Cimego,'
Energia elett., (1963), 40, pp. 226-234.
43. BouvARD, M., and MOLBERT, J.: 'Calcul de Ia cheminee a etranglement de
Ia chute Isere Arc,' Houille Blanche, (1953), pp. 260-281.
44. MosoNYI, E.: Water power development, vol 2, high head plants, etc.
(Hungarian Acad. Science, 1960).
45. PAYNTER, H. M.: 'Methods and results from M.I.T. studies in unsteady
flow,' J. Boston Soc. civ. Engrs, (1952), 39, pp. 120-165.
46. PAYNTER, H. M.: 'Electrical analogies and electronic computers: surge
and water hammer problems,' Trans. Am. Soc. civ. Engrs, (1953), 118,
pp. 962-1009.
47. PISTILLI, G., and SAVASTANO, G.: 'La risoluzione dei problemi di oscilla-
zione di massa con l'impiego della calcolatrici elettroniche,' Energia
elett., (1960), 37, pp. 986-993, 1156-1166.
48. G., and SAVASTANO, G.: 'The study of level's oscillations in cylin-
PISTILLI,
drical surge tanks with the digital differential analyzer of the University
of Naples,' Cong. int. Ass. hydraul. Res., (1961), pp. 775-789.
49. SABUAK, R., and DORAN, W. H.: 'Hydraulic analysis of a double surge
shaft system by digital computer,' Cong. int. Ass. hydraul. Res., (1961),
pp. 639-655.
50. CHRISTIE, I. F.: 'The use of analogue computers for civil engineering
10 Theory of Mass Oscillation
10.1 Fundamental equations

The theory of the oscillation of the whole body of water in the pipeline
and surge tank is based upon three fundamental equations: the dynamic
equation, the equation of continuity and an equation giving the upward
velocity of the water surface in the surge tank.
Some assumptions are made to simplify the theory. We consider the
pipeline as rigid and the water as incompressible. The column of water
in the surge tank itself is neglected and the velocity head in the pipeline
and tank is usually neglected too.
(a) The dynamic equation. Fig. 10.1 shows a pipeline with a single surge
tank of any shape. Newton's second law can be applied to the water
column AB. At any time during the oscillation, whether the oscillation
is caused by the closure or the opening of the valve, the mass acceleration
is equal to the forces acting. That is,
pLA ~v = (the force acting on the column at A) + (the

t component of the weight of the column in
the direction of the pipeline axis) - (the
force acting on the column at B) + (pipe
losses).
The force acting at A is due to the elevation of the water surface in the
reservoir above the pipeline inlet, modified by the entry loss. When the
flow in the pipeline is away from the reservoir the pipeline velocity is
positive. The entry loss reduces the head, and hence the force, at A. When
water flows back into the reservoir, v is negative and the entry loss causes
the head in the pipeline to be greater than that due to the static head.
108
Theory of Mass Oscillation 109
The force acting at B depends on the elevation of the water level in the
surge tank and also on the losses occurring at the entry to the surge tank,
where there may be a restrictive throttle. Hence
dv
pLA dt = pgA{(HA ±entry losses)+ (Hn- HA)- (Hn + y
+ throttle losses) ± (pipe loss)} (10.1)
The sign of the losses depends upon the direction of flow, the loss
always being in the direction of flow. Flow from the reservoir to the
turbine along the pipeline is taken as positive, and upward flow in the
FIG. 10.1
surge tank is positive. The sign of the losses is always correct if we use the
modulus form, hF = FpvjvJ and hT = FTuJuJ, where u is the upward
velocity of the water surface in the surge tank.
Fp is the pipeline friction coefficient and hF is the total head loss in the
pipeline between the reservoir and the base of the surge tank. The loss
hp is made up of:
(a) entry losses;
(b) pipeline friction;
(c) secondary losses caused, for example, by screens and shut-off valves.
(b) and (c) together form the 'pipe loss'.
FT is the throttle friction coefficient related to the upward velocity of the
water surface in the surge tank, u. The throttle loss, hT, is made up of:
(a) head loss at any restriction, such as an orifice plate, at the base of
the tank;
(b) head loss at the tee-junction;
(c) head loss at the expansion as water enters the surge tank. hT may have
different values for upward and downward movement of the water
(for example, for a throttle of the type shown in Fig. 9.3, page 103).
Substituting these values for the losses in equation (10.1), cancelling HA
and H B, and dividing through by the weight of water in the pipeline
(pgLA), we get the first fundamental equation, the dynamic equation,
Ldv
- - + y + Fpvlvl + Fpuiui
g dt
= 0 (10.2)
(b) Continuity. The second fundamental equation is obtained by satis-

fying the conditions of continuity at B. The flow to B from the pipeline
equals the flow away from B up the surge tank and down the penstock.
That is,
vA = uAs + Q (10.3)
(c) Water surface velocity. The third fundamental equation gives

the upward velocity of the water surface in the surge tank in terms of the
elevation of the water surface there above the water surface in the
reservoir, that is,
dy
U=- (10.4)
dt
Combination of the fundamental equations gives second order differ-
ential equations, which can only be solved in special cases_I,2,a
10.2 Simple undamped oscillation

One case which may be solved directly is the oscillation following instanta-
neous (or 'sudden') complete closure in a frictionless system having a
single simple surge tank. Q, Fp and FT are all zero, and As is constant,
and the fundamental equations become,
Ldv dy
gdt + y = 0; vA = uAs; u = dt
Asdy
Hence v=--
A dt
LAs d 2y
and ---+y=O
g A dt 2
If we writeLr for L/g andAr for As/A, this becomesLrAr(d 2y/dt 2) + y = 0.
This is a linear, homogeneous, second order differential equation, whose
general solution is
y = 27T
A cosT t · T
+ B sm 27T
t, where T 1st · d of osc1·n atton.
· h e peno ·
When the oscillation starts, t = 0, and the water surface in the tank is
at the same level as the water surface in the reservoir because there is no
friction. Therefore, when t = 0, y = 0 and A = 0.
B is the amplitude of the undamped oscillation, which is given the
symbol y*, and
. 27T
y =y* sm-yt
If vo is the velocity in the pipeline when t = 0,

27T
v=vocosTt
By differentiation, h = vov(Lr/Ar) (10.5)

T = 27Tv(LrAr)
Fig. 10.2 shows y and v plotted against t
FIG. 10.2
10.3 Logarithmic equations

Sudden complete closure of the valve produces oscillations which often
determine the size of a surge tank. Consequently a simple method of
obtaining surge heights under such conditions is valuable.
The frictionless oscillation investigated in Section 10.2 is of little direct
practical application, since friction losses always occur. However, the
logarithmic method which we now consider is simple and can be applied
quickly, producing theoretical surge heights which are close to those
observed in practice. The method may be used for simple or throttled
tanks, provided that the cross-sectional area of the tank, As, is constant.
Equations developed by Cole4 and Mosonyi5 for a simple (unrestricted)
tank may be written as
Yl - yo = ln _{3_ (10.6)
{3 {3- Yl
where fJ is the damping factor.
LA 1 Lr
[5=--=--
2FpgAs 2 Fp Ar
For throttled tanks6 equation (10.6) becomes
J1 - yo = In fJ - hTo (10.7)
fJ fJ- J1
where hTo is the throttle loss at time t = 0, i.e. immediately after the closure
when the full initial pipeline flow Avo is instantaneously diverted into the
surge tank through the throttle.
fJ = LA =~ _L_r_
2{Fp + FT(A/A 8) 2 }gAs 2 F8 Ar
where Fs = Fp + FT/Ar 2•
Subsequent surges, as shown in Fig. 10.3 may, be obtained by applying
the equation
IYnl + IYn-11 = ln fJ + IYn-11 (10.8)

fJ fJ- IYnl
The sign of y changes for each half-cycle.
ht
I
I
I
I
I
I
FIG. 10.3
Equations (10.6), (10.7), and (10.8) can be solved by trial and error,
using graphical interpolation if required.
Example 10.1
An unrestricted surge tank 10 m diameter is at the downstream end of a 2000 m
long pipeline whose internal diameter is 2·5 m. With steady flow of 30 m 3/s the
level of the water surface in the surge tank is 18·22 m below the level of the
reservoir.
Determine the height of the first two upsurges and the first downsurge if the
valve at the turbine shuts completely and instantaneously.
Surge tankA
L =20oo rn
Lr =203·94
Ar= 16
SOLUTION
d= 2·5m A= 4·909m2
Ar = 16
d.= 10m As= 78·54m2
L = 2000 m Lr = 2000/9·807 = 203·9
yo= -18·22m vo = Qo/A = 30·0/4.909 = 6·112m/s
-yo 18·22
FP = vo 2 = (6.11 )2 = 0·4878
1 Lr 203·9
{J = ZFPAr = 2 X 0·4878 X 16 = 13-065 m
To find y1, we use equation (10.6)
Yl- yo= In _P_. ie y1- 18·22 = 13·065
fJ {1 - y1' . . 13-065 13-065 - y1
(LHS) (RHS)
30·22
Try y1 = 12·0 m LHS = 13- 065 = 2·314
13·065
RHS = I n - - = In 12·27 = 2·507
1·065
29·22
Try y1 = ll·Om LHS = 13- 065 = 2·237
13-065
RHS =In--= In6·326 = 1·845
2·065
Interpolating as Fig. 10.4 (a), y1 = 11·65
29·87
Try y1 = 11·65 m LHS = 13- 065 = 2·286
13·065
RHS =In--= In9·233 = 2·223
1·415
30·02
Try y1 =11-8m LHS = 13- 065 = 2·298
13-065
RHS = I n - - =In 10·328 = 2·335
1·265
Interpolating as Fig. 10·4 (b), y1 = 11·75
29·97
Try y1 = 11·75 m LHS = 13-065 = 2·294
13-065
RHS = I n - - = In 9·935 = 2·296
1·315
(a)
1·845
11·0 11·65 12·0

y,
11·65 11·75 11·80
Fm.10.4
Try y1 = 11·74m LHS = 29 "96 = 2·293

13-065
13-065
RHS = I n - - = In 9·842 = 2·287
1·325
i.e. y1 = 11·75 m
To obtainy~ we use equation (10 8) lynl + IYn-II =In fJ + IYn-II

' . ' fJ fJ - lyn I
. J2 + 11·75 24·81
yn-l = y 1 = 11·75 m, so the equatwn becomes 13- 065 = In 13 .065 _ y 2
19·03
For y2 = 7·28 m LHS = 13·065 = 1-456
24·81
RHS = In--
5·785
= In 4·289 = 1·456
But y2 is negative, i.e. y2 = -7·28 m.
To obtain ya, we use equation (10.8), which with yn-1 = y2 = -7·28 m
becomes
ya+ 7·28 20·34
13-065 - 13-065 - ya
12·58
Try ya = 5·30 m LHS = 13 .065 = 0·963
20·34
RHS = In--
7·765
= In 2·619 = 0·963
i.e., ya = 5·30 m.
Non-dimensional form. Equation (I0.8) may be written as
y; + lnii -~~ = ln (1 + 1Ynp11)- 1Yp11 =Zn-1

or Xn + ln II - Xnl = ln (1 + Jxn-11)- Jxn-11 = Zn-1 (10.9)
where x = yf{J.
For each value of Z there are two roots which differ in absolute value
and which give the limits of two subsequent surges. One of the roots is
positive and gives the initial greater surge; the other root is negative and
gives the smaller surge at the end of the half cycle.
For the first surge, Zo can be found by considering the friction in the
pipeline and any throttle losses.
For a simple tank equation (10.6) may be written
X1 + ln (1 - X1) = Xo = Zo
yo
that is, Zo=-
{J
For a throttled tank equation (10.7) becomes
{J- hTo
X1 - ln = xo = Zo
{J- y1
which may be expressed as
X1 + ln (I - X1) = Xo +k = Zo (10.10)
{J- hTo
So ln = k - ln (1 - XI)
fJ- YI
and k = 1 {{J - hTo fJ - YI} = 1 {J - hTo
n fJ - YI . fJ n fJ
and substituting in equation (10.10)
Z = yo +1 {J - hTo
o fJ n fJ
When hTo is greater than fJ
Zo = yo + ln hTo - fJ
fJ fJ
Solution by successive approximations. Equation (10.9) (i.e. Xn
+ ln (1 - Xn) = Zn-1) may be solved by successive approximations
x2 x3 x4
ln (1 - x) = -x - 2 - 3 - 4 - ...
x2 x3 x4
so Zn-1 = - 2 - 3 - 4 - ...
If x' is an approximate value of yj{J, then the corresponding value of Z
will be Z' where
l(x')2 + i(x')3 + l(x')4 + ... = -Z'
If the deviation of x' is b so that the correct value of x = (x' + b),

then the correct value of Z is given by
i(x' + !5)2 + i(x' + !5)3 + l(x' + !5)4 + ... = -Z
If 8 is the error, i.e. 8 =Zn' - Zn-l, then
8 = o{(x') + (x')2 + (x')3 + ... } + lo2{1 + 2x' + 3(x')2 ...}

+ lb2{1 + 3x' + 6(x')2 + ... } + ..
or 8
x'
= 1 - x' 0
1
+ 2 (1
1
- x')2 2 + ° 1
3(1
1
°
- x')3 3 + ...
o
If the error is small, and is small compared with x', then we can neglect
o
terms of beyond the first order and
0 = (1- x')8
(10.11)
x'
If the error is larger, second order terms can be included to give the
quadratic
1 1 x'
2 (1 - x')2 b2 + (1 - x') b- 8 = 0
Only the positive root has significance when x < 1, so
b = {[(x')2 + 2s]t- x'}(l - x') (10.12)
If s is negative and is so large compared with x' that 2s > -(x') 2 ,
equation (10.12) is complex; the linear equation (10.11) should therefore
be used.
We can choose an arbitrary limit of s for the use of the quadratic form.
For example we might decide to use equation (10.11) if2s < O·l(x') 2 and
equation (10.12) if2s > O·l(x')2.
For the first trial, x' can be made equal to xo if Z > -1. However,
as x--+ 1 the variation of Z is very great for small changes of x, and it
is therefore desirable to start with a value of x which will result in rapid
correction.
The values of x corresponding to large negative values of Z are
z -1 -2 -3 -4 -5
X 0·84142 0·94753 0·98134 0·99322 0·99752
Assuming a linear variation of xfZ between these limits, we can interpolate
to obtain a trial x' from the known Zo.
Example 10.2
Use successive approximations to find the height of the first two upsurges and
the first downsurge following sudden complete valve closure.
Surge tank A
L, "203·94
A," 16
SOLUTION Zo = xo =yo= -l 8'22

= -1·3956
13-065
f3
Try x' = -xo = +1·3956. Then Z1' = x1' +In II- x1'l
Z1' = 1·3946 + In (0·3946) = 1·3946 - 0·9300 = 0·4646
e = Z1' - Zo = 0·4646 - ( -1·3946) = 1·8592
i5 = e(l - Xi') = 1·8592( -0·3946) = _ 0 .5260
The correction
Xi 1 1-3946
The next value of Xi' = 1·3946 + ( -0·5260) = 0·8686
Xi'= 0·8686; Zi' = 0·8686 + ln(0·1314) = -1·1606
B = Zi'- Zo = -1·1606- (-1·3946) = 0·2340
ij = 0·2340 X 0·1314 = 0·0 354

0·8686
Xt' = 0·9040; Zi' = 0·9040 + In (0·0960) = -1·4390
B = Zi' - Zo = -1·4390 - ( -1-3946) = -0·0444
B = -0·0444 X 0·0960 = _ 0·0047

0·9040
Xt' = 0·8993; Zt' = 0·8993 +In (0·1007) = -1-3957
B = Zi' - Zo = -1·3957 - ( -1·3946) = -0·0011
i5 = -0·0011 X 0·1007 = -0·000 1

0·8993
Xi'= 0·8992; Zt' = 0·8992 +In (0·1008) = -1·3948
B""'O
Therefore
Yi =Xi x {3 = 0·8992 x 13·065 = 11·75 m
To obtain x2 (and hence y2), Zi = In (1 + Xi) - Xi

= In (1·8992) - 0·8992
= 0·6415 - 0·8992 = -0·2577
Try X2 1 = Xi = 0·8992
(Note that the values of x and y change sign for each half cycle, and that we
are finding, in effect, lx21)
x2' = 0·8992; Z2' = x2' +In (1 - x2') = 0·8992 +In (0·1008) = -1-3948
B = Z2 1 - Zi = -1·3948- (-0·2577) = -1·1371
i5 = -1·1371 X 0·1008 = _ 0 . 1277
0·8992
The next value of X2 1 = 0·8992 + (-0·1277) = 0·7715
Continuing in the same way, we can tabulate
X2 1 Z2' B i5
0·7715 -0·7048 -0·4471 -0·1324
0·6391 -0·3801 -0·1224 -0·0619
0·5700 -0·2740 -0·0163 -0·0123
0·5577 -0·2581 -0·0004 -0·0003
0·5574 -0·2577 0
Therefore y2 (which is negative) = -Jx2J x fJ = -0·5574 x 13-065 = -7·28 m
Working on, we obtain xa = 0·4053
ya = xa X fJ = 0·4053 X 13-065 = 5·30 m
Successive approximations of this form can easily be carried out on a digital
computer with great accuracy.
10.4 Logarithmic curves

The function Z = x + ln (1 - x) used in equation (10.9) is independent
of the dimensions of the oscillating system and can be plotted as Fig. 10.5.
l
+2
I
+I
z -XI 0 Xz XI
I
0
~
~---
/
~- p
Zo I
-2 A B
-2 -I 0 +I +2
X
FIG. 10.5
As we have already found, for the first surge
Zo = xo + ln {1,8 -/Tol}
When hTo is greater than ,8 the positive curve B-B, for x > 1·0, is used.
The first (maximum) upsurge y 1 is obtained from the positive curve
A-0 (or B - B if hTo > ,8) by finding the abscissa x corresponding to the
known value of Z 0 • For example Zo in Fig. 10.5 gives point P1o which
determines x1, as shown.
To obtain the next surge height, y2, we locate Z1 by drawing the abscissa
at -x1 to the negative curve C-0 at N1, as shown in Fig. 10.5. Then we
project across to the positive curve A-0 at P2. x2 is the abscissa at P2.
Thus N1 and P2 are the roots of Z1.
The construction of the complete surge pattern can be simplified by
drawing the mirror image of the positive curve on the negative side, as
shown in Fig. 10.6.
(a)
X
x,o0-405 x4o0·319
x3 x4
0
Negafive c1urve :
~N2
r-r
~z3 o-0·065
0·1
~3 Z2o-0·114
c~
0·2
-----ffi-0·258 z
P2 I -0·3
~trror 1m age) posif1ve curve 04
A /
-0·7 -0·6 -0·5 -0-4 -0-3 -0·2 -0·1 0
-0·5
(b)
FIG. 10.6
Example 10.3
Use logarithmic curves to find the height of the first seven surges in surge
tank A following sudden complete valve closure.
SoLUTION See Fig. 10.6

yo -18·22
Zo = xo = - = - - - = -1·395
fJ 13-065
In Fig. 10.6 (a) the ordinate Zo = -1·395 cuts the positive curve A-0 at P1,
giving the abscissa x1 = 0·899.
Hence y1 (which is positive) = flx1 = 13·065 x 0·899 = 11·75 m
The abscissa x1 = 0·899 cuts the negative curve C-0 at N1, giving the ordinate
Z1 = -0·258. Z1 in turn gives point P2 on the positive curve A-0. Since
point P2 gives x < 0·7, we can go to Fig. 10.6 (b) for greater accuracy.
Surge tank A
Locating P2 at the intersection of Z1 = -0·258 and the curve A-0, we obtain

the abscissa x1 = 0·557.
Hence y2 (which is negative) = - fJx2 = -13-065 X 0·557 = -7·28 m.
Continuing from X2 = 0·557 we get point N2 on curve C-0, where Z2 =
-0·114. Hence we go to Pa on A-0 yielding xa = 0·405
ya (positive) = {Jxa = 13·065 X 0·405 = 5·30 m
Example 10.4 shows how the logarithmic method is applied to throttled

surge tanks.
Example 10.4
Surge tank B is similar to surge tank A except that there is a 1·8 m diameter
throttle at the base giving a loss of 50 u 2 • The pipeline and steady flow conditionr
are the same as for tank A.
Determine the height of the first two upsurges and the first downsurge following
complete valve closure.
Surge tank B
Lr=203·94
Ar= 16
SOLUTION
Fp = 0·4878 s2fm; F8 = FP + FT/Ar 2 = 0·4878 + 50/16 2 = 0·6831 s 2/m

Lr 203·94
f1 = 2F8 Ar = 2 X 0·6831 X 16 = 9"3296 m
50 X (6·112)2
hTo = FTuo 2 = FTvo 2 /Ar 2 = 16 2
= 7·2961 m
yo {1- hTo -18·22 9·3296 - 7·2961

Zo = {i +In -{1- = 9·3296 +In 9·3296
= -1·9530 + ln0·2181 = -3·4759

Zo < -1 For the first trial obtain x' by interpolation
i.e. = 0·9813 + (0·9932 - 0·9813) x 0·476
x'
= 0·9813 + 0·0119 X 0·476 = 0·9813 + 0·0057 = 0·9870
Z1' = x1' + In (1 - x') = 0·9870 + In 0·0130 = -3·3553
e = Z1' - Zo = -3·3553 + 3·4759 = 0·1206
-0·1 (x1') 2 < 2e <-(x1) 2; therefore use quadratic equation (10.12).

The correction
X1' = 0·9885; Z1' = -3·4764; e = -0·0006
2e < 0·1 (x1')2: correction by equation (1 0.11)
0 = e(l - Xl') = -0·006 X 0·0115 = Q·QQQQ
X1' 0·9885
Hence y1 = 0·9885 X 9·3296 = 9·222 m
To obtain x2, Z1 = In (1 + x1) - x1 = -0·3011
For the first trial assume x2' = x1 = 0·9885, and then proceed as follows:-
x' Z' e Eqn. 0
0·9885 -3-4759 -3-1747 10·11 -0·0370

0·9515 -2·0750 -1·7739 10·11 -0·0904
0·8611 -1·113I -0·8119 10·11 -0·1309
0·7302 -0·5798 -0·2787 10·11 -0·1030
0·6272 -0·3595 -0·0584 10·12 -0·0378
0·5894 -0·3008 +0·0003 10·11 +0·0002
0·5897 -0·3011 0·0000
Hence y2 = -0·5897 X 9·3296 = -5·501 m

To obtain xa, Z2 = ln (1 + x2) - x2 = -0·1261; hence
0·5897 -0·3011 -0·1750 10·11 -0·1218

0·4679 -0·1630 -0·0369 10·12 -0·0462
0·4217 -0·1259 +0·0002 10·11 +0·0003
Hence xa = 0·4220 and ya = 0·4220 x 9·3296 = 3·9366 m.

When hTo > {J, Xl is greater than 1. When -yo/ fJ is also greater than 1 and
Zo is greater than -1, there is no difficulty in proceeding from x 1' = -yo/{J,
as in Example 10.4. However, if Zo < -1, it is necessary to choose a more
accurate first trial for x1', because of the rapid change of Z with small changes of
x as x ->-1.
The first trial may be obtained by interpolation in the same way as when xis
slightly less than 1, but in this case, where x is slightly more than 1, the
corresponding values of x and Z are:-
z 0 -1 -2 -3 -4 -5
X 1·27846 1·12003 1·04748 1·01799 1·00669 1·00247
Example 10.5
Use the method of successive approximations to determine the first two upsurges
and the first downsurge following sudden complete valve closure in surge tank C.
Surge tank C
051 ~ 30 m 3 /s
SOLUTION FT = KT. (~;)

2g
2
; As=
Ar
(d•)
dT
2 = 100; FT = 509·85s 2/m
- FT - 509·85 - 2
Fs - Fp + Ar2 - 0 488 + (l 6) 2 - 2 479 s /m
0 0
Lr 203·94
fJ = 2FsAr = 2 X 2·479 X 16 = 2' 570 m
vo 2 (6·112)2
hTo = FTuo 2 = FT- = 509·85 - - - = 74·388 m
Ar2 (16)2
z =yo= 1 !fJ- hTol = -18·22 1 71·818 = _ 3.7582

0 fJ n fJ 2·570 + n 2·570
For the first trial interpolate between the values of x for Z = -4 and Z = -3.
Hence x1' = 1·00942.
x' Z'
1·00942 -3·6552 0·1030 Quadratic -0·00092

1·00850 -3·7585 0·0003 Linear 0
Hence y1 =
1·0085 x 2·570 = 2·59 m
Z1 =In (1 + x1) -x1 = -0·3111
For the first trial assume x2' = (2 - XI) = 0·9915, since X2 is always less than
unity.
0·99150 -3·7752 -3·4641 Linear -0·02973

0·96176 -2·3022 -1·9911 Linear - 0·07916
0·88260 -1·2596 -0·9485 Linear -0·12616
0·75645 -0·6560 -0·3449 Linear -0·11103
0·64541 -0·3914 -0·0803 Quadratic -0·04944
0·59597 -0·3103 +0·0082 Linear + 0·00056
0·59653 -0·3111 0
Hence y2 = -0·5965 x 2·570 = -1·53 m

Z2 = In (1 + x2) - X2 = -0·1287
For the first trial assume xa' = X2 = 0·59653
0·59653 -0·3111 -0·1824 Linear -0·12338

0·47314 -0·1677 -0·0390 Quadratic - 0·04804
0·42542 -0·1287 0
Hence ya = 0-4254 x 2·570 = 1·09 m.
10.5 Stability
If the oscillations set up by valve movement are damped, as in Fig. 10.7,
the system is stable.
FIG. 10.7
When constant power is required, the flow is reduced at upsurge and
increased at downsurge since power is the product of head and flow. We
will examine this type of system in Chapter 12. Under certain conditions,
as when the area of the surge tank is too small, the amplitude of the oscilla-
tions increases. For example, the second downsurge may be lower than
the first. The system is then unstable.
According to the conditions derived by Thoma, 7 the minimum area of a
simple tank for stability is ATh (Thoma's critical area) defined as
LrA
ATh=--- (10.13)
2FpHf
Ht is the head available at the turbines at final steady condition, i.e.
Ht = Hs- Yt·
The derivation of Thoma's condition of stability is based on oscillations
of small amplitude.
When deciding on the minimum cross-sectional area of a surge tank it
is usual to introduce a factor of safety, n, so that As ~ n ATh, n being
greater than 1. The factor of safety may be between 1·5 and 1·8, although
larger values have often been used. Stability must be satisfied for all
portions of a variable area tank, including any narrow shaft connecting
expansion chambers.
Modifications to Thoma's critical area

(a) For oscillations with large amplitude Jaeger 8 suggests the factor
of safety n*,
n* = As
ATh
= (1 + 0·482 Y*)
Ht
(10.14)
(b) The velocity head in the pipeline at the base of the surge tank
improves the stability-i.e. it makes the minimum or critical area
less. In some tanks the cross-sectional area of the pipeline is reduced at
the base of the tank with a venturi-type contraction, like that shown
in Fig. 10.8. This increases the velocity head and helps the stability.
FIG. 10.8
If AB is the pipeline area at the base of the tank, the minimum
surge tank cross-sectional area may be reduced by the factor
Fp
whereB=--
(:J2
Fp +B 2g
(c) Stability is affected by variation of efficiency of the turbines. If, for
example, the load is increased and the efficiency rises with increase
of load, then the stability is increased. In the range beyond maxi-
mum efficiency, increase of power results in a drop of efficiency
and stability is reduced. See Fig. 10.9. If the surge tank has to deal
with such power increases beyond the duty point, the minimum
area must be greater than the Thoma critical area.
FIG. 10.9
(d) The statism of the electricity network and the mechanical statism
of the turbine governor also influence the stability. The power
station supplied by the hydraulic system with the surge tank under
consideration may feed a network which is also fed by other stations
without surge tanks. Often thermal plants, including nuclear
stations, are included in the same network as a hydro-electric
station. In these circumstances the stability is increased.
If the power supplied by the hydro plant is less than a third of the
total power supplied to the network, there is no danger of instability,
whatever the area of the surge tank.
(e) When the oscillations in a differential tank are very small, the water
level in the riser and the water level in the tank remain close, and
the total area of the riser and the tank may be considered for
stability purposes. However, when the oscillation is large the
amplitude in the riser will be greater than that in the tank. For
stability the effective area is larger than the area of the riser but not
so great as the sum of the riser and the tank area.
(f) When a system includes multiple tanks Thoma's analysis can be
adapted provided the oscillation is small. In the case of two tanks,
one upstream of the turbines and the second on the tailrace tunnel
(Fig. 10.10), the critical area of the upstream tank is greater than it
would be if there were no downstream tank.
J
FIG. 10.10
(g) Except in the case of simple tanks, stability may best be tested by
computing surge heights by finite difference methods as in Chapter
12. If ya is less than y1 for all expected flows, the tank is stable. If
ya is greater than y1 at any expected change of flow, the tank is
unstable.
REFERENCES
1. KRYLOFF, N., and BOGOLIOBOFF, N.: Introduction to non-linear mathematics
(Princeton Univ. Press, 1943).
2. MILNE, W. E.: Damped vibrations (Univ. Oregon, 1923).
3. REISMAN, A., and SILVERS, A.: 'On a non-linear differential equation common
to several branches of hydraulics,' J. Hydrology, (1967), 5, pp. 171-178.
4. COLE, R. S.: 'The surge-chamber in hydro-electric installations: methods
of calculation,' Instn civ. Engrs, (1927), selected engng papers No. 55.
5. MosoNYI, E.: Water power development, vol 2, high head plants, etc.
(Hungarian Acad. Science, 1960).
6. PicKFORD, J. A.: 'Surge tank design by logarithmic curves,' Wat. Pwr,
(1965), 17, pp. 397-400.
7. THOMA, D.: Ziir Theorie des Wasserschlosses bei selbsttiitig geregelten
Turbinenaulagen (Oldenbourg, 1910).
8. JAEGER, C.: Engineering fluid mechanics (Blackie, 1959), p. 239.
9. EscANDE, L., and HuRON, R.: 'Stabilite de deux chambres d'equilibre
respectivement solidaires des canaux d'amenee et de fruite,' Houille
Blanche, (1953), pp. 647-654.
a
10. MEYER, R.: 'Conditions analogues celles de Thoma pour une installation
hydroelectrique ayant une cheminee d'equilibre a l'amont et une autre a
l"aval des turbines,' Houille Blanche, (1953), pp. 640-646.
FURTHER READING
EscANDE, L.: 'The stability of throttled surge tanks operating with the electric
power controlled by the hydraulic power,' Hydraul. Res. (1963), 1, pp. 4-13.
EscANDE, L., and CLARIA, J.: 'Abaques caracteristiques pour Ia stabilisation
des cheminees d'equilibre par emploi de resistances Iiquides,' C. r. Acad.
Sci., Paris, (1961), 253, pp. 599-602.
JAEGER, C.: 'De la stabilite des chambres d'equilibre et des systemes des chambre
d'equilibre,' Schweiz. Bauztg, (1943), 122, pp. 255-257, 297-300, 314-317,
323-325.
JAEGER, C.: 'Present trends in surge tank design,' Proc. Instn mech. Engrs,
(1954), 168, pp. 91-103.
JAEGER, C.: 'A review of surge tank stability criteria,' J. bas. Engng, Trans. Am.
Soc. mech. Engrs, (1960), 82, pp. 765-783.
KAICHEV, P.: 'De la stabilite hydraulique des chambres d'equilibre dans cer-
tains cas complexes,' Houille Blanche, (1960), pp. 678-689.
MosoNYI, E., and NAGY, L.: 'Stability investigations by computer,' Wat. Pwr,
(1964), 16, pp. 312-314.
PAYNTER, H. M.: 'Electrical analogies and electronic computers: surge and
water hammer problems,' Trans. Am. Soc. civ. Engrs, (1953), 118, pp. 962-989.
STUCKY, A.: Cours d'amenagement des chutes d'eau. Chambres d'equilibre
(Ecole Polyt. Univ. Lausanne, 1951).
VALVIS, P. G.: 'Location of a surge tank in relation to speed regulation,' Wat.
Pwr, (1967), 19, pp. 463-468, 503-506, 514.
ZIENKIEWICZ, 0. C.: 'Stability of parallel-branch and differential surge tanks,'
11 Finite Difference Methods I
11.1 Outline of the methods

With the aid of electronic computers we can easily calculate the behaviour
of oscillating water in even the most complex surge tank systems by using
finite difference methods. Consequently these methods are of great
importance.
The principle is one of arithmetic, step-by-step or iterative integration.
The basis is the fundamental differential equations which we obtained in
the last chapter as equations (10.2), (10.3) and (10.4). Small 'steps' of
time, I:J.t, are taken, and the variation of other parameters during this
time interval can be calculated. These parameters vary during time I:J.t by
incremental amounts, which can be expressed as functions of I:J.t. Thus
the variations of Q, v, u andy during time I:J.t are I:J.Q, I:J.v, !:J.u and !:J.y.
Since the variation of any one parameter during time I:J.t depends on
the variation of other parameters, the mean value of these other parameters
during time I:J.t may be used. In addition, the cross-sectional area of the
surge tank may vary with elevation, and if the variation is continuous (as
in a conical tank) the mean value of As may be taken.
The flow taken by the turbines may vary with time, and it may be
dependent on the head, H, as when a governor controls the power to be
taken by the turbines. Power is proportional to flow multiplied by head,
and the head at the turbines depends on the water-level in the surge tank.
The available head also affects the flow when flow-control is by change of
flow-area as where a spear-valve is used with a Pelton wheel.
The mean values of H, As, Q, v, u andy during time I:J.t are given by
the symbols H, As, Q, v, ii and ji. Several parameters are changing at the
same time and trial methods are therefore necessary to determine the
mean values. This is usually tedious and in most cases a better method
is to reduce the time interval, M, and to take conditions at the beginning
of the step as prevailing throughout. The error resulting from this assump-
tion obviously reduces as I:J.t becomes smaller. Escandel has introduced a
129
method, which we will consider in Section 11.6, where mean values
of rapidly-changing parameters are taken with initial values of other
parameters; this results in improved accuracy without the need for trial.
11.2 Fundamental finite difference equations

Using the notation described above, equations (10.2), (10.3) and (10.4) are
rewritten as
~v
Lr ~t + y + Fpvlvl + FTiilul = 0 (11.1)
_ ii As+ Q
v=---= (11.2)
A
- ~y
u=- (11.3)
~t
It should be noted that where the tank area A 8 varies with the elevation,
FT also varies.
Equation (11.2) and (11.3) may be combined to give
~y - Q
ii = -Ar
/).t
+-
A
(11.4)
where
11.3 Successive trials (Pressel's method) 2

Equations (11.1) and (11.4) are used with the assumption that there is a
linear change of y and v during a short time interval ~t. Therefore
ji = y + !~y and v = v + t~v, y and v being the values of the beginning
of the interval.
For instantaneous complete closure with a simple tank, Ar is constant,
FT = 0 and Q = 0. During the first quarter-cycle equations (11.1) and
(11.4) become
~t
~v = -{y + i~Y + Fp(v + t~v) 2 } ( 11.5)
Lr
~t
~y = (v + i~v)-
Ar
(11.6)
Finite Difference Methods I 131
An arbitrary value of tl.t is chosen. Then, for each step, corresponding
values of tl.y and tl.v are estimated. The accuracy of these estimates is
then checked by substitution in equations (11.5) and (11.6); if the estimate
is found to be inaccurate, the procedure must be repeated. When satis-
factory values of tl.y and tl.v have been found, y and v for the next step are
obtained from the equation
Yn = Yn-1 + tl.y Vn = Vn-1 + tl.v
11.4 Simple arithmetic method

In this method the values of the variables other than t are taken as those
which occur at the beginning of the time interval !lt. Thus, for any step,
y, v, u, As, Fp and Q are taken as the initial values. Equations (11.1) and
(11.4) become
(11.7)
vA- Q
tl.y = = tl.t (11.8)
As
Provided the time interval tl.t is sufficiently small, this method is

reasonably accurate, and has the advantage of being straightforward.
In this section and in Chapter 12 we consider a number of cases of
increasing complexity.
(a) Sudden complete closure-simple tank

Example 11.1
Find the first upsurge in large tank A following sudden complete closure.
Surge tank A
· - - · - - · d5 =10m·-·
A5 =78·54m 2
L, = 203·94
A,= 16
SOLUTION
In Table 11.1, M is taken as 10 seconds, and therefore in the first quarter-
cycle equations (11. 7) and (11.8) become
10 10
Av = -(y + 0·4878v2) X - - and ily=-v
203-9 16
TABLE 11.1
Column 2 4 5 6 7 8
I!J.t I!J.y y FPv2 = y + FPv2 I!J.v
0·625v 0·4878v2 = 0·04905 X
column (7)
seconds seconds m/s m m m m mfs
Line Line
1 0·0 vo = 6·111 YO= -18·22 18·22 0·00 1
2 10·0 3-82 -0·000 2
3 10·0 6·111 -14·40 18·22 3·82 3
4 10·0 3·82 -0·187 4
5 20·0 5·924 -10·58 17·12 6·54 5
6 10·0 3·70 -0·321 6
7 30·0 5-603 -6·88 15·32 8·44 7
8 10·0 3·50 -0·414 8
9 40·0 5·189 -3·38 13-14 9·76 9
10 10·0 3·25 -0·479 10
11 50·0 4·711 -0·13 10·83 10·69 11
12 10·0 2·94 -0·524 12
13 60·0 4·187 +3-81 8·55 11·36 13
14 10·0 2·62 -0·557 14
15 70·0 3·630 5·43 6·43 11-86 15
23 110·0 1·208 12·28 0·71 12·99 23

24 10·0 0·75 -0·637 24
25 120·0 0·571 13·03 0·16 13-19 25
26 10·0 0·36 -0·647 26
27 8·8 0·32 -0·571 27
28 128·8 0·000 13·35 28
Thus by taking IJ..t as 10 seconds, which involved 13 steps, we found the

value of y1 to be 13·35 m.
If !J..t = 5 seconds, there are 27 steps and y1 = 12·52 m
If IJ..t = 0·05 seconds, there are 2692 steps and y1 = 11·76 m
The percentage error of yr, when the simple arithmetic method is com-
pared with y1 obtained by the logarithmic method (y1 = 11·75 m) is
listed in Table 11.5. The calculation can be continued beyond Jl· The
water level in the surge tank now falls, dropping below the reservoir level.
The water coming from the surge tank is unable to pass down the penstock
to the turbines because the valve is completely closed. It therefore goes
back along the pipeline to the reservoir with a negative velocity v. The
maximum downsurge is reached when the velocity in the pipeline and the
surge tank again become zero.
Table 11.2 gives the first few steps to calculate the first downsurge y2.
You will notice that when vis negative the pipeline friction Fpvlvl is also
negative.
TABLE 11.2
tJ..t v ~y y FPDJvJ y + FPvJvl ~v
128·8 0·000 13-35 0·000 13·35

1·2 0·00 -0·073
130·0 -0·073 13-35 -0·00 13·35
10·0 -0·05 -0·655
140·0 -0·728 13-30 -0·26 13·04
10·0 -0·46 -0·640
150·0 -1·368 12·84 -0·91 11·93
10·0 -0·86 -0·586
160·0 -1·954 11·98 -1·86 10·12
The error resulting from too large increments of time, !::it, becomes
greater with surge heights after the first, as you can see from these figures.
e is the percentage error compared with the heights obtained by the
logarithmic method.
tJ..t e for y1 e for y2 e for ya
10 13-6 32·9 51·9

5 6·6 15·3 23-8
0·5 0·66 1·50 2·28
0·05 0·08 0·16 0·24
Comparison for other values of !::it is given in Table 11.5.
""""'
v v Surge tank A
-10
"
'"' v
Eoi~O >~ v
I
1'"'-. v
---r--1
""'-"'"
I",
I 10 100 1000"' 10.000
-0 IT-I --~-;-;---~----=-~,·
n number of steps v
FIG. 11.1
The importance of choosing a suitable time interval !1t cannot be too
strongly emphasized.
The error found for y1 following sudden complete closure in surge
tank A is shown in Fig. 11.1. It can be
seen that s is approximately equal to 100/n,
where n is the number of steps.
If the calculation is done by slide rule
or by the use of logarithmic tables, the
work involved in obtaining accurate surge
A ~ 0·7854xDP 2
heights is very tedious and time-consuming.
AS ~ 0·7854x DS 2
AR~AS/A
There is a high risk of errors in the arith-
LR L/G
~
metic. Electronic digital computers make
¥~~~6~vo 2 light work of these calculations and are free
from the possibility of arithmetic errors.
For example, the calculation of ya in steps
of 0·05 seconds for surge tank A mentioned
above involved about 10 000 steps: t for
ya is 499·045 seconds. Each step, of course,
requires the calculation of v, y, Fpvlvl,
y + Fpvlvl, !1v and !1y, but this is a simple
matter for the computer, which can quickly
deal with the more complicated problems
discussed in Chapter 12.
Fig. 11.2 shows the flow diagram for a
Fortram programme for the most simple
case-the first upsurge following sudden
complete closure in an unrestricted surge
DTR DT V/DV
~ X tank of uniform cross-sectional area.
DYR~ DTRxV/AR
DVR~-DTRxSUMILR
+
T ~ T DTR
Y~Y+DYR (b) Sudden complete closure: throttled tank of
constant area. We can easily allow for the
v~V+DVR
effect of a restriction at the base of a surge

tank by including the term Fpulul in the
FIG. 11.2 calculation of !1v by equation (11.7).
Example 11.2
Use the simple arithmetic finite difference method to obtain the height of the
first upsurge in surge tank C following sudden complete valve closure.
SoLUTION FT
(~:r
= KT -2g- = 509·85 s2/m
The calculation is shown in Table 11.3.
Surge tank C
Yst=-18·22m
dr= lm
Kr=I·O
051 =30 m3/s
TABLE 11.3
SUM
flt fly y flu u v FPvlvl FTuiui (4) +
(8) +
(9)
Line (I) (2) (3) (4) (5) (6) (7) (8) (9) (10)
1 0·00 YO= -18·22 0·382 vo = 6·11 18·22 74·39 74·39

2 10·00 3-82 -0·288
3 10·00 -14·40 0·154 2·46 2-96 12·09 0·65
4 10·00 1·54 -0·002
5 20·00 -12-86 0·152 2·43 2-89 11·78 1-81
6 10·00 1·52 -0·006

7 30·00 -11·34 0·146 2-35 2·68 10·94 2·28
8 10·00 1-47 -0·007
9 40·00 -9·87 0·139 2·23 2·43 9·92 2·48
10 10·00 1·39 -0·007

11 50·00 -8-48 0·132 2·11 2·17 8·87 2·56
* * *
* * * * • • •
* * *
39 190·00 2·61 0·017 0·27 0·04 0·15 2-80
40 10·00 0·17 -0·009
41 200·00 2·78 0·008 0·14 0·01 0·04 2-83
42 9·17 0·09
43 209·17 2-87
In line I columns (1), (4), (7) and (8) are the same as the corresponding
parameters in Table 11.1.
Column (6) is obtained from the equation u = vo/Ar.
In column (9) FTuiul = 509·85 X (0·382) 2 = 74·39 m.
Column (10) = y + Fpvlvl + FTuiui.
In line 2 column (5) is obtained from column (10), line 1
column (3) is obtained from column (6), line 1
In line 3 column (4) is the sum of column (4) line 1 and
column (3) line 2
column (6) is the sum of column (6) line 1 and
column (5) line 2
column (7) is obtained from column (6); v = uAr.
11.5 Jakobsen's methods

For a simple tank and instantaneous complete closure, equation (11.5)
may be expressed in a different form as
Lrflv = -{y + !fly+ Fp(v!v! + !v!flv + !flv 2)}flt

If !flv2 is neglected, this becomes
Lrflv = -{y +!fly+ Fp(v!v! + !v!flv)} flt

When the value for fly given by equation (11.6) is substituted, we obtain
Lrflv = - {y + h(v!vl + lvlflv) + l(v + !flv) ~:} flt
Collecting all terms in flv in the left-hand side,
or (11.9)
The value of flv obtained from equation (11.9) is used in equation (11.6)
to obtain fly.
Table 11.5 shows the Jakobsen Method to be more accurate than the
simple arithmetic method by comparing the surge heights for tank A.
11.6 Escande's method1

If you examine Fig. 11.3 you notice that at
the beginning of oscillation there is a rapid
change of y and a much more gradual
change of v. On the other hand when
maximum upsurge is approached v changes
y
quickly and y changes gradually. Similar
alterations in the rate of variation of y and v
occur every quarter-cycle.
In his D.P. method (Methode Difference
FIG. 11.3 Finie) Escande used this characteristic by
taking the initial value of the slowly-changing parameter and the mean
value of the rapidly-changing parameter. Thus
when
1;1>1~v~' ji = y + !L\y, v=v
when
~~vi> 1;1. v = v + !L\v, ji=y
The method can be applied to all types of surge tank problem, including
the more complicated cases considered in Chapter 12. However, we have
used the Escande Method for sudden complete closure of an unrestricted
tank only. You may like to adapt the simple arithmetic method used in
Chapter 12 to allow for Escande's modification.
Example 11.3
Use Escande's Method to determine the height of the first upsurge in Surge
Tank D following sudden complete valve closure.
SOLUTION
TABLE 11.4
y
Line (1)
/';.t
(2) (3) (4)

/';.y
(5)
y
(6) (7)
FPvlvl y+FPvlvl
(8) (9)
/';.v
(10)
111
(II)
1~"1
(12)
1 0·00 5·000 -10·00 10·00 0·00
. . .
2 5·00 2·50 0·000 0·250 0·000
3 5·000 2·50 -8·75 10·00 1·25 -0·062
4 5·00 4·938 -7·50 9·75 2·25
25 40·00
•
2·453 6·74 2·41
•
9·15
. .
26 5·00 1·23 -0·457 0·182 0·186
27 2·224 1·11 6·74 1·98 8·72 -0·436
28 45·00 2·017 7·85 1·63 9·48
• •
0·01
•
10·00
40 65·00 0·113 9·994
41 5·00 0·06 -0·500 0·006 4·424
42 1·13 0·013 -0·113
43 66·13 0·000 10·007
Take M = 5·0 seconds, and tabulate as in Table 11.4.

Line 1 gives the initial conditions. The term for pipe friction FPDo 2 balances
the term for the elevation of the surge tank water level yo and consequently in
line 2 ilv is zero. Column 5, line 2 is ily = vo X ilt/Ar = 5·00 X 5·0/10·0 =
2·50. Column 11 in line 2 is ily given by column 5, line 2, divided by y given by
column 6, line 1, i.e. ily/y = 2·50 -10·0 = -0·250. Similarly column 12 in
line 2 gives ilv from column 10, line 2, divided by v from column 3, line 1;
ilvfv = 0·0/5·0 = 0·0. Iily/yl > I vfvl and therefore in line 3, y = y + tily = -
10·00 + t x 2·50 = -8·75; v = v = 5·00 m/s. Column 8, line 3 is still 10·00
since v = v, but column 9, line 3 is now y + Fpv 2 , = 1·25 and ilv in column 10,
line 3 = -(j + Fpv 2)ilt/Lr = -1·25 x 5·0/100 = -0·625 m.
The table is continued in this way until line 26, where I ily/yl < I ilvfvl.
Therefore, in line 27, y in column 7 is the same as y in column 6, line 25, i.e.
6·74 m. v in line 27, column 4, is the sum of v in column 3, line 25, and half of
ilv in column 10, line 26, i.e. 2·453 + t (-0·457) = 2·224 m/s. For the
remainder of the table, I ily/yl < I ilvfvl, soy= y and v = v + tilv.
In line 41, ilv = -0·500, so at timet= 70 seconds v = v + ilv = 0·113-
0·500 = -0·387. Therefore by t = 70 seconds the water in the pipeline is
moving backwards towards the reservoir, and the maximum upsurge has been
passed.
The reduced time interval ill = 1·13 seconds in column 2, line 42, is obtained
by proportion. ilv in column 10, line 41, is reduced to -0·113 m/s in column
10, line 42, so that v in column 3, line 43, is zero.

Calculation of pipe friction is usually based on the 'v2-law', i.e.jis constant
in the Darcy equation HF = (4fL/d)(v 2f2g). This is true only for high
Reynolds Number and rough pipes-the 'rough turbulent' zone is shown
on Fig. 11.4. The Colebrook-White formula 4 •5 covers the whole range of
0 020
20
50
'- 1>0 ~
c
Q "»0 tOO
"
~
u ' /v,...6
.E .Q
200 "§
0
c
""""J'
500 "'
~
u
0·005
"'"'
c
..c
Q;
1000 CJ>
:J
0
u 0 004 2000 0
a:
5000
0·003 10000
20000
10 3 10 4
lieynolds number, Re
FIG. 11.4
turbulent flow, and the Wallingford charts and tables6,7,s are based on
this formula. Wallingford charts and tables can be used to compute the
friction loss in the pipeline.
Alternatively the loss may be taken as proportional to vn where 'n' has
a value other than 2. 9,lo A typical equation is HF = Fp'vLBl or HF =
Fpv!v!o.st. Finite difference methods may be used to calculate the surge
height using such values of hF.
TABLE 11.5. Comparison of finite difference methods
!J.t Simple arithmetic Jacobsen Escande !J.t

n y s n y s n y s
seconds steps m % steps m % steps m % seconds
10·00 13 13·351 13-6 14 11·821 0·72 13 12·587 7·15 10·00

5·00 27 12·521 6·59 27 11·767 0·17 27 12·261 4·37 5·00
2·00 67 12·053 2·61 68 11·751 0·04 67 11·964 1-85 2·00
Yl 1·00 134 11·900 1-31 135 11·749 0·02 134 11-858 0·94 1·00
0·50 269 11·824 0·66 270 11·748 0·01 269 11·803 0·48 0·50
0·50 673 11·778 0·27 674 11·748 0·01 673 11·769 0·19 0·20
0·10 1346 11·763 0·14 1347 11·748 0·01 1346 11·758 0·10 0·10
0·05 2692 11·756 0·08 2692 11·748 0·01 2692 11·752 0·05 0·05
10·00 31 -9-678 32·9 32 -7·311 0·39 31 -9·112 25·1 10·00

5·00 63 -8·397 15-3 64 -7·299 0·23 63 -8·211 12·7 5·00
2·00 158 -7-724 6·06 159 -7·284 0·01 158 -7-670 5·31 2·00
Y2 1·00 317 -7·502 3·01 318 -7·284 0·01 317 -7-479 2·68 1·00
0·50 635 -7·392 1·50 636 -7·284 0·01 635 -7·381 1·35 0·50
0·20 1589 -7·327 0·61 1590 -7·284 0·01 1589 -7-322 0·55 0·20
0·10 3179 -7-305 0·31 3180 -7·284 0·01 3179 -7-302 0·27 0·10
0·05 6358 -7·295 0·16 6359 -7·284 0·01 6358 -7·293 0·13 0·05
10·00 49 8·044 51-9 51 5·326 0·59 49 7·550 42·6 10·00

5·00 99 6·557 23·8 100 5·305 0·18 99 6·417 21·2 5·00
2·00 249 5·786 9·31 250 5·297 0·03 299 5·751 8·64 2·00
Y3 1·00 498 5·538 4·58 500 5·296 0·02 498 5·522 4·28 1·00
0·50 997 5·416 2·28 999 5·296 0·02 997 5-408 2·14 0·50
0·20 2495 5·344 0·91 2496 5·296 0·02 2495 5-340 0·8() 0·20
0·10 4990 5·320 0·46 4992 5·296 0·02 4990 5-318 0·43 0·10
0·05 9981 5·308 0·24 9983 5·296 0·02 9981 5·306 0·22 0·05
The error s is that when y is compared with the logarithmic method, i.e.
y1 = 11·747 m; y2 = -7·283 m; ya = 5·295 m
REFERENCES
1. ESCANDE, L.: "Methodes nouvelles pour Ie calcul des chambres d'equilibre·
(Dunod, 1950).
2. JAEGER, C.: Engineering fluid mechanics (Biackie, 1957), p. 212.
3. JAKOBSEN, B. F.: 'Surge tanks,' Trans. Am. Soc. civ. Engrs, (1922), 85,
pp. 1357.
4. CoLEBROOK, C. F., and WHITE, C. M.: 'Experiments with fluid friction
in roughened pipes,' Proc. Roy. Soc. A, (1937), 161, pp. 367-381.
5. COLEBROOK, c. F.: 'Turbulent flow in pipes, with particular reference to
the transition region between the smooth and rough pipe laws,' J. Instn
civ. Engrs, (1939), 11, pp. 133-156.
6. ACKERS, P.: Resistance offluids flowing in pipes and channels, Hydraul. Res.
Paper No. 2 (H.M.S.O., 2nd Edn., 1963).
7. ACKERS, P.: Tables for the hydraulic design of storm-drains sewers and pipe-
lines, Hydraul. Res. Paper No. 4 (H.M.S.O., 1963).
8. PICKFORD, J. A.: 'The use of Wallingford Charts and tables for pipeline
problems,' Civ. Engng pub!. Wks. Rev., (1965), 60, pp. 1761-1764.
9. BLAIR, J. S.: 'New formulae for water flow in pipes,' Proc. Instn mech.
Engrs, {1951), 165, pp. 75-87.
10. GmsoN, A. H.: 'The investigation of the surge-tank problem by model
experiment,' Proc. lnstn civ. Engrs, (1924), 219, pp. 161-173.
12 Finite Difference Methods II
12.1 Modifications
In Chapter 11 we saw how various finite difference methods may be applied
to the mass oscillations which follow sudden complete valve closure in a
straightforward cylindrical tank with or without a throttle. The same
methods-Pressel, Simple Arithmetic, Jakobsen and Escande-can be
used for more complex cases.
In this chapter we deal with some modifications and use the Simple
Arithmetic Method throughout. The methods of Pressel, Jakobsen and
Escande can equally be used for these further cases, but the computation
becomes more complicated. The modifications are of two types: different
flow conditions and different types of tank. The variations are:
Flow conditions: Sudden complete

. 1 opening
Su dden partla va1ve movement
}sec wn 12.2
1.
Linear change of flow Section 12.3

Constant valve opening Section 12.4
Constant power output Section 12.5
Flow from several sources}
Section 12.6
Flow from external source
Tank variations: Variable area tank Section 12.7

Spilling tank Section 12.8
Differential tank Section 12.9
Branching tank Section 12.10
Tail-race tank Section 12.11
Multiple tank Section 12.12
141
12.2 Flow to turbines during oscillation: sudden valve movements
Q in equation (10.3) is no longer equal to zero when there is flow to the
turbines during mass oscillation. The continuity equation is
vA = uAs +Q
For sudden valve movements Q becomes Q1 at the start of the oscillation
and the continuity equation becomes
vA = uAs + Q, (12.1)
Equation (12.1) applies to complete opening, when Qo = vo = 0, and also
to partial movements, when both Qo and Q, are pre-determined positive
flows.
Example 12.1. Sudden complete opening

Determine the minimum water level in surge tank A when the valve suddenly
opens to give a steady flow of 30 m 3 fs.
SOLUTION Take M = 10 seconds
TABLE 12.1. Surge tank A sudden complete opening
SUM
I !:1t L1y y L1u u v Fpvjvj (4) + (8)
Line (I) (2) (3) (4) (5) (6) (7) (8) (9) Line
1 0·00 0·00 -0·382 0·000 0·00 0·00 l

2 10·00 -3-82 0·000 2
3 10·00 -3-82 -0·382 0·000 0·00 -3-82 3
4 10·00 -3-82 0·012 4
5 20·00 -7·64 -0·370 0·187 0·02 -7-62 5
6 10·00 -3·70 0·023 6
7 30·00 -ll·34 -0·347 0·561 0·15 -ll·19 7
8 10·00 -3-47 0·034 8
9 40·00 -14·81 -0·313 1-110 0·60 -14·21 9
10 10·00 -3·13 0·044 10
ll 50·00 -17·94 -0·269 1·806 1·59 -16·35 ll
* * *
* * * * • •
* * *
12 100·00 -26·28 -0·026 5·696 15·83 -10·45 12
13 8·11 -0·21 13
14 108·11 -26·49 14
15 110·00 -26·48 0·006 6·208 18·80 -7·68 15
16 10·00 0·06 0·024 16
17 120·00 -26-42 0·030 6·61 21·15 -5·27 17
18 10·00 0·30 0·016 18
19 130·00 -26·12 0·046 6·84 22·84 -3·28 19
20 10·00 0·46 0·010 20
21 140·00 -25·67 0·056 7·01 23·93 -1·74 21
Finite Difference Methods II 143
Surge tank A
Lr' 203·94
Ar' 16
When the valve opens, at first the whole of the flow to the turbines comes
from the surge tank. Therefore by continuity,
Q/ + uoAs = 0
Qj 30
uo = - - = - - -
78·54
= -0·382m/s
As
At time 0 Oine 1) there is no flow in the pipeline and the water level in the surge
tank is the same as in the reservoir, i.e. h1 = yo = 0, and all columns except
column 6 (uo = -0·382) are zero.
~y in line 2 is obtained from the equation ~y = u M and ~u is zero since the
SUM in column 9 of line 1 is zero.
In line 3, y = -3·82 is the sum of y = 0 in line 1 and ~y = -3·82 in line 2;
u = -0·382 is the sum of u = -0·382 in line 1, and ~u = 0 in line 2; v in line
3 is obtained from the continuity equation (12.1), i.e.
uAs + Q/ -0·382 X 78·54 + 30

v = A = 4·909
In line 5 v = uA, + Q1 = -0·370 x 78·54 + 30 = 0.187 m/s

A 4·909
The table continues for the downsurge until t = 108·11 when y = -26·49 m.
After this time the level rises. Therefore yt = - 26·49 m.
Example 12.2. Sudden partial closure

In surge tank A* the flow is suddenly reduced from 30 m 3 fs to 15 m 3/s. Deter-
mine the maximum water level in the surge tank.
(Note: the system with surge tank A* is the same as with surge tank A except
that the friction coefficient Fp = 0·5, and hence yo = -0·5vo 2 = -18·68 m).
Surge tank A*
d5 = 10m
· - - · - - · A5 =78·S4m 2.
SOLUTION Take M = 10 seconds

TABLE 12.2
SUM
t /::;.( y !::;.y u v /::;,v FPvlvl (3) + (8)
Line 0) (2) (3) (4) (5) (6) (7) (8) (9) Line
1 0·0 -18·68 0·191 6·112 18·68 0·00 1

2 10·0 1·91 0·000 2
3 10·0 -16·77 0·191 6·112 18·68 1·91 3
4 10·0 1·91 -0·094 4
5 20·0 -14·86 0·185 6·018 18·11 3·25 5
6 10·0 1·85 -0·159 6
7 30·0 -13-01 0·175 5·858 17·16 4·15 7
At time 0, the initial flow (Qo = 30 m 3/s) is passing down the pipeline and
vo = Qo/A = 6·112 mjs. The flow to the turbines is suddenly reduced to
Q1 = 15 m 3 js. Therefore by continuity,
Qf + uoAs = voA
uo = voA- Qt = 30- 15 = 0·191 mfs
As 78·54
In line 2, !::;.y = uM = 0·191 X 10 = 1·91 m, and /::;.v = -SUM X M/Lr
= 0. In line 3, v is still 6·112 m/s since /::;.v was zero in line 2. By continuity
u remains at 0·191 mjs since neither v nor Q has changed. In line 4, /::;.v = -
SUM X M/Lr = -1·91/203·92 = -0·094 mfs; v in line 5 is therefore
6·112- 0·094 = 6·018 mfs. By continuity u = (vA - Qt)/As = (6·018 X
4·9088 - 15)/78·54 = 0·185 mjs. The computation continues until y1 = -
1·72m.
12.3 Linear change of flow

The operation of the valve may be operated so that the flow to the tur-
bines retards or accelerates at a uniform rate. If TM is the time taken for
the valve movement and t is any time during the movement (i.e.
0 < t < TM), then
Q = Qo + (Qt- Qo)t/TM
Using finite difference methods we note that in time 1:1t the flow changes
by an amount 1:1Q where 1:1Q = (Qt- Qo)l:1t/TM.
Example 12.3
What is the maximum surge height in surge tank A when the flow is reduced
from 30 m 3 /s to zero in 200 seconds, the valve closure being performed in such a
way that there is uniform retardation.
Surge tank A
cf,' 10m
. - - . - - . As; 7S·54m 2
d'2·5m
A' 4·909m 2
Lr; 203·94
Ar' 16
SOLUTION Let M = 20 seconds

TABLE 12.3
SUM
t M Q y ~y u v ~v Fpvlvl (4) + (9)
Line (I) (2) (3) (4) (5) (6) (7) (8) (9) (10) Line
1 0·0 30·0 -18·22 0·000 6·112 18·22 0·00 1

2 20·0 0·00 0·000 2
3 20·0 27·0 -18·22 0·038 6·112 18·22 0·00 3
4 20·0 0·76 0·000 4
5 40·0 24·0 -17-46 0·076 6·112 18·22 0·76 5
6 20·0 1·53 -0·065 6
7 60·0 21·0 -15·93 0·110 6·037 17·78 1·85 7
* * * * * * * *
21 200·0 0·0 6·05 0·160 2·566 3·21 9·26 21
* * * * * * *
27 251·0 0·0 11·75 0·000 0·000 0·00 11·75 27
In line 3, at time t = 20 seconds, the flow has been reduced to

Q = Qo + (Qt - Qo)t/T,u = 30 + (0 - 30)20/200 = 27 m 3/s
v has not changed, since in line 2, ~v = M x SUM/Lr = 0.
. . vA- Q 30- 27
By contmmty, u = ~ = 78 _54 = 0·038 mfs
The tabulation continues until line 27, when u = 0 and y1 = 11·75 m.

12.4 Constant valve opening

The flow to impulse turbines such as Pelton wheels is usually controlled
by spear valves. When the load demanded from the set alters, the valve
may suddenly be moved to a new position, and the area available for the
passage of water is then different from the original area.
The flow through the valve is given by the equation derived from
Bernoulli's theory, Q = CdAov(2gH), where Cd is the coefficient of
discharge of the valve. The head at the valve is equal to the difference
FIG. 12.1
between the level of the water surface in the surge tank and the level of
the valve, less friction losses in the penstock and throttle losses at the base
of the tank. Neglecting penstock losses, and with H 8 as the static head
over the valve (Fig. 12.1)
Q = CdAov(2g)v(Hs + y + FTuJuj)
When the oscillations have subsided, Q = Qt and u = 0, so
Qt = CdAoy(2g)y(Hs + Yt)
Yt is always negative.
Consequently, at any time during the oscillation
Q may be calculated independently for each step. If LlQ is required, as
when the tank is overflowing (Section 12.8), it may be taken as the differ-
ence between Q at the end of the step (at time t + Llt) and Q at the
beginning of the step (at time t).
Alternatively, the flow may be related to the flow Q8 which would pass
through the valve if the head on the valve were equal to the static head H 8 •
For a simple tank with FT = 0, the equation corresponding to equation
(12.3) is
Qs
and LlQ ~- Lly
2Hs
Example 12.4
The valve to the turbine in the hydro-electricity plant using surge tank A* can
be set to give a required flow under steady conditions. The static head over the
valve is 50 m.
If the flow is initially 30 m 3 fs and the valve setting is altered so that there will
be a steady flow of 15 m 3 /s, calculate the maximum surge height.
Surge tank A*
• d5 = 10m
· - - · - - · A 5 =78·S4m 2.
A=
r. 16
SoLUTION Take M = 10 seconds.

Since the tank is unrestricted, FT = 0, and equation (12.3) becomes
Q = Qr J( Hs
Hs
+ y)
+ Yt
Qr 15
Qr = 15 m 3 /s; Hs =50 m; vr = - = - - = 3·056 mfs
A 4·909
:. yr = -FpVr 2 = -0·5(3·056) 2 = -4·67 m
Therefore Q = 15 ) ( 50 + Y)
45-33
Tabulating, we obtain,
TABLE 12.4
SUM
t l1t Q y ~y u v ~v Fplvlvl(4) + (9)
Line (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Line
1 0·0 12-47 -18·68 0·223 6·112 18·68 0·00 • 1
2 10·0 2·23 0·000 2
3 10·0 12·91 -16-45 0·218 6·112 18·68 2·23 3
4 10·0 2·18 -0·109 4
5 10·0 13·32 -14·27 0·206 6·002 18·01 3·75 5
6 10·0 2·06 -0·184 6
* * * * * * * * * *
21 100·0 15·26 -3·08 0·055 3·982 7·93 4·85 21
22 10·0 0·55 -0·238 22
* * * * * * * * * *
28 10·0 0·11 -0·178 28
29 140·0 15-47 -1·80 0·000 3·149 4·96 3·16 29
* * * * * * * * * *
320·0 14·86 -5·53
* * * * *
500·0 15·04 -4-44
At time 0 (line 1) the water level in the surge tank is at y = -18·68 m,

corresponding to the pipeline friction head loss with Q = 30 m 3/s. The head
over the valve is therefore Hs + Yt = 50 - 18·68 = 31·32 m. The valve is
altered to give a steady flow of 15 m 3/s under a head of Hs + yJ, i.e. 45·33 m.
However, the head available is only 31·32 m, and therefore
Q = 15 J( 31·32)
- -
45·33
= 12·47 m 3 /s
30m3/sis passing down the pipeline and the difference between 30 and 12·47 m 3/s
enters the surge tank, and by continuity,
u = vA - Q = 30 - 12·47 = 0·223 mfs
As 78·54
In line 2, ~y = uM = 0·223 x 10 = 2·23 m, and ~v is zero since SUM in
line 1 is zero.
In line 3 y = -18·68 + 2·23 = -16·45
Hence Q = 15 J( 50- 16·45)

45 .33 = 12·91 m 3/s,
and u = 6·11 X 4·909- 12·91 = 0 .218 mfs

78·54
The tabulation is continued to the maximum surge level y1 = -1·80m when
II= 140 sec.
If the calculation is continued further, y2 = -5·53m when 12 = 320 seconds,
y 3 = -4·44m when 13 = 500 seconds.
12.5 Constant power output

A power station has to provide electricity in accordance with the require-
ments of the grid, and the changes of water flow are secondary to changes
in the power output of the turbine/generator sets. Therefore the altera-
tions of waterflow, which cause the mass oscillation we are studying are
likely to be determined by the turbine requirements rather than by the
need for constant gate opening or constant acceleration of the water.
The power obtained from the flow of water through the turbines is
given by
P = 17pgQH
where 17 is the efficiency. In British units the horsepower is given by
hp = O·ll31]QH where Q is measured in cusecs and H in feet. In SI
units the power is measured in watts and P = 9807nQH, where Q is
measured in m 3 js and H in metres.
If the new power after a change is Pt then
Pt = 980717fQtHt = 98071JtQt(Hs + Yt)
At any time during the oscillation, provided penstock friction is
neglected,
P = 9807nQ(Hs + y + FTuJuj)
If the power output is to be maintained at P1 from time t = 0, then
during the oscillation
Q=0 171 ~+~ =0 171 ~

1] (Hs + y + FTuJuj) 'Yj (Hs + y + FTujui)
When we use finite difference methods, Q may be calculated inde-
pendently for each step and ~Q taken as the difference between initial and
final flows for the step.
For a simple tank with constant efficiency
Ht Qt
Q = Hs + y . Qf and ~Q = Ht ~y (12.4)
Example 12.5
A hydro-electricity plant with surge tank A* is governed to give the required
power under steady conditions. The static head over the turbines is 100m.
If the flow is initially 30 m 3 fs and the power requirement is altered so that it
can be satisfied with a steady flow of 15 m 3/s, calculate the maximum surge height.
Assume the efficiency is constant.
Surge tonk A*
d5 = 10m
· - - · - - · A5 =78·S4m 2•
SOLUTION Take llt = 10 seconds

Q = __!!j_ Qt = Hs + Yt Qt
Hs + y Hs + y
Qt = 15 m 3 /s; Hs =lOOm; Vf = 3·056 m/s; Yt = -4·67 (as Example 12.4)
Therefore Q = 95·33 X 15
100 + y
Tabulating, we obtain,
TABLE 12.5
SUM
t llt Q y 11y u v 11v FpvJvl (4)+(9)
Line (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Line
1 0·0 17-58 -18·68 0·158 6·112 18·68 0·00 1

2 10·0 1·58 0·000 2
3 10·0 17·25 -17·10 0·162 6·112 18·68 1·58 3
4 10·0 1-62 -0·078 4
5 20·0 16·92 -15·48 0·162 6·034 18·21 2·73 5
6 10·0 1-62 -0·134 6
* * * * * * * * * *
21 100·0 15·03 -4·86 0·082 4·366 9·53 4·76 21
* * * * * • * *
37 180·0 14·54 -1·67 -0·007 2·857 4·08 2·41 37
* * * * * *
370·0 15·20 -5-89 0·000 3·137
* • * * *
560 14·94 -4·26 0·000 3·038
At time 0 (line 1) the water level in the surge tank is y = -18·68 m.

The head on the turbine is therefore Hs +yo= 100- 18·68 = 81·32 m.
The governor is set to give the required power with a steady flow of 15 m 3/s
when the head is 95·33 m. The flow taken by the turbines at time t = 0 is
therefore
95·33
Q = - - x 15 = 17·58 m 3/s
81·32
By continuity, = vA- Q = 30- 17·58 = 0_158 /
u As 78·54 ms
In line 3, - 95·33 - . 3
Q- 100 _ 17. 10 x 15- 1725m /s
The tabulation is continued to the maximum surge level of y1 = -1·67m when
11= 180 seconds. Further computation yields y2 = -5·89 m when 12 =
370 seconds, and ya = -4·26 m when ta = 560 seconds.
Consecutive valve moments. The design of some surge tanks is based on

superimposed oscillations, as mentioned in Section 9.3. Computation of
y by finite difference methods presents no difficulty.
12.6 Flow from several sources

(a) Flow from external source into surge tank. Occasionally water flows
directly into a surge tank. The supplementary flow may be intentional or
accidental. If the surge tank is suitably located, water may be abstracted
from a secondary intake, as shown in Fig. 12.2. Accidental flow occurs if
FIG. 12.2
surface run-off enters the tank following rainfall or an unexpected under-

ground stream is encountered during construction. With Qe designating
the flow from the external source, the continuity equation becomes
vA + Qe = uAs + Q
(b) Multiple pipelines to base of tank. The turbines may be fed by a

number of reservoirs, each of which has a separate pipeline to the base of
the surge tank as in Fig. 12.3 (a). If there are n pipelines and the water level
is the same in all the reservoirs, the fundamental equations are
dva
Lra dt + Y + FPaValvaJ = 0
dvn
Lrn dt + Y + FPn VnlvnJ = 0
AaVa + AbVb + . . . AnVn = AsU +Q

6 24PP·
Reservoir c
Power station
FIG. 12.3 (a)
Finite difference methods may be used even when the reservoirs are at
different levels, but trial methods are necessary to compute the flow in each
pipeline at each step and the calculation becomes very protracted.
(c) Branching pipeline. When the system is supplied by several reservoirs

which feed branching pipes, as Fig. 12.3 (b), the fundamental equations
become
dva
Lra dt + YJ + FPaValval = 0
dvb
Lrb dt + YJ + FPbVblvbl = 0
dvc
Lrc dt + Y- YJ + FpcVclvcl = 0
AaVa + AbVb = AcVc = AsU +Q
FIG. 12.3 (b)

For conditions (b) and (c) the system may be replaced by a single
equivalent pipeline whose friction coefficient must be estimated.
12.7 Variable area tank

In the fundamental continuity equations, vA = Asu + Q, the cross-sec-
tional area As must correspond to the elevation of the water surface. Where
the changes of cross-section are abrupt, as in a tank with expansion
chambers, it is merely necessary to use the appropriate value of As corres-
ponding toy at each step, as in Example 12.6. With conical tanks the area
changes during the step and the dynamic equation becomes
FT also varies since it is proportional to (A 8 ) 2 , i.e. (FT = kT/2g)(A 8 /Ar)2.
Example 12.6
Determine the maximum surge height in surge tank E for sudden valve closure
from a steady flow of 30 m 3 fs.
Surge tonkE
d, =l2 m
40m
20m
Reservoir level
-10m
-20m
d = 2·5m -30m
A= 4· 909m2
L = 2000m
SOLUTION Take t:..t = 10 seconds, and tabulate.
TABLE 12.6
SUM
(4)+
t l1t !1y y ds A, u v !1v FpvjvJ (10)
Line (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Line
1 0·0 -18·22 6·0 28·27 1·061 6·112 18·22 0·00 1
2 10·0 10·61 0·000 2
3 10·0 -7-61 8·0 50·27 0·597 6·112 18·22 10·61 3
4 10·0 5·97 0·521 4
5 20·0 -1·64 8·0 50·27 0·546 5·591 15·25 13-61 5
6 10·0 5-46 0·667 6
7 30·0 3·82 5·0 19·64 1·231 4·924 11·83 15·65 7
8 10·0 12·31 0·767 8
9 40·0 16·13 5·0 19·64 1·039 4·157 8·43 24·56 9
10 10·0 10·39 1·204 10
11 50·0 26·52 7·0 38·48 0·377 2·953 4·25 30·77 11
12 10·0 3·77 1·509 12
13 60·0 30·29 7·0 38·48 0·184 1·444 1·02 31·31 13
14 9·41 1·73 1·444 14
15 69·41 32·02 7·0 38·48 0·000 0·000 0·00 32·02 15
With so large a time interval 11t, considerable errors occur. For example,
the surge tank diameter is taken as 6·0 m during the whole of the first ten
seconds, while the water surface rises to -7·61 m, which is 2·39 m above the
enlargement aty =-10m.
If t:..t is reduced to 1 second, y1 becomes 26·07 m.
If t:..t = 0·1 second, y1 = 25·44 m.
12.8 Spilling tank

A spillway is sometimes provided at the top of a tank to limit upsurge and
to save construction costs. Spilling may occur around the whole perimeter
or through an opening of limited width. Referring to Fig. 12.4, the
discharge over the spillway is
Qs = Cabha~ = Csha~
where Ca is the coefficient of discharge for the spillway, b is the width, ha
the head causing the discharge and Cs = Cab. ha = y - HsP where
HsP is the height of the spillway above the static water level.
During spillage the upward velocity of the water surface u ( = dy/dt) is
different from the velocity of the water in the tank below the spillway Us.
Then by continuity
+ Cshd~
AsUs ~ AsU
and Av ~Asu + Cshi + Q (12.5)
Differentiation of equation (I2.5) leads to
-dv
dt
= -AI {As-
du
dt
+ 2-3Csha~-
"dha + -dQ}
dt dt
dv = 2_ {As du + ~ C hatu + dQ}·

8 (I2.6)
dt A dt 2 dt
-Spillway-
-+-'S=-ta_tic l::.. . ev-'-'el_ _
---Surge t a n k - - -
FIG. 12.4
The momentum equation is

dv
Lr dt + Y + hvlvl + FTuslusl = 0
Substituting dvfdt from equation (12.6),
Lr { du
A As dt
dQ} + y + Fpvlvl + FTuslusl = 0
+ 23 Cshatu + dt
and
du I {3
- = - - - Cshatu
dQ
+- A
+ Lr
-(y + Fpvlvl + FTUslusl)
}
dt As 2 dt
Transferring to finite difference terms we obtain the equations required
for the computations
flu= - flt {~ C 8halu + flQ + A (y + Fpvlvl + FTuslusD} (12.7)

As 2 Lr
1 Q
V = A{AsU + Csha 312 +Q} = ArUs +A (12.8)
flha = fly = uflt (12.9)

These equations are only satisfactory when both Cs and IJ..t are small.
An alternative method of dealing with conditions during spill is to calcu-
late the change of water level from equation (12.9), v from the momentum
equation and u from equation (12.8), as shown in example 12.7. Even
with this method it is important to take small increments of time. If IJ..t
is large, the first increment !J..y above the spillway level may give a water
level such that the spilling flow far exceeds the flow (u 8 A 8) coming up the
surge tank.
Example 12.7
A spillway in surge tank F is set at a level 5 m above the static level, otherwise
the arrangement is the same as surge tank A. The spillway coefficient is Cs =
100 mH s- 1 • What is the maximum height of the surge following sudden complete
closure from a steady flow of 30 mfs?
Surge tankF
Lr= 203·94
Ar=l6
SoLUTION Take /:!.t = 1 second.

The calculation proceeds normally until y is greater than 5 m.
TABLE 12·7
SUM
t Ll.y y Ll.v v Fpvlvl (3) + (6) us ha Os u
Line (I) (2) (3) (4) (5) (6) (7) (8) (9) (10) (II) Line
1 72·0 4·8804 3·519 6·04 10·92 0·2199 0·2199 1
2 0·2199 -0·054 2
3 73·0 5-1003 3·465 5·86 10·96 0·2166 0·1003 3·177 0·1761 3
4 0·1761 -0·054 4
5 74·0 5·2764 3·411 5·68 10·96 0·2132 0·2764 14·534 0·0282 5
6 0·0282 -0·054 6
7 75·0 5·3046 3-357 5·52 10·82 0·2096 0·3046 16·794 -0·0038 7
8 -0·0038
9 76·0 5-3008
i.e. yt=5·30m
In line 2,
!:!.y = u/:!.t = 0·2199
In line 3,
v 3·465
Us = - = - - = 0·2166
Ar 16
Qs = Cshal-5 = 100(0·1003)1 · 5 = 3·177 mfs
Qs 3-177
u =Us- As= 0·2166- 78 .54 = 0·2166- 0·0405 = 0·1761 mfs
In line 4,
~y = u~t = 0·1761 X 1·0 = 0·1761
12.9 Differential tank

During upsurge the computation is the same as for a simple tank until the
water level reaches the ports. Thereafter the oscillation passes through
the following stages:
Outer tonk Riser Outer tonk

oreoA; oreaA'
FIG. 12.5
(a) Fig. 12.5. The water surface continues to move upward in the riser,
but some water passes through the ports to the outer tank, or expan-
sion chamber. The flow through the orifices is q', and the following
equations may be obtained:
q' = ±CaAoy(2gl(y- y')l) (12.10)
vA- Q = q + q' (12.11)
and q' /).t = As' f).y' (12.12)
(b) Fig. 12.6. The water level in the riser reaches the spillway and over-
flows, and water is still passing into the outer tank through the ports.
Equations (12.10) and (12.11) still apply but the rise of level above
the riser is now found by using equations (12.7) and (12.9).
FIG. 12.6 FIG. 12.7
(c) Fig. 12.7. The outer tank is completely full and the water level
becomes the same across the whole tank which behaves as a simple
tank. Further increase of level is in accordance with the equation
vA = (A' + As')u + Q (12.13)
The surge reaches its maximum height as Us becomes zero.
(d) Fig. 12.8. With negative flow Us at the base of the tank, the level
drops in the riser and some flow passes through the ports. For a
J~~/ t
, ff ,'
~00000~
(negative)
-·- ·--·--·--
!u,
(negative)
FIG. 12.8 FIG. 12.9

time the spillway discharges water from the outer tank to the riser.
The same equations as in stage (b) are used with reversed signs.
(e) Fig. 12.9. During further downsurge the water level in the riser
falls more rapidly than that in the outer tank. Flow from the outer
tank is in accordance with equation (12.10) and q' is negative
because (y - y') is negative. Crt may be different for flow into or
out from the riser. Equations (12.11) and (12.12) are also valid, but
care must be taken with the signs.
To enable the tank to operate satisfactorily during downsurge the water

stored in the outer tank must be released quickly. On the other hand a
differential tank controls upsurge better if the flow to the outer tank is
limited during upsurge. Satisfaction of these contrary requirements can be
effected by either providing flap valves, which only open during downsurge,
or by fairing the ports so that the discharge coefficient is larger for flow out
of the outer tank.
12.10. Branching tank

If the tank has two separate water surfaces, as in Fig. 12.10, there is a
secondary oscillation, and the equations become as follows:
For the main tank
vA = AsUs + Q
c
v-
L0 is the mean length of the secondary oscillating water column

between A and C
FIG. 12.10
6A
For the secondary oscillation of the column between the two surfaces
dvb
Lrb dt- (ya- Yc) + FpbVbJvbJ = 0
dye
Acdt = Abvb
dya
AsUs = Aa dt + AbVb
dya dya dye
Hence vA = Q + Aadt + Abvb = Q + Aadt + Acdt.
12.11. Tail race tank

The fundamental equations still apply with appropriate change of sign.
Referring to Fig. 12.11 we see that the momentum equation becomes
D..t
D..v = +-
Lr
(y- FpvJvJ - FpuJuJ)
u
-T
y
_L_·-·-·-·-·
v area A
FIG. 12.11
The continuity equation is
Av = Q- Asu
12.12 Multiple tanks

Escande has derived a great variety of equations for systems containing
more than one tank.
For two tanks there are two possible arrangements.
(a) Both tanks upstream of the turbines. The continuity equations are:
D..t
D..ya =- (AaVa - Abbb) (12.14)
Asa
(12.15)
The dynamic equations are
!:!..t
!:!..va = - - (ya + FpaValval + FTaUa!Ual) (12.16)
Lra
!:!..t
!:!..vb = - - (yb- Ya
Lrb
+ FPbVblvbl + FTublubi) (12.17)
(b) Upstream and downstream tanks. For continuity, equation (12.4)

still holds, but equation (12.15) becomes
!:!..t
!:!..yb = - (Q - Abvb)
Asb
Similarly the dynamic equation for tank a (equation 12.16) is still true,
but equation (12.17) becomes
!:!..t
!:!..vb = +-
Lrb
(yb- FPbVblvbl - FTulul)
13 Approximate Methods
13.1 Relative values-simple tank

Calame and Gadenl introduced relative values as follows:
Dimensions of length are related to the amplitude of undamped oscilla-
tion,>'* = voy(Lr/Ar) (equation 10.5) giving the ratios )'r = yfy* and
Fr = Fpv 2fh·
Velocities are related to the corresponding velocity at steady flow con-
ditions or zero time. For valve closure Vr = vfvo and u, = ufuo, uo being
the upward flow at zero time.
Time is related to the period of undamped oscillation T = 2?Ty(LrAr)
(equation 10.15) giving the ratio tr = tfT.
(a) Sudden complete closure. The initial steady flow Qo is diverted to the
surge tank at zero time, so vo = Aruo. By continuity at any time during
the oscillation v = uAr and hence Vr = Ur andy* = uoy(LrAr).
For an unrestricted tank during the first quarter cycle the dynamic
equation (10.2) becomes
Lr dvfdt + y + Fpv2 = 0
From equations (10.3) and (10.4), with Q = 0, we know that
dv du du dy du
-dt = Ar-
dt
= A dy
r - . - = Ar . U-
dt dy
du
Therefore LrAr U dy + y + Fpv2 = 0
Substituting LrAr = (y*fuo) 2 and dividing through by>'* we obtain
dur
Ur -d + )'r + Fr = 0
Yr
162
Approximate Methods 163
Fro is the value of the non-dimensional ratio Fr at time t = 0, and
Fro= Fpvo 2/y*, so Fr = Fpv 2jy* = FroVr 2 = FroUr 2
dur
and Ur-d
Yr
+ Yr + FroUr 2 = 0
d(ur) 2
or -dyr- + 2FroUr2 = -2yr
The solution of this first order linear differential equation, with initial
condition tr = 0, Yr = -Fro and Ur = 1, gives
Ur = [1 - 2FroYr- exp { -2Fro(yr + Fro)}]/2Fro2
Maximum surge height occurs when ur = 0 and

1 - 2FroYr- exp { -2Fro(yr +Fro)}= 0
Series solution of these equations gives 2
2 1 4 1
Yrl = 1 - 3Fro + 9F ro2 + 135 Fro3 + 270 Fro 4 + .
The following approximations have been suggested by Eydoux,a

Sopwith2 and Sutton :4
(a) Yrl = 1 - iFro + !Fro2 = (1 - !Fro) 2 (13.1)
(b) Yrl = 1- iFro (13.2)
(c) Yrl = 1 - 0·6Fro (13.3)
Calame and Gaden gave an approximation for the first downsurge

following complete closure, which only applies if Fro is very small. It is
Yr2 = -1 + 2Fro (13.4)
They also prepared charts, based upon their graphical method described
in Section 14.2, to give the relative values for the first two upsurges and
the first downsurge. The curves are shown by full lines in Fig. 13.1, which
also gives the approximate equations.
Elsden5 has suggested an empirical rule for the first downsurge:
1
Yrz= - - - - (13.5)
7
1+3Fro
0·1
FIG. 13.1
(b) Sudden complete opening. At the instant of opening, the final steady
flow Qf is assumed to come wholly from the surge tank, so Qf = -uoAs.
When the oscillation has ceased, Qf comes wholly from the pipeline and
Qt = VfA. By continuity, at any time during the oscillation vA = uAs +
Q, which may be written
Qf Qj
v-
Vf
= u-
Uo
+ Qf or Vr = Ur + 1
It will be noted that Vr is related to the final steady velocity in the pipeline;
vo is zero.
F,,
0 0·1 0 2 0·3 0·4 0 5 0 6 0·7 0·8 0·9 1·0
"'
I ·00
~00~ . I
I
~9~ I ·04
1:; c ;;-....
~G ~
"l.lrv~ I
~
1·0
" ['.. I
0·8 ~~ ~~ -1--1--1· 10
"'\. ~
1-- """ -
-~ ~R~~ ·12
0·6 1-- ~
~ -Broken line
~---~~
I'\
shows equation
0·4 ·14 13.6
\ 'i:. I
~ I·15
~.-;..s;r
0·2
1\. co,
""'.:,.
0
' 9"' c
'\ <SG'
-0·2 ,cv~
..J,.
::<>
-0·4
f'.._
-0·6 "' !'--

-0·8
"r--..
-........~--.
.........
-1·0
0 0 ·1 0·2 0·3 0·4 0 ·5
F,,
0 ·6 0 ·7 0·8 0·9 ""'I 0
FIG. 13.2
No integration of the differential equations is possible. However,

Calame and Gaden applied their graphical method, which uses relative
values, to obtain curves shown in Fig. 13.2 and the approximation
Yrl = -1 - O·l25Fr! (13.6)
where Fr1 is the value of the non-dimensional ratio Fr for final steady flow.
Equation (13.6) conforms to the curve for first downsurge when
Fr1 < 0·8. Braun suggested an alternative formula,
Jr! = -0·5Fr! - y (0·25FrJ2 - 0·81Frf + 1) (13.7)

13.2 Other approximate equations for simple tanks

An equation which applies to both sudden closure and sudden opening is
(13.8)
which was derived in 1908 by Johnson, 6 who also obtained an approxima-

tion for the time at which maximum upsurge occurs,
(13.9)
These equations apply to sudden partial or complete opening and closure.
For sudden opening from no-flow, a formula was derived by

Forchheimer7 which may be expressed as
-y1 = +O·l78Fpvi + J[o·03l7(FPVJ2)2 + ~: v12]

(13.9)
-y1 = -o·l78yt + -vr<o·l78yt) 2 - 2PYtl
where .B is the damping factor used in Chapter 10.
Example 13.1
Use the approximate equations given previously and Calame and Gaden curves,
to find the height of the first two upsurges and the first downsurge following sudden
complete valve closure in surge tank A.
Surge. tank A
· - - · - - · ds=lOm·-·
As=78·54m 2
L,=203·94
A,= 16
SOLUTION
Y• = vo J( Lr) =
-
Ar
6·112 )(203-9)
--
16
= 21·82; -yo 18·22 0·8350
Fro=-=--=
Y• 21-82
FIRST UPSURGE
Equation 13.1 Jrl = (1 - !Fro)2 = (0·7217) 2 = 0·5209; y1 = 0·5209 X 21·82
= 11·37m
Equation 13.2 Jrl = 1 - !Fro= 0·4433; y1 = 0·4433 X 21·82 = 9·67 m
Equation 13.3 Jrl = 1 - 0·6Fro = 0·4990; y1 = 0·4990 X 21·82 = 10·89 m
Equation 13.8 Vf = 0
(yl + 18·22)2 = 12·75( -6·112)2 + 0·48782( -6·1122)2
= 476·1 + 332·1 = 808·2
Jl = v(808·2) - 18·22 = 28·43 - 18·22 = 10·21 m
C and G Curve For Fro= 0·8350, Jrl = 0·539
y1 = 0·539 x 21-82 = 11·75 m
FIRST DOWNSURGE
C and G Curve For Fro= 0·8350, Jr2 = -0·329; y2 = -7·18 m
1
Equation 13.5 Jr2 = - = -0·3392· y2 = -7·40 m
1 + 1·948 '
SECOND UPSURGE
C andG Curve Jr3 = 0·240 J3 = 5·23 m
The following table gives the percentage error compared with the logarithmic
solution
Jl J2 ya
SECOND
FIRST UPSURGE FIRST DOWNSURGE
UPSURGE
Log 13·1 13·2 13·3 13·8 Curve Log Curve 13-5 Log Curve
11·75 -3·2 -17·2 -7·3 -13-1 00 -7-28 -1·4 -1·6 5·30 -1·3
A number of other approximate methods have been developed. 8 • 9 •10 • Several

have been summarized by Pearsall. Most are semi-graphical methods involving
the replacement of the friction term hF = FPvlvl by a linear expression hF = kv
and are neither simple nor accurate.
13.3 Classified diagrams for throttled tanks

Two additional relative values are introduced to allow for throttle losses
r is the relative head loss where r = hT/y* = rour 2
z is the piezometric head at the bottom of the chamber below the
throttle, and Zr is the corresponding relative value:
Zr = zfy. = Yr + rourluri
For sudden complete closure, the fundamental dynamic equation
becomes
dur
Ur -d + Yr + Frovrlvrl + rouriurl = 0
Yr
or, for the first upsurge, where uris positive,
Yr + (Fro + ro)ur 2 (13.10)
Ur
If pipeline friction is neglected equation (13.10) becomes
dur dur
Ur-d + Yr + rour 2 = 0 or Ur-
dyr
+ Zr =0
Yr
Fro
FIG. 13.3
Since at the start of oscillation Yr = 0 and Ur = 1; this may be integrated
as
(13.11)
At the maximum upsurge Ur becomes zero and the relative height of the
first upsurge Yrl = Zrl = 1/y2. The rise is therefore reduced to 1/v2 of
that in a similar surge tank without throttle.
Classified diagrams. Using relative values with their radial graphical

method (Section 14.2) Calame and Gaden prepared 'Classified diagrams'
which give the height of the first surge.
Fig. 13.3 shows the diagram for sudden complete closure, and Fig.
13.4 shows that for sudden complete opening.
0'!> 1 z 0 =1·65
0
,, 1- - L - -
I
r - -'--I--t-
1 1 IL
"i'/ z, 0 =!·6 I I I 1
/ I
- -zr-;,-F;;"'o-- ~~--I-,-
1 I I
I
- - - ~~0- - ...1 - -
I
L -l- ..l - n-
I
1 "'
I I1 I I I
- -/- - - '0 f- - -~- - I_ -f
I z,o=i·3 o"' I I I I
/
..../- -z, -1·2 /~~~ - -I /_ - I-!- -
I 0=
j_ - - - -
1·2 1 oI I I
- - - - - - ~!---L-1--J --
1·I Z,.o=l·l ,'?1 I I I
/ 'l}r I
--"/---r--A--t--'- -1- - -
I
·0 //
/ Zro=i·O
/
1
"-"'
I
I
I
I I S
- -.zr - - "---- -o
=09; ,;
I_ -
1
+--
-1--
I
0
- - .J~ - - _3~ - _I_ - I

/-'Z,o=O·B I I
/ /
/
:,r- - -- -=-~-,
0·1 0·2 0·3 0·4 0·5 0·6 1·0

F,,
FIG.l3.4
Example 13.2
Use the classified diagrams to determine the height of the first upsurge following
sudden complete valve closure and the first downsurge following sudden complete
valve opening.
Surge tank B
a;= !Om
--·~~ 78·54 ~2
SoLUTION
r
Sudden complete closure
As in Example (13.1), vo = 6·112 m{s; Y• = 21·82 m; Fro= 0·8350
hTo = FTuo2 = FT ( ~: = 50 X {6·112/16) 2 = 7·296 m
hTo 7·296
ro = - = - - = 0·334
Y• 21·82
From the classified diagram, Fig. 13.3 with
Fro= 0·835 and ro = 0·334, Jrl = 0·423
Therefore Jl = Jrl X Y• = 0·423 X 21·82 = 9·23 m
Sudden complete opening

Frt = 0·835; ro = 0·334
From the classified diagram, Fig. 13.4, Jrl = -0·982.
Therefore y1 = -0·982 x 21·82 = -21·43 m
The significance of the lines and zones on the classified diagrams is as

follows:
Line S-S is for an ideal throttle, which is such that there is constant
piezometric head at the bottom of the tank. That is,
z =constant
zr = constant
zo =yt
Zro = Yrl
This is shown in Fig. 13.5 (a) for absolute values and Fig. 13.5 (b) for
relative values. The ideal throttle is considered below.
Absolute values Relative values
FIG. 13.5
Zone A is when the throttle is too strong and the piezometric head at
the bottom of the tank drops during upsurge.
zo> y1
Zro > Yrl
Figs. 13.5 (c) and 13.5 (d) show this.
Zone B is when the throttle is too weak, as shown in Figs. 13.5 (e) and
13.5 (f). The piezometric head rises during upsurge
zo <yl
Zro <Yrl
Zone 0, limited by line L-L in the diagram for sudden opening, indicates
that the motion is dead-beat, i.e. the water level drops to the final steady
level Yt without oscillation.
The ideal throttle The piezometric head at the base of the tank is constant.
(a) Neglecting pipeline friction
During the first upsurge
Zr = Yr + rour 2 = constant
Using equation (13.11) and the maximum upsurge Yrl = Zrl = 1y2, we
obtain
1
Zr = Zrl = V2 = Yr + ro{l - y(2)yr}
1 1
from which ro = V2 = 2 y2
(b) Allowing for pipeline friction, Calame and Gaden found that for
small values of Fro the throttle friction should be:
(i) at the start of motion ro ~ !v2 + !Fro

(ii) at maximum upsurge r1 ~ !v2 + !Fro
Johnsonl 2 and Richl3 derived charts and equations to give the area of the
ideal orifice for sudden closure and sudden opening.
Example 13.3
Find the diameter of the 'ideal orifice' for surge tank A, if the loss coefficient
of the orifice kT = 1·0.
Surge tonk A
d,: !Om
· - - · - - · A5 :78.·54m2
d=2·5m
A =4·909m2
Lr:203·94
Ar=16
SOLUTION
(a) Neglecting pipeline friction ro = h/2
hTo
ro = - and y* = 21·82m
Y*
Therefore hTo = 1!.. = 21 "82 = 15·43 m
y'2 y'2
hTO = kT VT0 2
%
and UTO = VO ~ = VO
AT
(!:..) 2
~
vo = 6·112 m/s; d = 2·5 m; dT = 1·0
Therefore dT = J(2·5)4 X 1·0 X (6·112)2) = .41(4·822) = 1·48 m

4 2 X 9·807 X 15·43 ·y
(b) Allowing for pipeline friction (Calame and Gaden)

At the start of motion ro ~ tv'2 + !Fro
Fro = 0·8350; therefore ro = 0·7071 + 0·2087 = 0·9158
Therefore hTo = 21·82 X 0·9158 = 19·98 m
dT = J(2·5) 4 X 1·0 X (6·112) 2) = .4I(3-9 20) = 1.41 m

4 2 X 9·807 X 19·98 V
At the end of motion

Therefore hTo = 21·82 x 1·333 = 29·09 m
dT = J(2·5)4 X 1·0 X (6·112)2) = .41(2·557) = 1·27m
4 2 X 9·807 X 29·09 V
Johnson's equations do not apply because the pipeline friction is too great.
It can be seen that for ideal throttling the size of the orifice has to
change during the oscillation. This, of course, cannot be done in practice.
Parmakian Charts
Two charts have been given by Parmakian14 to give y1 following
sudden complete closure and sudden complete opening.
REFERENCES
1. CALAME, J., and GADEN, D.: 'Theorie des chambres d'equilibre,' (Gautier-
Villars, 1926).
2. PEARSALL, I. S., 'A survey of surge tank design theories' NEL Report No. 56,
(1962).
3. EYooux, D.: Hydraulique generate et appliquee (Bailliere, 1921).
4. SUTTON, B. A.: 'Series solution of some surge tank problems,' Proc. Instn
civ. Engrs, (1960), 16, pp. 225-234.
5. ELSDEN, 0.: Surge Chambers, Chapter xviii of 'Hydro-electric engineering
practice,' Ed. J. Guthrie Brown (Biackie, 3nd Edn., 1964).
6. JoHNSON, R. D.: 'The surge tank in water power plants', Trans. Am. Soc.
mech. Engrs, (1908), 30, pp. 443-474.
7. FoRCHHEIMER, P. H.: Hydraulik (3rd Ed., 1930).
8. JURNEY, W. H.: 'Surge tank analysis', Tech. Mem. U.S. Reclaim. Serv.,
(1946), No. 632.
9. PRAsrL, F.: 'Wasserschlossprobleme,' Schweiz. Bauztg, (1908), 52, pp. 271-
277, 301-306, 317-320, 333-336.
10. WARREN, M. M.: 'Penstock and surge-tank problems,' Trans. Am. Soc. civ.
Engrs, (1915), 79, pp. 238-271.
11. STUCKY, A.; Cours d'amenagement des chutes d'eau. Chambres d'equilibre
(Ecole Polyt. Univ. Lausanne, 1951).
12. JOHNSON, R. D.: 'The differential surge tank,' Trans. Am. Soc. civ. Engrs,
(1915), 78, pp. 760-784.
13. RicH, G. R.: Hydraulic transients (Dover, 1963).
14. PARMAKIAN. J.: Waterhammer analysis (Dover, 1955).
FURTHER READING
BINNIE, A. M.: 'Approximate methods in surge tank calculations,' Proc. Camb.
phil. Soc. math. phys. Sci., (1946), 42, p. 156.
DuRAND, W. F.: 'On the control of surges in water conduits,' Trans. Am. Soc.
mech. Engrs, (1912), 34, pp. 319-363.
14 Graphical Methods
Before the introduction of electronic digital computers, the accurate use

of finite difference methods was very time-consuming and tedious. Conse-
quently a number of 'time-saving' graphical methods were developed. In
practice these methods all take quite a long time to carry out, and they
cannot compare in accuracy with numerical methods performed by
computers.
14.1 Schoklitsch methodl

This is a graphical representation of arithmetic integration, the equations
being the same as those used in Chapter 11, i.e.
dv = -f(y + Fpv!v! + FTu!ui)dt (14.1)
or dv = --(y + hF + hT)dtfLr
and dy =(~_g) dt (14.2)
Ar As
In the Schoklitsch Method the value of yin equation (14.1) is taken at

the end of the step dt, whereas in the Simple Arithmetic Method y is
taken at the beginning of the step. Consequently, surge heights found by
the two methods differ slightly for the same value dt.
Fig. 14.1 shows the first few steps of the determination of surge heights
for Surge Tank A with sudden complete valve closure. The arbitrary
value at dt is 10 seconds.
By continuity, when the valve is completely closed the total volume of
water flowing from the pipeline in time dt all passes to the surge tank.
This volume is shown by curve A as Vp = Avdt, where vis the abscissa
which is also used with curve C; the positive direction is towards the left.
7 175
Curve B shows the volume of water in the surge tank. In Fig. 14.1 the
origin is taken at the initial water level with steady flow, yo = -18·22 m,
and so Vs represents the volume of water in the surge tank above the
steady flow level. Consequently Vs = ~~ Vp.
Surge tank A
· - - · - - - . d5 =10m_ _ _
As= 78·54m2
L, = 203·94
A,= 16
1400 12001000 800 600 400 200 o20 k 0 10 20 30 40 50

~ m3 300D:T?r~-hh-o---- I seconds
FIG. 14.1
If we assume that throughout time f!.t the pipeline velocity is what it

was at the start of tJ.t, then the water level rises by tJ.y = vf!.t/Ar since
f!.Vs = Vp = Avf!.t and also f!.Vs = fJ.yAs.
Therefore, at the end of the first interval f!.t (i.e. at t = 10 seconds) the
level of the water in the tank is at Y<lO>· This is carried across from curve B
to the main graph, C. On C a curve F-F is drawn showing hp against v.
Graphical Methods 177
Positive hF is drawn downwards using the same scale as they-curve in B.
Positive v is drawn towards the left, using the same scale as in A.
By equation (14.1), with hT = 0 because the tank is unrestricted, ~v = -
(y + hF)M/Lr. From the original conditions shown as point F, a line
is drawn at a slope of - M/Lr. This gives the variation of v withy, and
the intersection of this line with Y<IO> at C10 gives ~v. In fact,
~V(0-10) = -(y<l0) + hF<Ol)~t/Lr
Finally, the values of y obtained at B are transferred across to D with
time t as the abscissa.
A complete diagram to find y1, the first upsurge, is shown in Fig. 14.2.
It could be extended to give subsequent surge heights although the lines
FIG. 14.2
would become confused as they would be traversing the same area as

those already drawn.
The Schoklitsch Method can deal with arrangements other than
complete sudden closure with a simple tank. For example, if there is only
partial closure, A is modified to give the final flow Q,. Constant Q1 is
shown by a horizontal straight line as in Fig. 14.3. For valve opening
Av~t lies below the line for Q1, and Vs decreases after the start of
oscillation.
If Q varies with time or with the elevation of the water surface, a
subsidiary curve is drawn which is then replotted on A with v as the
abscissa.
With a variable area tank, such as surge tank E, the slope of the tank
volume curve varies, as shown in Fig. 14.4.
Schoklitsch diagrams can also be used for spilling and differential tanks.
Throttle losses in a restricted tank may also be included.
Since the Schoklitsch method is fundamentally the same as numerical
solution by finite differences, the time interval !lt affects the accuracy of
A v:p
Outflow= o,l::;t
FIG. 14.3
the result. The height of the first upsurge y1 following sudden complete
closure in surge tank A is as follows:
!lt Yl
10 seconds 10·95 m
5 seconds 11·36 m
2 seconds 11·59 m
1 second 11·67 m
By the logarithmic method y1 is 11·75 m.

As with numerical finite difference methods, mean values can be used. 2
A number of ways of doing this have been suggested, including one
proposed by Schlag3 ·4 which is similar to Escande's D.F. Method. It
employs mean values of y at the start of the oscillation and mean values
of v when !lvfv > !lyfy.
Bouvard and Molbert5 modified the Schoklitsch method by plotting y
against Q instead of y against v.
Surge tank E
~=12m
40m
d5 = 7m
20m
d5 =5m
Reservoir level
ds =am
-10m
d5 = 6m
-20m
d5 =4m
d = 2·5m -30m
A= 4· 909m2 d5 =15m
5000 4000 3000 2000 1000 0
L = 2000 m 051 =30 m3/s V. m3
FIG. 14.4
14.2 Calame and Gaden radial method
Using relative values as in Chapter 13, Calame and Gaden 6 developed
their radial method which is sketched in Fig. 14.5. The axes are Vr and yr,
and the radius of curvature is p.
For sudden complete closure the initial conditions are given by point A1,
where Vr = Ur = 1 and Yr = -Fro; dvrfdyr = 0; the tangent is hori-
zontal and the radius of curvature for the first arc is p1 = -1 so that the
FIG. 14.5 FIG. 14.6

centre of curvature for the first arc is at C1, the vertical projection of A1
on the yr-axis. A short arc is drawn from point A1 to point A2. From
A2 a horizontal line is drawn to meet the pipe friction curve at F2.
From F2 a vertical line is projected onto the Yr-axis at Pz.
The centre of curvature of the next arc, A2A 3, is at C2, at the intersection
of A1C1 and A2P2. The next arc AaA4 has its centre at Ca, at the inter-
section of A2P2 and AaPa, and so on. The curve A-A is continued by a
series of short arcs until it meets the Yr axis, when Ur = Vr = 0, giving the
abscissa Jrl·
Pearsall4 has suggested a method using constant time intervals and
mean values of Ur, as shown in Fig. 14.6. Arc A1A2 is drawn from the
initial point A1 as before, with p = -1 and the centre of curvature at C1.
From the fundamental equations of Chapter 10 we know that for
complete closure v = Ar dyfdt. If we use relative values this becomes
Y* dyr
VrVo = A r - -
T dtr
Substituting the values of y* and T from equation (10.5) we get
Vr = Ur = Ar ..!_ uoy(Lr/Ar) X dyr = _.!.._ dyr

Uo 27r(LrAr) dtr 277 dtr
Replacing the differentials by small finite increments,

fltr flyr
-=-=tancp
1/277 Ur
In Fig. 14.6, B1B2 represents l:lyr(l-2) and M2B2 represents Ur(l-2J·

Therefore a line is drawn from B1 with a slope of tan cp and point M2
is found such that M2A2 = tflur(l-2)· This fixes A2, the end of the first
arc. From A2 a horizontal line is drawn to meet the friction curve at F2,
and F2 is projected down to the Yr-axis at P2. Hence the centre of curvature
of the arc A2Aalies on the intersection of A1C1 and A2P2 at C2 as before.
FIG. 14.7
B2 is next obtained as the projection of M2 and A2 onto the Yr-axis. A
line B2Ma is drawn at slope tan t/J and MaAa is made !Aur(2-3), and
the curve is continued.
Fig. 14.7 shows the construction of a curve {yr against vr) for the first
quarter-cycle, and also a curve {yr against tr) which is the form usually
required.
Escande7 has developed a similar method, 'methode semi-graphique', in
which real values of v are plotted against real values of y.
Radial graphical methods can be used for complete opening, for partial
valve movements, for conditions where the flow Q is not constant, and for
throttled tanks.
14.3 Surge height trace by analogue computer

An electronic analogue computer may be used to trace the variation of
y with t.s Fig. 14.8 shows a typical circuit diagram. It is for a throttled
tank of uniform cross-sectional area where FTu and FTo differ, and
where Q, is constant.
- ve reference voltage + ve reference voltage
FIG. 14.8
The recorders shown on the diagram give the following traces against t:
Rl gives the surge height, y
R2 gives the pipeline velocity, v
R3 gives the upward velocity in the surge tank, u.
The accuracy of the traces may be increased by making the maxima as
near full-scale as possible. Multiplying factors, which are not shown in
Fig. 14.8, are introduced throughout to give near maximum voltage.
REFERENCES
1. ScHOKLITSCH, A.: Hydraulic structures (ASME, 1937).
2. SoRENSEN, K. E.: 'Graphical solution of hydraulic problems,' Trans. Am.
Soc. civ. Engrs, (1953), 118, pp. 61-77.
3. JAEGER, C.: Engineering fluid mechanics (Blackie, 1957).
4. PEARSALL, I. S.: 'A survey of surge tank theories,' N.E.L. Report No. 56,
1962.
5. BouvARD, M., and MoLBERT, J.: 'Methode graphique pour le calcul des
cheminees d'equilibre,' Houille Blanche, (1950), pp. 535-554, (1951),
pp. 356-371.
6. CALAME, J., and GADEN, D.: Theorie des chambres d'equilibre (Gautier-
Villars, 1926).
7. EscANDE, L.: Methodes nouvelles pour le calcul des chambres d'equilibre
(Dunod, 1950).
8. CHRISTIE, I. F.: 'The use of analogue computers for civil engineering
FURTHER READING
BOURDON, C.: 'Contribution a !'etude des chambres d'equilibre a etranglement
par la methode graphique,' Genie civ., (1960), 137, pp. 360-364.
ESCANDE, L.: 'Manouvres rythmiques pour une cheminee deversant a entrang-
lement avec !'influence de la hauteur de chute dans le cas d'un orifice,·
C. r. Acad. Sci., Paris, (1961), 252, pp. 2339-2342.
LEINER, Dr.: 'Ermittlung der Schwingungen im Wasserschloss,' Ver. Deutscher
Ing. Z., (1965), 69, pp. 1637-1642.
PEARSALL, I. S.: 'Comparative experiments on surge tank performance,' Proc.
lnstn mech. Engrs, (1963), 177, pp. 951-970.
WooD, F. M.: 'Graphical treatment of non-linear friction in water-hammer
and surge study,' Trans. Engng lnst. Can., (1958), 2, pp. 33-42.
ZieMAN, B.: 'Methodes nouvelles pour le calcul des cheminees d'equilibre,'
Houille Blanche, (1953), pp. 580--598.
15 Open Channel Surge
15.1 Unsteady flow in open channels

Any change of flow in an open channel causes a wave to be propagated
from the point where the change is started. The wave is 'translatory';
all particles of water in any cross-section move together and may be
assumed to remain in one plane. Open channel waves may be divided
into two broad categories depending on whether the vertical component
of the acceleration of particles of water is negligible or not. If the vertical
component is small, there is said to be gradually varied unsteady flow;
a large vertical component is associated with rapidly varied unsteady
flow.
Examples of gradually varied unsteady flow are flood waves in rivers
and the change of flow resulting from slow operation of controlling
structures such as gates and sluices.
With rapidly varied unsteady flow or 'surges', an abrupt change of
surface level occurs as the surge moves along the channel. Surges are of
four types, depending on whether they move upstream or downstream
and whether the water level behind them is higher than in front (a positive
surge) or lower (a negative surge). Positive surges have stable fronts, and
negative surges are unstable.
The four types of surge are these:
1. Backwater surge, Fig. 15.1 (a). This is a positive wave moving
upstream and is due to a sudden reduction of downstream flow, as when
turbine gates close and set up a surge in a headrace channel, or when a
landslide blocks a river bed. A tidal bore is a particular type of backwater
surge. Backwater surges are treated in detail below.
2. Flood surge, Fig. 15.1 (b). This is also a positive surge, but moves
downstream. It is caused by a sudden increase of upstream flow, as in a
tailrace channel of a hydro-electric scheme when the turbine demand
increases. Flood waves are especially important in underground power
184
Open Channel Surge 185
stations, where the tailrace may be in a tunnel which under steady flow
conditions is only part full, acting as an open channel. Sudden gate
opening can cause a flood surge which fills the tunnel completely.
Flood surges occur in navigation canals when lock gates are suddenly
opened. Failure of a dam causes a flood surge in the valley below. When
the earth dam at the South Fork reservoir in Pennsylvania failed in 1889
a surge over 100ft high left the dam site and moved rapidly, though with
less height, towards Johnstown.1· 2
3. Drawdown surges, Fig. 15.1 (c) are negative and are caused by an
increase of flow downstream. They occur in headrace channels of hydro-
electric schemes and may determine the maximum rate at which the
~ --==-----
(a) (b)
(c) (d)
FIG. 15.1
turbines can accept load. Another cause of drawdown surges is the

operation of siphons, as in a siphon overflow from a combined sewer.
4. An ebb surge, Fig. 15.1 (d) is negative and results from a reduction
of upstream flow. When the sluice of a branch distributing channel is
suddenly opened there is an ebb surge in the main channel downstream
of the junction.
With both types of positive surge (backwater and flood) even small
changes of flow may produce large and dangerous surges. The speed,
or 'celerity', of a surge increases with depth. A gradual change of flow
may propagate a series of small waves each of which is travelling over
deeper water than its predecessor. The small waves therefore overtake
each other, combining to form a large surge.
This is what happens when a bore or eagre is formed in an estuary
with large tidal variation. The bore in the River Severn is a well-known
phenomenon, and bores are also formed, especially at spring tides, in
the River Trent and around the Bay of Fundy. The Hangchow bore on the
Chien Tang River in China reaches a height of nearly 30 feet.
Surges may be steep-fronted, as shown in Fig. 15.1, or they may be
undular, as shown in Fig. 15.2.
FIG. 15.2
Surge diagrams like those used for waterhammer have been used by
Bergeron and others for the analysis of open channel surges.
A special case is a monoclinical rising wave. The wave front travels
downstream with constant celerity. Many flood waves in natural channels
can be approximated to a monoclinal rising wave. Fig. 15.3 shows the
A is the cross-sectional area
~
:::;; ~A
~
FIG. 15.3
notation used. By continuity, the flow into the front of the wave is equal
to the flow out, so
(c- v1)A1 = (c- vz)Az and
15.2 Backwater surges

The rise of water associated with a backwater surge occurs in two stages.
The advancing surge-front gives an abrupt change of water surface
level. Behind the front the water rises at the downstream end of the
channel. This continues until the arrival of the negative wave moving
downstream; the negative wave is reflected by the reservoir when the
positive backwater surge reaches the head of the channel. Sometimes a
spillway is provided at the downstream end of the channel to prevent the
continuing rise of level after the surge-front has passed upstream.
In a non-rectangular channel the height of the surge front is greater
at the sides than in the middle; the height at the sides may be twice the
mean height. This increases the danger of overtopping the channel
banks.
Equations to determine the surge celerity, c, and the height of the surge
front, h, may be obtained by applying the principles of momentum and
continuity.
Using the notation of Fig. 15.4, taking the mean width of the channel
as the wavefront as B and assuming that the centre of gravity of the front
is !h below the downstream level, we obtain from Newton's second law
the following equation,
from which C
gh
= (VI- ( 1 +-
Bh) -VI (15.1)
vz) 2A!
FIG. 15.4
and
Substituting this value of vz in equation (15.1) we obtain
c = gh Bh)
v _ v1A1 - cBh ( 1 + 2A - v1
1 A1 + Bh 1
or c = J(g(A! + Bh)(2Al + Bh)) - VI (15.2)

2A1B
For a rectangular channel equation (15.2) becomes

At the downstream end of the channel where the flow is reduced and
the surge begins,
cBh = Q1- Qf (15.4)
Qf is the final steady flow out of the channel. For example for a head-race
channel in connection with a hydro-electric plant, Qf is the new steady
flow demanded by the turbines.
The initial surge conditions may be found by using equation (15.2) or
equation (15.3) with equation (15.4) to obtain c and h. Trial methods are
employed.
Example 15.1
The discharge from the downstream end of a 20ft wide rectangular channel is
suddenly reduced from 1000 cusecs to 200 cusecs. If the initial normal depth is
8·9 ft, determine the surge height and celerity at the downstream end of the
channel.
SOLUTION
QI = 1000 ft 3 /s; Q2 = 200 ft 3/s; therefore QI - Q2 = 800 ft 3 /s
QI
B =20ft; di = 8·9 ft; therefore AI= 178 ft 2 ; VI= A= 5·62 ft/s
From equation (15.3),

C = J(g(di + h)~I + th)) _VI= J
(32·2(8·9 -8~~(8·9- i;h)) _ 5.62
= 1·90v[(8·9 + h)(8·9 + !h)] - 5·62 (15.3a)

From equation (15.4) cBh = QI - Q, = 800 ft 3/s; ch = 40 ft 2 /s
Tabulating the solution of these equations for trial heights we get
h c from 15.3 (a) c from 15.4
2ft 1·90y(10·9 x 9·9) - 5·62 = 14·13 ft/s 40/2 = 20 ft/s

3ft 1·90y(11·9 X 10·4) - 5·62 = 15·58 ft/s 40/3 = 13-33 ft/s
2·66 ft 1·90 v'(11·56 x 10·23) - 5·62 = 15·06 ft/s 40/2·66 = 15·04 ft/s
This is reasonably accurate, soh = 2·66 ft; c = 15·05 ft/s.
Behind the surge front the water surface may be assumed to be hori-
zontal for a short channel with little friction. For long steep or rough
channels the water behind the front is assumed to rise parallel to its
previous surface.
Fig. 15.5 (a) shows the surge front at two points as it moves upstream
from a to b. When the front is at b the surface downstream of a has
risen by r above the level of the surface when the front was at a.
If the surge front moves from a to b in time !:lt, the distance moved,
!:ll, is equal to c!:lt, where cis the mean celerity of the surge.
b 0
(a)
(b)
b 0
(c)
FIG. 15.5
As the front moves from a to b the net inflow builds up the surge
between a and b and also causes a general rise r over the whole surface
area downstream of a. This downstream surface area, which includes
the area of any forebay, we call AD. Referring to Fig. 15.5 (b) we see
that by continuity
AI
At(Q1- Qz) =-:- (Q1- Qz) = lBAI(ha + hb + r) + rAD
c
Hence (15.5)
Note: If the side slopes are very flat the second term in the denominator
of equation (15.5) may be corrected to read l(Baha + Bbhb + B,r) where
Ba, Bb, and Br are the widths of the channel at the mean heights of the
surge front at A, the surge front at B and the general rise r respectively.
From Fig. 15.5 (c) it can be seen that
ha + r + izAI = hb + hAl
therefore Al=r+ha-hb (15.6)
it -iz
At the downstream end of the channel the slope of the surface behind
the surge front, iz, is negligible compared with the original hydraulic
gradient, h. The water surface behind the front can be considered as
horizontal.
Further upstream iz may be obtained from the friction loss and the
difference of velocity head between two sections, i.e.
. HF(ab) Vzb 2 - Vza 2
lZ(ab) = (Al)ab - 2g(Al)ab
Example 15.2
The channel of Example 15·1 has a forebay with a surface area of 20 000 ft 2 in
which complete regain of velocity head is assumed. The bed slope of the channel
is 1/2500.
Calculate the time for the surge to pass through the forebay and the distance
along the channel at which the surge height is 2·4 ft. What will then be the height
of the surface in the forebay above the initial steady level?
SOLUTION
V1 2 (5·62) 2
The original velocity head hv = 2g = 2 x 32.2 = 0·49 ft
We can assume that the mean water level in the forebay for the original steady
flow is 0·25 ft above the level at the entrance.
While the surge is passing through the forebay the flow entering the forebay
from the channel continues at its original steady rate of 1000 cusecs. The
outflow is at the new reduced rate of 200 cusecs. There is therefore a net inflow
of 800 cusecs which builds up the level of the water in the forebay until it is at
the surge height of 2·66 ft calculated in Example 15.1.
The time taken for the surge to pass through the forebay is equal to the time
required to raise the level from 0·25 ft to 2·66 ft above the original water level
at the entrance. Therefore the time is
(2·66 - 0·25) X 20 000 d
800 = 60 ·2 secon s
At the point where the surge height is 2·4 ft the celerity is found from equation
15-3 (a), i.e.
c = 1·90y'[(8·9 + 2·4)(8·9 + 1·2)] - 5·62 = 14·69 ft/s
Equations (15.5) and (15.6) are used to obtain rand Mas follows:
For the section under consideration, An = 20 000 ft 2 ; ha = 2·66 ft; hb =
2·4 ft; B = 20ft; (Q1 - Q2) = 800 ft3/s.
The mean surge celerity c = -!(15·05 + 14·69) = 14·87 ft/s and
Ql - Q2 = 800 = 53·80 f
c 14·87 t
Since the initial flow was at normal depth, h = s = 0·0004, i2 may be neglected
in this downstream length-i.e. the water level behind the surge front is assumed
to be horizontal.
20 OOOr r + 2·66 - 2·40
Therefore !!./ =
53·80 - ! X 20(5·06 + r) 0·0004
20 OOOr r+ 0·26
3·20- lOr 0·0004
By trial, r = 0·10 ft and M = 900 ft
The surge height is 2·4 ft when the surge front is 900ft from the forebay
entrance.
The height of the surface in the forebay is ha + r = 2·76 ft above the initial
steady level.
The process may be repeated for other surge heights to obtain a back-
water surge curve, as Fig. 15.6. It can be seen that overtopping of the
:··· ... ·········
s
lJrfoce
:·-·-.=:=::.:.:.::::::.:::.:.:.::
!
···········
Pror;1
e lJflcJer ..
lfl'liof
steoo
Y flow
Outlet end
FIG. 15.6
banks is most likely near the downstream end of the channel. The level
will rise further when the surge front is reflected from the reservoir until
the negative reflected surge (an ebb surge) reaches the downstream end.
The treatment of a backwater surge which has been described may also
be applied, with appropriate change of sign, to other types of open channel
surge.
REFERENCES
1. HIMMELWRIGHT, A. L. A.: 'The Johnstown flood,' Harper's Mag., (1933),
167, pp. 443-455.
2. O'CoNNOR, R.: Johnstown: the day the dam broke (Alvin Redman, 1959).
FURTHER REFERENCES
CHOW, V. T.: Open-channel hydraulics (McGraw-Hill, 1959) pp. 554-572.
ELSDEN, 0., and ALBERY, A. C. R.: Canals and free-flowing channels, Chapter
XV of 'Hydro-electric engineering practice,' Ed. J. Guthrie Brown (Blackie,
2nd Edn, 1964).
HAws, E. T.: 'Surges and waves in open channels,' Wat. Pwr, (1954), 6,
pp. 419-422.
SANDOVER, J. A., and ZIENKIEWICZ, 0. C.: 'Experiments on surge waves,'
Wat. Pwr, (1957), 9, pp. 418-424.
SELLIN, R. H. J.: Flow in channels (Macmillan, 1969).
list of Notations
Waterhammer (chapters 2-8)

A pipe cross-sectional area
Ao open cross-sectional area of partly-open valve
As horizontal cross-sectional area of surge tank
a velocity of pressure wave
. CaAov'(2g)
B valve openmg factor = A
Cc coefficient of contraction
Cr reflection factor
Ct transmission factor
c pipe end-fixing coefficient
D outside diameter of pipe wall
d inside diameter of pipe
E Young's Modulus of elasticity
e thickness of pipe wall
F variable pressure head of wave travelling away from point of origin
f variable pressure head of wave travelling back towards origin
f Darcy friction coefficient
g gravitational acceleration
H surge pressure head
Ho steady pressure head
h pressure head above normal at any time
ht head loss due to friction
hL total head loss
hvp head equal to vapour pressure and partial pressure of released gases
I moment of inertia of pumping set
hydraulic gradient
K bulk modulus
L total length of pipeline
193
I length of pipeline from reservoir to intermediate point
N speed of pump (rev/min)
p power of pump
p pressure increase corresponding to head h
Q flow in pipeline
q flow into surge tank
T total time of valve moment
t time at instant under consideration
v volume
v velocity in pipe
X distance along pipe from downstream end
w weight density
z2 ratio of total head to initial head
'
Allievi's pipeline constant
A angle of surge lines = tan- I!!:.._
gA
. a
rabo-
g
period of pipe or critical time 2 ~
a
')! Poisson's ratio
p mass density
7" time interval for surge diagram
7" proportion of full gate opening
characteristics on surge diagram
Surge tanks (chapters 10-14)

A cross-sectional area of pipeline
AB cross-sectional area of pipe at base of surge tank
Ar surge tank area ratlo
. = A As
As horizontal cross-sectional area of surge tank

AT cross-sectional area of restriction in throttled surge tank
ATh limiting cross-sectional area of surge tank for stability
d inside diameter of pipeline
ds diameter of cylindrical surge tank
dp diameter of restriction in throttled surge tank
Fp friction coefficient for pipeline such that yo = -Fpvo 2 or
Yf = -FpVf2
Fr Calame and Gaden parameter for pipeline loss= Fpv 2fh
Fs friction coefficient for pipeline and throttle = Fp + FT/Ar 2
List of Notations 195
friction coefficient for throttle such that hT = FTuJuJ
gravitational acceleration
head over turbine
static head over turbine (with no flow)
head loss between reservoir and surge tank= FpvJvl
head loss due to throttle
friction coefficient for throttle such that hT = kTuTJuTJ/2g
length of pipeline from reservoir to surge tank inlet
. f . .
length ratiO L
Lr o p1pehne = -
g
n friction index
n stability factor of safety
Q flow in penstock
hT
r Calame and Gaden parameter for throttle loss = -
h
period of mass oscillation, neglecting friction
total time of valve movement
time, measured from start of motion studied
Calame and Gaden parameter for time = f
u upward velocity of water surface in surge tank
Ur Calame and Gaden parameter for surge tank velocity = .!:!.. or .!:!..
Uo Uf
UT upward velocity through throttle

v velocity of water in pipeline, positive towards surge tank
Vr Calame and Gaden parameter for pipeline velocity = ~ or .!:.

Vo Vf
w weight density
X logarithmic parameter for water level in surge tank = ~
y water level in surge tank above static level of reservoir (i.e. measured
positive upwards)
Yr Calame and Gaden parameter for water level = L

y*
amplitude of oscillation neglecting friction
total head over base of surge tank = y + hT
z
Zr Calame and Gaden parameter for total head = -
h
{3 damping factor =!~
FsAr
.1. increase during time .1-t
p radius of curvature in Calame and Gaden graphical method
p mass density
Subscripts
o initial steady conditions
f final steady conditions
s static conditions-with no flow
st steady conditions
1 conditions after first quarter-cycle of oscillation
2 conditions after further half-cycle
Suprascript bar
average values during time interval .1-t
Open channel surge (chapter 15)

An downstream surface area
A1 original area at any cross-section
A2 total area under surge front
B width of channel at half wave front
c absolute speed of propagation of surge
d1 original water depth
d2 depth under surge front
g gravitational acceleration
h height of wave front
hydraulic gradient
Q1 original flow along channel
Q2 flow behind wave front
Qf final downstream flow
v1 original velocity at any cross-section
v2 mean velocity in surge front
p mass density
Index
ACCIDENTS 98 Blair, J. S. 85, 140

Ackers, P. 140 Bleifuss, D. J. 106
Addison, H. 41 Bogolioboff, N. 127
air in solution 15 bore 184-185
vessel 78-81 Borot, G. 85
Albery, A. C. R. 192 Bourdon, C. 183
Allievi, L. 2, 3, 26, 31 Bouvard, M. 106, 178, 183
classical equation 37, 58 branching pipes 65-68, 152
formula 22 surge tank 159-160
altitude 14 Bratfisch, A. E. 96
American Society of Mechanical Engin- Braun, E. 165
eers 2 bulk modulus, K 15, 20, 21, 28
American Water Works Association 49, by-pass for pump 81-83
69
amplitude of undamped mass oscillation CALAME, J. and Gaden, D. 2, 4, 101, 105,
111 162, 163, 165, 169, 172, 173, 180, 183
analogue computer-see computer charts 163-165
Angus, R. W. 2, 4, 16, 69, 85 radial graphical method 180-182
approximate methods of surge tank Cameron Highlands h.e. scheme 101
analysis 162-174 Carstens, M. R. 85
Braun 165 Cartwright, K. 0. 96
classified diagram 167-170 Cates, W. H. 17
Elsden 163 Causon, G. J. 96
Eydoux 163 coefficient of compressibility-see bulk
Forchheimer 166 modulus
Johnson 166 channel, headrace 1, 184-185, 188
Parmakian charts 173 tailrace 184-185
simple tank 162-167 Chappy, A. 3
throttled tank 167-173 characteristic curve 54
arithmetic integration-see f.d. methods Chow, V. T. 192
atmospheric pressure 14 Christie, I. F. 107, 183
Claria, J. 128
BACKWATER SURGE 184, 186-192 classified diagram 167-171
Bata, G. L. 106 closed surge tank 102, 104
Bergeron, L. 2, 4, 55, 69, 186 closure-see valve movement
diagram-see surge diagram Cole, R. S. 111, 127
Bergeron, P. 85, 96 Colebrook, C. F. 139
Binnie, A. M. 41, 83, 86, 106, 174 Colebrook-White friction formula 138
Biro, C. 3 Combes, G. 69, 85
197
compressible fluid 6 Elsden, 0. 105, 163, 173, 192
computer 2, 129 end conditions 32-41
analogue 104, 182-183 surge tank 87
digital 104, 134 Enever, K. J. 31
conical surge tank 102 enlargement-see pipe
concrete, reinforced, pipe 28 equivalent pipe 15, 153
consecutive valve movements 151 Escande, L. 2, 97, 105, 106, 127, 128,
constant power output 125, 149-151 129, 136, 139, 160, 182-183
continuity at base of surge tank 110 D. F. method 136-139, 178
contraction in pipe 35 semi-graphique method 182
H-Q diagram 53 Evangelisti, G. 105
at base of surge tank 125 Evans, W. E. 85
Contractor, D. N. 41 expansion chamber 102
control of surge 78-83 expansion joint 24
Cowen, W. 105 external source to surge tank 151
Crawford, C. C. 85 Eydoux, D. 163, 173
critical areas for surge tank stability 125
curves, logarithmic 119-121 F-WAVE, /-WAVE 27
factor of safety for stability 125
DAM FAILURE 185 faired throttle 103
damping factor, {J 112, 166 f.d.-finite difference method
dead-beat motion 171 feeder tank 81-83
dead end 33 Ferrand, G. 3, 105
H-v diagram 51 finite difference methods for surge tank
surge diagram 57 analysis 129-161
Defazio, F. G. 86 branching tank 159-160
density of liquids 15, 21, 29 constant power output 149-151
design chart for air vessels 80 constant valve opening 146-148
D. F. method 136 differential tank 157-159
diagram, H-v 50-52 Escande D. F. method 136-139
H-Q 52-54 flow from external source 151
surge 54-96 flow from multiple reservoirs 151-153
diameter of pipe, change 34-35 friction 138-139
Dickinson, J. C. 105 fundamental equations 129-131
differential surge tank 2, 102, 103 Jakobsen's method 136, 139
f.d. method 157-159 linear change of flow 144-145
Schoklitsch graphical method 178 multiple tanks 160-161
stability 126 partial valve movements 143-144
digital computer-see computer simple arithmetic method 131-135,
discharge to atmosphere 37, 44 141-161
diverter 98 spilling tank 154-157
Donsky, B. 86 successive trials 130
Doran, W. H. 107 sudden complete opening 142-143
Dorsch, R. G. 3 tail race tank 160-161
drawdown surge 1, 185 throttled tank 134-135
Dubin, C. 4 time interval 132-133, 139
Due, J. 85 variable area tank 153-154
Durand, W. F. 105, 106, 174 Flesch, G. 3
dynamic equation for mass oscillation flood surge 184
108 flywheel 76
Forchheimer, P. H. 166, 173
EAGRE 185 forebay 190
ebb surge 185 Fox, J. A. 3
Edison, G. 106 fracture of pipe or pump 1, 98
efficiency of turbine 126 Francis turbine 98
elasticity, pipe walls 21-25 Franke, P. G. 105
Index 199
frequency of pipeline 1 hydro-electric schemes 1, 98-99
friction-see pipe and throttle waterhammer 87-97
Fullard, S. F. 106
fundamental equations IDEAL THROTILE 170-172
f.d. methods 129-131 impulse turbines 87
mass oscillation 108-110 incompressible water column theory 6-
waterhammer 26-28 16
inertia 6-16
GADEN, D.-see Calame & Gaden pump 75-76
gas in solution 15, 29 Innertkirchen surge tank 104
gate valve 12 instability-see stability
Gaulhet, A. 86 intermediate time and position, surge dia-
Gayed, Y. K. 85 gram 58-60
Gerrard, R. I. 105 Isere-Arc surge tank 104
Gibson, A. H. 4, 31, 41, 105, 140 iterative integration-see f.d .methods
Gibson, N. R. 49
Gibson, W. L. 106 JAEGER, C. 3, 31, 86, 97, 105, 125, 127,
Goldman, G. 0. 31, 85 128
governing of turbine 100, 129 Jagger, B. K. 106
stability 126 Jakobsen, B. F. 106, 136, 139,
gradually varied unsteady channel flow f.d. method 136, 139
184 Johnson, R. D. 2, 4, 105, 106, 166, 172,
graphical analysis for surge tanks 175- 173, 174
183 Johnstown disaster 185, 192
analogue computer 182-183 Joukowski, N. 2, 3, 31
Bouvard & Molbert method 178 law 22, 37, 42, 43
Calame & Gaden method 180-182 Juhasz, Professor du 40
Escande method 182 junction 35-36, 152-153
Pearsall modification 180 H-Q diagram 52-54
Schlag modification 178 surge diagram 65-68
Schoklitsch method 175-179 Jupillat, R. 97
graphical analysis ofwaterhammer 50-97 Jurney, W. H. 174
H-Q diagram 52-54
Hv- diagram 50-52
surge diagram 54-96 KAICHEV, P. 128
-see also surge diagram Kamel, M. Y. M. 85
Gray, C. A. M. 69 Kaplan turbine 98
Kennedy, J. F. 4
HAGLER, T. W. 85 Kennison, H. F. 17, 31
Halcrow, W. T. 106 Kerr, S. L. 16, 17, 31, 96, 97
Halliwell, A. R. 31 Kittredge, C. P. 86
Halmos, E. E. 3, 31 Knapp, R. T. 86
Hangchow bore 185 'knocking' 5
Harding, D. A. 3, 69 Kryloff, N. 127
Hawkins, P. 97, 105
Haws, E. T. 192 LAC BLANC-LAC NOIR ACCIDENT 2
Hayashi, T. 3, 69, 106 Leiner Dr 183
head, limiting 14 length, equivalent 15
headrace, 1, 184-185, 188 Levin, L. 101, 105
Himmelwright, A. L. A. 192 Lightner, C. 3
Hudson, W. 106 limiting pressure 14
Hunter, J. K. 106 Linton, P. 31, 85, 86
Huron, R. 127 liquids, properties 21
hydraulic losses 68 Livingstone, A. C. 16, 85, 86
-see also pipe friction and throttle load rejection 98
losses load to turbine 1, 100-101
logarithmic method for surge tank analysis Paynter, H. M. 3, 97, 106, 128
111-124 Peabody, R. M. 17
curves 119-121 Pearsall, I. S. 31, 69, 105, 167, 173, 180-
equations 111-119 183
non-dimensional form 115 modification to radial graphical method
successive approximations 116-119 180-181
throttled tank 121-124 Pelton wheel 87, 98, 129
Lundgren, C. W. 85 penstock 99
Lupton, H. R. 2, 4, 16, 31, 69, 86 period of mass oscillations, T 110
of pipe 19
MAAHS, A. J. 17 Perliter, S. 86
Madich, P. B. 106 Pickford, J. A. 106, 127, 140
Mae surge tank 104 pipe
Maione, U. 102, 105 anchoring 24
Marchal, M. 3 branching 65-68, 152
Marples, E. I. B. 69, 86 constant, ~ 37
Marston, F. A. 17 contraction 34, 35, 53
mass oscillation 1, 100, 108-124 end condition 32
maximum pressure, 14, 45 enlargement 34, 35, 53
mean values equivalent 15, 153
f.d. methods 129-130, 136-138 expansion joints 24
graphical methods 178, 180-181 fracture 1, 2, 98
methode semi-graphique 182 frequency 1
Meyer, R. 127 friction
Milne, W. E. 127 mass oscillation 100, 101, 109, 138--
minimum permissible head 14-15 139
model studies 102 surge diagram 54, 68
modified bulk modulus 28 waterhammer 41
Molbert, J. 106, 178, 183 junction, 35, 65-68, 152
monoclinal rising wave 186 period 19
Mosonyi, E. 106, 111, 127, 128 reinforced concrete 28
multiple pipelines 151-152 rigid 6-16, 20-22
f.d. method 160-161 series 15-16, 60, 65
surge tanks 102, 104 short 6
stability 127 tapered 16
Murillo, J. 105 wall, elasticity 14, 21, 34
thickness 23, 28, 34
NAGY, L. 128 Pistilli, G. 106, 107
negative surge 184 Poisson's ratio 21
notation 2, 193-196 positive surge 184
numerical methods-see f.d. methods power taken by turbines 129
constant 149-151
'OBSERVER', TRAVELLING 55 Prasil, F. 174
O'Connor, R. 192 Pressel's f.d. method 130
opening-see valve movement pressure, atmospheric 14
oscillation-see mass oscillation -discharge diagram 52-54
superimposed 101, 151 limiting 14
waterhammer 37-39 vapour 14-15
-velocity diagram 50-52
PAGDIN, B. C. 17 properties of liquids 21
parallel pumps 83 pump, by-pass 81-83
Parmakian, J. 4, 31, 69, 86, 173, 174 characteristic curve 54
partial change of flow failure 11
inertia pressure 10-11 fracture 1
mass oscillation 142-144 inertia 75-76
waterhammer 46, 49 parallel 83
Index 201
pump (contd.) Schnyder, 0. 2, 4, 55, 69, 85, 86
reciprocating 84-85 diagram-see surge diagram
stopping 1, 5 secondary intake to surge tank 151
H-v diagram 51-52 Sedijatmi, R. M. 106
surge diagram 70--85 Sellin, R. H. J. 192
submersible 80 separation 71-74
pumping main 70-85 series pipes 15-16
surge diagram 60-65
RADIAL GRAPHICAL METHOD for surge Severn bore 185
tank analysis 180--182 sewage 15, 29
Ransford, G. 69 Sharp, B. B. 85
rapid closure-see valve Shelson, W. 106
rapidly varied unsteady channel flow 184 Silvers, A. 127
reaction turbine 87, 98 Simon, Miss 0. 3, 31
reciprocating pump 84-85 simple arithmetic f.d. method 131-135,
reflected wave 32-39 141-161
H-v diagram 51 simple surge tank 102
reflection factor 35 size of pipe 29
reflux valve 78 slow valve closure-see valve
Reisman, A. 127 solution of gas in water 15
rejection of load 98 Sopwith, D. G. 163
relative values, surge tank 2, 162 Sorensen, K. E. 183
relief valve 98 sound, speed 28
reservoir, reflection 32-33 South Fork reservoir 185
H-Q diagram 55 Soverzene surge tank 104
H-v diagram 51-52 spear valve 87
hydro-electric scheme 99 spilling surge tank 102-104
resonance 1 f.d. method 154-157
restriction-see throttle stability 124-127
effect on waterhammer 36-37 statism 126
H-Q diagram 54 step-by-step integration - see f.d.
H-v diagram 51 methods
retardation, constant rate Stepanoff, A. J. 86
inertia pressure 9-10 Stevenson, D. 41
mass oscillation 144-145 stereogram 40
waterhammer 47 stopping of pump-see pump
reverse flow 76 strain energy 22-25
Rich, G. R. 106, 172, 174 Streeter, V. L. 3, 69, 86
Richards, T. R. 85 Strowger, E. B. 97
rigid pipe 6-16, 20--22 Stucky, A. 128, 174
rigidity factor 28 successive approximation 116-119
riser in differential surge tank 103 trials 130
Rossens surge tank 104 sudden valve closure-see valve
rotary valve 14 superimposed oscillation 101, 151
surge control 78-83
SABLJAK, R. 107 surge diagram 2, 54-96
Sandover, J. A. 192 branched pipes 65-68
Sarginson, E. J. 17 by-pass 81-82
Savastano, G. 106, 107 comparison with Allievi 58
Schlag, A. 178 hydraulic loss 68
Schoklitsch, A. 175, 183 intermediate times and positions 58-60
graphical method 175-179 pumping mains 70--84
increments of time 178 reciprocating pump 84-85
mean values 178 reverse flow 77-78
partial closure 177-178 separation 72-74
variable area tank 177, 179 series pipes 60--65
surge diagram (contd.) turbine (contd.)
surge tank 87-96 efficiency 126
uniform pipe 55-57 f.d. method 146-151
surge tank
computer 104 UNDAMPED MASS OSCILLATION 110-111
design criteria 100 Urbain, E. A. 86
model 102
purpose 98-99 V2-LAW 8
stability 124-127 vacuous space 15,32
types 102 Vallentine, H. R. 16
-see also closed, differential, multiple, valve
simple, spilling, tail race, throttled, constant opening 146-148
variable area surge tanks discharge to atmosphere 37, 44
surge tank analysis-see approximate gate 12-13
methods, f.d. methods, graphical partially closed 36
analysis rotary 14
surges, superimposed 101 spear 87
Suter, P. 3 valve movement
Sutton, B. A. 163, 173 mass oscillation 98
Swaminathan, K. V. 31 closure 110-124, 130-139
consecutive 151
TAILRACE 184-185 opening 142-143
surge tank 160-161 partial 143-144
tapered pipe 16 sudden 142-144
temperature 14-15, 29 waterhammer 5, 11-14, 48-69
Thackrah, D. G. 86 closure 6-12, 18-25, 33, 37-40, 42-
Theorie des chambres d'equilibre 2, 4 51, 87
thick-walled pipe 28 opening 10-12, 46, 49, 87
Thirriot, C. 86 rapid 42-46
Thoma, D. 125, 127 slow 6-10, 46-48
critical area 125 sudden 18-19, 31, 50-52
Thomasson, P. G. 3 Valvis, P. G. 128
throttle in pipe 36 vaporization 14
H-Q diagram 54 vapour pressure 14-15, 21
surge diagram 68 variable area surge tank 102
throttled surge tank 100-103 f.d. method 153-154
classified diagram 167-171 Schok!itsch graphical method 177, 179
faired throttle 103 stability 125
f.d. method 134-135 Veggeberg, J. M. 3
ideal throttle 172-173 velocity head 125
logarithmic method 112, 121-124 velocity of waterhammer wave-see wave
throttle losses 101, 109, 112, 134 venturi contraction 125
waterhammer 93-94
tidal bore 184-185 WALLINGFORD CHARTS AND TABLES 138
Tongland surge tank 104 Warren, M. M. 174
Torrell, P-A. 106 water, properties 15
transients 1 waterhammer 1-12
translatory wave 184 Allievi's theory 2, 37
transmitted wave 34-37 elastic theory 18-31
transmission factor 35 fundamental equations 26-28
'travelling observer' 55 incompressible theory 6-16
Trividac, A. 97 surge diagram 54-96
Tucker, D. M. 85 Water Power 97, 106
tunnel in rock 28 wave
turbine 87, 98, 129 flood 186
loading 1, 100-101 monoclinal rising 186
Index 203
wave (contd.) Wood, D. J. 3
oscillation 37-40 Wood, F. M. 183
reflection 32-37 Woodward, A. C. 3, 85
translatory 184 Wylie, E. B. 3
transmitted 34-37
velocity 21, 28-30 YOUNG, G. A. J. 85
weather 15 Young's modulus 21
White, C. M. 139
Widmann, R. 105 ZAOUI, J. 69
Wilson, J. N. 16, 85, 86 Zieman, B. 102, 105, 183
Wolfe, T. F. 17 Zienkiewicz, 0. C. 97, 105, 128, 192

Sediment Erosion in Hydro Turbines 1

Caricato da

Informazioni sul documento

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Sediment Erosion in Hydro Turbines 1

Caricato da

Copyright:

Formati disponibili

Sediment Erosion in Hydro

Hari Prasad Neopane

This thesis is submitted to

i HPN PhD Thesis, NTNU

HPN PhD Thesis, NTNU ii

iii HPN PhD Thesis, NTNU

PREFACE AND ACKNOWLEDGEMENTS

HPN PhD Thesis, NTNU iv

v HPN PhD Thesis, NTNU

HPN PhD Thesis, NTNU vi

3.4 Turbine Design .............................................................................................................3-11

vii HPN PhD Thesis, NTNU

APPENDIX A HYDRAULIC TURBINE ROTOR REPORT AT BEP………………………………...A-1

APPENDIX C PAPER I………………………………………………………………………...C-1

HPN PhD Thesis, NTNU viii

ix HPN PhD Thesis, NTNU

HPN PhD Thesis, NTNU x

U U + u m/s f (VPN ) Function of velocity of particle ‐

U Reduced peripheral m/s Re , p Particle Reynolds’s number ‐

xi HPN PhD Thesis, NTNU

ρ Density of fluid kg/m3 λ Turbine coefficient ‐

ω Angular velocity red/s ϕ Diffusion of turbulence stress W/kg

σ Normal stress N/m2 τ Shear stress N/m2

ε Turbulence eddy dissipation m2/s3 Ω Speed number ‐

η Efficiency ‐ β Blade angle degree

γ Impact angle rad γ 0 Maximum Impact angle rad

μ Dynamic viscosity Kg/N‐m μ eff Effective viscosity kg/N‐m

π Constant ‐ κ Turbulence kinetic energy m2/s2

α0 Maximum Guide vane angle degree α Guide vane/ Impingement degree

HPN PhD Thesis, NTNU xii

xiii HPN PhD Thesis, NTNU

1-1 HPN PhD Thesis, NTNU

Main river Small river Major river Total power Economically

Sapta Kosi 3.6 18.75 22.35 10.86

Sapta Gandaki 2.7 16.95 20.65 5.27

Southern 1.04 3.07 4.11 0.88

Country Total 10.84 72.45 83.29 42.13

HPN PhD Thesis, NTNU 1-2

1.1.2 Sediment erosion problems

1-3 HPN PhD Thesis, NTNU

1.2 MOTIVATION FOR THIS RESEARCH

HPN PhD Thesis, NTNU 1-4

1-5 HPN PhD Thesis, NTNU

1.3 OBJECTIVE OF THIS STUDY

1.4 STUDY METHODOLOGY

HPN PhD Thesis, NTNU 1-6

1.5 SCOPE OF STUDY

1.6 OUTLINE OF THE THESIS

1-7 HPN PhD Thesis, NTNU

2.2 MECHANISMS OF SOLID PARTICLE EROSION

2-1 HPN PhD Thesis, NTNU

Cutting Fatigue Brittle Melting

Cyclic Non‐ cyclic Loss of fluid

2.3 ABRASIVE AND EROSIVE WEAR

HPN PhD Thesis, NTNU 2-2

2.3.1 Mechanism of abrasive wear

2-3 HPN PhD Thesis, NTNU

2.3.2 Mechanisms of erosive wear

2.3.2.1 Cutting erosion

HPN PhD Thesis, NTNU 2-4

2.3.2.2 Surface fatigue

2.3.2.3 Plastic deformation

2.3.2.4 Brittle fracture

2.4 CONTROLLING PARAMETERS FOR EROSIVE WEAR