Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
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
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
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.
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.
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
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
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
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
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 ‐
Greek Symbols
ηh Hydraulic efficiency ‐ φ
Viscous stress N/m2
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
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
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.
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)
Karnali and
3.5 32.68 36.18 25.1
Mahakali
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.
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).
(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.
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.
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
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.
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.
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.
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).
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
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
Penetration Plastic
of deformation
cutting edge to
failure
Figure 2‐1 Mechanisms of solid particle erosion
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.
(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.
(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)
abrasion or cutting mechanism. The material is removed by scouring or scrapping by
sharp edges of the particles forming short track‐length scars.
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)
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.
1
FD = ⋅ C D ⋅ AP ⋅ ρ ⋅ U 2
2 Equation 2‐1
FL = C L ⋅ AP ⋅ ρ f ⋅ U 2
Equation 2‐2
π
FB = ⋅ d P ⋅ (ρ P − ρ f ) ⋅ g
3
Equation 2‐3
6
FR = m ⋅ (− 2 ⋅ ω ⋅ U P − ω ⋅ ω ⋅ rP ) Equation 2‐4
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.)
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.
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.
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.
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.
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).
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
Figure 2‐11 Effect of flow on erosive wear (Stachowiak and Batchelor, 2006)
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).
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.
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.
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.
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,
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.
⋅ (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.
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.
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
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.
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.
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.
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.
(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.
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
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
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.
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).
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.
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.
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.
• 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.
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.
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.
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
⎝ ⎠
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.
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
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.)
400
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
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.
case of a Francis turbine (Thapa and Brekke, 2004). These two forces are given by the
following equations.
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.
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
(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.
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.
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
determined that spherically shaped particles had higher settling velocities than particles
with other shapes.
Velocity of particle in swirl flow Velocity of particle for different size & shape
15 13
12
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
Particle velocity [m/s]
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
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
(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
0.2
0.1
0 10 20 30 40
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.
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.
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
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.
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
⋅ 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
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.
⎛ 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
⎝
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.
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.
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
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.
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.
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
[ 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
⎧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.
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.
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
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
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
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.
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
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.
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.
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.
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.
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
(a) Volume flow rate and guide vane angle (b) Efficiency and guide vane angle
Figure 7‐1 Hydraulic performance of turbine
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.
(a) Stay vanes with covers (b) Suction side of stay vane
Figure 7‐2 Predicted erosion pattern on stay vanes
(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.
(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
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
1 1
0.8 0.8
[kg/m2s]
[kg/m2s]
0.4 0.4
0.2 0.2
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
(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]
0.4 0.4
0.2 0.2
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
(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
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
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
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.
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.
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.
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
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 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.
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.
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
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
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.
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.
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.
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.
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.
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.
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.
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.
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
5. Blade Loading Charts
Chart 1 Blade Loading at 20% Span
Chart 2 Blade Loading at 50% Span
Chart 3 Blade Loading at 80% 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
Figure 9 Velocity Vectors at 50% Span
Figure 10 Velocity Vectors at 80% 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
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
Turbulence Model = SST
Turbulent Wall Functions = Automatic
S1 Passage Fluid Water
Buoyancy Model = Non Buoyant
Domain Motion = Stationary
Turbulence Model = SST
Turbulent Wall Functions = Automatic
S2 Passage 2 Fluid Water
Buoyancy Model = Non Buoyant
Domain Motion = Stationary
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
Conservative Interface Flux
R1 Periodic 1 Side PER2 Passage 3 Interface
Turbulence = Conservative
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 =
R1 to R1 Conservative Interface Flux
R1 PER2 2 OUTBlock Interface
Periodic Side 2 Turbulence = Conservative
Interface Flux
Mass And Momentum =
Conservative Interface Flux
R1 R1 to S2 Side 1 INFLOW Passage 3 Interface
Turbulence = Conservative
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
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
Flow Regime = Subsonic
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
Conservative Interface Flux
S1 Periodic 1 Side PER1 Passage Interface
Turbulence = Conservative
1
Interface Flux
Mass And Momentum =
S1 to S1
Conservative Interface Flux
S1 Periodic 1 Side PER2 Passage Interface
Turbulence = Conservative
2
Interface Flux
Mass And Momentum =
Conservative Interface Flux
S1 S2 to S1 Side 2 OUTFLOW Passage Interface
Turbulence = Conservative
Interface Flux
Wall Influence On Flow = No
S1 S1 Blade BLADE Wall
Slip
Wall Influence On Flow = No
S1 S1 Hub HUB Wall
Slip
Wall Influence On Flow = No
S1 S1 Shroud SHROUD Wall
Slip
Mass And Momentum =
Conservative Interface Flux
S2 R1 to S2 Side 2 OUTFLOW Passage 2 Interface
Turbulence = Conservative
Interface Flux
Mass And Momentum =
Conservative Interface Flux
S2 S2 to S1 Side 1 INFLOW Passage 2 Interface
Turbulence = Conservative
Interface Flux
Mass And Momentum =
S2 to S2
Conservative Interface Flux
S2 Periodic 1 Side PER1 Passage 2 Interface
Turbulence = Conservative
1
Interface Flux
Mass And Momentum =
S2 to S2
Conservative Interface Flux
S2 Periodic 1 Side PER2 Passage 2 Interface
Turbulence = Conservative
2
Interface Flux
Wall Influence On Flow = No
S2 S2 Blade BLADE 2 Wall
Slip
Wall Influence On Flow = No
S2 S2 Hub HUB 2 Wall
Slip
Wall Influence On Flow = No
S2 S2 Shroud SHROUD 2 Wall
Slip
4. Tabulated Results
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
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
5. Blade Loading Charts
Chart 1: Blade Loading at 20% Span
Chart 2: Blade Loading at 50% Span
Chart 3: Blade Loading at 80% 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
A-11 HPN PhD Thesis, NTNU
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
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
A-13 HPN PhD Thesis, NTNU
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
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
Figure 9: Velocity Vectors at 50% Span
Figure 10: Velocity Vectors at 80% Span
A-17 HPN PhD Thesis, NTNU
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
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
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
4. Tabulated Results
Table 6 Performance Results
Table 7 Summary Data
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
3. Physics Report
Table 4: Domain Physics for 685 SST without particle after corecting_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
Turbulence Model = SST
Turbulent Wall Functions = Automatic
S1 Passage Fluid Water
Buoyancy Model = Non Buoyant
Domain Motion = Stationary
Turbulence Model = SST
Turbulent Wall Functions = Automatic
S2 Passage 2 Fluid Water
Buoyancy Model = Non Buoyant
Domain Motion = Stationary
Table 5: Boundary Physics for 685 SST without particle after corecting_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
Conservative Interface Flux
R1 Periodic 1 Side PER2 Passage 3 Interface
Turbulence = Conservative
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 =
R1 to R1 Conservative Interface Flux
R1 PER2 2 OUTBlock Interface
Periodic Side 2 Turbulence = Conservative
Interface Flux
Mass And Momentum =
R1 R1 to S2 Side 1 INFLOW Passage 3 Interface Conservative Interface Flux
Turbulence = Conservative
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
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
Flow Regime = Subsonic
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
Conservative Interface Flux
S1 Periodic 1 Side PER1 Passage Interface
Turbulence = Conservative
1
Interface Flux
Mass And Momentum =
S1 to S1
Conservative Interface Flux
S1 Periodic 1 Side PER2 Passage Interface
Turbulence = Conservative
2
Interface Flux
Mass And Momentum =
Conservative Interface Flux
S1 S2 to S1 Side 2 OUTFLOW Passage Interface
Turbulence = Conservative
Interface Flux
Wall Influence On Flow = No
S1 S1 Blade BLADE Wall
Slip
Wall Influence On Flow = No
S1 S1 Hub HUB Wall
Slip
Wall Influence On Flow = No
S1 S1 Shroud SHROUD Wall
Slip
Mass And Momentum =
Conservative Interface Flux
S2 R1 to S2 Side 2 OUTFLOW Passage 2 Interface
Turbulence = Conservative
Interface Flux
Mass And Momentum =
Conservative Interface Flux
S2 S2 to S1 Side 1 INFLOW Passage 2 Interface
Turbulence = Conservative
Interface Flux
Mass And Momentum =
S2 to S2
Conservative Interface Flux
S2 Periodic 1 Side PER1 Passage 2 Interface
Turbulence = Conservative
1
Interface Flux
Mass And Momentum =
S2 to S2
Conservative Interface Flux
S2 Periodic 1 Side PER2 Passage 2 Interface
Turbulence = Conservative
2
Interface Flux
Wall Influence On Flow = No
S2 S2 Blade BLADE 2 Wall
Slip
Wall Influence On Flow = No
S2 S2 Hub HUB 2 Wall
Slip
Wall Influence On Flow = No
S2 S2 Shroud SHROUD 2 Wall
Slip
4. Tabulated Results
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
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
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.
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
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:
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
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,
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:
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.
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)
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)
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.
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
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.
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)
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 Absolute velocity, reduced ‐ Ω Speed number ‐
value
D Diameter m p Number of particles 1/m2
E Energy m2/s2 ρ Density kg/m3
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
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.
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
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
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
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
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
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
350
should be almost in the radial direction 400
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
increase if the operating head increases.
Particle velocity [m/s]
11.5
The size of the particle is inversely 6mm ran. sand
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
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
with a given diameter would stay at the 0.3
orbit of the injecting radius until either Steel particle 5mm size
Trian. ceramic 6mm size
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 ‐
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.
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
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
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)
⎧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.
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
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
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
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
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
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
Index 197
* 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
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
-+--==·-·-·-·-.II£ I
I
I
I
I
I
I
f
I
I
I
I!I"t-·-·-·-
I
L__v=_o_:._o_=_o_ _--.+.-----__.l
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"·
Introduction to Waterhammer 9
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
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
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
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
Introduction to Waterhammer 11
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.
(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
12 Analysis of Surge
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)
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
Therefore
i; =
J[
v 4fL/d
{(A/AoCc)-1}2 + 4fL/d
J
Introduction to Waterhammer 13
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
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
14 Analysis of Surge
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
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.
for section 2:
16 Analysis of Surge
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
f
-sect<on 2 - 1 - Section 1 - 1
Ar
1----L2- - - • + • - - -
FIG. 2.7
LnAl
.+-
An
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.
Introduction to Waterhammer 17
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
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)
20 Analysis of Surge
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,
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
Elastic Theory for a Simple Pipe 21
Rocks: Granite 50 7
Limestone 55 8
Quartzite 40 6
Sandstone 3 0·4
Schist 14 2
--------------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
24 Analysis of Surge
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)
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)
Elastic Theory for a Simple Pipe 25
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.
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)
Static level
·-·-·-·-·-·-·-·-·
FIG. 3.4
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)
For other types of pipe and tunnel the 'modified' bulk modulus is as
follows:
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
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.
30 Analysis of Surge
SOLUTION
(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
(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)
2·1 X ton)
) = 1100m/s
=; J(;)
=; Jc(~ ~ fH))
for a rigid pipe (3.5)
h
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
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
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)
h,
Before
H0 +h, i Ho p Ho
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.
fr = CrFi; F2 = Fa = CtFi; Ct - Cr = 1
36 Analysis of Surge
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)
Valve
closes
FIG. 4.7
40 Analysis of Surge
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
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
!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
(b)
T l __ _
I h
~~I_..___--' •
I = 0 T fi- fL+ T T1me
FIG. 5.1
Example 5.1
When the valve at the end ofa 3900 m long pipeline closes in4i seconds the velocity
is reduced as follows-
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.
44 Analysis of Surge
2L 2 X 3900
SOLUTION p, = -a = 1300
= 6 seconds
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
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.
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
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
2L-x
t----
a
and - hmax - - T - - due to the f-waves
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.
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
( ~ 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.
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
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
(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.
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
54 Analysis of Surge
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.
Q Q
FIG. 6.5
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
56 Analysis of Surge
(a}
Oo
A 8
(b)
H
'fes
I
(c)
H
A o
(d) H I liz+<
I I
81 I
FIG. 6.6
Graphical Analysis 57
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.
58 Analysis of Surge
(a)
(b) 1j;B2
H I
/ A! -.\fAI
-----t:', A3
R ''
FIG. 6.7
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.
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
(a)
A ==::::::;===~c==~2=~B
(b)
FIG. 6.10
(a)
A========~========~c~==~2==~B
2
...s.
..L fLt : a
1
(b)
FIG. 6.11
A====~====~c~~+~~2======~B
j_ fL = .!:-L j_ fL =..!:L
2 I a1 2 2 a2
FIG. 6.12
Graphical Analysis 65
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.
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
Graphical Analysis 67
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
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
68 Analysis of Surge
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)
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.
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,'
J. Instn Wat. Engrs, (1953), 7, pp. 87-125.
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
(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
72 Analysis of Surge
~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.
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.
Waterhammer in Pumping Mains 73
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-
74 Analysis of Surge
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.
Waterhammer in Pumping Mains 75
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.
76 Analysis of Surge
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
(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
78 Analysis of Surge
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.
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
80 Analysis of Surge
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
Waterhammer in Pumping Mains 81
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.
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.
82 Analysis of Surge
(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
Waterhammer in Pumping Mains 83
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.
84 Analysis of Surge
£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.
Waterhammer in Pumping Mains 85
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,'
Trans. Am. Soc. mech. Engrs, (1956), 78, pp. 1297-1304.
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,'
Trans. Am. Soc. mech. Engrs, (1937), 59, pp. 661-668.
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.
86 Analysis of Surge
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,'
J. Instn Wat. Engrs, (1953), 7, pp. 87-125.
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
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.
~:8~:--·-l~
L,
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).
90 Analysis of Surge
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.
(c)
Ho
I
I
I
FIG. 8.3 (c)
(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.
92 Analysis of Surge
e• ...,.
(a)
(b)
(c)
FIG. 8.4
(a)
\
(b)
FIG. 8.5
94 Analysis of Surge
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.
(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.
96 Analysis of Surge
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.
Waterhammer in Hydro-electric Supply Pipes 97
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
FIG. 9.1
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
Introduction to Surge Tanks 101
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.
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).
JL
Variable area tonk
Simple tonk
JL
Differential tonk
JL
Throttled lank Spilling tank
n
.Closed tonk
JUL
Multiple tanks
FIG. 9.2
f---------811. diam.--------1
FIG. 9.3
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
Introduction to Surge Tanks 105
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,'
Proc. Instn mech. Engrs, (1963), 177, pp. 951-970.
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.
106 Analysis of Surge
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.
Introduction to Surge Tanks 107
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
problems,' Proc. Instn civ. Engrs, (1963), 25, pp. 267-286.
10 Theory of Mass Oscillation
(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,
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).
110 Analysis of Surge
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)
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 ·
Theory of Mass Oscillation 111
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
FIG. 10.2
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
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.
Theory of Mass Oscillation 113
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
114 Analysis of Surge
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
Fm.10.4
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)
116 Analysis of Surge
{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'
0 = (1- x')8
(10.11)
x'
Theory of Mass Oscillation 117
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
(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
Theory of Mass Oscillation 119
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.
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
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.
120 Analysis of Surge
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.
Surge tank A
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
122 Analysis of Surge
SOLUTION
For the first trial assume x2' = x1 = 0·9885, and then proceed as follows:-
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
- 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
x' Z'
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.
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
Theory of Mass Oscillation 125
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.
(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
126 Analysis of Surge
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.
Theory of Mass Oscillation 127
(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.
128 Analysis of Surge
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,'
Proc. Instn mech. Engrs, (1956), 170, pp. 265-271.
11 Finite Difference Methods I
~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
~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
(11.7)
vA- Q
tl.y = = tl.t (11.8)
As
Surge tank A
· - - · - - · d5 =10m·-·
A5 =78·54m 2
L, = 203·94
A,= 16
132 Analysis of Surge
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
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.
""""'
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
134 Analysis of Surge
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
SoLUTION FT
(~:r
= KT -2g- = 509·85 s2/m
The calculation is shown in Table 11.3.
Finite Difference Methods I 135
Surge tank C
Yst=-18·22m
dr= lm
Kr=I·O
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)
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.
136 Analysis of Surge
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.
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 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
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.
Finite Difference Methods I 139
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.
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.
140 Analysis of Surge
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:
SUM
I !:1t L1y y L1u u v Fpvjvj (4) + (8)
Line (I) (2) (3) (4) (5) (6) (7) (8) (9) Line
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.
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.
SUM
t /::;.( y !::;.y u v /::;,v FPvlvl (3) + (8)
Line 0) (2) (3) (4) (5) (6) (7) (8) (9) Line
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.
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
SUM
t M Q y ~y u v ~v Fpvlvl (4) + (9)
Line (I) (2) (3) (4) (5) (6) (7) (8) (9) (10) Line
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
Finite Difference Methods II 147
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
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
148 Analysis of Surge
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
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
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.
150 Analysis of Surge
Surge tonk A*
d5 = 10m
· - - · - - · A5 =78·S4m 2•
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
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.
FIG. 12.2
Reservoir c
Power station
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.
dvb
Lrb dt + YJ + FPbVblvbl = 0
dvc
Lrc dt + Y- YJ + FpcVclvcl = 0
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
154 Analysis of Surge
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.
-dv
dt
= -AI {As-
du
dt
+ 2-3Csha~-
"dha + -dQ}
dt dt
-Spillway-
---Surge t a n k - - -
FIG. 12.4
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
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
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
Finite Difference Methods II 157
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
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)
158 Analysis of Surge
(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).
(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)
vA = AsUs + Q
c
v-
FIG. 12.10
6A
160 Analysis of Surge
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.
u
-T
y
_L_·-·-·-·-·
v area A
FIG. 12.11
The continuity equation is
Av = Q- Asu
(a) Both tanks upstream of the turbines. The continuity equations are:
D..t
D..ya =- (AaVa - Abbb) (12.14)
Asa
(12.15)
Finite Difference Methods II 161
The dynamic equations are
!:!..t
!:!..va = - - (ya + FpaValval + FTaUa!Ual) (12.16)
Lra
!:!..t
!:!..vb = - - (yb- Ya
Lrb
+ FPbVblvbl + FTublubi) (12.17)
(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
2 1 4 1
Yrl = 1 - 3Fro + 9F ro2 + 135 Fro3 + 270 Fro 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
164 Analysis of Surge
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.
Approximate Methods 165
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'.._
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,
(13.8)
(13.9)
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
Approximate Methods 167
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
Fro
FIG. 13.3
Approximate Methods 169
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.
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
FIG.l3.4
170 Analysis of Surge
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 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
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.
172 Analysis of Surge
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:
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*
Approximate Methods 173
Therefore hTo = 1!.. = 21 "82 = 15·43 m
y'2 y'2
hTO = kT VT0 2
%
and UTO = VO ~ = VO
AT
(!:..) 2
~
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).
174 Analysis of Surge
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
or dv = --(y + hF + hT)dtfLr
and dy =(~_g) dt (14.2)
Ar As
Surge tank A
· - - · - - - . d5 =10m_ _ _
As= 78·54m2
L, = 203·94
A,= 16
FIG. 14.1
FIG. 14.2
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
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
180 Analysis of Surge
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. 14.7
182 Analysis of Surge
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.
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.
Graphical Methods 183
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
problems,' Proc. Instn civ. Engrs, (1963), 25, pp. 267-286.
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
~ --==-----
(a) (b)
(c) (d)
FIG. 15.1
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
from which C
gh
= (VI- ( 1 +-
Bh) -VI (15.1)
vz) 2A!
FIG. 15.4
and
c = gh Bh)
v _ v1A1 - cBh ( 1 + 2A - v1
1 A1 + Bh 1
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
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.
Open Channel Surge 189
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
190 Analysis of Surge
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.
Open Channel Surge 191
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
192 Analysis of Surge
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
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
194 Analysis of Surge
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
y water level in surge tank above static level of reservoir (i.e. measured
positive upwards)
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