Sei sulla pagina 1di 220

Dynamic behaviour of hydraulic structures

Part C Calculation methods and experimental investigations

P.A. Kolkman T.H.G. Jongeling

Delft Hydraulics

The three manuscripts (parts A, B and C) were put in book form by Rijkswaterstaat in Dutch in a limited edition and distributed within Rijkswaterstaat and Delft Hydraulics. Dutch title of that edition is: Dynamisch gedrag van Waterbouwkundige Constructies. Ten years after this Dutch version Delft Hydraulics decided to translate the books into English thus making these available for English speaking colleagues as well. Because of in the text often referred is to Delft Hydraulic reports made for clients (so with restrictions for others to look at) we decided to limit the circulation of the English version as well. However, the books are of value also without perusal of these reports. The task was carried out by Mr. R.J. de Jong of Delft Hydraulics. Translation services were provided by Veritaal (www.veritaal.nl). Delft Hydraulics 2007

Delft Hydraulics

Table of contents Part C List of symbols Part C 1 2 3 INTRODUCTION ..............................................................................................................1 ANALYSIS OF DESIGN AND CONDITIONS ...............................................................5 CALCULATION METHODS FOR DYNAMIC BEHAVIOUR OF STRUCTURES IN WATER ..............................................................................................................................9 3.1 Considerations for calculating in the frequency domain or in the time domain 9 3.2 Calculation of the added water mass for two-dimensional situations (ignoring wave radiation) .......................................................................................................12 3.2.1 Assessment of the frequency range in which the added water mass is not frequency-independent anymore ......................................................................12 3.2.2 Added water mass calculated with potential theory .........................................13 3.2.3 Assessment of added water mass based on schematized flow pattern .............14 3.2.4 Examples of complete calculations ..................................................................15 3.3 Calculations in the frequency domain ..................................................................20 3.3.1 The single (degree of freedom) mass spring system ........................................21 3.3.2 The single (degree of freedom) mass spring system in water ..........................24 3.3.3 Response of a double (degree of freedom) mass spring system (direct method) ..........................................................................................................................26 3.3.4 Modal Analysis in case of a double (degree of freedom) mass spring system.28 3.3.5 General formulation of a system with multiple degrees of freedom ................30 3.3.6 Coupled systems with structure components and fluid components................34 3.3.7 Examples of coupled systems with structure components and fluid components ..........................................................................................................................35 3.4 Calculations in the time domain using the indirect method ...............................43 3.4.1 Modal Analysis and impulse-response method in case added mass and damping are frequency-independent ................................................................43 3.4.2 The impulse-response function in case added mass and damping are frequencydependent..........................................................................................................44 3.5 Calculations in the time domain with the direct method ....................................48 3.5.1 General .............................................................................................................48 3.5.2 Response of a single mass spring system to an external load ..........................48 3.5.3 Coupled systems with structure components and fluid components................51 4 CALCULATION METHODS FOR WAVE IMPACTS .................................................61 4.1 General ....................................................................................................................61 4.2 Impulse theory ........................................................................................................62 4.3 The linear shock wave model.................................................................................64 4.3.1 Wave impact against a rigid wall .....................................................................64 4.3.2 Rigid wall and air-water mixture......................................................................65 4.3.3 Wave impact against a compressible wall........................................................66 4.4 The non-linear shock wave model.........................................................................67 4.5 The flow-pressure model (ventilated shocks).......................................................70 4.6 The air compression model....................................................................................71 4.6.1 The linear air compression model ....................................................................72 4.6.2 The non-linear air compression model .............................................................73 4.7 Numeric calculation of the pressure function (in time) in case of a wave impact ..................................................................................................................................76 4.8 Extrapolation of results from a scale model to prototype values .......................77 i

Delft Hydraulics

4.9 Influences on impact load due to a responding structure .................................. 79 SCALE MODELS ........................................................................................................... 81 5.1 Introduction............................................................................................................ 81 5.1.1 General ............................................................................................................. 81 5.1.2 Investigation strategy for a real project............................................................ 82 5.2 Scale rules and scale effects in case of vibration investigations and investigations of wave loads .................................................................................. 83 5.3 Classification of scale models for vibration and wave impact investigations... 90 5.3.1 Classification with respect to reproduction of geometry ................................. 90 5.3.2 Classification with respect to reproduction of dynamic properties.................. 92 5.4 Possible critical aspects of dynamic models......................................................... 94 5.5 Verification of the model technology.................................................................... 95 5.6 Measuring system and data processing.............................................................. 101 5.6.1 General ........................................................................................................... 101 5.6.2 Instrumentation .............................................................................................. 102 5.6.3 Monitoring and registration of measuring signals ......................................... 103 5.7 Elaboration of measuring results........................................................................ 104 5.7.1 General ........................................................................................................... 104 5.7.2 Statistical elaborations ................................................................................... 105 5.7.3 Elaborations in the time domain .................................................................... 107 5.7.4 Elaborations in the frequency domain ........................................................... 110 6 EXAMPLES OF SCALE MODELS FOR DYNAMIC INVESTIGATIONS .............. 119 6.1 Rigid model with flowing water, for vibration investigations.......................... 120 6.2 Rigid model for investigation of wave load........................................................ 123 6.3 Single (degree of freedom) mass spring system for vibration investigations; translatory............................................................................................................. 130 6.4 Single (degree of freedom) mass spring system for vibration investigations; rotating.................................................................................................................. 138 6.5 System with multiple degrees of freedom in case of floating gate ................... 141 6.6 Multiple (degree of freedom) mass spring system for response investigations in case of flow and waves ......................................................................................... 145 6.7 Continuous-elastic model for vibration investigations ..................................... 148 6.8 Continuous-elastic model for investigations of wave loads and vibrations .... 152 7 INVESTIGATIONS OF PROTOTYPE STRUCTURES.............................................. 159 7.1 General.................................................................................................................. 159 7.2 Vibration measurements ..................................................................................... 160 7.3 Wave impact investigations................................................................................. 163 7.4 Elaboration of measuring results........................................................................ 164 7.5 Experiences with respect to vibration and wave impact measurements......... 164 8 REFERENCES............................................................................................................... 171 8.1 Delft Hydraulics Reports (in Dutch) .................................................................. 171 8.2 Other reference material ..................................................................................... 177 APPENDIX I ......................................................................................................................... 181 DEDUCTION OF THE RESPONSE FUNCTION IN THE TIME DOMAIN FROM THE TRANSFER FUNCTION IN THE FREQUENCY DOMAIN.................................................... APPENDIX II ........................................................................................................................ 187 SCALE RULES AND SCALE EFFECTS FOR INVESTIGATIONS OF DYNAMIC BEHAVIOUR OF SCALE MODELS......................................................................................... APPENDIX III....................................................................................................................... 199 5

ii

Delft Hydraulics

DESCRIPTION OF A CALCULATION PROGRAM FOR THE DETERMINATION OF THE ADDED WATER MASS FOR A FLAT ROD WITH RECTANGULAR SECTION (STRIP) IN WIDE WATER IN CASE OF TRANSLATORY AND ROTATING VIBRATIONS .............................................................................................................................. INDEX BY SUBJECT (PART C)..........................................................................................207 Biographies of the authors......................................................................................................212

iii

Delft Hydraulics

List of symbols Part C a ai an A A Ap B c cc cw cw C Ca1 Ca2 CA d D D Da ei f f F F Fr Fw Fw F0 Fo g G h hc h1 h2 H(f) i I k k Kl Kw K kw L Lw LC iv = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = mean distance between water front and structure (m) amplitude factor of the ith natural vector (1, .i, ..n) (m) amplitude of the nthe vibration peak (m) surface area (m2) flow area beneath the gate (m2) cross-section area of tube (m2) beam of ship (m) damping constant (Ns/m) celerity compression wave in the structure (m/s) damping due to water, also called added damping (Ns/m) celerity compression wave in water (m/s) damping matrix (Ns/m) Cauchy number related to the spring stiffness Ca1 = k /( v 2 L) Cauchy number related to the module of elasticity Ca2 = E /( V 2 ) discharge coefficient related to the opening of the gate (-) initial thickness of enclosed air cushion (m) draught of ship (m) diameter of tube (m) damping number = c /( VL2 ) ith natural vector (n=1, ..i, .n) excitation frequency of the flowing fluid (1/s) vector of the force acting on the nods of the finite elements (N) force (or load) (N) force vector (N) Froude number = V / gh or V / gL load caused by flowing fluid (N) amplitude of a periodic load caused by flowing fluid (N) amplitude of force (N) floating force (N) gravitational acceleration (m/s2) weight (N) water depth or depth of structure under water (m) culvert height (m) upstream water depth downstream water depth transfer function in the frequency domain index indicating the ith Eigenvector impulse (vector) (Ns) spring constant (N/m) slamming coefficient compression module of air (m2/N) compression module of water (m2/N) stiffness matrix (N/m) hydrodynamic stiffness, or added water stiffness (N/m) length, in particular length of tube (m) representative length of added water mass (m) perimeter of the tube cross-section (m)

Delft Hydraulics

Lculvert L1 L2 m mw M Ma n n n N p p0 q ql qr Q Q0 Q' r R Re s S S t T T u v v v0 v0 V V V' Vs w We x x x y y Y Y

= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =

length of culvert (m) length of a tank (m) length of connecting culvert between tank and outer water (m) mass (kg) added water mass (kg) mass matrix (kg) mass number = m /( L3 ) n-direction is perpendicular to a surface number of cycles of natural oscillations scale factor; an index refers to the quantity involved a number, indicating distance in N L pressure (N/m2) initial pressure (N/m2) discharge per unit width (m2/s) discharge of left gate per unit width (m2/s) discharge of right gate per unit width (m2/s) discharge (m3/s) permanent part of discharge (m3/s) time dependent part of discharge (m3/s) complex number (real part is related to damping, the imaginary part is related to the frequency) (radials/s) radius, half of internal tube diameter (m) Reynolds number = vD/ (-) damping term for a damping force which is proportional to the square of the velocity of the vibration movement (Ns2/m2) stiffness matrix (N/m) Strouhal number = fL/V (-) time (s) duration (s) vibration period (s) velocity component in direction x (m/s) velocity component in direction y (m/s) velocity vector (m/s) velocity of vibrating object (m/s) initial water velocity (m/s) volume (m3) (reference) fluid velocity (m/s) time depending part of fluid velocity (m/s) ship velocity (m/s) vector of (linear independent) displacements of nodes (m) Weber number = w LV 2 / coordinate in horizontal direction displacement in direction x amplitude of vibration in direction x coordinate in horizontal direction (perpendicular to x direction) displacement in direction y amplitude of vibration of object (m) displacement vector (m) v

Delft Hydraulics

Y z z z

= = = = = =
= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =

amplitude of harmonic movement y(t) (m) coordinate in vertical direction water surface variation (m) dy/dt (m/s) or water surface variation in case of oscillation (m) ratio of potential and velocity vstructure (m) ratio between increase of gate discharge (per unit width) and water surface variation (m/s) phase angle (radials) air percentage air percentage after the front of the shock wave relative damping Poisson constant gap height (m) time independent part of gap height (m)
energy loss or energy transfer per period (Nm) difference in water depth between upstream and downstream (m) hydraulic head difference (index refers to over what location) (m) time dependent part of hydraulic head difference (m) mesh distance (m) pressure difference (N/m2) time step (s) discharge per length unit L and per unit width (m/s) step in potential (m2/s) ratio between wall velocity and velocity of incoming water (m/s) velocity potential (m2/s) phase angle of the ith natural frequency (1I, .n) (rad) root-mean-square (RMS) value of displacement (m) density of a fluid (kg/m3) density of structure material (kg/m3) discharge coefficient or contraction coefficient mean value of displacement in direction x (m) kinematic viscosity (m2/s) loss coefficient (index refers to what related) surface tension (N/m) RMS value of displacement in direction x (m) impact duration (s) time displacement (s) angular frequency of the vibration (rad/s) ith natural frequency of the vibration (1.i, .n) (rad/s) natural (angular) frequency of the vibration


* 0 E h H H' L p t q i c x x if n

vi

Delft Hydraulics

INTRODUCTION

Part C of this manual offers an overview of the calculation methods and experimental investigations for the benefit of designing new structures in a responsible way, improving existing structures and deepening our knowledge of dynamic phenomena. The investigations themselves will often require the input of specialists. This part aims to provide the designer with some understanding of the available methods of investigation. Two warnings in advance: This part goes into matters more deeply than the first two parts and indicates limitations; it is therefore more concerned with background information. Although the reader may occasionally get the impression that everything is known, the available methods of investigation are far from perfect. Many data are lacking and each approach has its limitations.

The methods of investigation that are available to test a design are: a. b. c. d. Analysis and desk research. Carrying out calculations in the frequency domain or in the time domain. Using scale models. Carrying out measurements of existing structures.

This order may be representative of the operating procedure in case of an analysis of a design for a structure that will be built, provided that method c, due to the costs and the duration of the investigation, will often only be used when methods a and b indicate the necessity for it. If the dynamic behaviour cannot be forecasted accurately, then method c will qualify earlier (both in time and in importance). Methods b and c will run simultaneously: the scale model serves to discover phenomena, and the calculation model will be started up according to circumstances, and will be calibrated with the available measuring results. Next, the calculation model will be elaborated with a number of variants under different conditions, and finally the selected variant will be verified again in the scale model. The order of methods a through d mentioned above is also followed in the chapter arrangement of this Part C. ad a. In order to carry out an analysis, naturally existing knowledge must be mobilized. Therefore a specialized descriptive bibliography is available, see in Delft Hydraulics A110. Flow-induced vibrations naturally do not only occur at hydraulic structures. Vibrations induced by flowing air or fluid (of wings, plating and other components, propellers, flow machines, flap gates etc.) may also occur in maritime engineering, aeronautical engineering and mechanical engineering. Analysis should preferably be carried out based on criteria that relate to the design, stiffness and damping. Using a checklist may serve as a starting point; for possible causes of vibrations, see for example Part A, Chapter 1. Refinements may be possible by checking the design on the basis of one or more criteria (for example as mentioned in Part A, Paragraph 4.4.6). 1

Delft Hydraulics

The evaluation of the design with respect to the occurrence of wave impacts is especially an evaluation of the shape of the design. Only when it is not possible to change the shape in such a way that these impacts may be avoided, it will be necessary to assess or calculate the effect on the structure. This part of the manual deals with the structural aspects of the design and the vibration and wave impact phenomena; most aspects with respect to vibrations and wave impacts have already been dealt with in Part A and Part B. ad b. As regards vibrations, Chapters 2 and 4 of part A have instigated calculating in the frequency domain. Chapter 3 of this part will discuss these matters more extensively. Calculating in the frequency domain is not useful for wave impacts. Calculating in the time domain is also introduced as a method in Chapter 3 of this part as well. Apparently, often too many basic data are unavailable in order to make a complete calculation useful. If the load is assumed to be known, it is however possible to calculate the response of the structure. For certain kinds of vibrations of gates and valves, the extent of self-excitation may be determined on the basis of calculations. Chapter 4 discusses various calculation methods for the determination of wave impacts in greater detail. ad c. There are many kinds of scale models that are used for the investigation of vibrations or response in case of wave impacts (overview or detailed models, completely rigid or completely elastic models, and everything in between). In some cases models in wind tunnels may be used as well. Each model offers possibilities of investigation, but also has its limitations. Scale models may have important scale effects. Scale models may be a very important tool in the design process for the investigation of dynamic phenomena. On the basis of the current state of affairs, scale models are in certain ways even more important than calculation models, as they may sometimes reveal unexpected phenomena. Often calculation procedures may only be set up on the basis of scale model observations. The calculation model serves to elaborate quickly with many variants in case of a systematic variation of conditions and design parameters. Chapter 5 provides an overview of the possibilities of scale models. Chapter 6 provides examples. ad d. Measurements of existing structures do not occur often, as in reality structures usually suffice reasonably well. In those cases there is no need regarding the structure itself to carry out more detailed measurements. But there is a number of structures that do not meet the requirements, giving cause for repairs and extra maintenance. Moreover, the supervisor needs to be continuously aware of the fact that vibrations or wave impacts may occur. In general, measurements of existing structures are indispensable, first of all as a means of verifying the calculation and scale model investigations that are used and secondly to put into perspective what might be accomplished with these methods. Practical experiences and measurements also give cause for revising the boundary conditions or the theoretical concept. Examples of this are:

Delft Hydraulics

observing cavitation and the ensuing occurrence of great dynamic loads; the growth of mussels and other kinds of fouling on structure components; the suction of air; damages of a gate edge, as a result of which vibrations are generated; the being out of action of gates, as a result of which the conditions at the other gates are more severe; the occurrence of high-frequency plate vibrations; these can never be forecasted adequately on the basis of scale model investigations in the design phase.

Prototype measurements are the final piece of the reviews of model and scale effects that are always present in scale models. Chapter 7 discusses a number of special aspects regarding the measurement of prototype structures. The following results are shown as examples of an interesting measurement in prototype that was carried out in January, 1995, i.e. the vibration measurements carried out at the gate of the tide lock at Ravenswaay. The results are rather surprising and as yet there is no definitive explanation for the cause of these vibrations.

Figure C1.1: Transverse waves upstream of the vertical lift gate in the tide lock at Ravenswaay, generated by vibrations of the closed gate. Figure C1.1 shows transverse waves that relate to the vibrations of the gate (80 m span) in the situation of the closed gate. The gate is of the same type as the one that is used in the storm surge barrier at Krimpen (Part A, Chapter 6, Example 6.2b), though in case of the tide lock at Ravenswaay a rubber bottom edge has been used to obtain a good seal. The vibrations of 3 Hz were horizontal and vertical combined, with the greatest acceleration 3

Delft Hydraulics

horizontal in the flow direction: 3 m/s2 (deflection approximately 10 mm). The waves are subharmonic standing transverse waves (see for a more detailed description of this type of transverse waves Part A, Paragraph 4.6) that could be observed on both sides of the gate. The fact that vibrations occurred when the gate was completely closed was surprising in itself, and it was only possible in case there was a leakage gap. The leakage was probably due to causes relating to the closing procedure or due to fouling on the sill. The vibrations were generated, starting from a hydraulic head of 2.5 m.

Delft Hydraulics

ANALYSIS OF DESIGN AND CONDITIONS

As Part C especially focuses on calculation methods and experimental investigations, only little attention is given to the general analysis of the design. Each design begins with a global analysis in which the expected wave and flow conditions are discussed and on the basis of which the structure may be determined more specifically and in greater detail. Preceding a detailed study and/or investigation of the dynamic behaviour of structure components that may be critical (such as gates and trash racks), first the local boundary conditions for those components need to be determined separately. Based on the hydraulic head of the structure, the local hydraulic head of the gate and the flow velocity at a trash rack need to be deduced. The local wave conditions need to be determined on the basis of the global wave conditions. For example, a different combination of water level of the hydraulic head applies to each opening percentage at a gate or valve. The wave height may vary strongly with the water level and will also be influenced by the discharge that flows through the structure. The requirement that a structure should not experience great dynamic loads, may lead to limitations in the operational management of the engineering structure. These points will be further discussed below, in which the aspects of vibrations and wave impacts will be treated separately. Regarding the local hydraulic head and the local wave conditions This especially focuses on the local hydraulic head, for as far as local wave conditions are concerned, in any event these will require a calculation using a standard calculation program. The local waves may be considerably higher and steeper than the incoming wave due to reflection, canal narrowing, sloping (ascending) bottom and local shapes that function like a hoop net. At a gate or a valve, the local hydraulic head may deviate considerably from the hydraulic head of the entire structure. As a result of this, the determination of the local hydraulic head first requires an impression of whether the inertia of the water in the culvert should be included in the calculation or not. If this is the case, the calculation of the local hydraulic head becomes very complex (see Delft Hydraulics Report R1506). The following procedure is used. First, the gate drag term, gate, or the coefficient of the flow, Ca, needs to be determined for each opening percentage. This is defined as:
H gate = gate
2 Vculvert 2g

(C2.1)

(in which H = hydraulic head, V = flow velocity, g = gravitational acceleration) and:

Q = Ca Agate 2 g H gate (with Q = discharge, Agate = flow-through opening of the gate)

(C2.2)

Given the closing program of the gate or valve, the drag term needs to be determined as a function of time.
5

Delft Hydraulics

Next, the movement equation of the water in the culvert needs to be analyzed. This equation is as follows:
H = H drag + H inertia
2 Vculvert Lculvert dVculvert = ( culvert + gate ) + 2g g dt

(C2.3)

By first assessing how the discharge is distributed over time when the last term (the inertia term) in Equation C2.3 is being ignored, it is possible to obtain a first assessment of the discharge over time. Consequently, a first estimate may be made of the magnitude of the inertia term, and whether this is important relative to the total hydraulic head. The term apparent hydraulic head is introduced: H apparent = H Lculvert dVculvert g dt (C2.4)

As the term dVculvert/dt is only negative during the closing of the gate, the apparent hydraulic head of the structure during closure is greater than the real hydraulic head. For this reason it may be advisable to select a slower movement velocity for the closing of a gate or valve, than for the opening. If the apparent hydraulic head that is calculated in this way is not much greater than the real hydraulic head (5% seems to be a reasonable limit in this case), then a dynamic calculation of the closing program may not be needed. If a dynamic calculation is chosen, this means a calculation in the time domain with a small time interval, t: (using Equation C2.3) each time interval is calculated as follows:
Vculvert = dVculvert t dt (C2.5)

The result may be translated into a discharge as a function of the gate position. For the first or the last trajectory in which the valve is nearly or completely closed, the gate drag coefficient gate becomes very great, even up to infinitely great. To circumnavigate this problem in the calculation, a constant discharge coefficient, Ca (related to the gate opening) is used for the last part of the closing program, as well as the assumption of a constant hydraulic head, a negligible resistance of the culvert and an increasing or decreasing gate opening linear with time. This results in an analytical solution. The dynamic equation for the discharge then looks like this: Q = Ca Agate 2 g (H H inertia ) (C2.6)

in which it may be proven by way of an indirect demonstration that if Agate changes linear with time, the discharge increases or decreases linear with time as well, in which the inertia hydraulic head becomes a constant. From Equation C2.6 it may be deduced that:

Delft Hydraulics

L dQ d(Ca Agate ) dQ = 2 g (H culvert ) gAculvert dt dt dt

(C2.7)

As all coefficients with time are constants, dQ/dt follows from the solution of a quadratic equation. From the calculation of the discharge history mentioned above, the local pressure differences across the gate may be deduced. If necessary, the absolute pressure is determined as well; this may be important in connection with the occurrence (or not) of air suction or cavitation. To be able to calculate better the local pressure at the gate or valve, the inertia of the water in the culvert is subdivided into inertia upstream and downstream of the valve. It also needs to be taken into account, that directly downstream of the gate, the pressure is extra low (relative to the pressure a little bit further down in the culvert). This relates to the contracted nappe that may give cause for the recovery of potential energy. In case of a shipping lock, the lock chamber is already partially filled before the gates of the filling culvert are fully open. This means that for each gate position there is a different normative external hydraulic head of the filling or emptying culvert. If the lock chamber is filled through multiple gates, then the situation with one or more gates out of action needs to be considered as well. For it is possible that the selection of the complete filling program strongly relates to the avoidance of vibrations, cavitation and air suction.

Regarding vibrations It is useful to assess the resonance frequencies as a function of the span of the structure; this may certainly influence the choice of the span. From the data concerning the Strouhal number (e.g. Part A, Paragraphs 4.2, 5.2, 5.3, 5.4 and 5.7) an impression may be obtained about the dominant excitation frequencies due to turbulence in the wake of the structure and of structure components. It may be verified whether the design of a gate or valve may give cause for bath plug vibrations or for other vibration modes with self-excitation. About this, see Part A, Chapter 7 to begin with. Sliding gates experience a greater damping in case of possible vibrations and are safer where vibrations are concerned. In case of gates and valves the choice of the type of seal (or the omission of these) is very important for the possible occurrence of vibrations. For this, see also Part A, Chapter 7. In case of gates, not only the flow through the gate opening plays a role, but also the remaining surrounding flow. The surrounding flow and the pressure distribution in the gate shaft are important for the gate load and the possible generation of vibrations. For a (storm surge) barrier in a tidal area it makes a lot of difference whether it is eventually decided that the barrier is closed around the time of the turning point of the tide or that it is closed after waiting until a critical situation is reached. The reason for the decision to close the gate around the time of the turning point may be the wish to avoid powerful forces during the movement of the gates, or to avoid the probability of dynamic behaviour or the limiting of translatory waves.

Delft Hydraulics

Regarding wave impacts In the following (and this essentially applies to the whole of Part C), periodic wave loads are not further discussed, as there is no interaction with the dynamic properties of the structure. Wave impacts may occur in those situations in which the wave (for example when it is reflected) is locked in and in situations in which the moving water surface hits a wall that runs parallel to this water surface. Even a water surface that slowly moves up and down horizontally may generate a wave impact, if it hits a horizontal ceiling. In the same way as discussed in Part B, there is a number of aspects that need to be taken into account when designing a structure. In connection with the avoidance of wave impacts, the most important aspect is the positioning of the structure relative to the waves that are present. In those cases in which waves hit the structure, the water needs to be capable of moving horizontally or vertically unobstructed. As horizontal wave propagation is not possible in case of a closed gate, vertical parts of the structure preferably need to be smooth; the presence of, for example, horizontal ceilings and closed working platforms in the wave zone must be avoided. Wave reflection due to a wall or a closed gate may be observable even from great distances; therefore a bridge deck at a certain distance from the wall may be in a danger zone with respect to the occurrence of a wave impact. Horizontal girders of a gate or gates that are located at a critical height, are in less danger of being loaded by wave impacts during the outflow of a discharge, as it is much more difficult for waves to reach the gate in case of flow. If a horizontal element (e.g. a platform or a horizontal girder) is unavoidable, it must be perforated and preferably be replaced by a lattice structure with rods that are sharp or rounded at the bottom, instead of flat. If waves are capable of penetrating into a culvert in which the gate is positioned, then it is advisable to design the gate shaft in such a way, that a connection is maintained between the shaft and the culvert at the side of the incoming waves. This prevents the wave from being locked in and the water from being slowed down too suddenly. To prevent rattling due to wave impacts in case of a barrier that is surrounded on both sides by the same water level, prestressed sliding supports may be used. The evaluation of the design requires specialized knowledge. Part A (flow-induced vibrations) and Part B (wave impacts) contain much of that knowledge in summary. As far as the possible sensitivity to flow-induced vibrations is concerned, we refer to Part A, Chapter 1, in which a classification of possible vibrations may be found, Chapter 4, which deals with various kinds of vibration mechanisms at gates, Chapter 5, which deals with flow-surrounded objects and Chapter 7, which discusses remedies. The evaluation regarding wave impacts is especially an evaluation of the design. If the design cannot be adapted in such a way that wave impacts may be avoided, the effect of an impulse load must be assessed more closely. An analysis of the dynamic properties of the structure is then required. About this, see Chapter 6 of Part B.

Delft Hydraulics

CALCULATION METHODS FOR DYNAMIC BEHAVIOUR OF STRUCTURES IN WATER

Calculations of the dynamic behaviour of structures in water are not carried out very often. On the one hand this is to do with the complexity of the calculations themselves and on the other hand it is to do with the many unknown factors and insecurity concerning the magnitudes that need to be introduced. Also, the mathematical-physical description of the interaction between fluid and structure is not always that easy. This means that the development of knowledge in the past has mainly been focused on the avoidance of dynamic phenomena. The possibilities of calculation methods however are increasing rapidly. Next to this, a lot of understanding and experience has become available thanks to all the effort spent on large-scale structures. This especially applies to discharge sluices and storm surge barriers in the delta area of the Netherlands. This chapter therefore discusses at length the various available calculation methods.

3.1 Considerations for calculating in the frequency domain or in the time domain
There are various methods to calculate the dynamic behaviour of dry structures in case of a given load. Not all methods are suitable for structures in water. First of all, there is always a need to find fundamental equations that accurately describe the problem. In case of vibration problems, inertia always plays a role, both of the vibrating masses and of certain fluid elements. By taking periodic solutions as a starting point, in which all terms include a sinus function or a complex e-power, the acceleration is replaced by a displacement, multiplied with -2, after which the sinus or e-power may be removed from all terms. Only when this direct approach does not yield sufficient results, it will probably be decided to change to a more complex calculation in the frequency domain or to a calculation based on a mathematical model in the time domain. Whether a calculation in the time domain may lead to results, depends on the question whether the extra mass, damping and stiffness terms that are generated by the water, depend on the frequency of the response or not. For situations in which the added water mass is frequency-dependent, the natural frequencies of a structure in water cannot be calculated directly. The fact of the matter is, that for this calculation the added masses need to be known, and these again depend on the natural frequency. Calculation is only possible through repetition. By assessing the natural frequency, the frequency-dependent added water mass (or the matrix expression for it) may be calculated; by using this, the calculation of the natural frequencies may be carried out again. The frequency-dependency relates to the radiation of waves that are generated by the vibrating or moving structure, and therefore to the presence of a free water surface. If the vibration is high-frequency or the movement is of short duration, then the radiated waves have a short wave length and the dynamic pressures that relate to the waves may only be observed across a limited part of the depth (relative to the free water surface). The influence of the wave radiation therefore disappears at higher frequencies. See further Paragraph 3.2.1. Below, the various calculation methods are summarized and briefly explained.

Delft Hydraulics

a. b.

c.

d.

Direct calculation in the frequency domain; this is especially adequate in case of periodic loads. As the frequency is known, there is no fundamental difficulty when the added water mass is frequency-dependent. Indirect calculation in the frequency domain and in the time domain, while using the Modal Analysis (the theory of eigenvalues). Random loads may be considered; using this method however is limited to those conditions in which the added water mass, damping and stiffness are not frequency-dependent. Indirect calculation in the time domain with the impulse-response function technique, especially adequate for calculation of the response to loads of short duration. With this method it is also possible to include frequency-dependent terms in the calculation. The method has been used in the calculation of forces due to the collision of ships against, as an example, fenders, fender walls, mooring posts; instead of an independent external load, the load was generated due to the compression of, for example, one or more fenders. Direct calculation in the time domain, adequate in case of impulse loads, but only applicable for situations in which the added water mass, damping and stiffness are frequency-independent. In case of a linear system it is possible to calculate directly in the frequency domain, as a linear system in case of a harmonic load also generates a harmonic response with the same period. Consequently, time-dependent factors (such as sin(t) or eit) may be removed from the equations by division, as a result of which the remaining equations may be solved directly. This applies both to a single and to a multiple system. = angular frequency, i = indication of an imaginary number, t = time. If there are damping terms (i.e. proportional to the vibration velocity), then terms with both sin(t) and cos(t) are formed. This means that two systems of equations are formed, from which the phase shift between load and response may be determined. This outcome may also be obtained through complex calculations. If the periodic load F(t) is written in terms of Feit, then the movement y(t) of each mass point may be written as or eit. The eit may be removed from all terms by division. This way of differentiating with time is the same as multiplying with i. If F is a real number, then the outcome for consists of a real and an imaginary part, which results in the phase shift between the harmonic force and the harmonic vibration movement. Here again, two systems of equations are formed, i.e. by equalization of all real terms and all imaginary terms respectively. Here, the two equations directly lead to a solution; this solution is referred to as the particular solution. As the superposition principle applies to linear equations, the general solution, i.e. that of a freely damping out vibration, may be added to the particular solution, thus complying with the initial conditions at the moment in time t = 0. Calculation of the free vibration however is difficult when the added water mass is frequency-dependent. Paragraph 3.2.1 discusses this in more detail, using an example. The direct method of solution is adequate in case a response diagram needs to be determined (Figure A2.3 in Part A). Modal Analysis is used in the frequency domain (response in case of harmonic load), but also in the time domain and it is applicable to systems with multiple degrees of freedom with a load that is distributed randomly over time. Paragraph 3.3.5 discusses this method in more detail. Here it suffices to give a brief summary of the various elements of this calculation method.

ad a:

ad b:

10

Delft Hydraulics

The response of the structure is considered to be factorized into components, eigenvalues or modes. Each mode has one eigen (or natural) frequency, one eigen (or natural) vibration mode and one eigne value for the load; these are univocally coupled to each other. The amplitude relative to the corresponding mode may take on different values. Each of the vibration modes corresponds to one of the free vibrations (therefore without external load) of the undamped system. There is an eigen value for the load that corresponds to each mode (in fact, the load distribution) that statically generates the same deflection as the one that relates to the corresponding vibration mode of the free vibration. The amplitude of the free vibration and that of the corresponding load initially are not fixed. It may be shown that this load distribution, regardless of the frequency in which this load makes contact, is always accompanied by the same vibration mode and that the response curve in the frequency domain is identical to that of a single mass spring system. By factorizing the actually present load into the corresponding eigenvalues of loads (each therefore with the right load distribution across the masses as well), the amplitude of each of the eigenvalues of the loads is now fixed. At this point it is possible to calculate the response in time for each eigenvalue of the load, in the same way as for a single mass spring system. The sum of these responses results in the eventual response of the structure to the load. In this calculation, the structure may also be damped, provided that the damping is equally distributed across the springs or across the masses, proportional to the magnitude of the spring stiffness or the mass concerned. ad c: Using the impulse-response function is a method for calculating the response to a randomly distributed load in case the response of the structure to a load of short duration (the so-called unit impulse) is known. This load may be a known external load or it may have been generated by impressions of an elastic element (for example during a collision); also, in case the spring is non-linear, it is possible to calculate the response. For structures in water, in which the added damping and mass may be frequency-dependent, it is of course not possible to determine an impulse response. For an object that is not deformable however, and that moves in the water, a solution has been found. Paragraphs 3.4.1 and 3.4.2 discuss in what way the response to the unit impulse may be determined. Calculating directly in the time domain means that the calculation occurs in small steps. The second law of Newton (F = m*a) applies to each mass. The force F that operates on the mass m and causes an acceleration a, consists of the external load, of spring forces and damping forces. In case of complex systems, calculating with this method requires a lot of time and in practice it is therefore less suitable for the determination of the response to a periodic load for an entire frequency range. One advantage is that in case of calculating directly in the time domain, non-linear components may be included.

ad d:

All calculation methods mentioned above relate to structures with an external load or a freely moving or vibrating structure (for example, after an impact has occurred). The structure is schematized into a system consisting of masses, springs and dampers. The water generates both the external load and the passive forces, such as the added water mass, water damping and spring stiffness. As the water itself may also include components with their own natural vibration period (in case of fluids these are referred to as oscillations, rather than vibrations), in some cases it may be better to describe the fluid together with the structure, as one system

11

Delft Hydraulics

with, naturally, multiple degrees of freedom. Examples of such systems have been included in this chapter (Paragraphs 3.3.6 and 3.3.7).

3.2 Calculation of the added water mass for two-dimensional situations (ignoring wave radiation) 3.2.1 Assessment of the frequency range in which the added water mass is not frequency-independent anymore
Wave radiation is generated when, in a situation with a free water surface, water is displaced in horizontal direction due to the vibration of the structure. If the wave length of the radiated wave relative to the water depth is small, then the influence of the wave radiation is small as well. In that case, the velocity at which the wave travels is small, so that it radiates little energy. The pressures in a vertical due to the radiated wave decrease exponentially, when the distance from the water level increases; this distance needs to correspond to the wave length. The influence of the wave therefore is also smaller when the wave length is smaller. This results in the fact that h/ may be taken as a characteristic parameter of the influence of the wave radiation, and that again turns out to be proportional to 2h/g. In this, is the wave length, h is the water depth, or the depth location of the structure relative to the water level, and is the angular frequency of the vibrating structure. The vibration may be regular or may be a response vibration generated by an impact. In order to investigate the influence of 2h/g on the added water mass (and on the added damping as well), a study has been done (Delft Hydraulics Report W254) concerning a gate that is part of a vertical wall, and in which the gate is vibrating horizontally. The results may be found in Part A, Figure A3.1. If the water depth and/or the frequency are so great that 2h/g > 10, then the CL value (a measure of the added water mass) appears not to experience any influence from the radiating wave. In case this influence is no longer present, then the CL value has also become independent of 2h/g, and therefore of the vibration frequency. The water damping due to wave radiation may also be ignored at these high frequencies. In the equation: if 2h > 10 (C3.1) g then the added water mass is frequency-independent and the water damping due to wave radiation may be ignored. Other investigations also seem to confirm this result (for example investigations of a ship oscillating in transverse direction, Fontijn (1975)). Theoretically it may be demonstrated that, also for very low values of 2h/g, the added water mass tends toward a constant, corresponding to the value found in case the water level is a rigid wall. For practical purposes this is not relevant, as dynamic phenomena no longer occur at these low frequencies. In reality, the application of Equation C3.1 means that structures such as gates and trash racks are sufficiently stiff (i.e. showing a sufficiently high ) for carrying out calculations with a constant water mass. In case of structures in a closed culvert, wave radiation plays no role, so that there too the added water mass is frequency-independent. In case of long cylinders positioned in flowing fluid or in case of floating structures, the movement frequencies may be so low that there is frequency-dependency. Also the 12

Delft Hydraulics

periodic wave excitation is situated in that part of the frequency range in which the added water mass is strongly frequency-dependent. In calculating the collision forces during the mooring of ships at fender structures, an added water mass and frequency-dependent damping also need to be taken into account.

3.2.2 Added water mass calculated with potential theory


In Part A, Paragraph 3.2, the concept of added water mass has been discussed. For a vibrating structure in drowned condition (structures that are below the water surface or in a culvert), the influence of wave radiation may be ignored. This also applies to structures that penetrate the water level, provided the vibration frequencies are high enough (see previous paragraph). The influence of the free water surface is represented by the condition of constant pressure, in case the wave radiation is ignored. Contrary to the situation with a rigid wall, in case of a free water surface a flow component may also be present that is at right angles to the water level. In case of still water, the flow that is related to the added water mass may be calculated as potential flow. In that case, the following preconditions apply to the outlines of the fluid area:

the velocity component at fixed walls, and at right angles to the wall, equals 0; the precondition that the pressure is constant applies to the free water surface, which translates into the precondition that the potential at that location equals 0; a vibrating wall is replaced by a fixed wall that is modelled with sinks and sources in such a way that the velocity component of the water that approaches perpendicular to the wall corresponds to the amount of water that is displaced due to the vibration movement.

In case of potential flow without wave radiation, the water flows across the entire area, in-phase and proportional to the velocity of the vibrating structure. The relation between vibration velocity and potential at the surface of the vibrating structure appears to be a measure of the added water mass. This relation is the same, both in case of a permanent flow (in which the water displacement due to the vibrating object has been replaced by sources that generate a permanent discharge) and in case of an oscillating flow. Potential flow is based on very simplified equations for the water, i.e. the requirement that there is irrotational flow and the continuity condition. If u, v and w are the velocity components in the directions x, y and z, and if p is the pressure, then it follows that:

p 1 u p 1 v , and = = t t x y
The (velocity) potential (x,y,z,t) is defined as:
u= x

p 1 w = t z

(C3.2)

(C3.3)

Similar expressions hold for the directions y and z. The continuity condition results in: 13

Delft Hydraulics

u v w 2 2 2 + + =0= 2 2 2 x y z x y z

(C3.4)

From Equations C3.2 and C3.3 it follows that:


p= + constant t

(C3.5)

because:
p 2 u = = x xt t

From a potential flow calculation, the potential along the wall of the vibrating structure may be found, according to:
vstructure gives S

(3.6a)

in which vstructure = the velocity of the vibrating structure, and S = the potential found at the wall of the structure. It also holds that:
vstructure results in t s t

(3.6b)

vstructure is the acceleration of the structure, and for the potential flow in Equation C3.5 it was t results in the pressure. found that t For that reason the ratio is: s = (3.7) vstructure

apart from fluid, it is also a measure for the relation between the pressure generated per unit of acceleration. After totalizing over the contours of the surface (and taking into account that only the force in the direction of the vibration is relevant), this therefore results in a measure for the volume of the added water mass. has the dimension of a length (m).

3.2.3 Assessment of added water mass based on schematized flow pattern


In Part A, Paragraph 3.2.6, a number of cases were discussed, in which a nearly closed gate was included as part of a wall. If the gate vibrates at right angles to the wall, a first assessment of the added water mass may be carried out by assuming that the water approaches and runs off radially. This is illustrated in Part A, see Figures A3.9 and A3.10. The sector angle at which the water approaches and runs off, is determined by the geometry; in case of a gate as part of an infinite wall, the sector angle even becomes 180. For two-dimensional real situations, the area in which the flow is coupled to the vibration of the gate is always limited. If not, an infinitely great added water mass would be found. The location of the free water surface relative to the structure provides a measure for the limitation of the flow area. 14

Delft Hydraulics

In case of flow-surrounded objects there is no radial approach flow or runoff and a method of assessment as indicated above therefore is not useful in those cases. For two-dimensional situations such as often encountered at gates and valves, a potential calculation may be carried out according to the relaxation method. First, a potential of = 0 is established (moreover, this might as well be any random value, even with an irregular distribution across the area). If the right conditions have been introduced along the edges, it is now possible to determine the eventual potential distribution by systematically determining, point by point, the local potential based on the potentials of the surrounding points that are temporarily kept at fixed values. By repeating this process again and again, the ideal potential is approximated more closely; this process is convergent. This calculation method is also suitable for flow-surrounded objects in wide water surfaces. The calculation method in its complete form may be found in Kolkman (1988), and will be presented in the following paragraph for cases with an infinitely long strip in wide water that oscillates at right angles to its plane. This example has been selected as it is simple, and the outcome may be verified directly. In the description of the calculation program that has been included in Appendix III, this method has been made more concrete. This calculation procedure is also very well suited when using a spreadsheet.

3.2.4 Examples of complete calculations


Below, two examples are presented, the results of which may be verified with the data from an analytical calculation: the flat strip in wide water in case of translatory and rotational movements (from Kolkman (1988)). This calculation method is discussed in great detail, as it lends itself to being widely used.

a. Flat strip in wide water in case of translatory vibration Figure C3.1 shows the points of calculation for the calculation of the strip.

Figure C3.1: The total grid with conditions at the outer contours and at the symmetry and anti-symmetry lines. 15

Delft Hydraulics

For reasons of symmetry and anti-symmetry it may suffice to calculate within one quadrant. The symmetry lines may be replaced by a fixed wall, and the anti-symmetry lines by a contour along which the pressure remains constant. The last precondition is identical to that in case of a free water surface. All contours, also those of the structure, are positioned in the middle between the points of calculation. The strip height is 2a. The vibration is represented by x = x sin(t ) . For the grid, a distance measure L is chosen, that is equal in directions x and y. Figure C3.2 shows the handling of a sloping surface that vibrates.

Figure C3.2: Discharges along the contours of a vibrating structure in case of a sloping wall. For the vibration velocity, a random value V0 is introduced. As each calculational node is representative of an area with height and width L, the edge discharge of the vibrating object equals: q = V0 L The calculation area consists of the first quadrant of Figure C3.1, see Figure C3.3. (C3.8)

16

Delft Hydraulics

Figure C3.3: The calculation area with definitions used. The dimensions of the area are H in horizontal direction and V in vertical direction. The calculation program uses the numbers NH and NV. The horizontal dimension therefore is H = NHL, and the vertical dimension is V = NVL. The strip height is NSL. In the calculation, the potential is represented by the magnitude . As already mentioned before, with this, multiplied with the density of the fluid, , after division by the vibration velocity, results in a measure for a part of the added water mass. For the central area, the continuity condition that is used for the adjustment of each of these points applies, saying that the small discharges that come in from below, from above, from the left and from the right, together must equal zero. For each of the small horizontal discharges, in case of an infinitely thin strip, it holds that: q = VX L = L = L (C3.9)

For the potential on the vibrating wall, relative to the potential of the calculational node nearby, it holds that:
= q 1 1 L = q L 2 2 (C3.10)

17

Delft Hydraulics

In reality this always appears to result in a potential that is too great, which is the reason why, after many attempts, it has been decided to introduce a strip with thickness 0.3 L (relative to the central axis). Therefore: = 0.2q0 (C3.11)

In case of the relaxation procedure described above, it appears to be possible to reach a faster convergence by using the so-called over-relaxation. This means that the change of potential before and after the adjustment, will still be subjected to a multiplication coefficient. The over-relaxation is defined as being this coefficient. Empirically, within the framework of this calculation procedure, it has been determined that the over-relaxation results in a substantial advantage as regards the number of times that the adjustment needs to be repeated. The over-relaxation should not be greater than approximately 1.7, otherwise the calculation process becomes instable. The calculation process is stopped by using a criterion for stopping, described with the concept of accuracy. By starting off with the situation in which all potentials are zero, and therefore also the added water mass mw is zero, after all points of calculation have been released once, a steadily improving assessment of the added water mass is obtained. By repeating this procedure n times, the average improvement that is obtained may be described as mw/n. The improvement that is obtained with the last repetition, is compared with the average improvement. This ratio is referred to as accuracy. See also Figure C3.4.

Figure C3.4: The applied criterion for stopping. The adjustment of the points does not occur in the same way for all points. Therefore Figure C3.3 presents different categories. CATEGORY 1: these are the central points, for which the continuity condition results in:

( above ) + ( right ) + ( under ) + ( left ) = 0 which results in:


18

(3.12)

Delft Hydraulics

= ( a + r + u + l ) / 4 in which = potential in the corresponding calculational node.

(3.13)

CATEGORY 2: the angular element top left. The left and upper limits both have as a precondition = 0. The four discharges in the calculational node together again result in zero: 2(0 ) + ( r ) + ( u ) + 2(0 ) = 0 (3.14)

The factor 2 results from the distance of the calculational node to the contour that is only L. Therefore, the corresponding potential difference results in a discharge that is two times greater. After elaboration, C3.14 results in: = ( r + u ) / 6 CATEGORY 3: the elements on the left, situated above the strip. For these, it is found in an analogous way, that: = ( a + r + u ) / 5 (3.16) (3.15)

CATEGORY 4: the elements at the left side of the vibrating strip (except for the angular point bottom left). At height z, the resulting discharge through the wall equals q(z). Based on the continuity condition, it is found that: ( a ) + ( r ) + ( u ) + q ( z ) = 0 which results in:
= {( a + r + u + q ( z )} / 3

(3.17)

(3.18)

Without further deduction, for the remaining categories only the results are presented. CATEGORY 5: the element bottom left. = ( a + r + q(1)) / 2 In this, q(l) is the discharge produced by the vibrating strip in the lower element. CATEGORY 6: the elements at the bottom, except for the angular points. = ( a + r + l ) / 3 CATEGORY 7: the point bottom right. = ( a + l ) / 4 (3.21) (3.20) (3.19)

19

Delft Hydraulics

CATEGORY 8: the point at the right, except for the angular points. = ( a + u + l ) / 5 CATEGORY 9: the element top right. = ( u + l ) / 6 CATEGORY 10: the elements at the surface, except for the angular points. = ( r + u + l ) / 5 Appendix III describes a calculation program, including the resulting outcome. In Part A it was found that for a long flat strip the added water mass equals the density of the fluid, multiplied with the volume of the cylinder concerned (Figure A3.3). In Appendix III it was found that the water mass in one quadrant is 0.799 R2 (with R = radius of the circle concerned), while in theory for the quarter circle this would have been 0.785 R2. The grid that was investigated contained 54 by 18 points, of which the (bisected) strip was 6 points high. When more points are used, the accuracy increases.
b. Flat strip in wide water in case of rotational vibration Also, a program has been developed that calculates the added polar water mass. For the greater part, this program equals that of the translatory strip, see Appendix III. The changes relate to the following points.

(3.22)

(3.23)

(3.24)

The discharges per element, qs(Nz) are not equal along the entire height of the strip, but increase with the distance to the bottom (for this, {*(Nz-)*L}*L is introduced (in the latter relation, Nz is a counting unit, in which NsL results in the strip height). The factor relates to the position of the centre of the element concerned; is the angular velocity of the vibration. To determine the moment, the pressure that is found is multiplied again by (Nz)*L*L.

The moment that is eventually found (divided by ), is made dimensionless with the strip height to the power of four. In Appendix III it is indicated that in case of a rotating strip, the theoretically expected added (polar) water mass in one quadrant should be /32 (= 0.0982), while the calculation program indicates 0.1046, this again in case of a grid of 54 by 18 calculational nodes and a (bisected) strip height of 6 points.

3.3 Calculations in the frequency domain


Calculations in the frequency domain are carried out in case of harmonic loads. If the harmonic load exerts a long-lasting influence, eventually a vibration is generated with the same frequency as that of the load, and with a constant amplitude. Precondition however is that the system is linear. If the external load makes contact at a certain moment in time, then there are also starting forces that damp out after a while. The eventually resulting vibration is 20

Delft Hydraulics

referred to as the vibration relating to the particular (or special) solution of the vibration equation, while at the beginning of the load there is superposition of the particular solution and the general solution. The general solution concerns the free (damping out) vibration that occurs after an initial load, that is no longer present after a while. This paragraph offers a brief summary of the calculation methods that are available to determine the response in case of periodic load of a linear system. The point of departure for the single mass spring system, is a system with (linear) damping; in case of the multiple system an undamped system suffices. See also Bouma (1976).

3.3.1 The single (degree of freedom) mass spring system

Figure C3.5: Diagram of a single (degree of freedom) mass spring system. In case of a periodic load, the system in Figure C3.5 may be described with the following equation: d2 y dy m 2 + c + ky = F0 eit (C3.25) dt dt For the symbols used, see Part A, Paragraphs 2.2.1 and 2.2.2. In this, it is assumed that the harmonic load F0 eit exerts a long-lasting influence and that the starting forces have damped out. The solution of Equation C3.25 in that case provides the particular solution. The natural frequency of this system results from the solution y = Y0 eint of the reduced equation, in which the damping and the external load have been omitted:

21

Delft Hydraulics

d2 y + ky = 0 dt 2

(C3.26)

For the natural (angular) frequency n it is now found that:

n =

k m

(C3.27)

The solution of Equation C3.25 may be presented in various ways: y = Yeit = Y0 ei (t + ) or: y = Y (cos + i sin )e
it

(C3.28)

Y is a complex number, Y0 is a real number. The phase angle, , may result from the real part R and the imaginary part I of the solution Y, i.e.:

= arctg ( I / R)
Filling in y = Yeit in Equation C3.25 for the particular solution, results in: Y= F0 m + ic + k
2

(C3.29)

(C3.30)

Multiplying the numerator and the denominator by the added complex value of the denominator, and introducing the natural angular frequency n and the relative damping ( = cn/2k or = c/(2mn)) results in Equation C3.31. The relative damping results from relating the damping to the critical damping. The latter is the damping exactly at the point in which the system that experiences decaying vibration has no periodic component anymore (compare Equation C3.37).
2 1 2 i 2 n kY n = 2 F0 2 1 2 + 4 2 2 n n

(C3.31)

Elaboration of y results in:


2 1 2 i 2 n n F y= 0 k 2 2 1 2 + 4 2 2 n n

(C3.32)

22

Delft Hydraulics

The phase angle may now be calculated directly using Equation C3.29. The amplitude of Y is obtained by totalizing the squares of the real and the imaginary part of Y and extracting the square root of this. This is not further elaborated here. The presentation of the response calculation in the frequency domain may be found in Part A, Figure A2.2. If the excitation has not exerted its influence for very long, but only begins at the moment in time t = 0, then also the general solution needs to be added to the solution of Y that was found (the so-called particular solution), in order to comply with the conditions at t = 0, i.e. that the deflection in that case, as well as the vibration velocity, are still zero. The general solution (i.e. of the so-called natural process) results from the equation for the free vibration: d2 y dy m 2 + c + ky = 0 (C3.33) dt dt The solution of this equation is a damped vibration that may be presented as: y = Ye rt = Y pt eit , in which still needs to be determined. The real part of r indicates the extent of damping out of the vibration, the imaginary part of r indicates the periodicity. Filling in Equation C3.34 in Equation C3.33 results in the following for r: (r = p + i ) (C3.34)

mr 2 + cr + k = 0
which results in: r= c c2 k 2 2m 4m m

(C3.35)

(C3.36)

The term under the radical sign results in zero exactly when (by definition) the damping is critical. Therefore the damping that was made dimensionless (the so-called relative damping), , is also defined in such a way that it equals 1 exactly when the damping is critical. c c = or (because equation C3.27) = (C3.37) 2mn 2 km and that is exactly the damping at which the response is periodic or not. If the term under the radical sign is negative (as the damping generally is small, this is usually the case), after filling in the value of the dimensionless damping and the natural frequency n, through Equations C3.27 and C3.37, this results in:
r = n in 1 2 The corresponding vibration of the natural process may now be written as: (C3.38)

23

Delft Hydraulics

y = Ya e nt e + in

1 2 t

+ Yb ent e +in

1 2 t

(C3.39)

As the vibration as described in Equation C3.39, together with the particular solution, needs to comply with the initial condition of t = 0, concerning which it is assumed that y = 0 and (dy/dt) = 0, it is important to know what the deflection is, and what the velocity of the general solution is at t = 0. These are respectively: yt =0 = Ya + Yb and
dy 2 2 = Ya (n + in 1 ) + Yb (n in 1 ) dt t = 0

(C3.40)

(C3.41)

The combination of Equations C3.32 and C3.39 needs to comply with the initial conditions. With this, all data are available for the calculation of the real and the imaginary part of Ya and Yb. This concerns solving a two times two linear system of equations. If the damping is non-linear, for the calculation of the particular solution an assessment method may be used that is valid for as long as the instantaneous relative damping is < 0.04. In case of this small damping, there is hardly any influence on the harmonic response movement (the development remains sinusoidal). In this, it may be calculated how great the energy dissipation per period is, in case of non-linear damping. This dissipation is then compared to the energy dissipation that results from the linear damping, c. The latter is easy to calculate and is (per period):

Elin = cY 2

(C3.42)

3.3.2 The single (degree of freedom) mass spring system in water


As already mentioned in Part A, Paragraph 2.2.3, for a vibrating object in still water, in which the damping is more or less quadratic relative to the vibration velocity (therefore when the damping force = s(dy/dt)|dy/dt|) (with s = quadratic damping coefficient), it is found that:
8 Equadr . = s 2Y 3 3 from which it then follows (with Elin = Equadr . ) that: (C3.43)

cequivalent =

8 sY 3

(C3.44)

The free vibration now damps out exponentially. But across a small time interval it may be approximated (provided the damping is not too great) by an exponentially damping out vibration. The equivalent damping may always be determined from a registration of the vibration, using the logarithmic decrement of consecutive amplitudes. By plotting out the

24

Delft Hydraulics

resulting damping as a function of the amplitude, it may be determined whether there is quadratic damping or another damping that is linear or not. In this way, for example, the Coulomb friction (with a constant friction force that operates in the opposite direction) is recognizable due to the fact that the enveloping curve of the vibration is a straight line (see Part A, Figure A2.3, Figure D bottom left). The calculation of the response as presented in the previous paragraph may also be used for structures in water. The water not only indicates the external load, but also causes loads that are coupled to the movement of the structure. For as far as the latter are coupled to the acceleration of the vibrating structure, these are referred to as added water mass, mw, if proportional to the displacement of the added stiffness, kw. For a periodically loaded structure, Equation C3.25 now changes into: (m + mw ) d2 y dy + (c + cw ) + (k + k w ) y = F0 eit 2 dt dt (C3.45)

As long as the added terms are frequency-independent (see Paragraph 3.2.1), nothing changes in the solutions presented in the previous paragraph. However, in case of structures in water with a free water surface, wave radiation may play a role. The added terms do become frequency-dependent in that case. The consequences of this have already been discussed in Paragraph 3.1. As the mass spring system as described in Equation C3.45 is linear, in case of a longlasting periodic load the frequency of the eventual response is fixed as well (this equals that of the load) and the added terms may be introduced right away. For randomly shaped structures the added terms may be calculated, also in the frequency range in which the added mass and the added damping are frequency-dependent. For this purpose, there is a calculation program that is coupled to the TNO-DIANA program (ref. Delft Hydraulics Report Q525). The natural frequency of the structure in water, albeit through a repetition procedure, may be calculated as well. Apart from that, the added water mass is not only frequency-dependent when a vibrating structure radiates waves. Also the already present flow may be influential. This is taken into account by presenting the added water mass as a function of the dimensionless frequency (the Strouhal number S = fL/V, in which f is the frequency of the vibration, L is a measure of length of the vibrating structure, and V is the approach flow velocity). The frequency-dependency of the added terms is an important factor, especially in case of circular cylinders, see Part A, Figure A5.12, from Sarpkaya (1978). In the figure on the left, the periodic forces that are in-phase with the acceleration of a forced oscillating cylinder are plotted out as added water mass coefficient. The fact that the added terms depend so much on the frequency and approach flow velocity relates to the flow instability that is present at circular cylinders, which also generates great periodic loads. In the proximity of the critical Strouhal number (i.e. the S value at which the maximum flow excitation occurs) S = 0.2, it appears that from 0.3 onward the added water mass remains more or less constant. Between 0.2 and 0.3 this added water mass may increase 1.5 to 2 times, and under S = 0.2 it even reverses its sign. In case of other shapes than circular cylinders it is found again and again, that the added water mass of structures in wide water surfaces is hardly frequency-dependent or flow-dependent, provided precondition C3.1 is met.

25

Delft Hydraulics

3.3.3 Response of a double (degree of freedom) mass spring system (direct method)

Figure C3.6: Diagram of double (degree of freedom) mass spring system. Figure C3.6 presents the diagram and the symbols used in a double mass spring system. Below, only the undamped system will be discussed. The system with damping presents no special problems in case of periodic load; the calculations however are much more elaborate. The dynamic equations for both masses in case of periodic load are: d 2 y1 m1 2 + (k1 + k2 ) y1 k2 y2 = F1eit dt d2 y k2 y1 + k2 y2 + m2 22 = F2 eit dt

(C3.46)

26

Delft Hydraulics

As the system is undamped, in case of periodic load the response of each of the masses is always in-phase with the load. Because of this, the time-dependent factor in the load (therefore eit or cos(t)) may be removed by division, and a relation remains between the amplitudes of the load and of the responses. The amplitudes of the magnitudes are indicated by capital letters. The equations then become:
(k1 + k2 m1 2 )Y1 k2Y2 = F1 k2Y1 + (k2 m2 2 )Y2 = F2

(C3.47)

Another way of presenting this is:


+ k1 + k2 k2 F1 m 0 Y1 2 1 * = + k2 0 m2 Y2 F2 k2 The short version of this is: (C3.48)

( K M ) Y = F
2

(C3.49)

in which K and M are matrices, and Y and F are vectors. The stiffness matrix K consists of four components that may be described as follows: a unit displacement y1 in case of a fixed second mass (therefore y2 = 0) results in a spring force that influences the first mass (-k1-k2) and a force (k2) that influences the second mass. If the first mass is fixed and the second mass experiences a unit displacement, then again there is a force that influences the first mass (k2) and a force that influences the second mass (-k2). As the terms move to the left side of the equation, their sign reverses. Further consideration of Equations C3.48 and C3.49 may explain these matters more clearly. The mass matrix M of a dry structure only includes diagonal terms. This changes in case of a structure in water. Filling in the load amplitude and the frequency in Equation C3.47 results in two equations in Y1 and Y2, from which these may be solved directly. This is the particular solution. In order to meet the conditions at t = 0, the general solution needs to be added to this again. For this, first the two resonance frequencies of the system need to be determined; these follow from: (C3.50) ( K M 2 ) Y = 0 The solution is found by equalizing the following determinant with zero:
k2 k1 + k2 2 m1 =0 k2 m2 2 k2

(C3.51)

This results in a quadratic equation in 2. Given that physically only positive values are important, it is found that there are two natural (angular) frequencies, i.e. 1 and 2. For each of these frequencies, the relation between the amplitudes Y1 and Y2 may be determined using Equation C3.46. Each of these relations may be represented vectorially with the vectors A and B. The absolute magnitudes of Y1 and Y2 may be represented by multiplying them with the amplitudes A and B (that may still be chosen freely). Analogous to what was found in

27

Delft Hydraulics

Equation C3.39 for the single mass spring system, the general solution may now be written as: y = Aa Ae + 1t + Ab Ae 1t + Ba Be +1t + Bb Be1t (C3.52)

With the amplitudes Aa, Ab, Ba and Bb it is possible to meet the initial conditions. This calculation method is also useful for structures in water in case of situations in which the added mass and damping are frequency-dependent. As the solution consists of a sum of periodic components, for each of these the appropriate added water mass may be introduced. In reality this procedure requires a lot of calculation. Therefore the direct method is in fact only used in order to find the particular solution.

3.3.4 Modal Analysis in case of a double (degree of freedom) mass spring system
As already discussed in Paragraph 3.1, Modal Analysis is used to determine the response of systems with multiple degrees of freedom. Also in the frequency domain, it is often rewarding to use this type of analysis. The response of the structure to a random external load is considered to be factorized into components that are referred to as eigenvalues or modes. Each mode has one natural frequency, one natural vibration mode and one eigenvalue of the load, that are univocally coupled to each other. The vibration mode is defined as the relation between amplitudes. The natural load is defined as the relation between loads, such as those that apply at each mass point (and in each of the directions in which the mass point may move freely). The vibration mode of each of the modes corresponds to one of the free vibrations (therefore without external load) of the undamped system. Each mode corresponds to an eigen value of the load (in fact, the load distribution) that statically has the same deflection as the load corresponding to the vibration mode concerned of the free vibration. The magnitude of the amplitude of the free vibration and that of the corresponding eigenvalue of the load however is not yet fixed. It may be shown that this load distribution, regardless of the frequency with which this load makes contact, is always part of the same vibration mode, and that the response curve in the frequency domain is identical to that of a single mass spring system. From this, it may again be deduced that this vibration mode is also relevant in case of a non-periodic natural load. By factorizing the actually present load into the eigenvalues of the loads concerned, the amplitude of each of the eigenvalues of the loads is fixed and the response in time may be calculated for each eigenvalue of the load, in an analogous way to that of a single mass spring system. The impulse response technique may also be used (to be discussed later in Paragraphs 3.4.1 and 3.4.2). The sum of these responses to each of these eigenvalues of the loads results in the real response of the structure to the load. For more details, see Bouma (1976). This procedure will be illustrated using a double mass spring system. Naturally, the equations presented in the previous paragraph may be used for this purpose. First, using Equation C3.51, both natural frequencies, 1 and 2, will be determined. Filling in these in Equation C3.50 results in the corresponding relations between Y1 and Y2 for each of the frequencies. From this, using Equation C3.49, and assuming that the load is static ( = 0), the relation between the amplitudes of the loads F1 and F2 is determined for each of the frequencies. As the magnitude of the amplitude may still be chosen freely, we will begin with F1 = 1. F2 is chosen in such a way, that the relation between y1 and y2 is correct 28

Delft Hydraulics

(i.e. equal to the relation that was found for the natural vibration). The loads F1 and F2 that correspond to each other are expressed in a vector F, and therefore this results in the loads F1 and F2, relative to the frequencies 1 and 2. Now the real force vector F may be factorized into F1 and F2 by stating that:

F1 + F 2 = F

(C3.53)

Next, the complete vibration calculation may take place. The fact of the matter is, that for each of the force vectors it holds that the deflection mode (the relation between y1 and y2) does not change, regardless of the question with what frequency, or with what timedependent function it makes contact. The response (the amplification factor), as with a single mass spring system, depends on the relation /n (i.e. the frequency of the load relative to the natural frequency concerned). The structure may also be damped, provided the damping makes contact itself parallel to the springs and is proportional to the spring stiffnesses. Another possibility is that each mass is coupled to the earth with a damper, in which each damper is proportional to the mass concerned. These two kinds of damping do not influence the vibration mode, because at each of the natural frequencies the mass forces (or rather, inertia forces) establish a balance with the spring force, and the spatial distribution of these forces is the same. Whether the distribution of the damping forces is coupled to the spring force or to the inertia force therefore does not make a fundamental difference. At each of the vibration modes, the mass force, the damping force, the spring force and the external force maintain the same relation relative to each other as would be the case with a single mass spring system. From the proportionality of the damping with either the springs or the masses, also the value of the relative damping may be calculated:

C 1 M 2n

or: (C3.54)

C n , = K 2

C = damping matrix

For structures in water, a Modal Analysis in which the superposition principle is being used time and again, is only useful within certain limits, as among other factors, the added water mass may vary with the frequency. It is however possible to calculate the added water mass for a forced vibration of the entire structure or for one of the surface elements. Using a repetition process, the natural frequencies may still be determined, but the response in case of a random different load may no longer be determined from the sum of eigenvalues of the loads. In case of wave impact response and vibrations generated due to flow, the method may still be used, as in that case the natural frequencies of the structure are so high that the added mass tends to become frequency-independent. However, for the calculation of the response in case of loads due to wind-generated waves, the frequencies of the wind-generated waves are so low that the added water mass is frequency-dependent: in that case it is not possible anymore to calculate with the method of Modal Analysis. The importance of this method is based on the fact that it shows that reviews concerning vibrations, as given in Part A for a single mass spring system, also retain their value for multiple systems and for real structures in which all elements are completely elastic (continuous-elastic systems).

29

Delft Hydraulics

3.3.5 General formulation of a system with multiple degrees of freedom


Generally speaking, a structure will have multiple degrees of freedom. In that case it is obvious to further elaborate the formulations toward a multiple mass spring system, even though that still is a strongly schematized representation of the actual system as well. A better schematization is obtained by dividing the real, continuous-elastic system into finite elements that are connected by way of nodal points. This method is referred to as the Finite Element Method (FEM). This method will be briefly discussed below, also in relation to the coupled structure-fluid problem. For more information, please look up the extensive range of references, such as Bouma (1976). The structure without water In case of the Finite Element Method, the continuous-elastic systems is divided into elements with finite dimensions. At the junctions of the elements (the nodes) connection preconditions (balance and compatibility) are stipulated. The equations however are formulated in a similar way to those of a multiple mass spring system. First, the undamped system is taken into consideration. The movement equations may be described as follows: d2w M 2 + Sw = f (C3.55) dt with

M S w f

= = = =

mass matrix stiffness matrix vector with linear-independent displacements of nodes load vector (load applying at the nodes)

The mass matrix and the stiffness matrix are set up on the basis of the mass matrices and stiffness matrices of the separate elements. When the mass is distributed across the nodes, the mass matrix that is formed is a diagonal matrix. In case elements with divided mass are being used, the element-mass matrices are filled, resulting in a band matrix as mass matrix. The stiffness matrix always has a band structure. A system with n linear-independent degrees of freedom has n eigenvalues and relative to this, n natural types of movement (modes). The free vibration w(t) of the system, which is generated, for example, after excitation, is a linear combination of all n natural vibrations: w(t ) = ai ei sin(i + i )
i =1 n

(C3.56)

with

ei i i ai

= = = =

ith vector (1,, i,n) ith natural frequency phase angle of ith natural frequency amplitude factor of ith natural frequency

The natural vibrations wi(t) of the system result from the homogenous system of differential equations: d2w M 2 + Sw = 0 (C3.57) dt 30

Delft Hydraulics

The influence of damping in this case has been ignored. Here again, generally speaking, it is assumed that damping, provided it is homogenously distributed, has no influence upon the eigenvectors ei. Moreover, the influence of the damping on the natural frequencies i is limited, as long as the damping is small. This is often the case with structures, so that damping in the natural frequency calculations usually does not need to be included. In the response calculations however, the damping is important. If the load is applied in one of the resonance frequencies, the damping is the only determining factor for the equilibrium amplitude. For a damped, linear system with external load f, the movement equations are as follows:

M
with: C = damping matrix

d2w dw +C + Sw = f 2 dt dt

(C3.58)

If there is damping present in the system, the natural vibrations are no longer independent. There is a power transfer between the various natural vibrations. This does not occur in the calculation model when there is a specific requirement upon the damping matrix C. Formulated in a general way, this is:
C = a.S + b.M

(C3.59)

in which a and b are constants. A damping matrix that is proportional to the stiffness matrix may represent the viscoelastic properties of materials very well in certain cases. This type of damping is referred to as Rayleigh damping. The response w(t) to a given external load f(t) that makes contact at the nodes at which the degrees of freedom have been defined, may again be calculated in different ways: a. b. using the eigenvectors (Modal Analysis). using the direct solution of the movement equations.

In both cases it is possible to calculate in the time domain and in the frequency domain. In case of method a, the point of departure is again that the response of a linear system to a random time-dependent load is considered to consist of a sum of eigenvectors. The extent to which the ith eigenvector ei cooperates with the response at a random moment in time t may be calculated, and is indicated by the participation function i(t). The response of the system is now presented as a linear combination of all n eigenfactors ei, multiplied with i: w(t ) = ai (t ) ei
i =1 n

(C3.60)

31

Delft Hydraulics

It is necessary to determine the eigenvectors in the first step. Each eigenvector corresponds to an eigenvalue of the load. The real load is factorized into eigenvalues of the loads. This results in the participation function i. This requires some elaboration; after this however, considerable advantages may be obtained in the calculation process, especially in case of multiple systems, as the resulting system of equations with the participation function i may also be disconnected. The result of the disconnection is that in the system of equations only diagonal matrices remain, as a result of which this system may be easily solved. For the disconnection, the property that the eigenvectors are independent (i.e. orthogonal) is being used. From a physical point of view, this means that there is no energy transfer between the natural vibrations. As already stated, in that case the damping needs to meet the specific requirement of Equation C3.59. If the damping does not meet that requirement, due to the fact that the distribution is not correct or due to the fact that the damping is non-linear, then an approximation is possible, provided the damping is not too great. The point of departure here is that the damping is so small, that it does not influence the vibration mode as such. A given harmonic natural vibration will experience a given, for example a non-linear, damping per mass point or per spring element, and this again results in an energy absorption per period. This energy absorption is a function of the amplitude of the vibration. The eigenvalue of the load relative to this natural vibration results in an energy supply per period. This energy supply will again be amplitude-dependent. The amplitude at which the energy absorption equals the energy supply is the sought-after equilibrium amplitude. In case of a dynamic analysis, usually it will be assessed which eigenvectors result in the greatest contribution to the response and which influence the damping has on these contributions. For this, the participation functions i(t) are determined in case of a harmonic load. In this, the excitation frequency is varied. The damping i corresponding to the natural frequency i may also be varied. This results in an impression of the maximum value of the participation function i as a function of natural frequency i and damping i. Constructors will strive to keep the lowest natural frequencies greater than the load frequency . In that case, especially the lowest natural frequencies are still important for the response, so that only a limited number of eigenvectors needs to be included in the calculation of the response. In case of wave impacts, the load duration is so short, that it is possible to include a large number of natural vectors in the calculation. Periodic loads (that may be considered as a sum of harmonic loads) allow for the quickest calculation in the frequency domain. In case of loads with a limited time duration it is possible to use the Fourier analysis. In the Fourier analysis, the load is transformed to the frequency domain. Next, the movement equations are solved in the frequency domain, after which the solution is transformed back to the time domain. For this procedure, it is possible to use the natural vectors of the system. The system of equations with the participation functions i as unknown magnitudes may also be solved in the time domain. This will be the case when there is a random nonperiodic load. For a certain moment in time t0, the solution of the system of differential equations needs to be given. This solution serves as an initial precondition for the system at moment in time t1. Next, the movement equations are solved by way of the time-stepping procedure. The magnitude of the time intervals is very important in this calculation method, as it influences the stability of the numeric calculation process.

32

Delft Hydraulics

In case of method b, the equations are solved directly, without using the natural vectors of the system. This method is less efficient, as it requires matrix elaborations that are very time-consuming. In case of non-linear systems the Modal Analysis technique is not useful, as it is based on the superposition principle. Non-linear systems may however be calculated with a direct solution method in the time domain. Non-linear damping and non-linear material behaviour, as an example, may be introduced into the equations. The system in still water and structure in case of periodic load For a structure in water, flow and waves are the most important sources of hydrodynamic excitation. As the structure moves and deforms due to the influence of these loads, so-called interaction forces are generated at the boundary surface of the structure and the fluid: the structure must displace fluid and as a result it generates extra pressures. These interaction forces must be included in the formulation of the calculation of the response. Therefore a coupling is necessary between the fluid system and the structure system. The coupling occurs by way of boundary conditions at the boundary surface of fluid and structure. The movement equations for the structure (linear system) may now be presented as follows: d2 w dw M 2 +C + Sw = f1 + f (C3.61) dt dt with: f1 = vector with interaction forces in the interface nodes, f = vector with hydrodynamic excitation forces. The vector f may also be determined separately, as by definition the response of the structure has no influence on these excitation forces (if this is the case however, they relate to vector f1). As the matrices M and C may be frequency-dependent in case of structures in water, the calculation needs to relate to periodic excitation. If this is not the case, it may still be assessed whether the response is more or less periodic, in which case the frequency of the response needs to be estimated first, next the matrices M and C need to be calculated and finally the frequency needs to be verified. A relatively simple formulation of the fluid movement is provided by the potential theory. This formulation does not include viscosity terms, as damping due to the viscosity of the water is being ignored, and is based on rotation-free flow. Damping due to radiating waves (surface waves and/or compression waves) may be included in this formulation, as well as added water mass (both effects are interaction forces). It is not possible to determine flow forces very accurately using this formulation, as the potential theory usually does not calculate the flow surrounding a structure very accurately. It is however possible to calculate wave forces, provided the structure is not too slender. There are developments toward using more complex formulations for the fluid in the structure-fluid-interaction problem. On the initiative of Delft Hydraulics, a component has been introduced into the calculation program DIANA of TNO-Bouw, that is based on the potential flow theory, which allows the calculation of a part of the interaction forces (added mass and damping due to wave radiation at the water surface) (see Delft Hydraulics Report Q525). Using this program it is possible to carry out natural frequency calculations, as well as calculations of the response at a given load. The added water mass matrix may be extracted 33

Delft Hydraulics

from this, as well as the damping matrix due to wave radiation. Both matrices have smaller dimensions than the structure matrices; the dimensions are related to the number of degrees of freedom in the interface nodes. Both matrices are also filled matrices (contrary to the diagonal matrix mentioned above), as a displacement in one of the degrees of freedom through the fluid generates pressure changes at the location of the other interface nodes. When wave radiation is included in the calculation (in DIANA, this is an option), both the added water mass matrix and the wave damping matrix are frequency-dependent. Without wave radiation, the added water mass matrix in the potential flow theory is frequency-independent, and the wave damping matrix equals 0. For as far as M and C are frequency-dependent however, calculations of the response in case of non-periodic load are still not possible. The assessment of these requires a very complex approach, in which an impulse-response function would have to be determined for each contact point between water and structure. See also Paragraph 3.4.2, in which this procedure has been used for a rigid object that moves freely in water.

3.3.6 Coupled systems with structure components and fluid components


The various calculation methods that were used in Part A are based on a harmonic movement of the structure, in which the forces generated by the fluid are analyzed in terms of added water mass, added water damping, added spring stiffness and negative damping. However, just like it is possible in case of a multiple mass spring system to calculate a freely damping out vibration, the water-structure system may be treated in a similar way. In this, it is even possible to calculate in the time domain. And it is also possible to observe the freely damping out vibrations that occur after an initial load. The flow inertia of the water plays an important role in a number of cases, see Part A, Paragraph 3.5. The most illustrative for this is the water that flows through a pipe. Just like in case of structures, in which the core of the dynamic behaviour is based on the behaviour of mass points that accelerate due to the external force and the internal forces of the system (springs and dampers), the core of the non-stationary flow behaviour of the fluid is based on the inertia of the flow in the pipe. For a pipe system with a number of shafts, as an example, the number of degrees of freedom is determined by the number of pipe pieces between the shafts themselves and between the shaft and the outer water.

34

Delft Hydraulics

Figure C3.7: Diagram of a system with a gate (two degrees of freedom), shafts and connecting culverts. In Figure C3.7 the number of degrees of freedom is five: two (due to the fact that the gate is capable of moving both horizontally and vertically) plus three (due to the fact that there are three pieces of culvert). All these degrees of freedom are more or less coupled, though a system like this will in principle also have five modes of vibration with five resonance frequencies. One or more of these modes may be instable, for example due to a bath plug vibration mechanism. The following paragraph presents examples, accompanied by the presentation of the equations, which are analyzed in general terms.

3.3.7 Examples of coupled systems with structure components and fluid components
This paragraph discusses the phenomenon of coupled systems on the basis of two examples that have already been presented in Part A: the first example (a) concerns the bath plug vibration and the second (b) the instable oscillations of the floating sector gates of the storm surge barrier in the Rotterdam Waterway (water level oscillations due to discharge variations generated by a gate position coupled to the water surface). a. The bath plug vibration Part A, Paragraph 4.4.2 discusses in great detail the so-called bath plug vibration. This vibration of the bath plug that is suspended from a spring/damper serves as an example of many other cases in which vibration is caused by feedback resulting from discharge variations due to the fact that a gate opening periodically varies in size with the vibration. As the discharge in the runoff pipe also generates an oscillation during the vibration of the bath plug, the total system has two degrees of freedom. The coupling is caused by the fact that the bath plug movement generates a periodic gap variation; the periodic change in resistance coupled to this causes both a discharge oscillation and a fluctuating head over the bath plug. 35

Delft Hydraulics

Figure C3.8: Diagram and definitions of the bath plug system. Figure C3.8 again shows the diagram of this calculation once again. The calculation as presented in Part A eventually includes the simultaneous solution of the following three equations: the equation for the gap discharge (completed with the water displacement due to the plunger effect of the vibrating bath plug); the pipe equation; the dynamic bath plug equation.

Another way of calculation that is based on a forced harmonic vibration of the bath plug, is also possible. This cancels out the third equation. If the movement is described as: y = Y0 eit (C3.62)

then the generated periodic forces may be distinguished into an in-phase force and an out-ofphase force relative to the forced movement. As the equations, similar to the calculation in Paragraph 4.4.2 of Part A, are linearised in that case (this is possible in case of a vibration with a very small amplitude), the forces are proportional to the amplitude of the movement. The fluctuating force may now be described as:

36

Delft Hydraulics

Fw = Fw eit

(C3.63)

in which:
Fw = Y0 A( ) + iY0 B ( )

(C3.64)

In order to determine A() and B(), the equations deduced in Part A for the pipe discharge are used: dV 2 dy 0 2 g dH d 2 y = + 2 2 g H 0 (A4.14) R R H 0 dt dt dt dt and the discharge of the pipe: H = L dV g dt

(A4.15)

On top of that, the discharge of the pipe results in an increase of the pressure downstream of the bath plug, and this means a decrease of the hydraulic head of the plug. The force then, that is exerted on the bath plug in the direction y, now becomes: Fw = g R 2 H Using Equations C3.65 and A4.15 in A.4.14 results in:
Fw + dFw dy d2 y = 2 L R 2 g H 0 L R 2 2 dt dt R 2 g H 0 dt 2 0 L

(C3.65)

In this, the expressions of Equations C3.62 and C3.63 may then be filled in again for the periodic displacement of the bath plug, y, and the hydrodynamic force, Fw. This eventually results in:
2 0 L2 3 R 2 L R 2 g H 0 2 g H 0 Fw L R 2 2 + 4 2 0 L2 2 (C3.66) = +i 2 2 Y0 2 0 L 2 0 L 1+ 1+ R 2 g H R 2 g H 0 0

Considering the expressions for A() and B() in Equation C3.64, these terms may be deduced directly from Equation C3.66. In Kolkman (1984) this is expressed in the following way: A( ) = and
2 2 LAp (1 + 2 0 LL2 3 / Ap ) c

1 + ( 0 LLc ) /( Ap 2 g H 0 )

(C3.67a)

37

Delft Hydraulics

B ( ) =

LLc 2 g H 0 {1 2 0 L /(2 g H 0 )}
1 + ( 0 LLc ) /( Ap 2 g H 0 )

(C3.67b)

In this, Ap is the surface of the cross-section of the pipe and Lc is the circumference of the pipe. It may also be deduced from Equation C3.67 which spring stiffness is critical for the occurrence or not of vibrations in the absence of a mechanical damping. This is when the water damping, and therefore B(), exactly equals 0. This results in:

2 =

2g H 0 0 L

(C3.68)

Now that the is known, also the A() has been fully determined. By definition, this is a negative (hydrodynamic) stiffness. It therefore holds (as 2 = stiffness/mass) that: k = m 2 + A( ) (C3.69)

As both the and the A() have been fully determined, it is also possible to determine the minimally required spring stiffness. It appears moreover, that A() equals the sudden negative hydrodynamic stiffness defined in Part A (see in Part A, Paragraph 4.4.2, the explanation of Equation A4.24). The result of Equation C3.68, combined with Equation C3.69 fully corresponds to what was found in Part A (Paragraph 4.4.2 and Equation A4.25) for the limit of stability. The advantage of the method discussed above is that, even if there is a mechanical damping, the minimally required spring stiffness may be determined directly.

b.

Instable oscillations in case of floating sector gates The second example concerns instable oscillations both of water and, coupled to this, the floating sector gates of the storm surge barrier at the Rotterdam Waterway; these matters have been discussed in Part A, Paragraph 4.5. The complete analysis has been presented by Bakker et al (1991). The phenomenon was originally discovered during a scale model investigation (see Chapter 6, Example 6.8 in this part) and was later analyzed theoretically.

38

Delft Hydraulics

Figure C3.9: Plan view and vertical cross-section of sector gates in the storm surge barrier in the Rotterdam Waterway with indication of the standing wave perpendicular to the main flow direction. As indicated in Figure C3.9, in the situation in which the gates were introduced afloat in the river, but not yet sunk down, severe gate movements were generated during a period coinciding (see front view) with the period of the standing wave at the downstream side between the abutments, perpendicular to the main flow direction. The amplitude of the gate movements was of the same magnitude as the amplitude of the standing wave. Upstream of the gates no wave movement was observed. In case of flow in the opposite direction (in that case, the downstream side is situated at the other side of the gates) something similar occurred, but in this case with the period of the second harmonic standing wave between the abutments. This will not be discussed any further below. Due to the rising and falling of the water level, the gates also appeared to be moving up and down; each time one gate went up, the other went down. The amplitude of the wave movement, crest-trough, could amount to a few metres (in case of a hydraulic head that was not greater than 4.5 m in any event), and the gates also moved up and down with more or less the same amplitude. At a later stage of the investigation, the design of the gate was changed in such a way, that it was not possible for these gate movements and water level oscillations to occur during the relevant conditions. For the analysis of the phenomenon, the point of departure was that the waves that were perpendicular to the main flow might correspond to the discharge characteristics of the gates, and that the discharge, due to the fact that the gates were more or less floating on the downstream water, was coupled to the rising and falling water level. For this reason, initially a diagram was drawn as indicated in Figure C3.10. 39

Delft Hydraulics

Figure C3.10: Diagram of the standing wave in an area, limited on both sides by a gate (discharges ql and qr are both related to the discharge that varies with time). Based on considerations regarding symmetry, it was possible to schematize the diagram of Figure C3.10 to that of Figure C3.11 again. This concerns a standing wave, limited on one side by a constant water level. It is assumed that the discharge per m, q, is coupled to the downstream water level variation at the location of the gate. The situation has been strongly schematized in such a way, that it would be possible to carry out a one-dimensional calculation.

Figure C3.11: Schematic presentation of the standing wave, caused by a fluctuating discharge. The schematic presentation is based on the fact that all discharge variation occurs at the location where the standing wave is reflected (therefore in the corners between abutment and gate), and that the waves do not radiate toward, and do not lose energy to the river. Indeed, in the scale model further down the river the standing wave was hardly noticeable. As only the lowest wave period is observed (of which the period is fixed), it has been decided to carry out an even stronger schematization, in which the water only has one single oscillation period. All flow inertia of the standing wave is considered to be concentrated in one (pipe) element and the water storage in one (basin) element. This schematization has been used to make it possible to arrive at one analytical solution. At a later stage the qualitative result has been verified using a more complete wave review.

40

Delft Hydraulics

Figure C3.12: Diagram with a basin-pipe system. The eventually resulting diagram therefore now includes one vessel with length L1 that is connected with the river through a culvert (length L2) (Figure C3.12). The culvert runs off the constant part of the discharge, q0, and on top of that a periodically varying discharge, qpipe. It is assumed that the water level inside the vessel may rise and fall, but does remain horizontal. The discharge underneath the gate consists of a constant part, q0, and a varying part, qgate. The water level upstream of the gate and downstream of the culvert are both assumed to be constant. The coupling of the discharge to the downstream water level is introduced into the equation as follows: q = Ca a 2 g (h1 h2 ) = q0 + qgate (C3.70)

In this, a is the gap height underneath the gate and Ca is the discharge coefficient. Assuming now that the gate opening is coupled in a certain way to the variation of the downstream water level, z, then it is possible to write for the discharge:

dq =

q q a q Ca a dh2 + dh2 + dh2 h2 a h2 Ca a h2 Ca = 0 ), then a

(C3.71)

Assuming now that Ca only varies very little with the gate opening (or Equations C3.70 and C3.71 may be combined into:

2g 1 a dq = Ca a dh2 + Ca 2 g (h1 - h2 ) dh2 2 h2 h1 - h2

(C3.72)

The relation between the discharge and the downstream water level is now very simplified and may be described as: 2g a 1 dq = dh2 , = Ca a dh2 + Ca 2 g (h1 - h2 ) (C3.73) h2 2 h1 - h2
41

Delft Hydraulics

In terms of qgate and z, combined with dq = qgate and dh2 = z, this results in:
qgate = z

(C3.74)

Based on Equations C3.72 and C3.73 together, it may be concluded that the factor that is a measure of the coupling of the discharge to the downstream water level, may become both positive or negative, depending on the hydraulic head and on the extent to which the gate movement is coupled to the downstream water. At a later stage, when investigating the instable gate behaviour, this factor turned out to be an important part of the investigation. may be strongly influenced by the shape of the cross-section of the gate. Among other things, the investigation used a rigid section model of the gate, in which the buoyant force and the discharge underneath the gate were determined as a function of both water levels (see Delft Hydraulics Reports Q1271 and 1278). In case of a change of the downstream water level, first the position of the floating gate must be calculated. In order to be able to calculate the water movement in the basin, next to Equation C3.74, also the equations for the pipe discharge and the detention of the basin need to be included.
The pipe discharge This was based on a pipe without friction, knowing that friction always generates a damping effect on the water level oscillations. The movement equation is:

L2

dV = gz dt pipe

(C3.75)

If the culvert height, hc, is also taken into account, then this is also the relation for the pipe discharge.
The water level oscillation in the basin

dz L1 = q pipe = V pipe hc dt

(C3.76)

If there would be no discharge underneath the gate, Equations C3.75 and C3.76 give the basic equation of the vessel-pipe system:
L1 d 2 z g + z=0 hc dt 2 L2

(C3.77)

This equation in z is identical with what was formulated for y in Paragraph 3.3.1 in case of the mass spring system. The natural frequency may now be determined directly:

n =

hc g L1 L2

(C3.78)

While translating the schematization of Figure C3.11 to that of Figure C3.12 it has been attempted to maintain the same period, as well as the same storage of water at an equal value of z. The culvert height was chosen to be equal to the water depth, a, in Figure C3.11. 42

Delft Hydraulics

All this resulted in a length L1 = L2 = 130 m (considered relative to the distance between the abutments of 360 m and the length of 400 m calculated from the standing wave period in the model). The equation for the basin may now also be written including the gate discharge, using Equation C3.74. This results in:
L1 d 2 z dz g + z=0 hc dt 2 hc dt L2

(C3.79)

This too is an equation that is completely analogous to what was found for a damped mass spring system (compare Equation C3.33). It appears that a positive gives cause to a negative damping. This means that the system in that case is instable and the oscillation amplitude after an initial generation will increase exponentially.

3.4 Calculations in the time domain using the indirect method 3.4.1 Modal Analysis and impulse-response method in case added mass and damping are frequency-independent
In case of a random load that varies with time, it is possible to calculate the response of a linear mechanical system, if it is known what the response is to an impulse of short duration with value I. The single mass spring system may serve to illustrate this. In case of a single mass spring system the impulse I causes the immediate generation of a velocity of the mass V = I/m, while at the moment when the impulse makes contact (t = 0), there is no displacement yet. From Equation C3.39 it appears that the vibration consists of a function with decrement nt, multiplied with a periodic function with angular frequency n(1-2). After an impulse of short duration, I, the free vibration may be written as:
y= I mn 1
2

ent sin n 1 2 t

(C3.80)

This may be formally deduced from Equation C3.39 combined with the initial conditions of Equations C3.40 and C3.41, in which Ya and Yb are complex magnitudes; it will also be noticed immediately that Equation C3.80 satisfies the set conditions at t = 0. Let us consider the situation at a moment in time t, in which the impulse makes contact at a moment in time t-, with > 0. The vibration in that case consists of what was caused by the impulse at moment in time (t-), in which it is taken into account that this occurred at a moment in time in the past, so that a free vibration was generated at a later stage. All small impulses that have occurred in the past need to be taken into account; varies from 0 to , or, if the load starts at t = 0, from 0 to t. Considering the fact that the impulse equals the instantaneous load, multiplied with the time duration dt, the total response results in: y (t ) =
0 t

F (t ) mn 1
2

e nt sin(n 1 2 )d

(C3.81)

43

Delft Hydraulics

This expression is referred to as the convolution integral. There are calculation programs by which, given the response to the unit impulse (therefore with I = 1), it is possible to calculate the integral for a random load function. For a linear time-independent system it is also possible in a formal sense, to deduce the expression for the convolution integral. Paragraphs 3.3.4 and 3.3.5 already discussed how Modal Analysis may be used for double and multiple (degrees of freedom) mass spring systems and for a continuous-elastic structure. The eigenvalues of loads may be determined by factorizing the actually occurring load. In that case it is of course possible to factorize a random, short-lasting impulse load into eigenvalues of impulse loads. For each of the eigenvalues of loads, the response to a unit impulse is known. Based on the solution that is found, it is again possible to calculate the response of a load that develops randomly with time. In case of the double mass spring system it was found that when using the matrix notation in the mass matrix, terms only occur in the diagonal (Equation C3.48). In case of water, the added mass terms are coupled; an acceleration of one mass in a certain direction will generate forces that are proportional to this acceleration on all masses in all directions. This also applies to the added damping and spring stiffness. Consequently, the method of solving the natural frequency, vibration modes and eigenvalues of loads remains the same. The natural load results in the same vibration mode, regardless of the frequency at which it makes contact. However, this only applies when the mass matrix does not change with the frequency. The method of Modal Analysis therefore may only be used effectively when the matrices for the added magnitudes are frequency-independent. For each of these eigenvalues of loads, the response of the system (which therefore as regards distortion matches the corresponding natural vibration mode) in the frequency domain is fully comparable to the response behaviour of a single mass spring system. If the static outward bending and the natural frequency are known, the amplification factor, A, results in the real distortion relative to the static distortion. See for the amplification factor Part A, Figure A2.2. It is however necessary to use the corresponding distortion, natural frequency and amplification for each eigenvalue of load.

3.4.2 The impulse-response function in case added mass and damping are frequency-dependent
In case of ships that collide with elastic structures, an approach has been developed in which, despite the fact that the added water mass and water damping are frequencydependent, an impulse-response method is being used. If the response of the ship to an impulse load is known, it is possible to calculate in the time domain again. Before the response to an impulse load may be determined, it is first of all necessary to assess what forces operate on the ship by the water, in case it is forced into a periodic oscillation. In this, it is taken into account that a ship, when viewed horizontally, has three degrees of freedom, and that each direction of movement may give cause for forces or moments corresponding to other degrees of freedom. These forces (the so-called transfer functions) need to be determined for the entire frequency range. In case the calculation methods or possibilities of testing are not adequate, it is necessary to carry out assessments. A distinction needs to be made between forces during the oscillations that are in-phase with the acceleration (the added mass forces) and forces that are in-phase with the oscillation velocity (the damping forces). The added water mass and damping are referred to as the hydrodynamic coefficients. Although the method with the impulse-response function is only used for the situation ship/fender wall, it will be briefly discussed here as it is interesting for understanding the 44

Delft Hydraulics

issue. The method may also be elaborated further into a system with multiple degrees of freedom (Fontijn, 1978 and 1988). Here the issue is limited to a ship with one degree of freedom: horizontally and in transverse direction. The impulse-response function needs to meet a number of preconditions. If the impulse-response function were known, it should be possible to use it in order to calculate the relation between force and movement in case of a periodic load. The response needs to result in exactly the same relation between force and movement that was also found in case of the forced oscillation. This information is used for the deduction of the impulse-response function. The difficulty in the elaboration is in the fact that sometimes only empirical data are available for the transfer function or that the theoretical data only apply to a (limited) part of the frequency range. Although the hydrodynamic coefficients are assumed to be known across the entire frequency range, at frequency = 0 certain preconditions need to be met as well. In the elaboration, this results in a singular point (only when this concerns the movements of a ship in the horizontal plane).

Fontijn (1978) has provided the results obtained for a ship that approaches in transverse direction (probably pulled by towboats), and next collides with an elastic fender wall. The collision itself will not be discussed any further here, except for the determination of the impulse-response function. Figure C3.13 shows data that have been determined in the frequency domain. These result from scale model tests, in which the ship was forced in different directions of movement (in the horizontal plane) into an oscillating movement and from calculations. In case of purely harmonic oscillations of the ship in transverse direction, the forces operating on the ship generated by the water have been divided into forces that are in-phase with the movement and forces that are out-of-phase. By dividing the in-phase force by the (instantaneous) acceleration of the ship and next by detracting the natural mass of the ship from the result, the added mass is obtained, which turns out to be a function of the angular frequency . This has been plotted out in the figure on the left (dimensionless) as a(). In order to make this added mass dimensionless it has been divided by the mass of the ship itself. The added or hydrodynamic damping results from dividing the out-of-phase force by the velocity of oscillation. This damping is represented by b(). The data of Figure C3.13 have been obtained through various methods: a. b. c. d. By carrying out scale model tests (see the measuring results). By using the long-wave theory (this approach is only valid in the low-frequency range and in case the water depth is not much greater than the draught of the ship). By using the potential theory as mentioned in Paragraph 3.2.2. In Fontijn (1975), results are given for the entire frequency domain. By adapting the results from c, by drawing a line across the points of measurement in a certain frequency range.

45

Delft Hydraulics

Figure C3.13: Added water mass and added water damping in case of forced oscillation of a ship in transverse direction (Fontijn, 1978). In this particular case (i.e. only in case of an uncoupled horizontal movement in the direction of the displacement) the impulse-response function is defined as the velocity that the ship obtains after a unit impulse; this is a function of time as, regardless of the influence of damping, the ship obtains a final velocity. The relation below was deduced (Fontijn, 1978): k (t ) = b( ) cos(t ) 1 2 d + (m + a(0)) 0 {m + a( )}2 2 + b 2 ( )

(C3.82)

46

Delft Hydraulics

It is recognizable that, when the water terms would not be present, the unit impulse would therefore generate a velocity that equals the impulse divided by the mass. The term a(o) also relates to the added water mass in case = 0. A similar mass may also be found in Part A, Figure A3.1. Figure C3.14 represents the impulse responses that relate to the various approaches.

Figure C3.14: The impulse-response function of a ship in case of pure displacement (transverse movement) (Fontijn, 1978). When using the long-wave theory it appears to be possible to obtain a reasonably accurate impression of the added water mass and damping. In Paragraph 3.5.3 it is indicated that, when using the long-wave theory, it is also possible to calculate directly in the time domain. When using the long-wave theory however, first it needs to be verified whether the frequency range in which the excitation is situated corresponds to the low frequencies of the theory. Yet another approach is possible for calculating in the time domain. This method is somewhat comparable to the impulse-response method, but it is based on the force equations in which the water forces are generated due to the fact that at a moment in time in the past, the ship had a certain acceleration and a certain velocity. Appendix I of this part presents the deduction of the corresponding memory functions. The remaining forces in the system are those of the classic equation of mechanics, in which the mass of the ship and the fender wall play a role, as well as the stiffness of the fender wall.

47

Delft Hydraulics

3.5 Calculations in the time domain with the direct method 3.5.1 General
Calculations in the time domain offer the following advantages. They make it possible to determine the response of a structure directly, in which the fundamental equations hardly require any elaboration. Non-periodic loads and loads of short duration do not generate extra complications. The equations used when calculating in the time domain may be easily changed and non-linear terms may be included completely.

There are also disadvantages however. Added water mass and water damping in principle are only known in the frequency domain and therefore must be elaborated toward the time domain. This may be complicated and requires specialized knowledge. In the high-frequency range, the frequency-dependency disappears and it is possible to calculate directly in the time domain. In case of vibrations it is necessary to calculate with small time intervals, while the calculation across many vibration periods must be carried out before an equilibrium vibration is reached. Many calculations are required in order to determine connections between magnitudes.

So far, very few calculations in the time domain have been carried out, and very little has been published about this subject. The following paragraph only discusses the single mass spring system. Multiple systems may be discussed in exactly the same way, which may also be seen in Paragraph 3.5.3, which discusses coupled structure-fluid systems. The core of the dynamics equations of structures is always the acceleration of the masses due to external forces (external load) and forces from within the system (spring and damping forces). In case of the presence of fluid, the inertia of the fluid is important as well; the acceleration of the flow is coupled to a pressure gradient or, in case of (pipe) flow, to a hydraulic head. As already mentioned in Paragraph 3.3.6, in case of a coupled fluid-structure system the number of degrees of freedom equals the sum of the number of degrees of freedom of the structure and the number of culvert parts in which the discharge may vary independently from other discharges.

3.5.2 Response of a single mass spring system to an external load


The basic equation for the single mass spring system is (see also Paragraph 3.3.1): m d2 y dy + c + ky = F (t ) 2 dt dt (C3.83)

48

Delft Hydraulics

in which m = mass, c = damping coefficient, k = spring stiffness, F(t) = external force that varies with time. This equation now needs to be solved in the time domain. For this, the initial conditions need to be known: t0, y0 and (dy/dt)0. It is important in what way, at each time interval t, the initial conditions t1, y1 and (dy/dt)1 change into t2, y2 and (dy/dt)2. To determine this, the auxiliary magnitude z is introduced: dy z= (C3.84) dt Now Equation C3.83 may be rewritten as: m dz + cz + ky = F (t ) dt (C3.85)

With this, Equation C3.83, which is of the second order, is changed into two coupled Equations, C3.84 and C3.85, of the first order. Through the first coefficient at a (random) moment in time t1 it may be calculated what the (new) values at moment in time t2 (a time interval t later) will become. t2 = t1 + t
dz z2 = z1 + t dt 1

(C3.86) (C3.87) (C3.88)

y2 = y1 + z1t

Preceding the elaboration of Equations C3.86 through C3.88 it is necessary to carry out a preliminary elaboration, which is the introduction of the core equation of the dynamic behaviour: dz F (t1 ) cz1 ky1 (C3.89) = m dt 1 Now it is possible to elaborate Equations C3.89 and C3.86 through C3.88 (in that order) one after the other for each time interval. At the end of each elaboration the clock needs to be put back, and the final values obtained become the initial value of the following calculation: t1 = t2 (i.e. the new t1 is made equal to the old t 2 ) y1 = y2 z1 = z2 The great advantage of calculating in the time domain is that it is very easy to introduce non-linear damping or spring stiffness. In that case, only Equation C3.89 changes. It is also possible to introduce the fact that the load is a function of y or dy/dt. 49 (C3.90)

Delft Hydraulics

Investigations have been carried out concerning the influence of the magnitude of the time interval on the calculation results in case of a single mass spring system, ending with the conclusion that in case of a time interval that is too great, the numeric error results in a certain resonant rise of the vibration. In case of a single mass spring system, this resonant rise may be compensated by introducing an extra damping, on top of the already existing damping. From a review of the dimension it may be deduced that the numeric error is expressed in an error in the logarithmic decrement, also referred to as the relative damping , that was defined in Paragraph 3.3.1, under Equation C3.30, as:

=
or:

cn 2k c 2mn c 2 km

=
or:

(C3.91)

This value is a univocal function of the relative time interval (i.e. the time interval that is made dimensionless by dividing it by the resonance period, T). By using the relation T = 2/n (in which the resonance-angular frequency n follows from Equation C3.27) and recalculating the value to a correction factor on the damping value c, it is found that:

corr . =

ccorr . t = 2 km 2 m / k

(C3.92)

Now it is systematically found, when comparing the calculation results with analytically known solutions, that ccorr. and t are proportional to each other in a linear way. Therefore: ccorr . = 2 km with = proportionality constant. From this, it directly follows that: ccorr . = k t Based on the comparison between the calculation results and the analytical solutions mentioned above, it appears that the proportionality constant equals 1. Therefore: ccorr . = k t (C3.95) (C3.94) t m/k

(C3.93)

With this, it is possible as an example, to apply small time intervals of only 0.03 times the resonance time, where this would normally have been 0.001 times the resonance time. This results in great time-saving. 50

Delft Hydraulics

It is of course also possible to introduce a refinement of the calculation method. A fourth-order Runga-Kutta method works very well, also in case of multiple mass spring systems. The Relation C3.95 that represents the correction on the damping coefficient, has not been tested for a system with multiple degrees of freedom. Calculations of the bath plug vibration in the time domain (for figure and symbols see Part A, Figure A4.8) have shown that ccorr. is not constant, and that it even depends on the gap width of the bath plug; this relates to the fact that the water generates an added water stiffness as well.

3.5.3 Coupled systems with structure components and fluid components


This paragraph presents four examples of situations in which water and structure constitute one composite system. a. b. c. d. a. The bath plug vibrations. The instable oscillations of floating sector gates. The instability of floating flap gates in case of wave load. A ship moored in transverse direction at a fender structure.

Bath plug vibrations In Part A, Paragraph 4.4.2, the bath plug vibration was analyzed, while Paragraph 3.3.7 of this part discusses in greater detail the equations, in which the structural and hydrodynamic elements are described as one single system. The same review is repeated here once again with respect to calculating in the time domain. The equations are now written somewhat differently however. For the sake of clarity the diagram of the bath plug is presented once again in Figure C3.15.

51

Delft Hydraulics

Figure C3.15: Simplified schematization of the bath plug system. The two magnitudes that vary with time are the pipe discharge, Qp, and the displacement of the bath plug, y. As the latter magnitude is determined by the forces that operate on the bath plug, initially only the acceleration of the bath plug is found (therefore d2y/dt2 = F/m). This second-order equation is again elaborated into two first-order equations by introducing: dy (C3.96) z= dt The equations for the magnitudes that vary with time now become, using: t2 = t1 + t as follows: z2 = z1 + (dz / dt )t y2 = y1 + z1t Q p 2 = Q p1 + (dQ p / dt )t (C3.98) (C3.99) (C3.100) (C3.97)

Before it is possible to carry out these calculations however, the magnitudes dz/dt and dQp/dt still need to be determined. 52

Delft Hydraulics

First, a number of auxiliary magnitudes are calculated: Qgap = Q p + Ap (dy / dt )

(C3.101)

(Ap = surface of the cross-section of the pipe),


Vgap = Qp

( + y )O

(C3.102)

(O = gap circumference, = initial gap width, = discharge coefficient),


V pipe = H gap = Qp Ap
2 Vgap

(C3.103) (C3.104)

2g

The force that operates on the bath plug in direction y is:


F = g H gap . Aplug

(C3.105)

with Hgap according to:


H pipe = H H gap
( Aplug
1 4

(C3.106)

D2 )

Now the two missing magnitudes may be calculated:


dz / dt = F / m

(C3.107)

and, based on the hydraulic head of the pipe, the flow acceleration and therefore the dQp/dt:
2 V pipe g 2 D dQ p / dt = H pipe 2 g Lp 4

(C3.108)

This calculation seems difficult, but when elaborating it, it turns out to be quite straightforward. When using these equations, the most difficult problems seem to be finding the right initial position (the value of the initial gap and the initial value of y), in which vibrations do occur, but the plug is not immediately sucked in. This problem becomes smaller when the pipe length is greater. Another problem is the very small time interval that is required. Here, it is appropriate to use the Runga-Kutta procedure. Instable oscillations in case of floating sector gates This example relates to investigations of the dynamic behaviour of the storm surge barrier in the Rotterdam Waterway, as already discussed in Paragraph 3.3.7. The phenomenon that was investigated has already been discussed in general terms in Part A, Paragraph 4.5. The situation has been presented once again in Figure C3.9. b.

53

Delft Hydraulics

In order to capture the phenomenon of transverse oscillation of the downstream water in a one-dimensional calculation model, the same schematization is used as for calculating in the frequency domain in which the same example was dealt with (Paragraph 3.3.7). Again, the starting point is the (bisected) system of communicating vessels (Paragraph 3.3.7 and Figure C3.12). The length of the vessel (L1) and the length of the pipe between vessel and downstream water (L2) together more or less correspond to half the distance between the abutments. Only the vessel is supposed to be filled and emptied due to the extra discharge, qd, that flows underneath the gate due to the oscillation of the gate; the permanent part of the discharge of the gate is ignored. The same applies to the pipe discharge (qp). Based on a two-dimensional scale model investigation, specially carried out for this purpose, it was determined how great the buoyant force was as a function of the water levels and the position of the gate. It was also known how great the discharge of the gate was per unit of gate length, again as a function of the water levels. The dynamic behaviour of the gate has been included in the calculations in the time domain, which means that the vertical acceleration of the gate is represented by: d 2 y F0 G = dt 2 m (C3.109)

(m = gate mass including an estimate (frequency-dependent) of the added water mass, G = natural weight, Fo = buoyant force, y = vertical displacement of the gate, all this per unit of gate length). Now, suppose that at a moment in time t1 everything is known: the water level rise in the basin, z (the water level upstream of the gate is assumed to be constant); the position of the gate, y; the vertical velocity of the gate w (or, similar to the previous paragraph w = dy/dt); the discharge. For the situation at a moment in time t2, a time interval t later, it is possible to calculate everything. The indices 1 and 2 are connected with moments in time t1 and t2 respectively for all equations following after this. z2 = z1 + qg1 q p1 Lbasin t (C3.110)

in which qg1 (the gate discharge at moment in time t1) may be determined on the basis of the water level in the basin, z1, and the gate position y1, at moment in time t1. Furthermore, for the vertical gate velocity w, it holds: w2 = w1 + Fo1 G t m (C3.111)

Fo1 follows from the water levels and the gate position at moment in time t1. For the vertical gate position it holds that: y2 = y1 + w1t (C3.112)

The acceleration of the water in the pipe, dVp/dt, is proportional to the hydraulic head of the pipe and the gravitational acceleration, g, and inversely proportional to the pipe length, 54

Delft Hydraulics

Lpipe. The hydraulic head here is the water level rise, z, of the basin. If the height of the pipe is defined as hc, then it is found that: gz (C3.113) q p 2 = q p1 + hc 1 t Lpipe Now all magnitudes are known at moment in time t2 and the calculation for the next time interval may be continued. The results of the calculations in the time domain showed a good correspondence with other calculations. At a later stage it was possible to include a large number of other components in these calculations, such as the variation of the cross-section of the immersion as a function of the downstream water level minus the flow-through opening underneath the gate. c. Instable behaviour of floating flap gates in case of wave load This concerns the situation of the floating flap gate barrier as described in Part A, Paragraph 4.5.2 and in Jongeling/Kolkman (1995). It was possible to describe the oscillations that occur with a communicating vessel system with a synchronically operating wave board; see Figure C3.16, taken from Figure A4.27 in Part A. The symbols of this figure are also followed.

Figure C3.16: Diagram of communicating vessels with a synchronic operating wave board.

55

Delft Hydraulics

Here again, the point of departure is the situation at a random moment in time t1, based on which the coefficients (of time) of the magnitudes that vary with time may be calculated. Using these coefficients, the magnitudes are determined again at a moment in time t2, a time interval t later. t2 = t1 + t The basic equations to obtain the magnitudes at moment in time t2 are:
dZ Z L2 = Z L1 + L t dt 1 dZ Z R2 = Z R1 + R t dt 1 dQ p Q p2 = Q p1 + t dt 1

(C3.114)

(C3.115)

(C3.116)

(C3.117)

The indices L and R relate to the basin on the left and the basin on the right respectively. In order to calculate dZ/dt and dQ/dt, the continuity equation and the movement equation are used. These calculations must precede the elaborations mentioned above. The auxiliary elaborations are: dZ L Q p1 + Vg1 (h0 + Z L1 )b = dt S byg
dZ R +Q p1 + Vg1 (h0 + Z R1 )b = dt S by g dQ p dt = g ( Z L1 Z R1 ) Ap 2 Lp

(C3.118)

(C3.119)

(C3.120)

In this, yg and Vg are the displacement and the acceleration of the wave board (toward the basin is positive) respectively. REMARK 1: As may be seen in the last set of equations, the surface of each basin is decreased by the wave board displacement multiplied with the width. In the calculation model it was found that, if this is not included in the calculation, the resonant rise of the water levels is two times greater, and that is an erroneous result. This has been verified at a later stage, using an analytical review. REMARK 2: The pipe is assumed to be frictionless, but for the calculation as such that is not necessary. In the last equation it might be possible to reduce the water level difference between the vessels with the friction head.

56

Delft Hydraulics

d.

Ship moored in transverse direction at a fender structure As already mentioned in Paragraph 3.4.2, it is possible to calculate directly in the time domain of the calculation of fender forces in case of a ship that collides in transverse direction with a fender structure, using the one-dimensional long-wave theory. It is not necessary in this case to use the impulse-response function technique. The long-wave theory may be used if the wave frequency is low; it is possible to use this due to the fact that the fenders usually are so weak that the ship is only slowed down gradually. It is however necessary to verify whether the response is sufficiently low-frequency or that the movement is sufficiently slow to be able to use this theory. Kolkman (1977) indicated as the upper limit of the frequency in case of a harmonic movement: g < 0.65 (C3.121) h In this, h is the water depth. In case of this condition, and in case of a periodic wave, the pressure amplitude at the bottom is only 20% lower than at the surface, while the celerity of the wave still is 12% lower than in case of the long wave. The conditions of the long wave are therefore assessed reasonably accurate. In case of a collision, the lower limit of the time duration Tmax (i.e. the time duration between the beginning of the collision and the moment in time at which the maximum force is reached) is: g Tmax > 2.5 (C3.122) h There still is a number of conditions that need to be met, which moreover in practice hardly results in any limitation: The ship needs to be of such a length that the discharge that flows around the bow and stern of the ship is small relative to the discharge that is coupled to the wave radiation, and The water depth underneath the ship, relative to the width of the ship, needs to be so small that there is pipe flow.

In case of long waves it holds that the discharge is distributed equally across the vertical.

57

Delft Hydraulics

Figure C3.17: The one-dimensional situation. Below, it is assumed that a ship that approaches in transverse direction (in reality the propulsion is generated by towboats, part of which push, and part of which pull the ship forward). The ship sails with a uniform velocity Vs-0. Initial drag forces are ignored, as these are compensated by the forces of the towboats. In this case there is also no wave radiation, as the water displaced by the ship is completely compensated by a reverse flow underneath the ship. As all effects that relate to the permanently moving ship are small, the same relations that are found for the relation between fender forces that operate on the ship and the velocity changes that the ship experiences, are also found in case of a ship that initially is at rest. Both left and right of the ship waves of equal height are generated and their sign is inversed. The water level difference results in an acceleration of the water underneath the ship and causes the transverse forces that operate on the ship.

Figure C3.18: The situation after the start of centric colliding. 58

Delft Hydraulics

At moment in time t = 0 (see Figures C3.17 and C3.18 for the agreements on the signs) it now holds that: y0 = 0 The ship velocity Vs is: Vs = Vs 0 The force that the water exerts on the ship is: Fw = 0 The discharge at the side of the ship is: q0 = 0 The water velocity underneath the ship is: u0 = Vs 0 D hD (C3.127) (C3.126) (C3.125) (C3.124) (C3.123)

Moreover, in case of permanency, for the wave height it holds that: z0 = 0 (C3.128)

The core of the calculation in the time domain is that the force operating on the ship is deduced from the hydraulic head of the ship and the spring compression, and because of that the acceleration as well, while the acceleration of the water column underneath the ship also follows from the hydraulic head. By integrating these factors in time, this results in the ship velocity and the water velocity underneath the ship. The other magnitudes, q and z, are coupled to this. The wave height z is coupled to the discharge q. As, in case of long waves, it holds for the celerity, c, that: c = gh (C3.129) and on the basis of continuity it may be deduced that:

cz = q
and it holds:
z= q gh

(C3.130)

(C3.131)

and therefore:

59

Delft Hydraulics

z=

Vs D + u (h D) gh

(C3.132)

Now it may be shown that, if a positive wave with height z is generated at the righthand side of the ship, a negative wave with height z is generated at the left-hand side. For the celerity of both waves is equal and there is no water that is withdrawn from the system or added to it. As the pressure distribution is hydrostatic in case of long waves, the force that the water exerts on the ship is found to be:

Fw = 2 gzDL
In this, L is the length of the ship. For the acceleration of the ship this now holds:

(C3.133)

dVs = Fw + elasticity ( y ) dt

(C3.134)

The elasticity may be a random function of the compression. A positive displacement y will result in a negative elasticity. For the water underneath the ship it holds (inertia water column underneath the ship) that: du 2 gz = B dt (C3.135)

For the sake of simplicity the flow resistance has been ignored here. Arithmetically, there is no problem in including this as well. In reality this means that in that case only the extra resistance in the equation for the collision condition (compare Equation C3.125) must be included. The last precondition is: dy = Vs (C3.136) dt Using Equations C3.132 through C3.136, it is possible to calculate the corresponding values at moment in time t2, one time interval t later, based on t, Vs, y and u.

60

Delft Hydraulics

CALCULATION METHODS FOR WAVE IMPACTS 4.1 General

Wave impact loads are dynamic loads to which the structure responds dynamically. Therefore the normal periodic wave load that is generated by wave reflection, wave diffraction and flow forces due to the water movement corresponding to waves is not discussed here. These forces are experienced by the structure as a quasi-static load due to the relatively high resonance frequencies. Part B discussed in detail the different ways in which wave impact loads may be generated. As far as load and response are concerned, the following elements may be distinguished.

The total impulse of the impact. The time-dependent wave impact load. The response of the structure in time. The interaction forces between the moving structure and the surrounding water. The possible influence (feedback) of the impact load due to the responding structure.

Not all these elements are (completely) accessible for calculation. This chapter mainly deals with the calculation of the wave impact load. For the calculation of the maximum in the impact load and, in some cases, the impact duration, various analytical models have been developed that will be discussed in the paragraphs below. These calculations naturally require input data, such as the velocity of the incoming water. These data generally need to result from scale model investigations. The analytical model that is used will have to be based on experience and physical observation. Calculation only makes sense when a number of input data is known, such as the velocity by which the water approaches the wall of the structure and the amount of enclosed air. For that reason, the calculation in reality will rarely be considered separately from using a scale model. As long as it is not known what phenomenon is dominant in a certain situation and for a certain type of wave impact (and at this point, little is known about this), the most unfavourable analytical model will have to be used. The same problem of choice occurs in case of interpreting the results of scale model investigations. As this directly relates to the calculation problems, this chapter concerning calculation methods will also discuss the interpretation of (scale) model results (Paragraph 4.8). Chapter 3 already discussed in detail the calculation of the response of a structure to a load that varies with time (such as a wave impact). In this, it is (implicitly) assumed that the load is not influenced by the response of the structure. The calculation method of the response without feedback is therefore not discussed any further at this point. The interaction forces between water and structure due to the response of the structure mainly consist of a hydrodynamic inertia force and a hydrodynamic damping force. The latter is a consequence of wave radiation, and is in fact only relevant in case of slow movements of the structure, i.e. when 2h/g < 10, with = angular frequency of the movement, h = characteristic dimension of the structure. About this, see Part A, Figure A3.1, where in the figure on the right, the pressure coefficient that relates to the wave radiation has been plotted out. For most hydraulic structures at which vibrations due to flow or wave impacts occur, 61

Delft Hydraulics

2h/g > 10 is the case, and this damping term may be ignored. The calculation of the added water mass has already been discussed in Paragraph 3.2.

4.2 Impulse theory


A wave that approaches the surface of a structure and causes an impact there, possesses a certain amount of movement or impulse. The impulse I is a vectorial magnitude and therefore has a direction. The impulse is defined as:

I = mv
with: m = the moving water mass v = water velocity (vector)

(C4.1)

At the moment of impact, part of this impulse, i.e. the component perpendicular to the surface of the structure (Ik), is reduced to zero; this means that the water velocity perpendicular to the surface is reduced to zero in a very short period of time. The force F required for this, must be generated by the structure, and according to the second law of Newton it equals:

F=

d(mv ) k dt

(C4.2)

with: m = the water mass involved in the impact v = velocity component perpendicular to the surface If the change of the impulse Ik as a function of time is known, the force development as a function of time is also known. In order to calculate the force F, it is first of all necessary to assess the impulse, therefore of m and v, at the moment of the collision. If the wave period and local wave height are known, it is possible to estimate the approach velocity v of the water surface with reasonable accuracy. The water mass m that is involved in the impact follows from a review of the added water mass (see also Paragraph 3.2). In this review, a unit acceleration perpendicular to the surface is exerted on the fluid edge, which is similar in size to the contacted surface, after which the inertia pressure that then occurs is calculated. When divided by the acceleration, this then results in the added mass. In case of great accelerations (in fact, in case of a short duration of the load, such as in case of wave impacts), the added water mass may be assumed to remain constant (i.e. independent of the magnitude of the acceleration); this makes the calculation of the impact load simpler. For a number of geometries, the added water mass is known, so that it is possible to make use of these immediately. The calculation method of Paragraph 3.2 is completely suitable for the determination of the added water mass that is relevant for a wave impact. Part A, Paragraph 3.2.2, Figure A3.4 presented the added water mass of a gate with height h1 that is situated on a sill. The complete water depth is h2. This situation is fully analogous to that of a wave front with height h1 that needs to be slowed down quickly, in which the water depth is h2. Next, the slowing-down time of the water (i.e. the time in which the velocity of the water surface is reduced to 0) needs to be estimated, as well as the velocity development in time. During an initial approach, the velocity development could be assumed to be linear. The slowing-down time depends on the possibility of distortion of the various elements involved 62

Delft Hydraulics

in the collision, such as the structure, the possibly enclosed air, the water itself and, in case of the presence of air bubbles in the water, the air-water mixture. The weakest element is the most determining factor of the slowing-down time: the weaker this element, the greater the slowing-down time. In reality it is difficult to estimate the slowing-down time, so that an accurate determination of the amplitude of the impact load is not really possible. In Part B, Paragraph 5.5 a cylinder that fell with a constant velocity on the water surface, penetrating the water surface, was discussed using this method. A review of the impulse however is important, as it is known on the basis of measurements in what range the impact duration for some specific geometries might be. Using this, it is also possible to determine the range of impact amplitudes for an estimated impulse. As the response of the structure mainly depends on the ratio between the impact duration and natural vibration time T (see also Chapter 6 of Part B), it is also possible to calculate the maximum possible response at the given impulse, assuming the unfavourable ratio of /T. In scale model investigations it is very important, as indicated below, that the magnitude of the impact impulse is measured. The impulse Ik follows from the integration of the impact pressure, p(t), across the contacted surface A and across the impact duration .

I k = p(t )dtdA = F (t )dt


A

(C4.3)

The pressure p(t) may be measured in the scale model as a function of time and place, but it is also possible to determine the total force F(t) separately or simultaneously. In a scale model that is based on the rules following from the invariance of the Froude number (see Chapter 5), it is now possible to scale up the impulse with the scale rule
3.5 nI = nL

(C4.4)

(nL = length scale). The scale rules for the composite magnitudes of the impulse, i.e. those for the impact pressure (or the integrated force) and those for the impact duration, however may not always be indicated just like that. The scaling of these magnitudes depends on whether air is present or not and whether the air may be compressed or not during the wave impact; in this, it is also important that for practical reasons the air pressure in the scale model usually is not scaled, as a result of which scale effects are generated. The scale rules therefore are not fixed in advance. Paragraph 4.8 discusses this in greater detail. The impulse that may be scaled directly may be used for an initial assessment of the effect of the wave impact on the structure. In case of a wave impact against a structure, it is normally the water that moves and the structure that stands still. The issue of the load becomes more complicated when both the water and the structure are moving. This occurs, as an example, in shipping (a heaving ship in a wave field) and is referred to as slamming. Imagine two bodies with an impulse (mv)1 and (mv)2 respectively, then in the special case that (mv)1 = -(mv)2, i.e. in case of an opposite movement, a situation of a collision will be caused that, due to considerations of symmetry, may be replaced by a situation in which each of the bodies collides with a fixed wall. In case (mv)1 -(mv)2, the collision force of the body with the, in an absolute sense, greatest impulse, will be smaller than in case of a collision with a fixed wall (as the slowing-down time dt increases); for the body with the, in an absolute sense, smallest impulse, the opposite applies. 63

Delft Hydraulics

Due to the equilibrium in the plane of the collision, the collision forces of both bodies are equal. Translated to the wave impact problem, this means that the slowing-down time dt is now determined by the distortion or the recession of both the water and the structure. As a result, estimating the slowing-down time dt is more difficult in that case then when it is only the water that moves. In shipping the slamming model (i.e. a flow-pressure model) is often used to determine the wave impact load (Fabula, 1957). In this model the relative velocity between ship and water is used. See Paragraph 4.3.

4.3 The linear shock wave model 4.3.1 Wave impact against a rigid wall
This model presupposes that the water approaches the fixed wall in perpendicular direction and is not capable of running off sideways (sufficiently). At the moment of impact the water is compressed and a shock wave is generated in the water that moves back from the wall. The front of the shock wave travels with the speed of sound in water cw (1430 m/s in pure water). It may now be deduced (see Kolkman, 1981) that:

p = pmax p0 = wcwv0
with: pmax p0 w cw v0 = = = = =

(C4.5)

maximum impact pressure pressure in the water before the impact (= patm at the free water surface) density (specific mass) of the water speed of sound in water approach velocity of the water front

This is the well-known hammer shock equation (for a single hammer shock problem). The Equation C4.5 includes:
cw = K

(C4.6)

in which K = compression module = -p * V / V (in this, p = pressure change, V = volume change, V = original volume). It may also be written as:

p = v0 K w

(C4.7)

In this, it is assumed that the density w is constant (which does not apply in case of a compression wave; the variation however is relatively small) and that the compression module is constant (therefore a homogenous fluid). The shock wave model as described above provides an absolute upper limit value for the impact pressure. In reality, the impact pressures will always be lower, first of all because the probability that the water surface is completely flat and makes contact with the structure in a parallel way (as a result of which the water is not capable of running off (sufficiently) sideways) is negligibly small, and also because there usually is air in the water, as a result of which the impact pressure is reduced.

64

Delft Hydraulics

4.3.2 Rigid wall and air-water mixture


In case there is air in the water, it may be asserted that the compressibility of the airwater mixture is determined by the air (this is the weakest element) and the density by the water. The compression module of air follows from the equation for the change of condition of an ideal gas:

pV = constant
with: p V = pressure = volume = Poisson constant (isothermal compression: = 1.0; adiabatic compression: = 1.405).

(C4.8)

The Poisson constant has a value that is between = 1.405 for adiabatic compression and = 1 for isothermal compression. In case of large air pockets, the compression develops more or less adiabatically. Differentiation of Equation C4.8 results in:

V dp + pV 1dV = 0
or, when pressure changes relative to the initial pressure, p0, are small: dp p = dV V

(C4.9)

p0 K = V V

(C4.9a)

When the air-water mixture is assumed to be homogenous with = air volume and (1 ) = water volume, then it more or less holds for the air-water mixture that:
K mixture =

mixture

p pV 1 = K= 0 V = (1 ) w

(C4.10)

(as, in case of a mixture, only the air part is compressed, the volume change due to a given pressure change is a factor smaller, and therefore the compression module is a factor -1 greater than in case of pure air). Using Equation C4.10, Equation C4.7 may be rewritten as:
p = v0

p0 (1 ) w

(C4.11)

When there is not a lot of air in the water (for example 0.1 percent) or when the air is distributed in small air bubbles across a certain volume of water, the compression of the air develops almost isothermal (=1), as in that case the heat may be quickly transferred to the water. In order to use the equation above, an estimate of the amount of air present in the water is required. The amount of air however may vary strongly for each situation; this, among 65

Delft Hydraulics

other things, depends on the previous history at the location of the impact zone (previous impacts and breaking waves are capable of introducing a lot of air in the water) and of the discharge of the water (flow or no flow). The maximum amount of air probably is a few percent; about this, reliable data are not available. Fhrbter (1996) has provided an expression for the influence of air on the water at the maximum impact pressure in a factor by which p, according to Equation C4.5, needs to be multiplied. He found that:

p = pmax p0 = wcwv0

1 K 1 + ( w 1) Ka

(C4.12)

with: Ka = compression module of air = p0 ~ 105 N/m2 for isothermal compression Kw = compression module of water ~ 2 109 N/m2 This equation indicates that the maximum pressure, calculated with the hammer shock equation C4.5, is reduced when the compressibility of the enclosed air bubbles is included. The equation results in impact pressure values that are closer to what was measured in reality than in case of the hammer shock equation. There is however a problem in making a reliable estimate of the air percentage in the water. One difficulty in this is also that the air is not always homogenously distributed and that the air percentage may vary strongly with time.

4.3.3 Wave impact against a compressible wall


In case of a wave impact against a compressible wall, a compression wave is generated both in the water and in the wall. The approach velocity of the water is v0; the compression velocity of the wall after the collision is established at v0, as a result of which the velocity change of the water is (1-)v0. The compression wave in the water has a velocity cw, the compression wave in the wall has a velocity cc. The pressures pmax in the contact surface of the water and the wall are equal and, due to an equal initial contact pressure p0 = patm, the pressure leap across the compression waves is the same as well: p = wcw (1 )v0 = wcw v0 with w = density (specific mass) of water c = density (specific mass) of structure 1 (C4.13)

After elaboration this results in: p = wcwv0 1+

wcw c cc

(C4.14)

It appears that the maximum impact pressure calculated with the hammer shock equation C4.5, by including the compressibility of the wall, is reduced. Relative to the reduction due to the presence of air in the water, the effect of a compressible wall however is generally negligibly small. In reality, structures (such as steel plate fields) will tend to deflect, rather than compress; in that case the equation above does not apply.
66

Delft Hydraulics

4.4 The non-linear shock wave model


In the linear shock wave model the point of departure is the fact that the density w and the compression module Kw of the water, and as a result the velocity of the compression wave cw, are constant. It is also assumed that the approach velocity v0 is small relative to the velocity of the compression wave cw. Moreover, in reality the latter is usually the case, even when air is present in the water (cw = 1430 m/s in pure water and cmixture drops to a lowest value of approximately 20 m/s in case of an air volume of 50%; in case of 1% air, cmixture = ca 100 m/s, see Allersma (1961) and Kolkman (1981)). In case the impact pressure is great relative to the initial atmospheric pressure, p0, the linear theory no longer holds. In general, for cases with air bubbles in the water, it is indicated below what changes there will be. At the location where the wave makes contact with the structure, an immediate pressure increase is generated. This pressure increase travels through the water (actually, the air-water mixture); the front of the pressure wave moves away from the structure, see also Figure C4.1. In the part of the water behind the shock wave front there is an increased pressure, and ahead of the front there is the original pressure. The air bubbles behind the front are compressed, as a result of which the air becomes stiffer; by comparison, the compression of the water is negligible. Due to the decrease of the volume of air bubbles, the specific mass of the mixture increases, but the relative increase is one order of magnitude smaller than the relative increase of the stiffness of the air. This also means that the shock wave velocity increases in case of increasing compression of the air.

Figure C4.1: Diagram for the calculation of a shock wave. If the pressure increase in the water behind the shock wave front is represented by p1 and the pressure ahead of it p0 (p0 is the initial pressure, slightly above the atmospheric pressure), then for the volume change of the bubbles from V0 to V1 it holds that: p0V0 = ( p0 + p1 )V1 or: (C4.15)

67

Delft Hydraulics

V1 = V0 (1 + p1 / p0 ) 1/ The volume of air in a unit volume of the mixture is therefore reduced to (1+p1/p0)-1/. The new air volume in the compressed mixture now becomes:

(C4.16)

(1 + p1 / p0 ) 1/ = 1 + (1 + p1 / p0 ) 1/
*

(C4.17)

The density of the mixture before compression approximately equals:

mixture = (1 ) w
and after compression it equals:

(C4.18)

mixture
and this may be presented as:

(1 + p1 / p0 )1/ = (1 ) w = 1 1 + (1 + p1 / p0 ) 1/
*

mixture = w
After writing down the mass balance:

1 1 + (1 + p1 / p0 ) 1/

(C4.19)

* v0 mixture = c( mixture mixture )

and the momentum balance:

p1 = mixture v0 (c + V0 )

for the relation of the extra pressure that is generated behind the shock wave front, p1, relative to the initial pressure ahead of the front, p0, it follows that:
p1 p0
1/ 2 p1 (1 ) wv0 1 1 + = p0 p0 2 (1 ) wv0 p1 is plotted out against . p0 p0

(C4.20)

This is plotted out in Figure C4.2. In this,

68

Delft Hydraulics

Figure C4.2: Graphical presentation of Equation C4.20 (the non-linear shock wave model), taken from Kolkman (1981/1992). If the term on the left in Equation C4.20 is developed using a Taylor sequence, and only the p linear term is included (therefore only valid for as long as 1 is small), then the term p0 1 p1 p ) . Eventually, the solution C4.11 is found again, i.e. the (1 + 1 ) changes into (1 p0 p0 solution of the linear shock wave model. Figure C4.2 also shows in what range the linear shock wave model is valid. Just like in case of the linear shock wave model, when using the non-linear shock wave model first an estimate needs to be made of the air volume in the water. In reality however, this is very problematic.
1

69

Delft Hydraulics

4.5 The flow-pressure model (ventilated shocks)


In case of the flow-pressure model, the maximum impact pressure pmax occurring during a wave impact is expressed in the flow pressure: 1 2 pmax p0 = k ( v0 ) 2 with: k p0 v0 = coefficient (in shipbuilding: slamming coefficient) = pressure before the shock (= patm) = approach velocity of the water (C4.21)

The coefficient k for a given geometry and at given hydraulic boundary conditions needs to be determined by testing. In some cases a theoretical estimate may be possible (for example in case of a circular cylinder, see also Chapter 5, Part B). In case of the flow-pressure model it is assumed that no air is enclosed between the water front and the structure; the water front therefore approaches the structure surface at a small angle (in reality this is nearly always the case), and there are no raised edges and the like that obstruct the outflow of air. It is also assumed that after the impact the water is capable of running off sideways quickly enough, as a result of which compression of the water is of secondary importance. The impact pressure is now determined by the inertia phenomena that occur, in connection with the blocking of the water. According to the second law of Newton, it is possible to express this as follows: d(mv) p= or: dt (C4.22) dv p= w L dt In this, it is assumed that the density of the water, w, is constant. The length measure L is a measure of the amount of water that needs to be blocked (m = wL). In this, L results from an added water mass review (see Paragraph 3.2). It is assumed now, that the velocity v0 is reduced to 0, therefore dv = -v0. If the water front makes contact with the structure at an angle, the slowing-down time, dt, may be considered proportional to a/v0, in which a is a measure for the average space between water front and structure. Therefore, for the impact pressure exerted on the structure it is found that:
2 pmax p0 :: w Lv0 / a

or:

1 2 pmax p0 = k ( w v0 ) 2

(C4.23)

The greater the angle between the water front and the structure, the smaller the coefficient k. The flow-pressure model is very well suited for making a first estimate of the impact pressure, provided it is plausible that no air is enclosed. In the latter case it would be better to use the compression model. A limitation of the flow-pressure model is the fact that for random geometries no (slamming) coefficient k is available. 70

Delft Hydraulics

Slamming coefficients are usually determined by way of drop tests. At the moment of the collision of the falling object with the water, a certain amount of water mass (the added water mass) is caused to accelerate; this is the same amount of water mass that is slowed down, in case of a collision of a moving water surface against the rigidly fixed object. Both experiments resulted in the same slamming coefficients, provided the added water mass is independent of (the magnitude of) the acceleration. This is the case when the acceleration is of short duration. For the added mass it does not matter whether both the water surface and the object are moving. The slowing-down time dt of the incoming water surface however is related to the difference in velocity between the object and the water surface, while the velocity v0 of the water surface that is to be reduced, is an absolute magnitude. The slamming equation C4.23 therefore cannot be applied to this particular case of load just like that. The relative velocity could be introduced into the equation; in that case it is however questionable to use slamming coefficients that have been determined on the basis of tests with objects that are dropped onto a still water surface. Bhattacharyya (1978) provides some results of pressure measurements of the keel plating of a ship in case of moderate wave conditions. When the measured pressures are expressed in vrel2, with vrel as relative vertical velocity between keel plating and moving water surface, an average value of 12.5 is found for the slamming coefficient.

4.6 The air compression model


When air is locked in and compressed between the water surface and the structure, it is possible to use the air compression model. This model does not only provide the maximum impact pressure, but it also provides information about the duration of the impact and about the frequency at which the compressed air oscillates. Places at which air may be easily enclosed are the corners between vertical and horizontal plates. I-beams on their side or vertical shafts with a ceiling are typical structure components at which air enclosure may occur. In this, the air will not be equally compressed across a large surface area; generally the water surface approaches the plate surface at an angle, as a result of which the enclosed air is forced into the corners. For the part of the plate at which the water first makes contact, the air compression model is not valid, but there the flow-pressure model needs to be used. The amount of air that the air compression model is concerned with, is one order greater than that in case of the shock wave model with an air-water mixture. Due to the small air bubbles, it is nearly always possible to take isothermal compression as a basis in the shock wave model (the heat is run off quickly to the surrounding water); in the air compression model, the air bubble is relatively large and the compression will develop adiabatically.

71

Delft Hydraulics

4.6.1 The linear air compression model


Bagnold (1939) was the first to set up an equation for the compression of the (enclosed) air in case of a wave impact. He based the equation on a piston model according to Figure C4.2; the elastic element is the air bubble, the piston itself consists of the water mass.

Figure C4.3: Piston model of Bagnold. The water mass m results from an added water mass review and equals wL per unit of piston surface, with L = added water mass length and w = specific mass of the water. The stiffness of the air is assumed to be constant (linear theory). This is only permitted in case of a small compression of the air bubble. The initial pressure p0 in the compression chamber equals the atmospheric pressure patm. According to Poisson: p (d x) = constant After differentiation, similar to what was found in Equation C4.9, it follows that: dp(d x) p(d x) 1 dx = 0 (C4.25) (C4.24)

In case of a small compression of the air bubble, x d and p p0 = patm. The spring stiffness k of the air per unit of surface of the piston therefore now is: k= dp p0 = dx d (C4.26)

The movement equation for the mass spring system may now be presented as:

w L

d 2 x p0 + x=0 dt 2 d

(C4.27)

72

Delft Hydraulics

with: x = (p-p0)/k The solution of the movement equation is:


p p0 = v0

p0 w L
d

sin

p0 t d w L

(C4.28)

This is a vibration that does not damp out. The maximum in the pressure and the period of the oscillation respectively result from:
pmax = p0 + v0

p0 w L
d

(C4.29)

and
T = 2 d w L p0

(C4.30)

When the linear air compression model is compared with the linear shock wave model in case of an air-water mixture, then it appears that both models result in the same pressure, if it holds that (compare Equation C4.11 with Equation C4.29) (1-)/ = L/d, i.e. when the ratio of the water volume and the air volume is equal. In reality the air volume of an air-water mixture usually is not higher than approximately 1%; the amount of air that is locked in, in case of a compression impact may be considerably greater. As a rule, the linear air compression model will therefore result in a lower pressure for an air-water mixture than the linear shock wave model.

4.6.2 The non-linear air compression model


In the non-linear air compression model great distortion of the air bubble is permitted. The expression for the maximum pressure is obtained by equating the kinetic energy of the water just before the impact with the distortion energy of the air at maximum compression or at maximum pressure. The kinetic energy at moment in time t = 0 just before collision is: Ekin = with A = surface area of the piston The distortion energy results from an integration of the sum of pressure and volume across the distortion trajectory. This results in (see also Kolkman, 1981): 1 2 w LAv0 2 (C4.31)

Edis

p Ad = 0 1

p 1 pmax max + ( 1) p0 p 0
1 1

(C4.32)

73

Delft Hydraulics

Remark: Although the complete deduction of Equation C4.32 is not presented here, roughly it runs as follows: Suppose the piston length is reduced from length d to length d1, then the x as defined in Figure C4.2 runs from the value 0 to d-d1. The pressure difference that the piston experiences en route is (p-p0). The energy that must now be overcome per unit of piston surface is:
Edis =
d d1

( p p0 )dx

Also Equation C4.25 now applies again, therefore: p(d x) = constant = p0 d These two equations result in the energy required to go from piston length d to d1. But d1 is also univocally coupled to the pressure again. Taking pmax for this pressure results in Equation C4.32. Equalizing Ekin to Edis results in the implicit equation:
1 1 p1 2dp0 p1 + 1 + ( 1) + 1 w ( L( 1) p0 p0

v0 =

(C4.33)

with: to p1 = pmax p0

74

Delft Hydraulics

Figure C4.4: Graphical presentation of Equations C4.29 and C4.33 (air compression model), from Kolkman (1981/1992). In Figure C4.4, taken from Kolkman (1981/1992), the relative pressure is: p1 p0 plotted out as a function of the magnitude:
2 v0 w L dp0

(C4.34)

(C4.35)

In this, the initial pressure p0 equals the atmospheric pressure patm. 75

Delft Hydraulics

As the axes are scaled logarithmically, for the linear air compression model Relation C4.29 will be presented by a straight line at a gradient of 1:2. The line for = 1.405 is displaced by a horizontal distance log 1.405 relative to the line = 1. In the diagram, Relation C4.33 results in curved lines. At the place where the lines of the linear model and the non-linear model converge, the more simple linear model may be used. The magnitudes on the horizontal axis of Figures C4.2 and C4.4 have approximately 1 the same significance. In Figure C4.2 the parameter represents the ratio of the water

part relative to the air part of the air-water mixture. In Figure C4.4 the significance of essentially the same.

L is d

4.7 Numeric calculation of the pressure function (in time) in case of a wave impact
In many cases it is not only important to know the magnitude of the impact pressure at the location where the water makes contact with the structure, but also to determine the pressure function (in time) in the fluid and alongside the structure. In this, the prime importance is the magnitude of the total load on the structure. But also the magnitude of the pressure gradient alongside the structure may be important. An example of this is a rock situated in a rock-fill structure (shoulder) at the base of a vertical breakwater: in case of a wave impact against the breakwater the rock, in case of a steep pressure gradient, may experience a sudden great force, due to which the rock is displaced. In case of more displaced rocks, the stability of the breakwater may eventually be endangered. In the next review, the starting point is that the maximum pressure pmax is already known, but the pressure function still needs to determined more accurately. Taking non-compressible water as a basis, the pressure function may be calculated using the following Laplace equation (Delft Hydraulics Report M1504):
2 p = 2 p 2 p 2 p + + =0 x 2 y 2 z 2

(C4.36)

This is a similar equation to that by which the added water mass is determined that is determining for the wave impact as discussed in Paragraph 4.2. The assumption that the water is incompressible, i.e. that pressure changes occur instantaneously across the entire fluid without any phase differences, is justified when the transit time of the compression wave alongside the structure is small relative to the impact duration. The boundaries need to meet the following preconditions: Fixed walls: Free water surface: Fluid edge: Impact surface: 76 p = 0; n = normal direction (Neumann condition) n p = 0 (Dirichlet condition) p changes into 0 at infinite p = pmax in the pressure-time diagram

Delft Hydraulics

4.8 Extrapolation of results from a scale model to prototype values


This paragraph about extrapolation of model results of wave impacts has been included in this chapter on calculation methods as it connects with the previous paragraphs and as calculation methods and scale model investigations are mostly used simultaneously anyhow. Chapter 5, in which scale models are discussed, will of course also refer to this paragraph, concerning the treatment of scale effects in more detail. The calculation models as discussed in Paragraphs 4.3 through 4.6 may only be used when a number of magnitudes is known. For the various calculation models these are: Shock wave without air (linear) (compare C4.14): w, Kw (which together result in cw), v0 Idem, with compressible wall (compare C4.14): w, c, cw, cc, v0 Shock wave with air (linear) (compare C4.12): w, cw, , , p0, Kw, K1, v0 Shock wave with air (non-linear) (compare C4.20): w, , , p0, v0 Flow pressure (compare C4.23): pw, v0, slamming coefficient Air compression (linear) (compare C4.29): w, d, L, , p0, v0 Air compression (non-linear) (compare C4.33): w, d, L, , p0, v0 In practice, this means that first a choice needs to be made concerning the calculation model and, next to that, many other data need to be known before it is possible to calculate the impact pressure. For all calculations, the velocity of the approaching wave front needs to be known; for this, carrying out scale model investigations is indispensable. When using a scale model it goes without saying that the wave pressure is measured as well. Except for the flow-pressure model, none of the calculation models gives cause to assume that the wave pressure may be translated to the prototype through the Froude scale or through another simple scale rule. Also, in a scale model investigation no additional data are made available that make it possible to carry out calculations in a more reliable manner. Without immediately searching for the final answer, it is however possible to carry out a sensitivity analysis, in which the qualitative influence of the various parameters on the impact pressure may be assessed. In this, it may be assumed that just before the wave impact occurs in a scale model, the geometry of the wave and the velocity of the wave front are represented more or less accurately (which, with respect to the detailed geometry of the wave front, does not have to be completely accurate). In the model a wave impact is measured now, in which all the factors mentioned above play a certain role. It is possible to assess for each of the calculation models discussed in the previous paragraphs, what the scaling of the impact is, if it is assumed that pw, , , cw and p0 do not change. Next, the influence of possible deviations of the factors just mentioned is determined qualitatively. The non-linear compression model, the results of which have been summarized in Figure C4.4, may serve as an example. Assuming that the length measures d and L are represented on the length scale, that pw and p0 do not change and that the squared water velocity, v02, has been reduced according to the length scale, then the translation of the model values of the magnitude on the horizontal axis to the prototype means, that this magnitude increases according to the length scale. According to the Froude scale (corresponding to what was found in case of the flow-pressure model), in which all pressures are translated on the length scale, the wave impact pressure should also increase with the length scale. See Chapter 77

Delft Hydraulics

5 and Appendix II for the significance of a model on the Froude scale. However, there is a value of the wave pressure that was measured in the model, i.e. as far as the scale model is concerned, the value on the vertical axis is known. The value of the parameter on the horizontal axis therefore may also be read. In that case the value of the parameter on the horizontal axis in prototype may therefore also be calculated, as a result of which the impact magnitude in prototype is found again. In other words, in this method it is not necessary to know the length of the water column, L, and that of the air column, d, in advance. When following one of the lines in Figure C4.4, then it may be seen that for wave pressures in the model that are not too great (in fact, in a 1:20 scale model there have never been pressures that were greater than 0.2 atmosphere), the line is (much) less steep than the slope 1/1, such as would have been found in case of an extrapolation on a Froude scale. Figure C4.2 graphically presents Equation C4.20 for the non-linear shock wave model as well. The results show the same development of the wave impact pressure as in Figure C4.4: for the order of magnitude of pressures as observed in a model, the translation to the prototype shows a much less steep ascending line than 1/1 again. After this type of calculation has been carried out, it is of course still possible to qualitatively assess the influence of changes in the values of and . It is very likely that the air bubbles in the prototype are relatively smaller than in a model, in which there is more heat exchange with the surrounding water, so that the adiabatic constant, , approximates the value 1 again. It is also probable that the air volume in a prototype is greater than in a scale model. Kolkman (1981/1992) has also carried out a similar assessment for other calculation models. The result of this is the table below. The scaling up to prototype measurements has been carried out on the basis of certain values of the wave pressure measured in a scale model. The pressures measured in the scale model and the pressures calculated in prototype are both expressed in metres of water column (mwc). measured in model flow pressure 1 mwc 2 mwc 3 mwc 40 mwc 80 mwc 200 mwc prototype values shock wave without air 6.3 mwc 12.6 mwc 32 mwc shock wave with air 8.4 mwc 20 mwc 80 mwc air compression model 9 mwc 28 mwc 178 mwc

In calculations like this, again and again it appears that the scaling up according to the flow-pressure model (ventilated shocks) results in the greatest prototype values. The flowpressure model corresponds to the Froude scale. It is therefore safe to take the Froude scale as a basis. The qualitative reviews of the other factors did not result in greater pressures. Conclusions: Unless there are clear reasons to deviate from this, the conclusion is that in case of translating from an impact pressure measured in a scale model to prototype, generally (for the sake of safety) the Froude scale rule should be applied. Another conclusion is that wave impact calculations cannot yet replace scale model investigations. Auxiliary calculations however are useful.

78

Delft Hydraulics

4.9 Influences on impact load due to a responding structure


Little is known about the effect of the structure responding to an impact load (feedback). As long as the structure is a single mass spring system, the effect of an impact of very short duration will be that the mass is caused to move, in which the amount of movement mV equals the impulse of the impact, I. The mass consists of the mass of the structure and the added water mass. The added water mass consists of the mass on the side from which the wave originates, and in addition, in case of a structure that is completely immersed in the water, the mass behind the structure. The greater the added water mass, the smaller the spring load becomes. This may be explained as follows. Theoretically speaking, the response of the structure to a load of short duration is an undamped vibration, in which the initial velocity equals Yn, with Y = vibration amplitude and n = natural (angular) frequency of the structure. If, for example, the total mass becomes two times greater, in case of the same load of short duration, then the initial velocity, and therefore also Yn, will be two times lower, while n becomes smaller with the square root from the mass (Equation C3.27). From this it is easy to deduce that, in case of a two times greater mass, the amplitude of the vibration decreases proportional to the square root of the total mass. In case of the calculation of the pressures, as a consequence of the response of the structure a decreased total pressure is generated due to wave impact included the added water mass at the front. During investigations at the discharge sluices in the Haringvliet, it was found (though not indicated in Delft Hydraulics Report M399) that the added water mass on the side of the wave of the structure needed to be calculated with the water level corresponding to the wave height after the impact. This is the water level that, after the impact, still generated a certain pressure increase (relative to the undisturbed situation) for a short while, see Part B, Figure B5.2. Based on physical model investigations (Witte, 1988) there are indications that another kind of influence due to the deflection of the structure is also possible. In measurements that were statically elaborated at one particular wave situation and varied stiffness of a wall, it was found that at an elastic wall the lower pressure peaks decreased by 25% (compared to observations at a stiff wall), while the high peak values do not decrease. At the stiff wall moreover, it was found that the lower pressure peaks coincide with a longer rise time. If the duration and the development of the impact are the same time and again, it should be possible to conclude from these two data that a decrease of the peak values occurs when the relation between impact duration and natural vibration time T of the structure is sufficiently great. That is the case therefore of an impact of a (relatively) long duration and/or a stiff structure. This result therefore is less probable, because it would mean that a stiffer model deviates more from a completely stiff structure than a weaker model. A more plausible explanation is that at higher pressure values, in which the water front approaches the wall faster, the air cannot escape so easily, so that in that case more (compressible) air is enclosed, as a result of which the stiffness of the structure itself has a lesser effect at the higher impact pressures. Instead of the instantaneous pressure, Miller (1980) has registered the pressure directly with time; this is a measure for the locally operating impulse load:

79

Delft Hydraulics

P = p (t )dt
0

(C4.37)

with = impact load, that remains unknown. This approach appears to be useful, if it is known that the natural vibration time of the entire structure or of the component concerned is long relative to the time duration of the impact. If this is not the case, then also the pressure history needs to be measured. As far as known, the effect of the deflecting structure on the impact load has not yet been assessed in calculations. The effect will strongly depend on the type of impact. The problem at a given real structure moreover, is that it has many degrees of freedom, and that, as an example, even the deflection stiffness of the plating of a gate is very important. The effect of compressible material (i.e. of a compressible structure), assuming that the shock wave model is valid, may be included through the speed of sound, see Paragraph 4.3.3. During the investigation of the gates of the discharge sluice at the Eiderdam (Delft Hydraulics Report M915), wave impacts that were registered on magnetic tape through linear filters, which were adjusted equivalent to the behaviour of a mass spring system, were registered once again. With this, next to a static analysis of the peak load and the impulse values of the impacts, it was possible to carry out a static analysis of the response of certain components of the gates as well.

80

Delft Hydraulics

SCALE MODELS 5.1 Introduction 5.1.1 General

A scale model is set up with the aim of obtaining some understanding of the processes that occur in prototype. These may be water models or structure models, whether or not combined with other elements such as salt/fresh water, sediment and soil properties. In case of investigations concerning vibrations and investigations concerning the response to wave impacts, usually a combination of a water model and a structure model is used. In order to obtain a univocal relation between properties of the model and the prototype, there are three preconditions that need to be met: Geometric uniformity; i.e. all length measures need to be scaled in the same way. The geometry includes the surroundings, the structure and the water level situation. Kinematic uniformity; i.e. all velocities in the model need to be reduced, relative to the corresponding velocities in the prototype. If the length scale and the velocity scale are given, the time scale is fixed. Dynamic uniformity; i.e. all forces in the model relative to the prototype need to be reduced according to the same force scale.

All preconditions need to be met simultaneously. Scale rules for hydro-elastic models (these are scale models in which the flow behaviour and the dynamic behaviour of the structure are modelled simultaneously) are obtained by verifying the dynamic equations of flowing water and of the structure concerning dynamic uniformity at a presupposed geometric and kinematic uniformity. It appears that it is not possible to meet the precondition that all terms need to be reduced simultaneously according to the same scale. Certain terms in the equations will have to be negligibly small to make reproduction in a scale model possible. What terms of the dynamic equations do or do not need to be scaled depends on the question what phenomena need to be reproduced in the model. The terms that are most important result in the so-called scale rules. The terms that are left out, in case they are not small enough to be ignored, give cause for scale effects. Once the scale rules have been established, the length scale (and therefore also the scale of the vibration amplitude and of other distortions), the velocity scale of the water and of the vibrations, the time scale and the scale of forces and pressures is also fixed. With this, all data resulting from measurements in the model may be translated to prototype values. The scale is defined as the relation between the corresponding values in prototype and model. A model at (length) scale 25 therefore is 25 times smaller than in reality in length, width and height. Scale effects cannot be quantified using the equations, as scale model investigations are only carried out for those cases in which the equations cannot bring about a solution. It is however possible to deduct certain characteristic values through the equations that are representative of the scale rules and of the scale effects. However great or small these values need to be in order not to give cause for important scale effects anymore, may only be determined empirically. 81

Delft Hydraulics

Next to scale effects, the scale model also has other defects and limitations. It only represents part of reality, as an example. And the geometry, as far as it concerns a future situation, is also not completely known; think as an example of the bottom condition of a river. These defects are referred to as model effects, in order to distinguish them from the scale effects mentioned above. This chapter is arranged as follows. Paragraph 5.2 offers a summary of scale rules for vibration and wave impact investigations (see especially the conclusions 1 through 6) and discusses possible scale effects. The deduction of the scale rules may be found in Appendix II. Paragraph 5.3 offers an overview of the types of scale models that are used. Paragraph 5.4 discusses the critical points that may give cause for defects. Paragraph 5.5 contains information about verification measurements. Paragraphs 5.6 and 5.7 discuss in greater detail the measuring system, the data processing and the elaboration of measuring results.

5.1.2 Investigation strategy for a real project


The investigation strategies vary strongly. They vary according to the weight of the project, the extent to which the structure type is known, experience and the risks involved in failure. Moreover, the available knowledge, means of investigation, time and money also play an important role. If there are only few possibilities for investigation, then recourse is taken to a known design and more conservative design assumptions will have to be made. Each investigation distinguishes a number of phases: a. b. c. Brainstorm phase with the aim of getting to know all problems and important design features. Systematic generation of variants that are tested under a great variation of circumstances. Verification of the chosen design.

In phase a, a scale model will be an important tool. If this is a straight gate or a slender body that is approached in transverse direction, then a two-dimensional model placed in a (narrow) flume will be sufficient. However, it is necessary that the model is capable of moving (vibrating) in multiple directions. If the shape of the gate or the rod varies across the length, then it needs to be considered investigating more cross-sections in the two-dimensional model (see Chapter 6, Example 6.3b). If couplings between vibration modes or vibration directions are to be expected, then this also needs to be the case in the model (Chapter 6, Example 6.6a). If a design is investigated, in which a great influence of the Reynolds number is to be expected, then initially a model needs to be used that has large dimensions (Chapter 6, Example 6.1a). If the flow pattern is strongly three-dimensional, then a more complete model will have to be used in the brainstorm phase as well (Chapter 6, Examples 6.5a and 6.7a). For the purpose of wave impact investigations in this phase it is sufficient to use a rigid model in which pressures or forces may be measured. It is however possible, also in case of a straight gate, to take account of waves that approach in oblique direction, and that may generate extra load at the corners near the pier and the abutment. In order to quickly measure a variation of circumstances, in case of vibration investigations, slow but continuously varying water levels are used, in which vibrations are 82

Delft Hydraulics

registered on a continuing basis. From this, the critical conditions that are followed in phase b are determined. In phase b, care must be taken to be very efficient. If phase a has resulted in sufficient understanding of the vibration mode, it may be attempted to develop a calculation method, by which the use of a scale model is reduced to a supplier of coefficients (Chapter 6, Example 6.5a). If however vibration investigations in a scale model are necessary, it may probably suffice to investigate only one vibration mode in a two-dimensional model. When optimizing the design of the structure, the investigation is limited to a number of critical conditions that were determined in phase a. It is relatively easy to change the design in such a twodimensional model. In phase c preferably a scale model is used that is as complete as possible. Sometimes this may be the same model that was used in phase a (Chapter 6, Examples 6.5a and 6.7a). This verification phase is important because the investigation of phase b is based on certain vibration mechanisms that were determined in advance under a limited number of conditions; in general therefore the scale model investigations are incomplete. If a Finite Element Method has already been applied to the structure, it makes the design of a fully elastic model easier. This also holds for the interpretation of measuring results. Reality may be different from the above. A three-dimensional scale model, as an example, is often not yet available in phase a. As more and more is known about the (limited) influence of elastic properties of the structure, there currently is a tendency to carry out wave impact investigations in rigid models only. In that case the distinction between phases a, b and c is less useful.

5.2 Scale rules and scale effects in case of vibration investigations and investigations of wave loads
The deduction of scale rules and possibly occurring scale effects are discussed in great detail in Appendix II. Here, only the conclusions are presented. The deduction of the scale rules on the basis of the hydrodynamic equations and the dynamic equations of the structure give cause for the necessity to equalize a number of dimensionless characteristic values in the model with those in the prototype. These characteristic values are: The Reynolds number:

Re =

VL

(C5.1)

In this, L is a representative length measure of the structure, V is the approach velocity and is the kinematic viscosity of the water ( = 1.2 10-6 m2/s for water of 14 C). The Reynolds number is the relation between the flow pressure (V2) and the viscous shear stress in the flowing water.
83

Delft Hydraulics

The Weber number: We =

LV 2

(C5.2)

In this, is the density of the water (1000 kg/m3) and is the surface tension (in case of air-water, this is = 0.077 N/m). The Weber number is the relation between the flow pressure and the pressure leap that is generated by the surface tension in case of a curved water surface. The Froude number:
Fr = V gL

(C5.3)

The squared Froude number represents the relation between the flow pressure and the pressure differences due to a difference in hydrostatic pressure in case of a leap or a change of the water level in case of geometric uniformity. The mass number: Ma = m L3 (C5.4)

In this, m is the mass of the structure or of a structure component. The mass number may be considered as the relation between the mass of the structure and the added water mass; the latter usually is proportional to L3. The Cauchy number (based on the spring stiffness): Ca1 k V 2 L (C5.5)

In this, k is the stiffness of springs to which the masses are coupled. The Cauchy number is the relation between the spring force in case of a compression or stretching on the length scale and the force that is exerted on the object by the flow pressure. The Cauchy number (based on the elasticity of the material): Ca2 = E V 2 (C5.6)

In this, E is the elasticity module of the material. This characteristic value represents the material tension relative to the flow pressure in case of a relative distortion of the material that remains the same (in model and prototype). The Strouhal number: S= fL V (C5.7)

84

Delft Hydraulics

In this, f is a frequency (either the excitation frequency, or the resonance frequency). As regards the excitation frequency, the requirement of a Strouhal number that remains the same follows directly from the requirement of kinematic uniformity (from this, it follows that the time scale equals the length scale, divided by the velocity scale). The requirement of reproduction of the Strouhal number based on the resonance frequency is already met, when the requirement of reproduction of the mass number and the Cauchy number is met. The Strouhal number therefore is a coefficient and therefore does not naturally belong in this series of characteristic values. The damping number: Da = c VL2 (C5.8)

In this, c is the damping coefficient of a mass-spring-damper system. The damping force in case of a vibration velocity that is scaled in the same way as the water velocities, here is related to the force due to the flow pressure. If this concerns a continuous-elastic structure (in which the entire structure may be distorted, contrary to a discretised system with a number of masses and springs), then this translates into the requirement that the relative damping, , in model and in prototype, needs to be the same. The requirement of simultaneous reproduction of all these characteristic values cannot be met. The simultaneous reproductions of the Reynolds, Weber and Froude numbers, conflict. Although for the reproduction of the Froude number the velocity in the model needs to be decreased by the square root of the length scale, reproduction of the Reynolds number in case of a water model however results in the fact that the flow needs to be increased in velocity on the length scale. It is difficult to find material, in which both the mass, the stiffness and the damping are right. This gives cause to making compromises in the choice of the scale; sometimes even the geometry of the model is adapted to obtain better elastic properties. If a choice is made for reproduction of the Froude number (the model is based on the Froude scale), due to the fact that the presence and the distortion of the free water level are strongly determining for the phenomena under investigation, it needs to be demonstrated that the erroneous Reynolds and Weber numbers do not result in major mistakes. This is the case when both characteristic values are big enough, both in the prototype and in the model. In a model situated in a tunnel, gravity plays no role; in that case the Froude number does not need to be reproduced and all kinds of coefficients may be determined as a function of the Reynolds number. The Weber number is especially coupled to phenomena with air bubbles and cavitation. These phenomena however are so complex, that a single scale model does not suffice. Below is a number of conclusions that have been taken from Appendix II. Conclusion 1: If the flow limitations in the model (a model set up as a tunnel model) are fixed, then the reproduction of the Reynolds number will probably determine the water velocity in the model. The viscosity, , is the same in the model and in prototype, as it is difficult to generate flow with a fluid other than water. The reproduction of the Reynolds number means that in the model the water velocity needs to be much greater than in the prototype. Often this is difficult to realize. If it is expected that the influence of the viscosity in the prototype is small (this is the case at hydraulic structures that are not 85

Delft Hydraulics

(or badly) streamlined), then it is possible to freely choose the flow velocity in a model, provided the Reynolds number does not go below a critical boundary. Concerning this critical boundary that depends on the shape of the structure, many references are available. Conclusion 2: If the flow is so strong that pressure differences cause a considerable distortion of the water level, in case of a model with a free water surface the Froude number needs to be reproduced. This also applies to waves. It is however necessary to meet the precondition that the Reynolds number in the model is big enough as well. If the Froude number is small, both in prototype and in model, then this does not need to be reproduced. If wave investigations are carried out in the model, then the Froude number always needs to be reproduced. A similar precondition as for the Reynolds number may also be formulated for the Weber number. The Weber number, LV2/ (with = surface tension), provided it is big enough, has no influence. The influence of the surface tension is that a pressure leap is generated between air and water that depends on the curvature of the water surface. Conclusion 3: In case of dynamic mass spring models, the scaling of the mass is independent of the velocity scale that is chosen. The mass number, m/L3, needs to have the same value in the model as in the prototype. Therefore it holds:
3 nm = n nL .

In this, n is the scale factor, while the index represents the corresponding magnitude. nL = length scale, n = scale of the density of the water. Generally speaking, in scale models it holds that n = 1, as also in case of the model, water is used to generate the flow. In case of the velocity scale that may be chosen freely (freely flow-surrounded objects, flow inside a tunnel etc.), the scale of the flow velocity, nv, and that of the spring stiffness, nk, are coupled through the Cauchy number, k/V2L. From that, it follows that:
2 nk = n nV nL

The damping number also needs to be reproduced, resulting in:


2 nc = n nV nL

In reality (for the sake of safety) most investigations are carried out in low-damped models. As the water damping may also be measured in the model, and if the real structure damping in prototype were known, it would also be possible to determine the total damping in prototype. In certain cases, when the vibration mechanism is known as well, it is possible to elaborate the influence of a deviating damping on the vibration amplitude that was measured. 86

Delft Hydraulics

In Part A, Chapter 3, some reviews focused on the added water mass, water stiffness and water damping. There, it was found that the added water mass is proportional to L3, the flow stiffness is proportional to V2L, and the flow damping is proportional to VL2. The model values resulting from this correspond to the scale factors that were formulated for the corresponding mechanical properties. The requirement that the scale of the resonance frequency corresponds to the reproduction of the Strouhal number (based on the resonance frequency), is automatically met. Conclusion 4: For a model of a continuous-elastic structure, in case of a velocity scale that may be chosen freely, the scale of the elasticity module equals the scale of the hydrodynamic pressures, and the density and the dimensionless damping of the structure material need to be the same as in the prototype. Therefore:
2 nE = n p = n nV

n ;struct . = n ; fluid
and

n = 1 To meet the last two requirements, in case of tunnel models in which there is a free choice regarding the flow velocity that may be established, the model is preferably built from prototype material, as a result of which the flow occurs with a velocity scale 1 (wind tunnel investigation of airplane vibrations). The advantage of using the prototype material for a model is that the internal damping is reproduced as well. Choosing another material is of course possible as well. As the requirement that the elasticity is represented on scale one is met, the deformation (L/L) that is measured in the model may immediately be used as prototype values. Conclusion 5: In case of reproduction of the dynamic behaviour of structures in models with water, in which the flow or the waves occur on the Froude scale, in the ideal case the elasticity module needs to be reproduced in the elastic model on the length scale. In a more general way this is formulated as: nE = n ng nL while it remains valid that: n ;struct . = n ; fluid and n = 1
87

Delft Hydraulics

Conclusion 6: In continuous-elastic models that have to meet the requirements of the Froude scale, it is pragmatic to build the model from synthetic material, in which it is accepted that in case of the model this results in a plate thickness that is too great. The scale for the thickness of the plating then becomes: nd = nL2 n ng / nE It may serve as a reminder that in case of a geometric reproduction of the plate thickness, the elasticity module of the model material on the length scale needs to be reduced. The factor by which the real elasticity of the model material deviates from this, is compensated by correcting the thickness with this factor. If (in case of wave impact investigations) also the plating stiffness needs to be right, the plating will get a reduced thickness: in that case additional reinforcements need to be mounted elsewhere in the model. Scale effects in case of vibration investigations If a choice is made for a model in which the Froude number is reproduced, then the scale effects corresponding to the chosen scale rules first of all relate to the incorrect reproduction of the Reynolds number and the Weber number. Generally speaking, if the Reynolds number and the Weber number are big enough in the model as well, then these will not give (much) cause for scale effects. How great these characteristic values need to be in that case, may for example be quantified using references concerning the drag terms (as a function of Re) and using the calculations relating to the influence of the curvature of the water level in case of waves and the behaviour of enclosed air bubbles with a small diameter (the diameter of air bubbles, as an example, is held to a maximum, as a result of which the bubbles fall apart and the bubble diameter in a scale model will not be scaled accurately). In case of dynamic models, it is difficult to meet the requirement of geometric reproduction in every detail; the measuring elements and the suspension should not influence the flow and in case of continuous-elastic structures a plating that is too thick is a factor that interferes with the model. Also, elements such as rubber sealing strips that compress inward and spring outward, are difficult to reproduce. The error in the plating thickness causes that the bending stiffness of the plating is also not represented accurately. For this reason, a solution of compromise needs to be found. A precondition that is difficult to meet is: n = 1 In a tunnel model in which the model consists of the same materials as the prototype, remains the same, as the damping is a material constant. In case of components that experience friction, such as hinges or rubber seals, the damping is not automatically to scale. In case of elastic models consisting of synthetic material, the material damping is too great. Also in the damping due to the fluid, scale effects may be expected. As far as the latter is concerned, the damping will be to scale in this case, if it is the consequence of the slip force of, for example, girders in turbulent flow, but not when the viscosity is normative (for example in case of a plate that vibrates in its plane: for then the slip stream depends on the boundary layer development). The damping is normative for the determination of the equilibrium amplitude in case of resonance. Generally speaking however, in case of the occurrence of resonance, the design 88

Delft Hydraulics

will have to be changed in such a way, that there is no resonance anymore. Even when the damping is not completely right, the model does give an indication concerning the occurrence or not of resonance. The damping in the model is preferably lower than in prototype. If no pure resonance occurs, then the damping is of much lesser importance. This may be seen when considering the response diagram of a single mass spring system (Part A, Figure A2.2). Also, the response to periodic excitation that is not pure (for example due to turbulence), is much less sensitive to the damping, as this excitation also consists of components with other frequencies than the resonance frequency. The damping is of secondary importance in case of the response to impulse phenomena, as the maximum amplitude occurs shortly after the impact, so that very little energy dissipation can occur. Scale effects in wave impact investigations As described in Part B, wave impacts are generated when the water is suddenly slowed down due to the presence of the structure. The amount of movement that is lost in the process, is coupled to and converted into the impulse value of the load (the magnitude of the force, integrated with time). In a scale model, in which the Froude number has been represented correctly (and this is necessary for investigations that involve waves), the impulse is reproduced correctly. The pressure distribution in space and time however, depends on the elastic properties of water, air and structure. Normative is the component that has the smallest stiffness. In scale models it is not possible to fully represent all elastic properties. For the effect of the erroneous stiffness of the air on the wave impact, see Paragraph 4.8, Delft Hydraulics Report M1335 and Ramkema (1978). Regarding the stiffness of the structure it is important to note that, also in case of continuous-elastic models, the local stiffness of the plating may only be reproduced when separate measures have been taken for that purpose. As appears from Conclusion 6, it normally suffices to have a model consisting of synthetic material, in which for plating and girders a material thickness is used that is too great. This means that the relative local bending stiffness is too great. Extra measures to make the plating stiffness right have been taken for the model of the sector gates of the outer gate of the discharge sluices in the Haringvliet (see Chapter 6, Example 6.8b). Generally, it is difficult to combine this with the reproduction of the stiffness of the entire structure and it requires a separate investigation (see Conclusion 6). In order to represent the stiffness of the air in a model on the Froude scale correctly, the compressibility (and therefore the initial air pressure) on the length scale should be reduced in the model. This is so difficult to realize that, so far, it has generally not been done. Moreover, the problem of small air bubbles remains: here the surface tension and the viscosity also play a role, and this for different reasons: the air pressure in the bubbles deviates from the ambient pressure (as a result of which the compressibility also changes), and the magnitude and the rise velocity of the bubbles are not reproduced (as a result of which the distribution and the magnitude of the bubbles deviates as well). In Delft Hydraulics Report M1057 tests have been described concerning wave impacts at a slope in case of a reduced air pressure. The results were not systematically different relative to those of the situation with a non-reduced pressure. It usually suffices therefore to carry out a global investigation of wave impacts, using a continuous-elastic model (response of the structure as a whole) and a detailed investigation in a stiff model, in which the wave pressures that are measured are translated to prototype values through special scale reviews. In Paragraph 4.8 it was indicated that, generally 89

Delft Hydraulics

speaking, for the sake of safety, the pressures due to the wave impact according to the Froude scale that were measured in a model need to be translated to the prototype. Other model effects Next to scale effects there is a large number of factors that determine whether the situation in a model is represented accurately. Among these are the accuracy of the dimensioning, the size of the area that has been modelled, the way in which water is introduced at the model boundary (flow distribution, flow direction, the extent of turbulence) and the way in which waves are generated etc. Furthermore, the disturbance of the flow pattern or vibrations by measuring instruments may also play a role. For specific cases concerning vibration investigations, error estimates have been made based on results from scale models (Delft Hydraulics Report Q1140). In case of vibration and wave impact investigations there is the additional question whether the wave shape is represented sufficiently accurate in every detail.

5.3 Classification of scale models for vibration and wave impact investigations
Below, a classification in categories is presented with respect to two aspects, i.e. the reproduction of the geometry and the reproduction of the dynamic properties.

5.3.1 Classification with respect to reproduction of geometry


Before the dynamic scale models may be introduced, it is first of all necessary to have an overview concerning the ways in which the geometry of a structure, together with its surroundings, may be reproduced. Tunnel model A tunnel model, in which a model of a structure is set up in a closed tunnel, may relate to a flow-surrounded object in wide water, to a structure that was also placed in a culvert or in a tunnel in prototype, and to a structure component that is submerged below the water so deeply that the water level does not need to be reproduced. Tunnel models are not often used in hydraulic structures due to the limited accessibility and the requirement of water proofness. Aspects of a tunnel model are: In a tunnel model it is possible to investigate both a total structure as well as a (twodimensional) section of a structure. The presence of tunnel walls means that the requirement of geometric reproduction are often not fully met. In a tunnel model gravity plays no role; the model may also be positioned on its side or upside down. As the contour of the fluid does not distort and as gravity does not influence the flow, the water velocity may be chosen freely. The flow pattern in that case does not change, except for possible scale effects due to the fact that the Reynolds number that was chosen was too low. All flow pressures and forces, both permanent and dynamic, increase or decrease proportional to the chosen scale of the flow pressure, V2. A.

90

Delft Hydraulics

At a later stage it will still be discussed that the reproduction according to scale of the dynamic properties of a structure in a tunnel model is very simple, due to the fact that the flow velocity in the model may be adjusted to the material properties of the structure, instead of the other way round. In case of a geometric model made of prototype material, if the flow generated is of equal velocity to the prototype value, all dynamic properties (therefore spring stiffness and mass) in the model are also correct. The result is that the time scale equals the length scale. Apart from that, the Reynolds number is not reproduced in this case. In a tunnel model the pressures may be reduced, as a result of which cavitation may also be generated in the model. Apart from this, this does not mean that cavitation and erosion due to cavitation are reproduced exactly according to scale. Moreover, cavitation falls outside the aim of this book. In a tunnel model it is only possible to increase the velocities to such an extent that in a scale model sometimes prototype values of the Reynolds number may be reached, or at least approximated. As long as only flow velocities, pressures and forces are measured in the model, and no vibrations, it is also possible to use a wind tunnel for the investigation of water problems, or to use a water model for the investigation of air flows. In order to reach high Reynolds numbers in a model therefore, special wind tunnel facilities with high flow velocity are available. See also Chapter 6, Example 6.1a. If vibrations due to water flow are investigated, it is not recommended to use wind tunnels. As the relation structure/fluid is not right, interaction phenomena are represented in a very incomplete way. A model with a free water surface and low flow velocity may also function as a kind of tunnel model. In that case, the water is almost level and may be considered as a flat tunnel wall. In a tunnel model like that, the major disadvantage that normally accompanies a tunnel model, i.e. the limited accessibility, is overcome. In this way it is possible to review gap flow underneath a gate in a model that is on its side; the free water surface is a wall of the section model in that case. Because of the possibilities mentioned above, a water tunnel model with a free water surface like this is often used in the preliminary design phase. In that case it is not a dynamic model. Because of the requirement of low velocities, the Reynolds numbers however are also low.

B.

Section models These are tunnel or flume models that are two-dimensional. In a flume with free water surface and a hydraulic head accross thenstructure, in case of reproduction of the hydraulic head on the length scale, the flow generated is automatically on the Froude scale. In this, the Froude number is reproduced (Equation C5.3). In case of a small hydraulic head and therefore low Froude numbers this is not necessary; in that case there is a certain parallel with the tunnel model, and, provided the hydraulic head in the model also remains small, the velocity scale may be chosen freely. A model on the Froude scale is of course suited for investigations with waves. Section models are used very often. As the number of geometric parameters relative to a three-dimensional reality is strongly reduced, the remaining parameters may be systematically varied. These models may serve the purpose of investigations (see Chapter 6, Example 6.3c), but they are also used often in the design phase of a project for the purpose of determining the dimensions and the design of the cross-section of the structure or of the details (Chapter 6, Examples 6.3a and 6.3b). 91

Delft Hydraulics

Section models are also used to determine the coefficients for a calculation model. Three-dimensional overview model or detail model This type of model has the advantage of a very complete reproduction of reality. The disadvantage, compared to a section model, might be the required space and other facilities, as well as the labour-intensive construction and investigation. Also the accessibility (consider, as an example, the placing of measuring instruments) may be a limiting factor. It is difficult to visualize the flow pattern in a three-dimensional model. It is also difficult to reproduce a culvert or a tunnel with gates that need to be tested dynamically. Consequently, often a combination is sought with a two-dimensional model or a three-dimensional model, together with a calculation model. However, even then there usually is a three-dimensional overview model as a final verification of the design. For dynamic investigations all three types of scale model mentioned above are used. C.

5.3.2 Classification with respect to reproduction of dynamic properties


The reproduction of dynamic properties may be realized in various ways. In order of (increasing) complexity, the following types of models may be used: a. Rigid model, in which forces and pressures are measured. It is essential for the measurement of forces, that the resonance frequency (which is determined by the stiffness of the dynamometer, the mass of the model and the added water mass), is high enough relative to the frequency range that is the focus of the investigation. The latter is different for each case. The pressure and force measurements in themselves do not provide complete information about the vibrations that may be expected. About self-excitation, nothing can be determined; this may only be determined once the structure is actually vibrating. The rigid model however, is often used because of its simplicity. It is best to use the model in combination with actual vibration models (as indicated in Chapter 6, Example 6.1a). For wave impact investigations a rigid model is adequate, in the sense that the pressures and forces that are measured hardly experience any influence of the elastic behaviour of the structure. There are however problems relating to the scale effects (compressibility of water and air, see also Paragraph 4.8) and the translation of the pressures that are measured to the spatial distribution of material tensions (see also Chapter 6, Examples 6.2a and 6.2b). Single mass spring model: this is a rigid model that is elastically suspended with one degree of freedom (vibration direction), in which it is possible to vary the direction of the vibration. The spring stiffness may also be varied. Within this type of model it is possible to distinguish: A model in which mass and spring stiffness are reproduced as well as possible. A model in which mass and spring stiffness are too great, but the resonance frequency is correct. With this, it is possible to investigate a number of important types of vibration after all (self-excitation, resonance vibrations), while it is also possible to determine the added water mass, water damping and water stiffness. Actively excited model: this is a stiff model, suspended from an excitator. It is only used for more general investigations (not connected to a specific project).

b.

c.

92

Delft Hydraulics

d.

e.

It is possible to use this model to determine coefficients, on the basis of which complete vibration calculations may be carried out. Multiple mass spring model: to be distinguished into: A model with one single mass, but with multiple springs, in which two or more vibration movements, each with one possibility of movement (for example translation and rotation in one plane), may occur simultaneously (investigation concerning interactions). It is essential that each degree of freedom is disconnected from the others, so that it is possible to establish the spring stiffness independently for each of the degrees of freedom, without crosstalk of the loads. The floating gate, as discussed in Chapter 6, Example 6.5a, also falls within this category. A model that is freely suspended by springs, in which the model is capable of vibrating freely in the three directions of movement in the horizontal plane. As the requirement of disconnection does not apply here, it is not possible to estimate in advance what vibration modes will be generated (see Chapter 6, Example 6.6a). A model with multiple masses and springs. Provided all relevant vibration frequencies and vibration modes have been determined, the first type has the advantage that the investigation relates to only one interaction phenomenon at the time. In case of the second and the third type, it is still possible that surprising discoveries may be made concerning unknown mechanisms of vibration excitation. Quantitative translation to the prototype however, is not always possible in those cases. Continuous-elastic model of a continuous-elastic prototype. This concerns models in which the spring stiffness and the mass, not only of the entire model, but also of all separate components (girders and plating) are included in the model as well as possible. As geometry and stiffness are completely coupled to each other in this type of model, the design of the structure that is to be created needs to be known almost completely before the model may be set up. A model like this therefore is mostly used as a final verification, after investigations of a section model have been carried out during the design phase. Examples of continuous-elastic models (also called elastically similar models) are presented in Chapter 6, Paragraphs 6.7 and 6.8.

The summary above cannot achieve that a choice is made concerning the question what needs to be done in each concrete case. About that, only a few general remarks may be made. For important structures sometimes multiple models are used. The rigid model, in which dynamic forces and pressures are measured, gives an indication whether self-exciting vibrations may occur. Rigid models may only serve to determine forces for as far as these are not influenced by the movement and the distortion of the structure. An analysis of the flow pattern however may sometimes indicate whether instability of the flow may be expected. In certain cases calculation methods for instable vibrations are available (see as an example Paragraph 4.4.5 in Part A, concerning galloping), for which the forces that are measured in a nonvibrating model in case of varied directions of approach flow are used as input. The continuous-elastic model has a stiffness that is strongly coupled to the design and therefore is not a very flexible means of investigation. Therefore it is mostly used as a final verification of the design, after other models have been used in the preliminary investigations for the determination of the main design. 93

Delft Hydraulics

If an analysis of the prototype indicates that one vibration mode is most sensitive to a possible vibration (for example a very rigid gate that is suspended from an elastic cable), then it will suffice to have a model with one degree of freedom. Some vibrations are only generated in a structure with multiple degrees of freedom; a scale model therefore needs to be set up in such a way, that the intended phenomena may occur. In case of wave impacts the weakest element is normative for the magnitude of the impact pressure. If this is the compressibility of an enclosed air volume or of the airwater mixture, then this compressibility is not to scale. During the investigation of the Eastern Scheldt storm surge barrier, a rigid model with a transparent wall was used, in order to be able to record on film the volume of an amount of enclosed air, with the aim of carrying out further calculations on that basis.

5.4 Possible critical aspects of dynamic models


Section models for dynamic investigations present the following questions: Is the two-dimensional flow pattern sufficiently representative? Has the dominant vibration direction been estimated accurately and is this perhaps in transverse direction to the axis of the flume? Is it sufficient to only consider the horizontal and/or vertical vibrations? As far as the flow pattern is concerned, is leakage on the sides acceptable and does this not generate any vibrations? Has the model been suspended well and, regardless of the intended vibration mode, are the stiffnesses great enough? If the suspension has components that are below the water surface, are these situated sufficiently outside the main flow? If the suspension is built inside a niche, in what way does this influence the added water mass and damping? Is the model guidance (which often consists of parallel springs and wires) such, that vertical vibrations do not generate parasitic horizontal vibrations and vice versa? Has the model not become too heavy, so that the spring stiffness needs to be exaggerated as well? For in that case only information is obtained about the extent of self-excitation and not about the response across the entire frequency range. Is the damping of the model low enough? What has been done in the model with the prototype seal that experiences outward spring and inward compression? Does the measuring instrument influence the flow or the vibration behaviour? Is the flume sufficiently vibration-free and stiff? Is it possible to make the spring so stiff that the model may also be used as a reliable dynamometer? Especially in case of models for investigations of dynamic wave loads, sufficient stiffness is very important.

A problem that may occur in case of a section model, is that the spring stiffness that results from the vibration frequency at which the investigation needs to be carried out, is so low that it is difficult to take up the stationary water pressure with this spring, or that it needs to be accepted that the model in this case is displaced from its original position. Continuous-elastic models present the following questions: 94

Delft Hydraulics

Considering the available choice of material (Perspex, Trovidur or another good thermo-setting pvc), is there ample choice of plate thickness? Does the temperature vary strongly? For this influences the elasticity in case of synthetic material. Has the scaling of the stiffness limited the geometry of the model? Is the material damping acceptable? For synthetic material this is five to ten times higher than for steel. In this, it needs to be remarked that it often appears that the prototype is often unexpectedly heavily damped as well, for example due to friction or distortion of rubber or prestressed bolt connections. Do the mass corrections that have been mounted influence the stiffness of the geometry? Is it possible for errors in the added water mass (due to a plating that is too thick) to be compensated by extra mass correction? Do the measuring instruments, and especially the resistance strain gauges, influence the stiffness and the geometry and do they not exert any damping force? Apart from this, is the model suited to carry out measurements using the resistance strain gauges (these have to be maintained at a constant temperature)? Sometimes an atomizer has been used with the aim of cooling.

A number of critical aspects of the continuous-elastic model may be overcome by thoroughly calibrating the model in dry condition concerning its bending and vibration properties. Especially when the prototype structure has already been calculated in detail as well, calibration may result in a lot of understanding (also in the prototype structure!). In case of a dynamic calculation even more verifications are possible. If a Finite Element Method has been used in the design, this may also be adapted in order to calculate the vibration properties of the model. Sometimes a comparison with the calibration of the model gives cause for adaptation of the model. Probably it is possible to decide on making more rough schematizations of the scale model, as a result of which it is possible to make better use of available thicknesses of plate material.

5.5 Verification of the model technology


Verification measurements as a consequence of a vibration investigation in a model have only occurred during the first project of Delft Hydraulics, in which elastic models were used. This investigation concerned the dynamic behaviour of the visor gates of the Hagestein Weir (Delft Hydraulics Reports M561 and M700). At a later stage also extensive vibration measurements have been carried out at the sector gates of the discharge sluice of the Haringvliet (Delft Hydraulics Reports M620 and M754), but the circumstances in model and prototype deviated too much to consider this as a verification of the model technology. Verification of wave impact investigations with prototype measurements is hardly possible. Measuring results are very spread out and, especially in case of prototypes, strongly depend on local wave characteristics. Moreover, wave impacts only occur during certain weather conditions, so that long-term measuring campaigns are required. In case of the discharge sluices of the Haringvliet, in prototype wave pressures and total forces operating on the sea gate were measured (Delft Hydraulics Report M754). Contrary to the model results found at the time (Delft Hydraulics Report M399), these were of very small magnitude. It had been forecasted that they would be smaller. In the model 95

Delft Hydraulics

exaggerated wind velocities were used at that time in order to generate irregular waves. As a result of this, the front end of the waves was also exceptionally steep and major impacts occurred. During the prototype measurements, land accretion had already occurred at the seaward side, so that also the wave height was smaller than forecasted. Below, the model investigations concerning the visor gates of the Hagestein Weir are briefly discussed. The visor gates of the Hagestein Weir serve the water-level control and discharge regulation of the Nederrijn River, and need to be for sustained periods of time, usually months on end, in a regulating position. The design of the weir was a new type, with light, weak, semicircular gates as its main characteristics. See for an overview of the weir and the crosssection of the gates, Part A, Paragraph 6.2, Example a. As there was no previous experience of such a structure, a detailed investigation into the dynamic behaviour of the gates was carried out. As this was the first investigation at Delft Hydraulics into the vibration behaviour of gates using vibration models, and the reproduction of continuous-elastic structures had only been developed very recently (by Delft Hydraulics, together with TNO-IBBC) as well, it has been attempted using models of different scales, to obtain some understanding of the scale effects and possible other model effects. After this, verification measurements have been carried out in the prototype. All investigations in scale models used the Froude scale. In summary, the scale model investigations included the following phases: Using the continuous-elastic model (1:20 length scale; 1:6 thickness scale), the a. vibration modes and resonance frequencies have been determined. See Figure C5.1 for the model and the excitator used. See also Example 6.7a. Because of the slender structure, in horizontal direction, the smallest resonance frequency was very low, i.e. dry 1.3 Hz and in water 0.6 Hz. Today, an investigation like that could be done using a Finite Element Method, provided it may be expanded with a module that also includes the added water mass. In a rigid model of a section (i.e. a section model) pressure fluctuations have b. been measured. The pressures indicated all frequencies between 0 and 10 Hz, so that always some vibrations should be expected. This was hardly influenced by varying the shape of the bottom edge. Bottom edge tests. In a model of a section (scale 1:6), suspended by springs, it was c. investigated whether in case of horizontal or vertical movement (at different resonance frequencies), resonance vibrations occurred. In case of the chosen profile of the seal at the bottom edge, these hardly occurred. The vibrations that were observed concerned the response to a kind of noise excitation, in a wide frequency band, most probably originating from the turbulence in the water downstream of the gates. In this model the local rubber stiffness had not been reproduced. Measurements under flow conditions in the elastically uniform model. The d. perforation of the bottom stiffening girder appeared to result in a considerable reduction of the vertical vibrations. The systematic measurements at various lifting heights and water levels did not result in important vibration amplitudes; for the horizontal movement the maximum amplitude was 7 mm, for the vertical movement 1.7 mm (prototype values).

96

Delft Hydraulics

e.

Verification tests in the section model as mentioned under c and in a new (section) model on a scale of 1:20. The already present 1:6 mode had the design that came closest to the actual structure; the 1:20 model had the cross-section that is similar to that of the continuous-elastic model. The flow-induced vibrations were compared, and likewise the damping by the water.

Figure C5.1: Elastically uniform model during excitation tests.

Figure C5.2: Real gate suspended by a strip. As the results under d. show a big margin of safety, and as the tests under e. gave no reason to expect big scale effects, the structure was judged to be sufficiently safe. In reality it appears that under special circumstances, resonance of parts of the plating occurs, that may be overcome by a change of the gate position. This could not be observed in the elastic model, as 97

Delft Hydraulics

the components of this model are much too stiff due to the plate thickness that is too great. At a later stage, measurements have been carried out at the actual gates (Delft Hydraulics Report Q322-I; this report focuses on the visor gates in the Rhine River at Driel), as the plate vibrations mentioned above increased more and more. The possible cause was a strong local distortion and damage of the rubber bottom edge. At a later stage reinforcements have been mounted at the bottom side of the gates.

Figure C5.3: Pulsator, used during prototype tests. Verification measurements: To verify the model technology, prototype measurements have been carried out1. the model has been set up again in order to reproduce all conditions of the measurements. The stiffness of the rubber side seal has been verified separately in case of dynamic load. This stiffness has been reproduced in the model. As the cables had not yet been mounted in case of the dry excitation tests, the gate was temporarily suspended from strips (Figure C5.2). This was useful for the verification, as the elasticity of the cable would otherwise dominate too much and the conclusions concerning this model technology would have been less reliable. Periodic excitation tests have been conducted with a hydraulic pulsator that made contact in the centre of the gate. This has been done for the horizontal-tangential and the horizontal-radial direction. These tests have been conducted both in prototype and in model. The model was in this case tested again under conditions that were as much as possible the same as those of the prototype measurements. Figures C5.4 and C5.5 show the results (at equivalent measuring points) during excitation in horizontal-tangential and horizontal-radial direction. The correspondence in frequency in both dry condition and in water is satisfactory. The amplitude amplification in the model was usually greater than in the prototype. The influence of the water on the resonance frequency was reproduced in the model reasonably well.
1

Carried out by the TNO insititute for mechanical engineering (IWECO).

98

Delft Hydraulics

Figure C5.4: Response of real gate and model gate, during horizontal-tangential excitation.

Figure C5.5: Response of real gate and model gate, during horizontal-radial excitation.

Figure C5.6 presents similar results during vertical excitation. The torsion stiffness of the model that is too great, the consequence of a strong schematization of the shape, results in 99

Delft Hydraulics

a resonance frequency that is too high. It was possible to forecast this reasonably well. The great damping of the water in the prototype is remarkable; during these measurements highfrequency vibrations occurred in the plating, which probably absorbed a lot of energy. Figure C5.7 presents the vibration registrations obtained during flow under normal operating conditions. Both in prototype and in model the gate was suspended from cables. The similarity between the results was satisfactory.

Figure C5.6: Response of real gate and model gate, during vertical excitation. The investigation has demonstrated that the model technology is reliable, provided that the difficulties caused by material damping in the model that is too great are taken into account in advance and provided that a lot of attention is given to details, such as elasticity and damping of rubber seals, friction and damping in hinges and elasticity of cables.

100

Delft Hydraulics

Figure C5.7: Recordings of vibration measurements of real gate and model gate, during flow conditions.

5.6 Measuring system and data processing 5.6.1 General


The purpose of measurements it to make phenomena visible, in this case the dynamic behaviour of the structure. In this, the storage and processing of data is very important. During the design process of the model it needs to be checked which magnitudes of the model need to be measured to obtain sufficient information for the interpretation and the translation of the processes that occur in the model. The model is set up in such a way, that these magnitudes may be measured, without the measuring instruments, cabling and support structures of measuring instruments essentially influencing the processes that need to be measured. Most instruments operate electronically and generate a signal that is proportional to the value of the magnitude that is measured. Using these instruments, a continuous signal (in time) is obtained. Some instruments (for example point gauges) are only read and therefore provide discontinuous information. On top of that, processes may be visually recorded on video or film (continuous) or using pictures (discontinuous). For the choice of measuring instruments, an estimate of the required measuring range and the frequency range of the recorders/transducers needs to be made in advance, so that it is possible to carry out measurements in the relevant frequency range with an accuracy that is as great as possible.

101

Delft Hydraulics

The basic ideas of the measuring system (electronic recorders) consist of the following components: measuring transducers with cabling and amplifiers; analogue verification, registration and processing instruments and/or; digital registration and processing instruments. The following paragraphs discuss the measuring system and the processing of measuring signals in greater detail.

5.6.2 Instrumentation
The most important and special aspects that emerge from measurements of vibration models result from the requirement that the instruments need to be small and that neither the measuring transducers and cabling, nor the flow, nor the mass, stiffness and damping of the model may be influenced. Consequently, the number of measuring points is usually kept to a minimum. It is important therefore, parallel to the scale model, to carry out a (dynamic) Finite Element Method calculation as well. With this, using the measuring results that were obtained at a limited number of locations, it is possible to get an idea of the total distortion. Vibration investigations often use the following kinds of measuring transducers: Resistance strain gauges: these are used for measuring relative distortions. Tensions and forces may be deduced from the values that were measured. These are suited for measurements in elastically uniform models. The requirement however is, that the combined stiffness of the resistance strain gauges and the corresponding water seal is not so great that, as a result, it influences the tension distribution in the model. Miniaturized resistance strain gauges are available. When using these on scale models consisting of synthetic material, attention needs to be paid to the fact that the heat that develops in the resistance strain gauges may be sufficiently removed (this may be a problem above the water level) and that the synthetic material and the glue are sufficiently waterproof in the long run. Dynamometers: if springs are also necessary in the model, the dynamometers are often integrated with these. It is important that the force is measured univocally in one direction without the influencing by forces that operate on the model in other directions. Pressure gauges: these have to be mounted precisely in the plane of the surface of the model, so that there are no air enclosures, and components of the flow pressure are not included in the measurements. Displacement transducers: contactless (for example acoustic) transducers are preferred, as these do not influence the movements of the model. Acceleration transducer: in order to avoid resonance in the transducers themselves (especially in case of impulse loads), the natural frequency of the transducers needs to be high. Damped transducers are available however; the damping operates in the natural frequency range of the transducer and at higher frequencies, as a result of which resonance may be avoided. Even then, the natural frequency still needs to be high to avoid the influence of phase shift in the relevant measuring range.

In order to determine the vibration mode minimally two acceleration or displacement transducers are required. One of the transducers is placed at a fixed position, preferably in the proximity of the antinode of the vibration that is expected, serving as a point of reference; the 102

Delft Hydraulics

other transducer is displaced along the structure. By comparing amplitudes and phase angles, a complete impression of the vibration mode may be obtained. Various suppliers have developed measuring and processing systems that make it possible to conduct such an analysis. The calculated mode of movement may then be presented on a screen, as an example. When working with acceleration transducers, the signal will usually be elaborated to displacements. In the end, it all focuses on the material tensions and those may be deduced from the displacements. This means that either the acceleration signal needs to be integrated two times, or the spectral values in the energy density spectrum need to be divided by the fourth power of the frequency. The latter, indirect method is much less accurate. When an analogue acceleration signal is recorded on magnetic tape and next is integrated, the noise of the tape is also integrated, causing a frequency-dependent error. In order to prevent this, the signal first needs to be integrated and only recorded afterwards. This problem does not exist in case of digital signals.

5.6.3 Monitoring and registration of measuring signals


Analogue signals Measuring magnitudes that are introduced as continuous-electronic (analogue) signals, need to be verified and registered. The registration occurs by way of so-called recorders (the signal is written down on paper) or by using tape recorders. If the signal is only written down on paper, follow-on operations are not possible anymore; in most cases this is not sufficient. The frequency range of the recorders must be sufficient for the frequencies that occur in the measuring signals. Writing down the signals during a test provides direct information about the functioning of the instruments, about the quality of the signal and about the development of the test, and is therefore strongly recommended. When a rape recorder is used for the registration of the signals, the outgoing signals of the tape recorder may be written down, so that it is also possible to verify the functioning of the tape recorder. A direct verification of the measuring signals may also be obtained by using oscilloscopes, voltmeters and the like. Analogue signals may be contaminated (for example by the hum of the electricity grid). It is also possible that spectral components of the signal are not relevant for the measurements. In those cases it is useful to change to the analogue filtering of the signals, as a result of which spectral components are removed. The filtering may cause a phase shift of (parts of) the signal; close attention to this is required. Filtering is sometimes necessary to avoid parasitic effects in the spectral analysis. Especially when the registration is of short duration, the analysis needs to be limited to a particular maximum frequency. See below. Analogue signals may be processed directly. For that purpose, various forms of processing equipment have been developed. The possibilities however are limited. Digital signals The digitalizing of the signals makes it possible for the data to be processed by computers: many advanced computer programs are available to process data files, to elaborate them into characteristic magnitudes, diagrams and the like and to combine them with other data. Data files (as an example in ASCII format) may be easily exchanged and may also be used as input for simulation programs. The (original) analogue signals are digitalized by taking samples of the signal at certain time intervals. The sample frequency (i.e. the reciprocal value of the sampling 103

Delft Hydraulics

interval) needs to be chosen sufficiently high, so that it may represent the original signal accurately. Theoretically, the sample frequency minimally needs to be equal to two times the highest relevant frequency (fmax) in the signal, but in practice this is three to four times fmax, in order not to experience any obstruction from the so-called aliasing (effect) that may occur around the Nyquist frequency (= bisected sample frequency). It may be better to filter the signals above fmax and to sample the signals with a sample frequency of 2fmax. This prevents aliasing and limits the amount of data. See also Bendat and Piersol (1971). Digitalized measuring files may be recorded as normal computer files.

5.7 Elaboration of measuring results 5.7.1 General


Time-dependent magnitudes may be deterministic or stochastic (random). For deterministic data mathematical relations may be indicated, making it possible that the process is described and forecasted exactly in time. An example of this is a periodic signal that consists of a finite number of sinuses. The magnitudes that are measured in a model however are nearly always of the stochastic type; this means that the measured magnitudes do not allow being forecasted in time. The measuring signals have a limited duration; for a correct interpretation of the phenomena that occur in a model, the measuring duration needs to be such, that the measurements are representative of the entire process. When the statistical properties of a measurement with a limited duration are representative of the entire process, this is referred to as a stationary ergodic process. The duration of the measurement therefore first of all depends on the type of signal. In case of a purely sinusoidal (deterministic) signal, measurements during one period are sufficient. In case of a stochastic signal, stationary ergodicity is required (apart from that, this is very difficult to demonstrate, but in model practice it is usually taken as a starting point); the duration of the measurement in that case especially correlates with the lowest frequency in the signal. The duration of the measurement that is chosen however, also depends on the kind of elaboration that will be carried out with the signal, as the statistical error for different elaborations correlates with the measuring duration each time in a different way. Generally speaking: the longer the duration of the measurement, the smaller the statistical error. Elaborations may be carried out in the time domain and in the frequency domain. Often statistical elaborations are carried out as well. Below, some elaborations that are carried out often will be discussed. For this, the findings of Bendat and Piersol (1971) have been used.

104

Delft Hydraulics

5.7.2 Statistical elaborations


Suppose, the time-dependent process x(t) is given, measured at the time interval (0,T).

Figure C5.8: Time-dependent process x(t). For this process, the following characteristics may be calculated. The average value x:
T

x = lim

1 x(t )dt T T 0

(C5.9)

The RMS value (root-mean-square value) x:


x = 1 2 x (t )dt T 0
T T

(C5.10)

The variant varx: 1 2 varx = { x(t ) x } dt T 0 (C5.11)

The standard deviation x:

x = varx

(C5.12)

The standard deviation x is mostly used as a measure for the magnitude of the fluctuation of the time-dependent process x(t). The magnitudes presented above are actually defined as the limit value for T , but in practice measuring files with a limited duration (assuming stationary ergodicity) suffice. It does follow that measuring files with a longer duration are usually more representative of a stationary process. Between the magnitudes mentioned the following relation exists:
2 = x2 + x2 x

(C5.13)

105

Delft Hydraulics

In case the average value x equals 0, then the standard deviation x equals the RMS value x. The average and the variation of the process x(t) may also be expressed using the probability density function p(x). The probability density function p(x) is the first coefficient of the probability function or probability distribution function F(x). The function F(x) indicates the probability that the stochastic variable x is greater or smaller than a certain value x*. By way of illustration, the probability function F(x)* = P{x x*} has been drawn in the figure below (P = probability), including the coefficient p(x) = dF(x)/dx, in which p(x) is the probability density. By definition, it holds: F() = P{x } = 1 and F(-) = P{x -} = 0. As the probability density is always positive, F(x) is a monotonously rising function. It may now be deduced that:

x =

x p( x)dx

(C5.14)

When the process x(t) has a normal (i.e. Gaussian) probability density function p(x), the average value and the standard deviation are sufficient to describe this stochastic process.

Figure C5.9: Probability function F(x) and probability density function p(x).

varx =

(x )
x

p ( x)dx

(C5.15)

Analogous to the variance of a single process x(t), for two time-dependent processes x(t) and y(t) the co-variance, covxy, may be determined. This is defined as:

cov xy =

( x )( y
x

) 2 p ( x, y )dxdy

(C5.16)

106

Delft Hydraulics

In this, p(x,y) is the combined probability density function of the processes x(t) and y(t). The extent of the correlation of the two processes is usually expressed by the correlation coefficient xy: cov xy xy = (C5.17)

x y

The value of this normalized magnitude is between 1 and +1. A value 0 indicates that the two processes are not correlated. All statistically independent processes are not correlated. The reverse is not always true, but for processes with a normal probability distribution it does hold that non-correlation implies statistical independence. A statistical elaboration that is often used is the determination of the exceedance distribution. The exceedance distribution, for example for a wave height measurement, indicates the probability that a certain wave height in the measuring file is exceeded. For this, the wave heights that are measured are arranged and classified (discretisation). The calculation of the exceedance percentage is based on the total number of waves in the measuring file. An example of an exceedance distribution is given in Figure B2.1. The exceedance distribution is an application of the probability distribution function F(x), as discussed above.

5.7.3 Elaborations in the time domain


In case of elaborations in the time domain, usually the aim is to characterize the process in time and to establish connections between signals. In a signal, as an example, it is possible to focus on periodicity or on impulse phenomena (impacts, wave impacts). Phase relations may be determined between various signals. It is also possible to focus on the extent of the resonant rise or damping out. Periodicity may be expressed by the auto-correlation function, Rxx(). For a signal x(t) at time interval T this is defined as follows (in fact, the limit value for T ): 1 Rxx ( ) = x(t ) x(t + )dt T 0 In this, is the time difference between the two processes x1 = x(t) and x2 = x(t + ).
T

(C5.18)

107

Delft Hydraulics

Figure C5.10: Auto-correlation function. The higher the value of Rxx(), the more plausible it is to have a periodicity with a period . For = 0 by definition Rxx = 2, and for non-periodic signals it holds that for the limit value at Rxx () = x2. Rxx may be both positive or negative. Assuming that the process is stationary, it holds that: (Rxx(0) | Rxx() | and Rxx() = Rxx(-) (mirror symmetry). Figure C5.11 shows the auto-correlation function for a noise-shaped signal (a signal with the energy distributed equally across the frequency band). A correlation function may also be determined by two processes x(t) and y(t); the result is referred to as a cross-correlation.

Figure C5.11: Auto-correlation function Rxx for a wide-band noise signal.

108

Delft Hydraulics

Figure C5.12: Cross-correlation measurement. The cross-correlation function Rxy() is defined as: Rxy ( ) = 1 x(t ) y (t + )dt T 0
T

(C5.19)

The cross-correlation function indicates the extent of correlation of two different processes. There is no mirror-symmetrical relation Rxy() = Rxy(-). However, it does hold that Rxy() = Ryx(-) and | Rxy() |2 Rxx(0).Ryy(0). When the processes x(t) and y(t) have an average value that equals 0 (x = 0, y = 0), instead of referring to the correlation function and the cross-correlation function, also the terms auto-covariance function or covariance function are used respectively. In a general sense, these may be defined as:
2 Cxx ( ) = Rxx ( ) x

(C5.20) (C5.21)

Cxy ( ) = Rxy ( ) x y

During measurements of impulse phenomena (for example of wave impacts), it is important to be able to detect the impulse phenomena in the measured signal. Impulse phenomena are characterized by a sudden steep gradient and often a large amplitude. It is

109

Delft Hydraulics

therefore possible to detect these by determining the first derivative of the signal. It is also possible that this is triggered by the exceedance of a given amplitude level. During vibration investigations, the extent of amplification or damping out is important. The extent of amplification and damping out in case of a linear system with viscous damping, may be expressed in the relative damping coefficient (see also Part A, Chapter 2). In order to determine the value of , it is sufficient to measure some of the amplitudes (an, an+1) of the resonant rise or damping out signal and to calculate as follows:

a 1 ln n 2 an +1

(C5.22)

Dynamic systems indicate a time-dependent output (or response) to a time-dependent input (or excitation). For a linear system with input x(t) and output y(t), y(t) may be expressed using the impulse-response function h() in x(t): y (t ) =

h(t ) x( )d

(C5.23)

The function h(t-) represents the response of the system at moment in time t to a unit impulse at moment in time . Due to the linearity of the system, the output y(t) may be found by superposition of the impulse response. The integration boundaries in that case run from = - to = -t. The integral is referred to as the integral of Duhamel and is a so-called convolution integral. Due to the property of commutation of convolution integrals, (C5.23) may also be presented as: y (t ) = h( ) x(t )d
0

(C5.24)

The function h() may now be considered as a weighing function that indicates the response of the system at moment in time t due to a unit impulse at moment in time (t-). When h() is known, y(t) may be calculated for any random input x(t). The lower boundary of the integral (C5.24) may, in more general terms, be put at -; as the response occurs as a consequence of an input, for < 0 this holds: h() = 0. A similar function exists in the frequency domain, i.e. the transfer function H(f), see below.

5.7.4 Elaborations in the frequency domain


Time-dependent processes may be characterized in the frequency domain by using their spectral components. A spectral representation that is often used, is the energy density spectrum. This spectrum is the Fourier transformation of the correlation function. The autospectrum Sxx is defined as:

110

Delft Hydraulics

S xx ( f ) =

xx

( )e i 2 f d = (C5.25)

xx

( ) {cos(2 f ) i sin(2 f )} d = (because of symmetry) 2 Rxx ( ) cos(2 f )d


The auto-spectrum is defined at the frequency interval - < f < and only includes a real component, as the imaginary component disappears, because: Rxx ( ) sin(2 f ) = Rxx ( ) sin(2 f ) The change of the integration boundary in (C5.25) is possible, because it holds that: Rxx ( ) cos(2 f ) = Rxx ( ) cos(2 f ) The spectrum is mirror-symmetrical, i.e. Sxx(-f) = Sxx(f), due to the fact that:

cos(2 f ) = cos(2 f )
Usually the one-sided spectrum Gxx is used, which also only has a real component and which is defined for 0 f < . This one-sided spectrum has the same energy volume as the double-sided spectrum Sxx:

Gxx = 2 S xx ( f ) = 2 Rxx ( ) e i 2 f d

(C5.26)

Figure C5.13 shows the difference between the one-sided spectrum and the doublesided spectrum.

Figure C5.13: One-sided and double-sided spectrum.

111

Delft Hydraulics

The auto-spectrum may also be calculated directly, without the use of the autocorrelation function, for the time-dependent process x(t) at the time interval T, using the Fourier transformation:

1 Gxx ( f ) = 2 T

2 2 T T x(t ) cos(2 ft )dt + x(t ) sin(2 ft )dt 0 0

(C5.27)

Here too, it holds that Gxx(f) is defined as the limit value for T . The spectrum shows how the energy is distributed across the frequency band; when there is periodicity, this may be observed in this spectrum because of the fact that the energy is concentrated in a narrow band. The average value of the time process x(t) may theoretically be read in the spectrum at 0 Hz, because:

x =

0+

xx

( f )df

(C5.28)

Practically speaking, this is of little significance, because the small integration area around 0 Hz approximates 0, so that the spectral value needs to move to infinite. In practice, calculated spectra for time-dependent processes with an average that is not equal to 0 do show a finite value at 0 Hz, but this relates to the magnitude of the frequency interval f: this is not 0. The average value may therefore not be deduced from the spectrum. For the RMS value it holds:
T

x = or:

G
0

xx

( f )df

(C5.29)

x = m0 with m0 = surface area of the spectrum.

(C5.30)

The energy density spectrum of the process x(t) is usually calculated for fluctuations around the time average (the average value is therefore filtered, i.e. artificially adjusted to 0). For a process with a time average that equals 0, it holds that x = x, so that the standard deviation x may then be calculated using (C5.29). An auto-spectrum in that case is also referred to as a variance spectrum. During wave investigations it is not customary to calculate the standard deviation of a wave spectrum. A characteristic magnitude of the wave height in a spectrum is the so-called significant wave height Hs. This approximately equals: H s 3.8 m0 and is defined as the average of the highest third part of the wave heights.
112

(C5.31)

Delft Hydraulics

For two time-dependent processes x(t) and y(t), using the cross-correlation function Rxy() by way of the Fourier transformation, the cross spectrum Sxy(f) may be calculated as follows:

S xy ( f ) =

xy

( )e i 2 f d

(C5.32)

This spectrum, defined for - < f < , different from the auto-spectrum Sxx(f), consists of a real part and an imaginary part. The imaginary part is not omitted this time, because: Rxy ( ) Rxy ( ) It holds that Sxy(-f) = Sxy(f)*, with Sxy(f)* as a complex addition of Sxy(f). This means that the real part of Sxy(f) is mirror-symmetrical and the imaginary part is anti-symmetrical. The one-sided variant Gxy(f), defined on the interval 0 f < , follows from: Gxy ( f ) = 2S xy ( f )
Or, written in a different way: (C5.33)

Gxy ( f ) = Cxy ( f ) iQxy ( f )

(C5.34)

with Cxy(f) as the real part (co-spectrum) and Qxy(f) as the imaginary part (quadratic spectrum) of Gxy(f). From both parts, the so-called magnitude spectrum |Gxy(f)| and the phase spectrum xy(f) may be deduced:
2 2 Gxy ( f ) = C xy ( f ) + Qxy ( f )

(C5.35) (C5.36)

xy ( f ) = arctg

Qxy ( f ) Cxy ( f )

The phase spectrum indicates for each frequency f the phase angle (in radials) between the signals x(t) and y(t), and in vibration investigations it is especially important to get to know the modes of movement. In the figure below, an example is given of the crossspectrum consisting of the magnitude spectrum |Gxy(f)| and the phase spectrum xy(f).

113

Delft Hydraulics

Figure C5.14: Cross-spectrum. For the processes x(t) and y(t) the extent of correlation may be deduced from the coherence function yxy2(f). This function is defined as:

(f)=
2 xy

Gxy ( f )

Gxx ( f ) G yy ( f )

(C5.37)

and is 1. If two processes are statistically independent, it holds that yxy2(f) = 0. When for a certain frequency f, xy2(f) = 1, then the processes x(t) and y(t) for that frequency are fully correlated. Now, imagine a given linear system with input x(t) and output y(t). In the frequency domain, analogous to the impulse-response function h() in the time domain, a frequency response function or transfer function H(f) may be defined, that represents the relation between input and output. Just like the cross-spectrum Gxy(f), H(f) is a complex number. When only the magnitude of the response is considered to be of interest, the following relation between input Gxx(f) and output Gyy(f) may suffice:

G yy ( f ) = H ( f ) Gxx ( f ) = H ( f ) H ( f )* Gxx ( f )

(C5.38)

The function | H(f) | is the frequency-dependent amplification factor. H(f)* is the complex addition of H(f). The transfer function H(f) however also contains phase information. The following relation holds:
114

Delft Hydraulics

Gxy ( f ) = H ( f ) Gxx ( f ) When H(f) and Gxy(f) are represented in polar notation:
H ( f ) = H ( f ) e i ( f )

(C5.39)

(C5.40) (C5.41)

Gxy ( f ) = Gxy ( f ) e Equation (C5.39) changes into:


Gxy ( f ) e
ixy ( f )

ixy ( f )

= H ( f ) e i ( f )Gxx ( f )

(C5.42)

In (C5.40) (f) is the frequency-dependent phase angle between x(t) and y(t). Equalizing the real and the imaginary parts in (C5.42) results in: Gxy ( f ) = H ( f ) Gxx ( f ) (C5.43) (C5.44)

xy = ( f )

It now appears that the phase angle (f) in the transfer function equals the phase angle xy(f) in the cross-spectrum. In order to determine a transfer function (amplification factor and phase angle), it suffices to calculate the cross-spectrum and the auto-spectrum of the input: the amplification factor follows from (C5.43) and the phase angle from (C5.44). The transfer function H(f) in the frequency domain and the impulse-response function h() in the time domain appear to be the Fourier transformation of each other: H ( f ) = h( ) e i 2 f d
0

(C5.45)

h( ) = H ( f ) ei 2 f df
0

(C5.46)

With the transfer function (or the impulse-response function) the properties of a linear system may be determined. For many systems the transfer functions are analytically determined and therefore it is possible to use them directly. A well-known example in mechanics is the single mass spring system with linear stiffness (k) and damping (c). A load is exerted on the system with a force F(t), the input, due to which it experiences a displacement y(t), the output. This system is schematically represented in Figure C5.15 and discussed below.

115

Delft Hydraulics

Figure C5.15: Single (degree of freedom) damped mass spring system. This system has already been discussed in Part A, Chapter 2. The transfer function of this system is: 1/ k (C5.47) H( f ) = 2 2 2 1 ( f / f n ) + ( 2 f / f n )

( f ) = arctg

2 f / f n 1 ( f / fn )2

(C5.48)

In this, fn is the natural frequency of the undamped system and is the relative damping. The maximum in the amplification factor | H(f) | is at a frequency: f r = f n 1 2 2 ( 2 0.5) (C5.49)

The frequency fr is the resonance frequency and appears to be a little lower than the natural frequency fn of the undamped system and also a little lower than the frequency fd at which a damped system experiences decaying vibration, after being hit: fd = fn 1 2 (C5.50)

116

Delft Hydraulics

The maximum in the amplification factor | H(f) | has a value of: H ( fr ) = 1/ k 2 1


2

( 2 0.5)

(C5.51)

In case of dynamically loaded systems it is customary to standardize the response (the displacement). The response in that case is divided by the displacement due to the load amplitude that is considered as a static force. This may be realized in the transfer function by dividing the amplification factor | H(f) | by | H(0) | = 1/k. An example of this normalized amplification factor may be seen in Figure A2.2 in Part A. The maximum in the normalized amplification factor now has a value: 1 ( 2 0.5) H ( fr ) = (C5.52) 2 2 1 When 2 << 1 it holds more or less that the maximum normalized amplification factor equals . When the input is not a force but, for example, a displacement, a movement velocity or an acceleration, then other transfer functions are found for the same mechanical system. The same applies to another output. For each input-output combination there is a unique transfer function. Based on a given energy density spectrum, a representative time function may be calculated. For this, the spectrum is divided into a number of frequency bands and the energy in these bands is elaborated into a sinus amplitude per frequency. Each sinus component is coupled to a stochastic phase angle. The sum of the sinus components results in the time function. This technique is used, among other things, for making an analogue control signal for a wave machine.

117

Delft Hydraulics

118

Delft Hydraulics

EXAMPLES OF SCALE MODELS FOR DYNAMIC INVESTIGATIONS

Overview This paragraph briefly describes a number of examples of scale models. The aim was to present a minimum number of examples, yet showing the major kinds of models, regarding the reproduction of the geometry and the dynamic properties for use of investigations into vibrations as well as wave impacts. The following models were selected. Rigid model with flowing water, for vibration studies (with respect to flow excitation) a. Grid gate; preliminary design storm surge barrier Eastern Scheldt. Three-beam model, five-beam model, both in a water tunnel, and three-beam model in wind tunnel. Rigid model for investigation of wave load a. Cooling-water intake near the coast; overview model. b. Lifting gate with upper girder storm surge barrier Eastern Scheldt; overview model of a part of the total barrier. Single (degree of freedom) mass spring system for vibration investigations; translatory a. Visor gates Hagestein Weir; section model (a model that is too heavy with adjustable vibration direction); resonance frequency continuously adjustable. b. Lifting gates storm surge barrier Eastern Scheldt; section model. c. Lifting gate (general investigation); section model clamped in wires. Single (degree of freedom) mass spring system for vibration investigations; rotating a. Sector gate in culvert (general investigation); overview model. System with multiple degrees of freedom in case of floating gate a. Floating sector gates storm surge barrier Rotterdam Waterway; overview model. Multiple degree of freedom mass spring system for response investigations with flow and waves a. Lifting gate in the Hartel barrier; section model. Continuous-elastic model for vibration investigations a. Visor gate Hagestein Weir; overview model. Continuous-elastic models for investigations of wave loads and vibrations a. Lifting gate in the storm surge barrier Eastern Scheldt; overview model of one gate with bridge girder. b. Sector gates discharge sluice in the Haringvliet; overview model of one set of gates with bridge girder.

119

Delft Hydraulics

6.1 Rigid model with flowing water, for vibration investigations


6.1a Grid gate (preliminary design storm surge barrier Eastern Scheldt) determination of dynamic load (through pressure measurements) section models with five and with three beams in water tunnel and in wind tunnel scale 1:1.9, 1:2.75, 1:3, 1:6.9.

See also: Part B, Paragraph 7.1. Aim of the investigation: Determination of resistance and dynamic pressures (the flow excitation insofar as this is independent of the movement of the structure itself), when the gate is completely or partially open. These investigations have been carried out in scale models that only represent a small part of the entire gate. Next to this investigation, carried out in closed tunnels, vibration investigations were also carried out in a continuous-elastic model and into wave impacts (Delft Hydraulics Report M1381). The latter two investigations were carried out in models of a complete gate with a free water surface.

Figure A: The grid gate in a caisson sluice (according to Delft Hydraulics Report M1338). Dimensions are prototype values. 120

Delft Hydraulics

Geometry and flow conditions: Figure A shows the design of a grid gate, to be placed in a sunken prefabricated caisson. The total grid has a span of 10 m and a height that varies from 10 to 20 m. Hydraulic head of up to 4 m (flowing) and 7 m (closed), in two directions. The gates are opened and closed by moving up one of the grids one beam height. The rounded-off shape of the beams relates to the discharge capacity; this needs to be great, as a little over 50% of the opening remains blocked by the beams. The rounded-off shape means, that during the investigation of the dynamic behaviour a big influence may be expected from the Reynolds number. In reality, fouling with mussels may strongly influence the flow, the resistance and the dynamic load.

Figure B: The two-beam model of the grid gate, tested in a water tunnel, and pressure transducers (installed later on). Investigation strategy: Due to the sensitivity for the Reynolds number, it was necessary to carry out investigations on a large scale. Because of that, only a few beams could be reproduced in the existing tunnels and flumes, and because of the transverse outflow not all gate positions could be investigated. This is why models were also set up on a smaller scale. The following models were set up: a. A 1:2.5 model consisting of a single beam with an open gate, standing upright in a flume, in which the velocities were so low that the water level could be considered as a tunnel wall. After determining the flow pattern and the resistance, the shape of the beam was changed once again. Two-beam model in a water tunnel (cross-section 0.9 x 0.5 m2, discharge 1200 l/s), scale 1:2.75, see Figure B. Gate open, respectively partially (up to 30%) closed. In this model the influence of fouling was also investigated (Figure C). Five-beam model (1:6.9) in the same tunnel as under b.

b. c.

121

Delft Hydraulics

d. e. f.

In the high-velocity wind tunnel of the National Aerospace Laboratory NLR The Netherlands a three-beam model was placed, scale 1:1.9. Tunnel cross-section 2 x 1.6 m2, velocities of up to 12 m/s. A continuous-elastic twelve-beam model of the entire gate, scale 1:23. A 1:3 continuous-elastic model has not been set up anymore, as finally this type of gate was not constructed.

Figure C: Tests with mussel fouling (two-beam model in water tunnel). Figure D: The five-beam model for investigations in the water tunnel. Result: The design would have been feasible, but mussel fouling has a big influence on forces and discharge. As it was expected that with this shape important scale effects would occur, both for the stationary forces as well as for the dynamic pressures no big differences were found at the various scales. The horizontal forces on the upstream beam were, in case of a fully open situation, a number of times greater than in case of a single cylinder in the flow. The critical Strouhal number (based on the approach flow velocity and the beam height) at which the excitation is maximum, is approximately 0.4. The small elastic model vibrated considerably and broke down; it was decided to connect the beams in the middle with a vertical rod.

122

Delft Hydraulics

Figure E: The three-beam model in the high-velocity wind tunnel. References: Delft Hydraulics Reports M1327-I, M1338, M1381-I, R1068 and De Jong and Nunen (1979).

6.2 Rigid model for investigation of wave load


6.2a Cooling-water intake at the coast rigid overview model, with waves determination of forces model placed in dynamometer scale 1:50

See also: Part B, Paragraph 7.5. Aim of the investigation: Measuring wave load on the total structure, as well as local pressures. The forces exerted on the tower have also been measured separately. Geometry and wave conditions: The casing with the three cooling-water intakes is situated under water. On top of it (in the original design) there is a tower, both for inspection and maintenance and for operating the gates. The tower which penetrates the water surface, experiences a quasi-static load due to the waves, but wave impacts also occur. Figure A shows the longitudinal section of the model of the intake with the tower. The width of the casing is 24 m, of the tower 16.4 m. Wave period 8-12 s., significant wave height 3-7.5 m.

123

Delft Hydraulics

Figure A: Longitudinal section of the model; waves are incoming from the right (all dimensions in m prototype). Dynamic properties of the model: As indicated in Figure B, the tower is mounted in a rigid five-component measuring system, that is connected with the substructure. The substructure in turn is mounted in a sixcomponent measuring system. The dynamometers are composed of components that were developed by Delft Hydraulics, in which parallel leaf springs were used that are fastened on two sides in end blocks. The end blocks may be displaced in one direction relative to each other (bending of the springs) and also rotate in one direction. The five-component dynamometer uses resistance strain gauges on the leaf springs; in case of the six-component dynamometer very stiff axial dynamometers have been used. In this model, the dynamometers have been enclosed in waterproof boxes, with rubber connecting strips to allow movement.

124

Delft Hydraulics

Figure B: Diagram of the six-component dynamometer of the total structure (H1 through H3 and V1 through V3 force signals in horizontal and vertical directions) and the five-component dynamometer of the tower.

125

Delft Hydraulics

Results of investigation: Maximum horizontal dynamic load on the entire structure 32,000 kN, and on the tower 29,000 kN. Vertically, this was 44,000 kN for the entire structure. After adapting the design (three towers instead of one), a reduction of 30 to 50% on the values mentioned above was obtained.

Figure C: The model during tests. References: Delft Hydraulics Report M1765. 6.2b Lifting gate with upper beam Storm surge barrier Eastern Scheldt rigid section model, with waves pressure measurements to determine wave load model, one and a half time the pier distance, in wave flume scale 1:50

126

Delft Hydraulics

See also: Part B, Paragraph 7.3 and Part C, Example 6.3b and 6.8a. Aim of the investigation: Orientation investigation concerning wave impacts on the gate and on the upper beam. This especially focused on the way wave impacts are caused (air enclosure effects and corresponding scaling problems), the order of magnitude of the impact loads, the effectiveness of the second screening plate at the seaward side, the reducing effect of the holes in the girders and finding the relation between wave impact pressures and wave parameters. After this, measurements in a continuous-elastic model have been used to assess the total impulse of the impacts.

Figure A: Design of measuring gate (with simplified design) and upper beam with pressure transducers. Geometry: The gate is positioned between a bottom sill beam and an upper beam. In the design that was investigated the pier distance centre-to-centre was 50 m, the flow-through opening was 45 m. See for the full cross-section Example 6.8a. Figure A shows the gate. The gate plating is situated on the Eastern Scheldt side. When it appeared that wave impacts occurred from the seaward side, in the model with the high plate girder gate in some tests on the seaward side an extra screening plate was mounted to reduce the wave impacts. A measuring gate was mounted in the model. The measuring gate covers approximately 65% of one gate; the glass wall of the flume intersects the gate (in order to be able to see the shape of the wave upon impact). Apart from that, a dummy gate has been mounted. Figure B shows the set-up of the complete model.

127

Delft Hydraulics

The number of holes in the horizontal girder is varied. During some tests the model was made transparent, in order to be able to record the air enclosure and the compression on film.

Figure B: Design of total model with measuring gate, dummy gate and intermediate pier. Dimensions in m prototype. Wave conditions and hydraulic head: Significant wave height 3.5 m with a period of 11 s, and a wave with height of 5.5 m and a period of 13 s. Water level seaward side +3 m, on the Eastern Scheldt side N.A.P. Measuring system: Pressure transducer in the plate girder and in the scale plating on the Eastern Scheldt side. Some pressure transducers in the upper beam as well. Dynamic properties of the model: Very stiff model that was fastened to the flume with a stiff dynamometer. 128

Delft Hydraulics

Results of investigation: Greatest pressure measured 500 kN/m2. This was on the second girder from the top. The average pressure on the gate plating, estimated on the basis of measurements: 130 kN/m2.

Figure C: Sequential photographs of one wave movement in the gate, hitting the upper plate girder. References: Delft Hydraulics Reports M1504, M1648 and Kooman et al (1980).

129

Delft Hydraulics

6.3 Single (degree of freedom) mass spring system for vibration investigations; translatory
6.3a Visor gates Hagestein Weir section model in flume with one degree of freedom with flow vibrations due to flow single mass spring system scale 1:6

See also: Part A, Paragraph 6.2a. Aim of the investigation: Next to the continuous-elastic model (scale 1:20) that was used in the brainstorm phase and as a final verification, this model was the most significant means of investigation to test the design of the cross-section (and especially the bottom edge) of the gate. As the vibrations were very small, in this model only the shape of the bottom edge of the gate has been varied. In those days (1956-1960), calculation methods for vibration calculations and Finite Element Method programs were not yet available. Especially horizontal and vertical vibrations were measured. During the elaboration, vibrations in the low-frequency range, in which the response is quasi-static, and in the resonance area have been differentiated. It was to be expected that, despite the fact that the weight of the gate in the model was too great, possible resonance vibrations would be represented to scale. For the purpose of scale investigations (there also was a 1:20 section model with the geometry equivalent to that of the continuous-elastic model), next to vibrations, the added water mass and the water damping have been measured. Some tests have been carried out with a (very flexible) side seal to check the influence of the lateral leakage on the vibrations. At a later stage this seal has been omitted in order not to run any risk that the damping would be too great.

130

Delft Hydraulics

Figure A: Section model, scale 1:6, visor gate of Hagestein Weir: view from aside. Geometry and flow conditions: Figures A and B present the views from aside and from above of the model. All dimensions are indicated in prototype values. The model is the reproduction of a part of the visor gates (see Part A, Chapter 6, Example 6.2a). The downstream water depth was approximately 4-6 m, the hydraulic head was maximum 3 m (prototype values).

Figure B: Same as Figure A: upper view

131

Delft Hydraulics

Dynamic properties: Mass: 325 kg (70,000 kg prototype) Vibration direction: from vertical to horizontal Frequency range dry: 0.8 to 2.8 Hz (prototype) The model was a factor 9 too heavy relative to the prototype. But as long as it concerned resonance vibrations and self-exciting vibrations, the results may be converted to a structure with a smaller weight. It is difficult to convert results from a section model to those of the entire gate; but if the section model is vibration-free, this may also be expected to be the case for the prototype gate. As may be seen in Figure A, one of the special features of the model is that both the vibration direction (with rods that have a point of rotation on both sides) and the spring stiffness are adjustable (using a rotating spring). Results: The shape of the bottom edge (sharp, semicircular or rectangular, always with a width of 10 cm) had little influence on the (weak) vibrations. In case of vertical vibrations the amplitudes varied up to 0.2 mm, horizontally up to 2 mm (prototype values). As no selfexciting vibrations occurred, it was possible to elaborate the vibrations in the resonance frequency, using the method as discussed in Part A, Paragraph 2.2.8. References: Delft Hydraulics Report M561-B and Kolkman (1963). 6.3b Lifting gates storm surge barrier Eastern Scheldt section model with one degree of freedom, with flow gate vibrations due to flow flow and wave forces on gates and beams detail model in flume single mass spring system scale 1:40

See Examples 6.2b and 6.8a. Aim of the investigation: Orientation investigation into vibrations of the lifting gates, also when the gate is closed, whether there are potential vibrations, due to flow in the leakage gaps. As the flow pattern is almost two-dimensional, in this phase a section model was sufficient. Measurements of flow load and wave load preceded this vibration investigation. Only in the final phase a continuous-elastic model was used, in which an available Finite Element Method calculation model of the gates was effectively employed.

132

Delft Hydraulics

Figure A: Vertical measurements with rigid suspension (gate seen from above and from the seaward side). Figure B: Vertical measurements with spring suspension. Geometry: See Example 6.8a for an overview of the structure with piers, upper beam and sill beam. The gate is positioned on the seaward side of these beams and may be moved up along the upper beam. The plating consists of arch barrels to reduce the torsion stiffness of the gate (important in case of unequal placement of the piers). One opening of the storm surge barrier has been set up as a model. A high gate and a low gate have been investigated. In the model the upper beam and the sill beam have been mounted in a stiff way. The gate was suspended by springs through trusses that protruded sideways from the section model. The springs were fastened to a dynamometer. The wall of the flume was provided with spacious recesses that could be separated for the larger part from the flume itself, so that the dynamometers remained outside of the flow. Tests have been carried out both with and without a plate on the seaward side of the gate (with a profile like a sheet pile wall) to prevent wave impacts against the horizontal plate girders. Moreover, at a later stage, contrary to the model, open frame girders have been used, as the wave forces in the proximity of a plate girder might become too great.

133

Delft Hydraulics

Blz. C: Measuring system for vibration in horizontal and vertical directions.

134

Delft Hydraulics

Flow conditions: Hydraulic head of 0 to 8.5 m. Downstream water level (Eastern Scheldt) varying between N.A.P. 1.5 m and +3m. Dynamic properties suspension system gate: Dynamometers in sequence with adjustable leaf springs. See Figure C. In connection with measurements at various gate positions, the entire force measurement system and the support system were fastened to a hoisting frame that could be secured in any position that was required. Sill beam and upper beam were provided with pressure transducers. The natural frequencies (dry conditions) were adjusted to 3.5 Hz for the vertical vibration measurements and to 6.5 Hz for the horizontal measurements. During the horizontal vibrations it was possible to generate rotational vibrations simultaneously.

Figure D: Suspension of model gate with extra plating at seaward side, seen from seaward side. Measuring system for horizontal (and rotational) movements.

135

Delft Hydraulics

Figure E: Detail of adjustable leaf springs and dynamometers at upper and bottom sides of the gate. Results: Self-exciting vibrations occurred during flow from the seaward side moving inland. In case of a closed gate, the flow through the bottom leakage gap was the cause. The amplitude of the vibration at the bottom of the gate was 1.6 times greater than at the top. In case of a half-open gate, the upper leakage gap gave cause for vibrations. Another vibration occurred when the upper girder was just below the water level. Reinforcement of the suspension was sufficient to eliminate the vibrations. References: Delft Hydraulics Reports M1424, M1494. 6.3c Lifting gate in open discharge sluice (general investigation) model clamped in wires (one degree of freedom), with flow vertical vibrations due to flow section model in flume single mass spring system scale: not applicable

See also: Part A, Chapter 6, Example 6.5f. Aim of the investigation: Systematic determination of the added water mass (as a means of comparison with the results of the mathematical approach), the water damping and the extent of self-excitation, all as functions of the water levels, the vibration frequency and the shape of the bottom edge.

136

Delft Hydraulics

Figure A: The model clamped in pre-tensioned wires, upper view and cross-section. Geometry and flow conditions: The gate model, situated in a 50 cm wide and 70 cm deep flume, was clamped in wires, in such a way that only vertical vibrations were possible. This method of suspension was chosen in order to obtain a minimum damping and a minimum crosstalk of vertical and horizontal load. Often, the problem is that the great horizontal load results in crosstalk in vertical direction, where the static and dynamic forces that need to be measured are small. The gate consisted of a vertical plate with at the bottom side a horizontally protruding lip in downstream direction. At a later stage other shapes of the edge were also investigated (Figure B). The stiff wires that guided the model, were mounted on one side only. In order to prevent that these would remain taut, regardless of the water load, opposite of the conducting wires, tension wires were mounted that were tightened with a spring. This results in a certain string effect with an additional stiffness; this stiffness however is not great. Conducting wires were also mounted in lateral direction. Dynamic properties: The mass of the gate was varied from 4 to 6 kg and the spring stiffness was strongly varied, so that the frequencies in the water varied from 2.5 to 9.5 Hz. Results: The extent of self-excitation appears to be a function of the Strouhal number (based on gap width and flow velocity in the gap) and this increases with the extent of suction due to flow along the bottom edge. Certain shapes of the bottom edge do not generate self-exciting vibrations, but a positive flow damping.

137

Delft Hydraulics

Figure B: Various shapes of bottom edges, in order of increasing degree of safety (against vibrations). References: Delft Hydraulics Reports M1322, M1490, Kolkman and Vrijer (1977) and Vrijer (1980).

6.4 Single (degree of freedom) mass spring system for vibration investigations; rotating
6.4a Sector gate in culvert model with one degree of freedom (rotation only), with flow rotational vibrations due to flow total model in tunnel single mass spring system scale: not applicable

Aim of the investigation: The aim of the investigation was to determine the added water mass, the water damping and the flow-induced vibrations in case of varied mass, mechanical damping and stiffness.

138

Delft Hydraulics

Figure A: Test installation. Geometry and flow conditions: Figure A shows the geometry of the sector gate (reversed Tainter gate) that was placed in a quadrangular culvert of 0.25 m, 2 m long at the upper side and 3 m long at the downstream side. The actual gate and the plating are situated at the downstream side, in order to prevent air suction through the gate shaft. The hydraulic head of the model could be increased up to 2 m. Mechanical aspects: The model was set up in such a way that the springs and the dampers could be mounted outside of the culvert. The axis was lead outward, in which a leakage gap of 1 mm all round was accepted. Figure A shows the test installation. Figure B shows the gate in detail and Figure C shows the special structure with the point of rotation. The latter will be explained in more detail. As the mechanical damping had to be completely independent of the hydraulic head, a knife edge structure was created, in which the knife was first swung out of the way. After that, the natural weight and the water load with spring forces were compensated in such a way that the axis was centralized. Next, the knife edge was put in place and a predetermined prestressing force was created using one of the springs. At a later stage it appeared moreover, that the force on the knife edge hardly influences the damping near the knife edge. For project-related investigations these days a graphite-lubricated nylon bearing is used that may also be used in water. With this, the need to lead the axis outside of the tunnel falls away.

139

Delft Hydraulics

Figure B: Geometry of reversed Tainter gate. Dynamic aspects: Mass of the gate varied from 4 to 10 kg. Variable spring, frequency range 5 to 15 Hz. Tests without and with linear mechanical damper. Results: There is no periodic excitation as long as the bottom edge is sharp and there is no protruding upper lip for sealing. The latter point was investigated (see Part A, Figure A4.14). In this model, the theory of response to noise excitation was tested (Part A, Paragraph 2.2.8).

Figure C: Knife edge bearing at the point of rotation. References: Delft Hydraulics Report S50-3 and Kolkman (1976).

140

Delft Hydraulics

6.5 System with multiple degrees of freedom in case of floating gate


6.5a Floating sector gates Storm surge barrier Rotterdam Waterway overview model with rigid gates with multiple degrees of freedom vibrations due to flow and behaviour in waves overview model model with (per gate) three degrees of freedom, rotating around ball hinge scale 1:60

See also: Part A, Paragraphs 4.5.1 and 6.2f. Aim of the investigation: Investigation of the dynamic behaviour of the floating gates in flow and wave conditions. Measurements of dynamic forces at hinge points, pull-push rods, bearing fenders and threshold blocks.

Figure A: Overview photograph of the gate model.

141

Delft Hydraulics

Geometry of the overview model: Figure B shows the situation with the sector gates in the docks and the situation with closed gates. The opening that needs to be closed by both gates is 360 m. The gate height is 21.5 m. Figure C shows the cross-section of the gate. The gates are sailed in afloat and sunk down. They are fastened radially through a point of rotation at the abutment and tangentially through a pull-push rod connected with a traction engine and a rack railway. The overview model includes both sector gates, the sill, the abutments and on both sides approximately 1.5 km of the Rotterdam Waterway. In the model, the sector gates have been executed as stiff structures, consisting of aluminium frames, filled with light-weight foam for modelling of the geometry (this, in order to be able to adjust the cross-section of the gate quickly). In the model only stationary situations have been investigated, in all possible combinations of gate and water levels. Instead of filling them with water, the gates in the model have been ballasted with lead blocks.

Figure B: Plan view of gates in open and closed conditions. Flow conditions: Flow from the seaward side: hydraulic head of the barrier dependent of closing procedure and remaining opening underneath the gates, ascending up to 4 m and in closed situation up to 6 m. Flow from the riverside: hydraulic head at the beginning of opening approximately 2 m, decreasing when the gates are forced up more. Waves conditions: At the seaward side the significant wave height is 0.75 to 2.5 m; the corresponding wave periods are 5 and 10 s respectively. Measuring system: Contactless, acoustic displacement transducers, above the gate ends for measuring of vertical gate movements. Dynamometer in the hinge points and the pull-push rods. Furthermore, various pressure transducers in gates and sill; dynamometer in fender underneath the gates and in parts of the sill.

142

Delft Hydraulics

Dynamic properties: Per gate three degrees of freedom in movement (three rotational freedoms in the hinge point). This results in the following main gate movements: almost vertical movement (heaving) and rotating around a radial axis (pitch movement). As during this process the gate becomes more or less submerged, the water functions as a hydrodynamic spring. Natural periods between 10 and 15 seconds. The third movement, in tangential direction, is determined for the greater part by the spring in the pull-push rod. In this, the natural periods are considerably lower and are approximately 3 s. In the model, the springs in the fenders and the pull-push rods have been executed with non-linear spring characteristics. Investigation strategy: The model was already available when the design was fixed in outline, but the shape of the cross-section still had to be optimized. During the initial phase of the investigation of this three-dimensional model, it was established that strong gate movements were generated relating to the transverse waves between the abutments (see Part A, Paragraph 4.5.1). These movements depended on the way in which the transverse wave takes the gate along and how this again influences the discharge underneath the gate. When this was also theoretically substantiated (Part C, Paragraph 3.3.7), calculation models were developed (Part C, Paragraph 3.5.3), in which coefficients concerning the extent of forcing up and discharge that could be established in a two-dimensional model. In this way it was possible to optimize the crosssection in the two-dimensional model. The final verification of the design occurred in the three-dimensional model again. Results: The investigation was very thorough and has played an important role in establishing the dimensions of the fastening structure and of the sill. The design of the cross-section of the gate has also been changed considerably, as the surprising result was found that both gates together, with the water as the connecting link, constitute a coupled system with coupled natural movements; see Paragraph 3.5.3, Example b.

143

Delft Hydraulics

Figure C: Cross-section of the gate. References: Delft Hydraulics Reports Q958-I, Q969, Q1140, Q1190, Q1271, Q1278, Q1293, Bakker et al (1991), Jongeling and Kolkman (1992).

144

Delft Hydraulics

6.6 Multiple (degree of freedom) mass spring system for response investigations in case of flow and waves
6.6a Lifting gate Hartel barrier section model with multiple degrees of freedom, with flow and waves vibrations due to flow and wave load section model multiple degrees of freedom in one plane scale 1:25

Aim of the investigation: Checking the vibration sensitivity of this design in case of flow-through underneath, in case of overflowing water and in case of a combination of both. As the vibration modes and the frequency range were not completely fixed yet, a choice was made for a model with three degrees of freedom in the vertical plane. This model was used in combination with a Finite Element Method calculation model for the gate, including the influence of the added water mass.

Figure A: Plan view of the large lifting gate. Dimensions in mm prototype. Geometry and flow conditions: As seen from above, the storm surge barrier consists of two lense-shaped gates that are suspended from lifting towers. The gates have a span of 98 and 49 m respectively. On the seaward side there is a retaining plating, on the riverside the lattice structure is open. The height of the gates is 9.30 m. When raised, the bottom edge is at N.A.P. +14 m and in closed, i.e. retaining, position this is at N.A.P. 6.30 m. The bottom is at N.A.P. 6.50 m, which results in a remaining gap of 0.20 m. Taking into account a sea water level of over N.A.P. +6 m, it is possible that there is 3 m overflow. As far as the cross-section is concerned, the investigation focused on a section model of the central part of the main gate; the vibration behaviour could be made to correspond to both the behaviour of the central section and that of the section at a quarter of the length. The investigation occurred in a flume with a width of 50 cm and a height of 70 cm.

145

Delft Hydraulics

Figure B: Cross-section of the gate, before and after the investigation. Dimensions in mm. (Remark: in this figure the seaward side is on the right of the gate). Dynamic properties: Because of the three degrees of freedom (horizontal, vertical and rotational) and the possible mutual couplings between these due to the flowing water, the dynamic behaviour is very complex. Moreover, the lowest frequency relates to the horizontal asymmetrical deflection. This has been calculated with the calculation model DIANA. For different situations during the closing process, combinations of the lowest natural frequencies that the gate may have, have been established in the model. For natural movements with the greatest deflection and/or rotation in the cross-section halfway the span, natural frequencies in wet condition have been established in the area 0.5-2.5 Hz, depending

146

Delft Hydraulics

on the position of the gate and the main direction of movement (horizontal, vertical, rotational). Results: As a result of the model investigation both the shape of the bottom edge and of the upper edge have been changed; for this, see also the cross-sections in Figure A. The bottom edge has been sharpened considerably, while the upper edge has been bevelled.

Figure C: Suspension of section model in the measuring frame (view from aside). The seaward side is on the left of the gate. References: Delft Hydraulics Report Q1500.

147

Delft Hydraulics

6.7 Continuous-elastic model for vibration investigations


6.7a Visor gate Hagestein Weir continuous-elastic mode, with flow vibrations due to flow overview model scale 1:20

See also: Part A, Paragraph 6.2a and Part C, Paragraph 5.5. Aim of the investigation: The aim of the investigation was to determine whether there was any danger of vibrations at this, at that time new, gate design, especially in the low-frequency range. It is characteristic of the visor gate that the lowest natural frequencies, both in dry and in wet conditions, are extremely low (from 0.5 Hz onward). Optimizing the cross-section occurred in a section model (see Example 6.3a). Calculation methods regarding vibrations were not available at that time (1956-1960).

Figure A: Elastically uniform model of the visor gate. Geometry and flow conditions: Part A, Chapter 6 shows the design of the weir (Example 6.2a). The weir consists of two openings of 48 m. Hydraulic head maximum 3.5 m. Bottom level at N.A.P. 4.5 m. Downstream water level varying from N.A.P. 0.5 to +2 m at a maximum upper water level of +3 m. The function of the weir is to regulate the upper water level. The arch is put under strain of tension by the hydraulic head. Consequently, a light gate could be selected, as a result of which the lowest natural frequency is low. It was this project, in which a continuouselastic model was used for the first time (see Paragraph 5.5). The cross-sections of model and prototype are different as regards plate thickness and reproduction of the frame girders. By emphasis, the scaling is based on the elastic properties of the gate as a whole (the lower harmonic vibrations therefore). In theory, the scale of a model consisting of synthetic material 148

Delft Hydraulics

with respect to the material thickness is roughly 1:6 (using the relation mentioned in Paragraph 5.2, Conclusion 6). Furthermore, the adaptations to make the model as simple as possible with available model materials, also play a role. The web plate of bottom and upper main girders are 15 mm thick in the prototype, and 4 mm in the 1:20 model. The thickness of the flanges in the prototype was variable across the length of the gate, and in the model a constant thickness of 4 mm was selected. The vertical plating in reality is 8 mm thick, in the model 1.5 mm. The frame girders in the prototype are each built from two angle sections with a width of 130 mm and a material thickness of 12 mm. In the model, a single strip of 4 mm thick and 20 mm wide was used for this purpose.

Figure B: Cross-section of the gate in prototype. Dynamic properties: The aim was to reproduce the dynamic properties of the prototype for the gate as a whole and for the suspension. Components such as frame girders and plating however have too much bending stiffness in the model. It is remarkable that all plate thicknesses in the model are relatively great. Figure B shows the cross-section in the prototype and Figure C shows the cross-section of the model made of synthetic material. In the latter figure it may be seen that the additional masses, necessary for this model technology, have been mounted for the greater part on the trusses and less on the beams, in order not to influence the stiffness of these too much. Figure D shows the lattice structure at the downstream side of the gate in prototype and in model.

149

Delft Hydraulics

Figure C: Cross-section of the Trovidur model (PVC). Results investigation: As a result of the investigation, in which none or hardly any vibrations were found, it was decided to continue the construction, the major change being the fact that an extra culvert would be mounted in the abutment, in order to improve the radial outflow and uniformity of the load on the gate. As the plate and the girders in the model were too stiff locally, the plate vibrations that were observed at a later stage, relating to the selected semicircular shape of the rubber bottom seal, were unfortunately not forecasted (see Part A, Chapter 6, Example 6.2a).

150

Delft Hydraulics

Figure D: View (from aside) of trussed girder in prototype and in model. References: Kolkman (1967). Also included in Kolkman (1976).

151

Delft Hydraulics

6.8 Continuous-elastic model for investigations of wave loads and vibrations


6.8a Lifting gate storm surge barrier Eastern Scheldt continuous-elastic model with stiffness outer plating also to scale with flow and waves vibrations due to flow without and with waves overview model continuous-elastic model with special outer plating scale 1:40

See also: Example 6.2b and 6.3b and further Part B, Paragraph 7.3. Aim of the investigation: The verification investigation in the continuous-elastic model focused on vibrations due to flow and wave forces. Apart from the continuous-elastic model that is described here, two section models were used as well: a stiff model for measuring wave pressures (see Example 6.2b) and a vibration model with one degree of freedom (Example 6.3b). A number of other models was available for the investigation of the design in relation to the dynamic behaviour. In the end, the investigations and reviews concerning scale rules and scale effects have lead to a change in the design, in which the horizontal plate girders have been replaced by a lattice structure made of round tubes.

Figure A: Schematic representation of piers, sill beam, upper beam and gate structures.

152

Delft Hydraulics

Figure B: The model of the low lifting gate with girders, seen from Eastern Scheldt (inner area). Geometry and flow conditions: Figure A gives an impression of the dimensions of the piers. The clear width of the opening between the piers is approximately 40 m. The opening is limited at the bottom by a sill beam and at the top by an upper beam. The gate height that closes the opening varies from approximately 5 to 12 m. The hydraulic head may theoretically increase up to 5 m, but the wave heights may rise very high (up to Hs = 4 m) as well. Dynamic properties: The continuous-elastic model gate at the moment of mounting was fastened to dynamometers through horizontal wires, so that next to tensions and displacements, the total load could be measured as well. The dynamometers were fastened again to a frame that moved up and down with the lifting of the gate. The (concrete) upper beam was also reproduced, using Trovidur in the continuouselastic model. As far as the gate is concerned, the plating thickness also needed to be such, that the bending stiffness was to scale, as this stiffness was a parameter for the torsion stiffness of the entire gate (normally, the plating of a continuous-elastic model is too stiff). During the design of the continuous-elastic model and the translation of the measuring results into tensions at critical points, a Finite Element Method program was used. This program was available at Rijkswaterstaat for the prototype gate and was completed with a Finite Element Method model of the gate with the elasticity, mass and dimensions of the added material of the Trovidur model. As a result of the comparison of the model gate stifnesses with the FEM model, adaptations were necessary of the scale model,but also of the FEM model; even basical improvements of the FEM model of the prototype gate could be made.

153

Delft Hydraulics

Figure C: Testing of the model gate with waves. Results: In case of the high gate (the deep situation), during flow from the seaward side and during an open position, a flow instability was found that gave cause for a low-frequency oscillation with a period of 15 s. When there are waves (and these always occur during this extreme situation), then this oscillation disappears. In case of the lower gate, if it is halfway open, a vibration of 3 Hz was observed (amplitude 11 mm), which also disappears again in case of waves. At this gate, in case of flow from the Eastern Scheldt in closed position, a vibration was measured with a frequency of 2 Hz and an amplitude of 7 mm. The model has also been used to measure wave impact responses and, using the measurements in stiff models as well, to draw conclusions in relation to local wave pressures. The responses that were measured, due to the relatively low natural frequencies, may be converted to prototype on the Froude scale.

154

Delft Hydraulics

Figure D: Overview of measuring frame. References: Delft Hydraulics Report M1561, De Jong et al (1980).

155

Delft Hydraulics

6.8b

Sector gates discharge sluice Haringvliet continuous-elastic model with stiffness outer plating also to scale; with flow and waves overview model continuous-elastic model specially adapted outer plating scale 1:20

See also: Part A, Paragraph 6.2C and Part B, Paragraph 7.4. Aim of the investigation: This model served as a verification investigation into the vibration behaviour and the response to wave impacts. In previous phases of the investigation, various models for vibration investigations, for investigations into vertical flow forces and wave forces (and local wave pressures) had been used. Parallel to this investigation in the continuous-elastic model, calculations were carried out with a Finite Element Method program (still being developed at that time), with an electric analogon and with analytical methods (based on the theory of eigenvalues).

Figure A: The elastically uniform model of gates and bridge girder. Geometry and flow conditions: See for the geometry and the conditions of the gates with the suspension Part A, Paragraph 6.2, Example c and Paragraph 6.5, Example c. A concrete triangular bridge girder spans 56.5 m. Points of rotation are fastened to this for the sector gates; these also serve to support the gate. The structure of trusses had a centre-to-centre distance of 14.8 m. Both the gates at the seaward side and at the landside are double-plated to facilitate movement under frozen conditions. The plating on the seaward side of each of the gates reduces the vertical wave load. Figures B and C represent the gate design in prototype and in model, from which the geometric differences may be deduced; these only relate to the plating thickness. The external geometry of the gates was maintained.

156

Delft Hydraulics

Figure B: Cross-section of sea gate in prototype. Dynamic properties: Both the gates and the concrete triangular (nabla) girder are scaled in mass and spring stiffness. In case of a continuous-elastic model in conditions of flow and waves on the Froude scale, the plate thickness is too great and consequently also the local deflection stiffness. As it was not known whether the deflection of the plating plays a role in case of wave impacts, the plate thickness has been adapted as well as possible to the actual stiffness. A calculation has been carried out of the stiffness in prototype including the small stiffening girders, and with a small adjustment, this total plate stiffness could be made consistent with the normal scale rules for the model plating. The local stiffness between the stiffening girders however remained much too great in the model. Results: Preceding the investigation in the continuous-elastic model, an investigation into wave impacts was carried out into a rigid model, resulting in a number of changes in the design (see Part B, Figure B7.4). The continuous-elastic model was only ready at a stage, in which the design phase of the structure had more or less ended, and therefore it especially served as a final verification. Next to local wave pressures, spud forces and total forces on the nabla bridge girder were determined. There were no important new elements that emerged. The wave pressures in this 1:20 model correspond to the measurements in the rigid model on a scale of 1:40. The wave impacts operate very locally; the spud forces were a factor 7 to 8 smaller than in case the maximum measured wave pressure had been used as a criterion. Measurements of spud forces roughly correspond to pressure measurements that were 157

Delft Hydraulics

integrated over the gate surface. Because of the wave load, the gate is only caused to experience a symmetrical vibration, which relates to the elasticity of the winch system; the distortion of the gate itself is subordinate to this. The (small) tensions measured in the bridge girder roughly correspond to the values that were calculated. Remark: Using the method of wave generation (regularly moving wave board in combination with a strongly exaggerated wind velocity) of that time, the wave steepness was too great, as a result of which the impacts that were measured were much too unfavourable. This was known during the investigations, but there was no better method available.

Figure C: Cross-section of sea gate in the model; dimensions in model values. References: Delft Hydraulics Report M399.

158

Delft Hydraulics

INVESTIGATIONS OF PROTOTYPE STRUCTURES 7.1 General


Generally speaking, prototype measurements may have very different aims, such as:

Detection of the cause of observed problems and the verification of possible remedies. Verification of existing calculation and model techniques. Additions to studies and investigations in the design phase. Checking of the specifications when completing the structure. Permanent monitoring of the behaviour of the structure and checking the circumstances.

Regarding the dynamic behaviour of structures it holds that the measurements relate to both the behaviour of the structure and the registration of the flow and wave conditions in which this behaviour occurs. Moreover, in case of a few structures measurements were not only carried out in case of flow-induced vibrations and wave impacts, but vibration modes and resonance frequencies of both the dry and the wet structure (still water) have also been determined separately. On the one hand this was done in order to verify the calculation and model technology on these points, on the other hand this was done in order to be able to better interpret the flow-induced vibrations and wave impacts. This especially focused on the visor gates of the Hagestein Weir and the segment gates of the discharge sluices in the Haringvliet, both mentioned in Paragraph 5.5. Measurements of prototype structures essentially are no different from measurements in scale models. What is different however, is the following: All operations are much more complex; it is more difficult to keep the measuring conditions under control. It is not always possible to find a fixed point outside of the structure, from where the structure may be artificially loaded or where a measuring instrument may be fastened. This gives cause to a different choice of excitation mechanisms and measuring instruments. The measurements are often spread out across a relatively long period of time. As it is desirable that the properties of the structure are known, in both dry condition and in still water, the measurements are often coupled to construction phases. Moreover, once the structure is operational, the normative conditions hardly ever occur. Sometimes special provisions need to be made regarding the structure, in order to mount the transducers and lead the cables through. Sometimes a fixed point outside of the structure needs to be created artificially, in order to be able to generate dynamic loads on the structure artificially. For this, an auxiliary structure may be necessary.

In case of prototype measurements it is very important that there are timely consultations, in which the period of the preparations and the measurements themselves, the task at hand and the activities of the designer/supervisor of the structure and those who carry out the measurements, are laid down. 159

Delft Hydraulics

As far as the organizing of the measurements is concerned, it is important that: the equipment is durable and reliable and is fastened in a very sound way; in case of measurements of sustained periods of time, the possible failure of equipment is taken into account, resulting in more transducers; when mounting pressure cells and resistance strain gauges, this is already provided for during the construction of the gates, so that at a later stage there is no need to damage the coating (maintenance or paintwork at a later stage appear to cause damage); cabling is strongly anchored and does not obstruct the activities of those who take care of the construction or finishing; there is a measuring cabin that is well-equipped with good climate control and a reliable energy supply with a constant voltage, and the measuring facility is habitable and contains sleeping places; during the measurements, good communication is possible (telephone) and in case of critical or exceptional circumstances, the measuring team may be called in at a moments notice; a scenario has been made which describes all operations; alternative scenarios need to be available in order to be able to communicate and respond fast during unforeseen circumstances.

It will be obvious from the above, that the preparations need to start a long time in advance. To ensure that the measurements do not take too much time, the elaboration is usually done at a later stage. These days computers are used for the processing of data. More and more operations may be done online however, as there are more facilities to exchange data quickly. Below, a number of specific points will be discussed regarding vibration measurements and wave impact measurements.

7.2 Vibration measurements


A. The determination of vibration modes and resonance frequencies For the determination of the vibration mode, it is necessary to mount three acceleration or displacement transducers on a number of cross-sections, in order to distinguish the vertical movement, the horizontal movement and the rotational movement of the cross-section. The best results may be obtained using artificial excitation. The excitation may be periodic or by way of a short impulse. In case of a short impulse, information is obtained across the entire frequency range simultaneously. In case of periodic excitation however, a better selectivity is obtained. The vibration mode during flow may also be determined through cross-correlation of the measuring signals.

160

Delft Hydraulics

Figure C7.1: Scaffold to excite the sector gates (in-situ) of the sluice Haringvliet from a fixed point. For the design of the sluice, see Section 6, Example 6.8b and for the excitator, see Figure C5.1. A few extensive vibration measurements were carried out with the aim of verifying the model technology. For this, it is required that the measurements are accurate and reliable. The contributions of experts from various disciplines proved to be very important, both in the field of equipment and for the processing of measurements in relation to the dynamic properties of the structure. Next to commercially available equipment, in addition special equipment was developed: specific prototype strain meters; resistance strain gauges were glued to a piece of (thin) steel plating under controlled conditions and sealed. This prefabricated resistance strain gauge was then glued onto the structure, and used for measurements; excitators; two types were used, of which one consisted of a recently (in those days) developed ship excitator with two excentrically rotating weights (this does not require 161

Delft Hydraulics

a fixed point outside of the structure that is tested), and the other from a specially developed hydraulic jack that was directed through a servo system on the basis of power or displacement. The ship excitator produces a force that is proportional to the excentricity and the square of the speed of rotation, and therefore is less useful in case of lower frequencies; absolute displacement transducers and acceleration transducers. These also do not require a fixed point outside of the structure under investigation. Both transducers consist of a mass that is coupled through a dynamometer to the structure at which the measurements need to take place. The elastic properties of the dynamometer together with the mass result in a certain natural frequency of the instrument. The force that is measured is determining for the acceleration of the structure, as long as the movement is low-frequency relative to the resonance frequency of the instrument. The instrument in that case is referred to as an acceleration transducer. If however the movement frequency is high relative to this resonance frequency, then the mass remains in its place and an absolute displacement transducer is created. A seismograph also operates according to this last principle. As the acceleration is proportional to the amplitude of vibration and the square of the frequency, it is less sensitive in the low-frequency range. It is not possible to take any measurements in the proximity of the resonance frequency of the transducer.

Before the vibration modes and resonance frequencies are measured, first a theoretical estimate needs to be made at the location where the antinode of the vibration that bends outward is situated. The structure is periodically loaded, preferably at this location. At that location, a vibration transducer is placed as well, serving as a reference for other measuring points. The lowest frequency and the mechanical damping of the structure may also be determined by rhythmically loading the structure by human force and allowing a freely decaying vibration. Measurements of flow-induced vibrations For incidental measurements there is portable equipment with velocity or acceleration transducers that may be fastened to the structure using clamps or magnets. The data are recorded on magnetic tape or by using a computer. In order to make it possible to carry out a reliable elaboration of results, it is important that the signal that is recorded is as large as possible and the noise is as limited as possible. This means that it is preferable that accelerations are first integrated once or twice, in order to prevent that the signal drowns in high-frequency noise. Filtration of the signal may also be useful; the signal that needs to be recorded in the remaining frequency range may then be amplified more. Remark concerning measurements for the verification of model investigations. If the aim of the prototype investigation is to verify a model investigation that was carried out earlier, experience shows that the model after the end of the prototype investigation usually needs to be set up all over again. For in case of prototype measurements, it is beforehand unknown what kind of hydraulic head will occur. A model investigation that was carried out earlier in the design phase of the structure usually concerns an extensive investigation at more or less extreme conditions and little investigation in case of more normal circumstances. The stiffness, the damping and the friction of components may strongly deviate from what was assumed during a model investigation that was carried out earlier. Especially the B.

162

Delft Hydraulics

stiffness and the damping of a hydraulic winch and of rubber components are difficult to estimate and to mount in the model.

7.3 Wave impact investigations


Wave impact investigations in prototype are of great general importance considering the uncertainty with respect to scale rules in scale model investigations and the calculation of wave impact pressures. Nonetheless, these kind of measurements have not been carried out very often. To start with, there are (at least in the Netherlands) very few locations in which wave impacts occur often and where the waves are more or less regular. And even at favourable locations (for measuring wave impacts), wave impacts only occur in case of certain combinations of wind direction, wind intensity and water levels. In case of incidental measurements, in which the measuring team only becomes active when there is a storm, it appears that forecasting of the occurrence of critical conditions is not very reliable. In case of measurements of long duration there are also problems: in order to obtain some understanding, many measurements need to be available at a relatively large number of places, as each impact is different, both regarding the interval of time and the spatial distribution. The wave impact measurements that have been carried out, generally have yielded less results than what was anticipated. The fact that the interesting signal is of short duration and, as mentioned, does not occur very often, implies that: automatizing of the measurements; preselection of the signal before it is recorded on tape; this too will have to be automatized. The occurrence of wave impacts is coupled to water level, wave direction and wave steepness. These data need to be registered as well, in case it is requested to determine the correlation of these physical magnitudes. Very near to the structure, the wave measurements always concern a combination of incoming and reflected waves. Statistic correlation results in a lot of spreading, as all these magnitudes play a role and the observation series is always limited in any event. In case of gates, the gate position and the discharge of the gate also need to be registered, as both influence the waves and whether or not impacts occur. Due to the great pressures and water velocities during wave load, extra attention needs to be paid to the fastening of measuring instruments and cables need to remain outside the reach of the waves. Similar to vibrations, for the verification of the model technology after the prototype measurements, the model needs to be available again to simulate the circumstances that occurred during the prototype measurements.

163

Delft Hydraulics

7.4 Elaboration of measuring results


Elaboration of measuring results is no different for measurements in the prototype from what was already established during scale model investigations. As all measuring signals are registered in digital form these days, the registration and elaboration has become much simpler and more reliable. Specific points during the elaboration of prototype measurements are: As the costs of measurements in prototype are high, it often suffices to have a limited number of instruments and a limited measuring program. If measurements cannot occur simultaneously with the operation of the structure, then the available time for the installation of instruments and for measurements is also limited. For the processing of the results there is an extra difficulty when the vibrations are located at places that are difficult to access (vibrations of a culvert gate or of a bottom edge of a gate that is positioned at a very deep location). Therefore it is also uncertain whether the measuring instrument is located at the best spot. For vertical gate vibrations sometimes measurements of the dynamic oil pressure in the lifting cylinder is sufficient. In that case it needs to be checked whether damping factors that have a great influence play a role, both with respect to the vibration itself and with respect to a possible interference of the measuring signal. Because of the factors mentioned above, it may be desirable to carry out theoretical analysis, in order to be able to obtain some understanding of the total behaviour of the structure, even from the limited measuring results. The elasticity of the lifting device is often not known very well and variable with the lifting height. This requires a separate analysis, and sometimes exactly for that reason extra measurements appear to be necessary. Measurements in which a gate is suspended from a fixed point, instead of a winch, offer the possibility of better determining the dynamic properties (mass, stiffness and damping) of the gate itself. The outward spring of rubber sealing strips causes insecurity concerning the established lifting height; this is important, as the measurements at small lifting heights are often the most interesting. Automatizing the wave impact measurements is difficult but necessary. The measuring signals always contain a lot of irrelevant information, with few impacts. Processing during the measurements is in fact necessary to free memory space or storage capacity. Processing wave impact measurements means that, apart from the analysis of registrations, statistic analyses of maximum peak size, impulse volume and rise time may be carried out. Before the elaboration of the measurements may begin, it must be clearly defined what the aim is. If it concerns the load of a specific structure, then at least the dynamic properties and the intensity need to be roughly known in advance.

7.5 Experiences with respect to vibration and wave impact measurements


Below a selection is presented of experiences of Delft Hydraulics regarding vibration measurements and wave impact measurements in prototype. Almost all of these projects were designed and supervised by Rijkswaterstaat. It is briefly indicated what kind of structure it is, what were the aims of the measurements, what measuring equipment was used and what 164

Delft Hydraulics

results were obtained by these measurements. It is also indicated in which reports the measurements have been reported. The overview concerns the following structures: a. Discharge sluice Haringvliet b. Storm surge barrier Eastern Scheldt c. Discharge sluice Brouwersdam d. Intake sluice Volkerak e. Storm surge barrier Krimpen f. Discharge sluice in the Afsluitdijk, Den Oever g. Northern abutment IJmuiden h. Weirs in the Lek River at Hagestein and Driel i. Shipping lock at Lith j. Weir at Sambeek k. Bottom-hinged gate at Neerbeek l. Tide lock at Ravenswaay m. Stop logs. a. Discharge sluice Haringvliet For data concerning the gates see Part A, Chapter 6, Examples 6.2c and 6.5a and Part C, Chapter 6, Example 6.8b. At the sector gates of the discharge sluice vibration measurements and wave impact measurements have been carried out. These measurements consisted of: determining the vibration modes and natural frequencies of the bridge girder (nabla girder) by excitation due to a ship excitator (excentrically rotating weights); measurements with acceleration transducers; determining the vibration modes and natural frequencies of the gates with a hydraulic pulsator; measurements with acceleration transducers and resistance strain gauges on the lifting structure; vibration measurements on the sea gate and the river gate in flow conditions; measurements with acceleration transducers and resistance strain gauges on the lifting structure; wave impact measurements and wave measurements; online processing with a computer; measurements with pressure transducers and resistance strain gauges on the lifting structure.

The measurements were reported in Delft Hydraulics Report M754. Wave impacts have only occurred sporadically over a measuring period spanning many years. The maximum pressure that was measured was approximately 50 kN/m2, half of which was related to the pressure that slowly varied with time. Vibrations occurred in various situations in case of a slightly greater lifting height of the gates (see Part A, Chapter 6, Example 6.2c). As the model investigation had concentrated on the shape of the rubber bottom edges, that were especially critical at very small lifting heights, these vibrations had not been forecasted. At a later stage, a considerable vibration was observed at one of the sea gates, due to leakage flow underneath the rubber bottom edge of the gate that was leaking locally due to a distortion of the bottom edge (probably due to the fact that during closure something had been 165

Delft Hydraulics

lying on the sill). Observations and measurements of this phenomenon have been reported in Delft Hydraulics Report R1510. b. Storm surge barrier Eastern Scheldt In Chapter 6, Example 6.8a, information may be found about the shape and location of the gate, the sill and the upper beam. Vibration measurements and wave impact measurements have been carried out at two lifting gates and one upper beam. These measurements have been carried out as part of the CONDITS project of Rijkswaterstaat. They were done using a measuring and data acquisition system, operated from the central control building of the storm surge barrier. The measurements consisted of: Determining the natural frequencies and vibration modes of the two gates, using acceleration transducers. Measuring vibrations and tensions in the gates and in components during the lowering of the gate in flow conditions, using acceleration transducers and resistance strain gauges. Measuring wave impact pressures on frame girders and horizontal plates of a gate, and measuring the response of the gates to wave impacts during the lowering of the gates and in retaining situation, using pressure transducers and acceleration transducers. Measuring wave impact pressures at the bottom side of an upper beam and the response of the beam in case of wave impacts, as well as the response during flow at the beginning of closing the gates and during the lowering of the gates. Measurements using acceleration transducers, pressure transducers and dynamometers in the supporting blocks.

The measuring system was reported in Delft Hydraulics Report Q298-I. Some measuring results were reported in Delft Hydraulics Report Q605. The analysis did not show strong vibrations or wave impacts during the storm conditions tested. Wave impact measurements at the bottom side of an upper beam resulted in pressures up to approximately 50kN/m2. The measurements were carried out just before and during the first part of closure of the gates in a situation with flow through the openings (hydraulic head approximately 0.15 m over the barrier). It is expected that greater impacts may occur. During the lowering of the gates in flowing water, as regards the gate as a whole, no significant vibrations were measured. In one of the round frame girders (in the final design, instead of closed plate girders, frame girders were mounted), a limited, fairly regular vibration was observed, which disappeared again when lowering the gate even further. The natural frequencies of the gates (main movement in dry conditions) appeared to be very similar to the natural frequencies of a continuous-elastic model of the gates. Below the water level there were major differences. These may be ascribed to the differences in design, as closed plate girders had been used in the model. c. Discharge sluice Brouwersdam For the verification of a model investigation (Delft Hydraulics investigation M1272), two culvert gates in the discharge sluice in the Brouwersdam were tested. These lifting gates were placed in parallel venturi-shaped discharge shafts with a length of 195 m. The gates are positioned completely below the water surface. In the culverts, flow is possible in two directions. Both gates were measured simultaneously, using acceleration transducers and portable registration equipment. The measurements have been reported in Delft Hydraulics Report R1347. The vibrations that were observed appeared again and again to be irregular

166

Delft Hydraulics

without resonance occurring. The amplitudes of the forces are very small relative to the static load; this holds for both the vibrations in horizontal direction as well as those in vertical direction. d. Intake sluice Volkerak For the shape of the gates, see Part A, Chapter 6, Example 6.2g. During the start of operation of the gate, it was observed that the lifting gates (30.8 m span) showed vibrations at small openings. As a result, vibration measurements have been carried out. As the rubber sealing edge at the bottom side of the gate was a possible source of the vibrations, measurements have also been carried out at a gate at which this seal had been removed. Measurements were carried out using acceleration transducers and portable registration equipment. At the gate with a rubber seal profile, high-frequency vibrations between 30 and 50 Hz were observed; at the gate without seal profile, these vibrations were not observed. The situation without rubber profile had also been tested earlier in a continuouselastic scale model, and in that case no vibrations occurred at the sharp edge either. The measurements have been reported in Delft Hydraulics Report Q322-II. e. Storm surge barrier Krimpen For the design see Part A, Chapter 6, Example 6.2b. Vibration measurements have been carried out at the second (secondary) lifting gate in retaining position with a small gap between the underside of the gate and sill and during lifting and lowering of the gate. The hydraulic head and/or the flow were first directed outward and then inward. The measurements were carried out with an acceleration transducer combined with a portable (vibration) analyzer. The aim of the measurements was to obtain some understanding in high-frequency and audible vibrations (a vibration frequency of 90 Hz was measured) at the bottom edge of the lifting gate in case of a small opening. The measurements were reported in Delft Hydraulics Report R1304. In Part A, Example 6.2b, the vibrations that were observed have also been described. These in-flow vibrations gave cause for further investigations, as they also occur at other structures. Previously, this was not recognized as such, and it was also not discussed in other sources. Discharge sluice in the Afsluitdijk, Den Oever Vibration measurements have been carried out at one of the sea gates (lifting gate with span 12 m) of the Simon Stevin sluice complex, with closed gates and during lifting and lowering. The hydraulic head and/or the flow were directed toward the Waddenzee. The measurements were carried out in order to verify a vibration that had been observed earlier in case of a closed gate. In that case there probably was a leakage flow. The measurements were carried out with four acceleration transducers and portable registration equipment. During the measurements, the vibration was not observed again. The measurements were reported in Delft Hydraulics Report Q322-II. Wave impact measurements were also carried out at two of the sea gates during storm conditions, with the gates partially lifted (storm position) and completely closed. Accelerations of the gate were measured at a number of points. The aim of the measurements was to check whether wave impacts occurred during the storm and whether impulse loads due to rattling in the gate recesses occurred. The acceleration measurements were carried out with portable registration equipment. The measurements were reported in Delft Hydraulics Report Q490. The impacts that were observed during the measurement especially occurred in case of a partially open gate; this relates to the placement of the girders. Moreover, in case of a small hydraulic head the waves cause rattling of the gates; this occurs less easily in case of closed gates. 167 a.

Delft Hydraulics

Northern abutment IJmuiden Using an experimental design as part of the vertical wall of the northern breakwater of the IJmuiden harbour, the suitability of a large-size wave impact pressure transducer was assessed. This type of pressure transducer was planned for wave impact measurements at the sector gates of the Haringvliet discharge sluices. This concerned a pressure cell that was translated to prototype, just like it had been used in earlier model investigations. The impact that was analyzed had a peak pressure of 180 kN/m2. The investigation was described by Geleedst and t Hart (1968). h. Weirs in the Lek River at Hagestein and Driel For the design of all weirs with visor gates in the Lek River, see Part A, Chapter 6, Example 6.2a and Part C, Chapter 6, Example 6.7a. A characteristic property of visor gates is that the local opening across the length of the gate varies in size at each position. At visor gates, in case of relatively small lifting opening, vibrations occur. The cause of this might be the design of the bottom edge. Vibration measurements at the visor gates of the weirs at Hagestein and Driel have been carried out, in order to obtain more understanding of the phenomena and come up with measures to prevent vibrations. The measurements have been carried out using acceleration transducers and portable registration equipment. The measurements and recommendations have been reported in Delft Hydraulics Reports Q1579 and Q322-1. When opening or closing the gate, the vibration zone is displaced; apparently there is a certain critical lifting height. Occasionally high-frequency vibrations have been observed (60 Hz) at an opening of approximately 5 cm; these however are not critical, as the gate never remains in that position. Tests have also been carried out in which one plate field at the bottom edge was reinforced; as a result of this, the vibrations were considerably reduced. i. Shipping lock at Lith After a revision of the wooden bottom seal of the lifting gate in the lower hydraulic head of the shipping lock in the Maas River at Lith, heavy vibrations were observed at small lifting heights. Vibration measurements were carried out using an absolute displacement transducer that was pushed against the structure by hand. The vibrations could be prevented by mounting intermittent steel strips at the upstream side near the bottom edge. The measurements were reported in Delft Hydraulics Report S50-1 and Kolkman (1980). Further data concerning the design and results of measurements are given in Part A, Chapter 6, Example 6.2e. j. Weir at Sambeek In the Stoney Weir gate in the Maas River at Sambeek, which consists of two parts, heavy vibrations were observed in the situation with the upper gate slightly lowered. In this situation there is a leakage gap between upper gate and lower gate. No measurements were carried out. Initially the supervisor was looking for measures to stabilize the overflowing nappe, by aerating the nappe. This however had no effect. The leakage flow turned out to be the cause of the vibrations. Temporarily, a seal of the leakage gap was mounted, as a result of which the vibration disappeared. This phenomenon has also been described in Kolkman (1980). See further Part A, Chapter 6, Example 6.2f.

g.

168

Delft Hydraulics

Bottom-hinged gate at Neerbeek At the bottom-hinged gate at Neerbeek (municipality of Neer, Limburg), vibrations were observed in case of an overflowing nappe. Measurements were carried out using a manually operated vibration transducer. Tests were carried out using nappe splitters (for aeration of the space below the nappe) and interference elements at the top of the gate; these were effective. The measurements were reported in Delft Hydraulics Report R251. Pictures have been taken of transverse waves in the upper water (see Part A, Figure 4.27). Tide lock at Ravenswaay In case of the situation of a closed gate, strong vibrations were observed at the gate. This concerned a straight wheel gate (80 m span), reinforced by a horizontal arched girder. A rubber edge was used as a bottom seal. The vibrations generated standing transverse waves, both at the upstream and at the downstream side. This concerns recent measurements (January 1995). A definitive explanation for the vibrations has not been found yet. There must be leakage flow. Probably the gate is distorted due to static (hydraulic head) load in such a way, that a bottom gap is caused. See also Chapter 1, Figure C1.1 with explanatory text. m. Stop logs. As a result of bad experiences in using circular stop log tubes, experimental investigations were carried out in a section model on a 1:16 scale at Delft Hydraulics concerning the flow pattern and forces during conditions corresponding to those that occur when lowered in flowing water. Also special design adaptations have been tested on the tubes in order to make it possible to place the stop logs in a controlled way. After that, similar stop logs were lowered into a discharge sluice (in prototype), in which it appeared that extra measures were needed that did not directly follow from the model investigations. In the laboratory a partial section was investigated on scale 1:1 then. The reason for the prototype behaviour was demonstrated: the separation of the flow, and therefore the pressure distribution around the tube, in case of round cylinders strongly depends on the Reynolds number. The extra measures concerned a welding layer on the tube at the location where separation occurred in the model; this was done to force the separation of the flow in prototype as well. The investigation has been described in Delft Hydraulics Report S50-5. In Kolkman (1980) this example has also been described. For more information, see Part A, Chapter 6, Example 6.7a. l.

k.

169

Delft Hydraulics

170

Delft Hydraulics

REFERENCES 8.1 Delft Hydraulics Reports (in Dutch)2

A110

Bibliograpgy of literature on dynamic behaviour of hydraulic structures (P.A. Kolkman)

M399 Part I t/m IV Golfaanval Haringvlietsluizen; verslagen modelonderzoek, 1960/1961, ir M.A. Aartsen, ir. E. Allersma. (Wave load on Haringvliet sluices; scale model) M561 Part A Vizierschuiven Stuw Hagestein; modelonderzoek dynamisch gedrag: schaalproeven, 1962, ir. P.A. Kolkman. (Visor gates Hagestein weir; scale model investigation dynamic behaviour) Part B Vizierschuiven Stuw Hagestein; modelonderzoek dynamisch gedrag: onderrandproeven, 1962, ir P.A. Kolkman. (Visor gates Hagestein weir; scale model investigation dynamic behaviour gate under edge) Part C Vizierschuiven Stuw Hagestein; ontwerp elastisch gelijkvormig model, 1962, ir P.A. Kolkman. (Visor gates Hagestein weir; design of elastically uniform model) M620 M700A Trillen Stuw Haringvliet; rapport modelonderzoek, 1961, ir. E Allersma. (Vibrations Haringvliet sluice; scale model investigation) Vizierschuiven stuw Hagestein, vergelijking dynamisch gedrag in prototype en model; prototype- en modelmetingen fase droog en fase stilstaand water 1968, ir P.A. Kolkman. (Visor gates Hagestein weir; comparison prototype and model; dry and wet gate (no flow) Buitenmeting Spuisluis Haringvliet; trillen door stroom, 1977, ir J.D. van den Bunt. (In situ measurements Haringvliet sluice; flow induced vibrations) Buitenmetingen Spuisluis Haringvliet; golfbelastingen, 1974, ir J.D. van den Bunt. (In situ measurements Haringvliet sluice; wave conditions) Buitenmetingen spuisluis Haringvliet; golfbelastingen. 1980, P. Bosland, ir J.D. van den Bunt, ir P. van Groen, ir A.C.M. Vermeer. (In situ measurements Haringvliet sluice; wave conditions)

M754 Part I

Part II

Part III

Reports can only be studied by others after Delft Hydraulics has received approval for this from the client involved. 171

Delft Hydraulics

Part IV

Buitenmeting spuisluis Haringvliet; evaluatieverslag, 1981, ir J.D. van den Bunt. (In situ measurements Haringvliet sluice; evaluation) Sielverschlsse Eiderabdmmungen; Wellenbelastungen; Filterversuche, 1968, ir G. van Staal. (Wave loads on Eider dam) Schaaleffecten bij golfklappen op een talud; verslag modelonderzoek, 1979, ir Th. van Doorn. (Scale effects wave impact on a slope; scale model investigation) Schuiven van de stroomsluis Brouwersdam; modelonderzoek krachten en trillingen. 1983, ir R.J. de Jong. (Gates Brouwerdam sluice; scale model investigation forces and vibrations) Toegevoegde watermassa en instabiele trillingen van schuiven met een verticale bewegingsmogelijkheid; Rapport C8 Toegepast Onderzoek Waterstaat (TOW), 1977, ir A. Vrijer. (Added mass and unstable vibrations of vertically movable gates) Stormvloedkering Oosterschelde; krachten en afvoercofficinten bij rooster schuiven; verslag modelonderzoek. 1978, ir R.J. van der Wal. (Storm surge barrier Eastern Scheldt; forces and discharge coefficients grid gates (alternative)) Part I Stormvloedkering Oosterschelde. Golfklappen op de schuif in de Oosterschelde caisson; verslag modelonderzoek, 1977, ir C. Ramkema. (Storm surge barrier Eastern Scheldt; wave impacts on gate in caisson (alternative). Stormvloedkering Oosterschelde. Golfklappen: een zuigermodel met samendrukbaar water; verslag modelonderzoek, 1979, ir C. Ramkema, ir C. Flokstra. (Storm surge barrier Eastern Scheldt; a piston model with compressible water (to simulate wave impact effects) Stormvloedkering Oosterschelde. Golfklappen: een literatuur overzicht en schaaleffecten in modelonderzoek;verslag bureaustudie, 1979, ir C. Ramkema. (Storm surge barrier Eastern Scheldt; a literature review on wave impacts and scale effects)

M915-II

M1057

M1272

M1322

M1327-I

M1335

Part II

Part III

M1338

Stormvloedkering Oosterschelde; trillingsgedrag van roosterschuiven; verslag onderzoek in een elastisch gelijkvormig model, 1979, ir R.J. van der Wal. (Storm surge barrier Eastern Scheldt; vibrations of a grid gate (alternative))

172

Delft Hydraulics

M1374-V

Belasting op remmingwerken en ducdalven door varende schepen; samen vattend rapport over rekenmethoden, Rapport C32 Toegepast Onderzoek Waterstaat (TOW), 1983, ir A. Vrijburcht. (Loads on berthing works and ducdalfs caused by sailing ships) Stormvloedkering Oosterschelde; reflectiecofficinten en overslag bij stalen roosterschuiven; verslag modelonderzoek, 1976, ir C. Ramkema. (Storm surge barrier Eastern Scheldt; wave reflection coefficients and wave overtopping in case of steel grid gates (alternative)) Stormvloedkering Oosterschelde; krachten en trillingen bij de hefschuiven in de pijlerdam; vooronderzoek met sectiemodel, 1978, ir R.J. de Jong. (Storm surge barrier Eastern Scheldt; forces and vibrations at lift gates between large piers (alternative)) Stabiliteitsgedrag van schuiven met diverse onderranden bij een verticale bewegingsmogelijkheid; verslag experimenteel en theoretisch onderzoek, Rapport C16 Toegepast Onderzoek Waterstaat (TOW), 1979, ir J. Uwland. (Stability of vertically movable gates with various designs for the under edges) Stormvloedkering Oosterschelde. Vooronderzoek m.b.v. een stijf sectiemodel naar stroom- en golfbelasting op dorpelbalken, bovenbalken en plaatliggerschuiven; vooronderzoek trillingsgedrag van de plaatliggerschuiven met behulp van een massa-veermodel, 1981, ir T.H.G. Jongeling. (Storm surge barrier Eastern Scheldt; pre-investiation with a stiff section model to flow and wave induced forces on sill beams, upper beams and plate girder gates; pre-investigation to the vibrations of the same gates with a single mass spring section model) Stormvloedkering Oosterschelde. Orinterend onderzoek naar golfklappen op de plaatliggerschuiven, sectie R15, loodrechte golfaanval en aanstroming; verslag modelonderzoek, 1982, ir R.M. Korthof. (Storm surge barrier Eastern Scheldt; pre-investigation to wave impacts on plate girder gates) Stormvloedkering Oosterschelde; pijleroplossing; onderzoek naar trillingsgedrag van plaatliggerschuiven en balken met behulp van een elastisch gelijkvormig model, 1981, ir H.G. Jongeling en ir H.W.R. Perdijk. (Storm surge barrier Eastern Scheldt; investigation dynamic behaviour plate girder gates and beams with an elastically uniform model) Stormvloedkering Oosterschelde; onderzoek naar toegevoegde watermassa's plaatliggerschuiven met behulp van een elektrisch analogon, 1982, ir C. Deelen. (Storm surge barrier Eastern Scheldt; investiagation to added mass of plate girder gates using an electric analogon)

M1381-I

M1424

M1490

M1494

M1504

M1561

M1582

173

Delft Hydraulics

M1648

Stormvloedkering Oosterschelde. Onderzoek m.b.v. een elastisch gelijkvormig model naar het responsiegedrag van de bovenbalken bij golfklapbelasting. Berekening van golfklapdrukken m.b.v. een wiskundig massaveersysteemmodel; verslag modelonderzoek en bureaustudie, 1981, ir T.H.G. Jongeling en ir H.W.R. Perdijk. (Storm surge barrier Eastern Scheldt; investigation in an elastically uniform model model to the resonse of upper beams to wave impacts; computation of wave impact pressures using a mass spring model) Stormvloedkering Oosterschelde. Verticale golfbelastingen op de vakwerkschuiven, loodrechte golfaanval; verslag modelonderzoek, 1982, ir R.M. Korthof. (Storm surge barrier Eastern Scheldt; vertical wave loads on truss girder gates)

M1723/M1687

M1765 Q298-I

Alto Lazio nuclear power plant. Wave forces on the cooling water intake structure; verslag modelonderzoek, 1982, ir R.M. Korthof. Conditiebewaking Stormvloedkering Oosterschelde; meetplan voor bovenbalken en schuiven (opmerking: deel II is gerapporteerd onder Q605), 1986, ir T.H.G. Jongeling. (Storm surge barrier Eastern Scheldt; measuring plan for in situ measurements at upper beams and Gates)

Q322 Part I Vizierschuiven in de Rijn te Driel; orinterende trillingsmeting aan de noordelijke vizierstuw; verslag buitenmeting, Rapport C44 Toegepast Onderzoek Waterstaat (TOW), 1986, ir T.H.G. Jongeling. (Visor gates in the Rhine (at Driel); in situ measurements vibrations) Part II Spuisluizen Volkerak, spuisluizen afsluitdijk den Oever; trillingsmetingen aan hefschuiven, Rapport C61 Toegepast Onderzoek Waterstaat (TOW), 1988, ir T.H.G. Jongeling. (Volkerak sluice and sluice den Oever; in situ measurements vibrations) Q490 Spuisluizen afsluitdijk Den Oever. Dynamisch gedrag hefschuiven bij golfbelastingen; verslag prototype metingen, 1990, ir H.W.R. Perdijk, ir T.H.G. Jongeling. (Sluice den Oever; dynamic behaviour gates during wave loads; in situ measurements)

Q525 Part 1 Dynamisch belaste hydraulische constructies; 2-d vloeistofmodel in DIANA voor de berekening van vloeistof-constructie interacties; afstudeerverslag, 1987, ir J.W.F. Wamelink. (Module in DIANA for 2D fluid structure interaction) Part 2 Dynamisch belaste hydraulische constructies; voorstudie naar de opzet van een numeriek vloeistof-constructie interactie model als onderdeel van het DIANA programmapakket; verslag bureaustudie, 1987, ir T.H.G. Jongeling, dr ir K.L. Meyer en ir F. Sas. (Module in DIANA for 2D fluid structure interaction) 174

Delft Hydraulics

Q605

Conditiebewaking Stormvloedkering Oosterschelde; meting tijdens hoogwater sluiting 14 febr.'89, eigenfrequentie schuif S13, 1989, ir T.H.G. Jongeling en ir J.T.M. van Doorn. (Storm surge barrier Eastern Scheldt; in situ measurements natural frequency gate) Venice barrier; study on the influence of the inclination angle and the gate side shape on gate response, 1988, ir T.H.G. Jongeling. Maeslantkering; onderzoek naar het responsiegedrag van de sectordeuren in een overzichtsmodel, 1989, ir T.H.G. Jongeling. (Maeslant barrier; dynamic behaviour of sector gates using an overview model) Venice barrier; study on gate response, forces in the gate supports and leakage discharge, 1990, ir T.H.G. Jongeling. Maeslantkering. Vervolgonderzoek naar het responsiegedrag van de sectordeuren in een overzichtsmodel; verslag modelonderzoek, 1990, ir T.H.G. Jongeling en ir J.J.A. van Huijstee. (Maeslant barrier; continuation of investigation dynamic behaviour sector gates using an overview model) Maeslantkering; onderzoek met behulp van een sectiemodel, optimalisatie vormgeving sectordeur; drukmetingen op drempel en sectordeur, 1990, ir T.H.G. Jongeling. (Maeslant barrier; investigation design sector gate using a section model) Maeslantkering; analyse van het responsiegedrag van de sectordeuren met behulp van een rekenmodel (met gebruikmaking van drukmetingen in sectiemodel), 1991, dr ir P.A. Kolkman. (Maeslant barrier; analysis of dynamic behaviour sector gates using a computation scheme and results of pressure measurements of a section model) Maeslantkering; onderzoek met behulp van een sectiemodel; optimalisatie vormgeving sectordeur; drukmeting op drempel en sectordeur; verslag modelonderzoek, 1990, ir T.H.G. Jongeling. (Maeslant barrier; optimization of the sdesign of the sector gates using a section model and pressure measurements on sille and gate) Maeslantkering; additioneel onderzoek voor de sectordeuren in een overzichtsmodel, 1991, ir T.H.G. Jongeling. (Maeslant barrier; additional investigation in an overview model)

Q744 Q969

Q1033 Q1140

Q1190

Q1271

Q1190

Q1278

175

Delft Hydraulics

Q1293

Maeslantkering; detailonderzoek in een sectiemodel; krachten en drukken ten behoeve van het responsiegedrag, stabiliteitsberekeningen; metingen van fenderkrachten en drukken bij landen, 1992, ir T.H.G. Jongeling. (Maeslant barrier; detail investigation in a section model; forces and pressures related to the dynamic behaviour; measurement of fender forces and pressures during landing of gate on sill) Europoortkering met open Beerdam, Stormvloedkering Hartelkanaal; trillingsonderzoek met behulp van een sectiemodel, 1993, ir G. van Driel en ir R.J. de Jong. (Storm surge barrier Hartelkanaal; investigation to vibrations using a section model) Trillingsmetingen Stuw Hagestein; rapport buitenmeting, 1993, ir R.J. de Jong. (Weir Hagestein (viso gate); in situ measurements) Stuwklep bij de Winkelmolenbrug in de Neerbeek; rapport trillingsonderzoek(prototype), 1963, ir M.A. Geleedst. (Weir gate Neerbeek; in situ measurements vibrations) Elastisch gelijkvormig model roosterschuiven; rapport voorbereiding onderzoek, 1977, ir H. Depeweg. (Storm surge barrier Eastern Scheldt; design of elastic similariy model grid gate (scale 1:3)) Trillingsmetingen aan de hefschuif noord van de stormvloedkering te Krimpen a/d IJssel; verslag buitenmetingen en bureaustudie, 1980, ir T.H.G. Jongeling. (Storm surge barrier Krimpen; in situ vibration measurements) Doorlaatsluis Brouwersdam; trillingsmetingen schuiven; verslag buitenmetingen, 1980, ir C. Deelen. (Sluice Brouwerdam; in situ vibration meaurements) Duwvaartsluizen in de Philipsdam; maximale krachten op schuiven; bureaustudie, 1981, ir C.R.M. Oudshoorn. (Shipping locks in Philipsdam); maximum forces on gates) Spuisluis Haringvliet, trillingsmetingen; verslag buitenmetingen, 1979, ing.W. Klinkenberg en ir A.C.M. Vermeer. (Sluice Haringvliet; in situ vibration measurements) Part I Nota (speurwerk-) onderzoek betreffende het dynamisch gedrag van sectorschuiven; opzet totaalonderzoek en verslag dempingsproeven, 1967, ir P.A. Kolkman. (Dynamic behaviour sector gates, investigation strategy and damping tests)

Q1500

Q1579 R251

R1068

R1304

R1347

R1506

R1510

S50

176

Delft Hydraulics

Part III

Nota onderzoek aan ronde schotbalken op prototypeschaal; onderzoek van trillingen bij waterbouwkundige constructies, 1968, P. Bosland en ir P.A. Kolkman. (Circular stop logs; vibration investigation) Investigation on rapidly varying forces on gates; visit to shipping lock at Lith. 1958, ir R. Hart. Trillingen van sluisdeuren, verslag rekenstudie; Rapport C4 Toegepast Onderzoek Waterstaat (TOW), 1976, dr ir J.C.W. Berkhoff. (Vibration of lock gates; computations)

Part IV W254

8.2 Other reference material


Allersma, E. (1959): "The virtual mass of a submerged sluice gate". 8th Congr. IAHR, Montreal, paper A23. Also Delft Hydraulics publication 18. Allersma, E. (1961): "The velocity at which sound travels in a mixture of air and water", WL-informatieblad X-7a. Bagnold, R.A. (1939): "Interim report on wave pressure research", Journal of the Institution of Civil Engineers, Vol.12, No. 1, London, England. pp. 202-226. Bakker, A.D., Jongeling, T.H.G., Kolkman, P.A. and Yan Shi Wu (1991): "Self-excited oscillations of a floating gate related to the gate discharge characteristics", XXIV IAHR Congress, session D, Madrid, pp. 423-432. Also Delft Hydraulics publication 462. Bendat, J.S. and Piersol, A.G. (1971): "Random Data; Analysis, Measurement, Procedures", Wiley-Interscience NY, USA. Bhattacharyya, R. (1978): "Dynamics of Marine Vehicles", Wiley-Interscience, NY, USA. Bouma, A.L. (1976): "Dynamica van Constructies, deel I t/m IV", handleiding bij het college b15 en b15a. Technische Hogeschool Delft, afd. Civiele Techniek. Campbell, I.M. and Weynberg, P.A. (1979): "Slam load histories on cylinders", in Int. Conference on Environmental Forces on Engineering Structures, Imperial College, London, England. Cummins, W.E. (1962): "The impulse response function and ship motions", Schiffstechnik, Bd. 9, Heft 47, pp. 101-109. Fabula, A.G. (1957): Ellipse-fitting approximation of two-dimensional normal symmetric impact of rigid bodies on water" in Proc. of 5th Midwestern Conference on Fluid Mechanics, pp. 299-315, Univ. of Michigan, USA.

177

Delft Hydraulics

Fontijn, H. (1975): "An approximative method for the determination of the hydrodynamic coefficients of a ship in case of swaying and yawing on shallow water" Rep. 75-4, Comm. on Hydraulics, Dept. of Civil Eng. Dept., Delft University of Technology, Delft. Fontijn, H. (1978): "The Berthing Ship Problem: Forces on Berthing Structures from Moving Ships". Rep. no. 78-2, Communications on Hydraulics, Dept. of Civil Engng., Delft Univ. of Technology, Delft. Fontijn, H. (1988): "Fender Forces in Ship Berthing, part I and II", proefschrift TU-Delft; ook Rep. no. 88-2, Communications on Hydraulic and Geotechnical Engineering, Delft Univ. of Technology, Delft. Fhrbter, A. (1966): "Der Druckschlag durch Brecher auf Deichbschungen", Mitteilungen Franziusinstitut Hannover, Heft 28, pp.1-206. Geleedst, M.A. en 't Hart, A.A. (1968): "Large Area Wave-Load Meter", ASCE-Journal Waterways and Harbours, Nov. 1968, WW4, pp.415-423. Also Delft Hydraulics publication 63. Haszpra, O. (1979): "Modelling hydroelastic vibrations", Pittman, London. Jong, R.J. de and Nunen, J.W.G. van (1979): "Excitation and vibration of a grid gate". In: Naudascher, E. and Rockwell, D. (editors), "Practical Experiences with Flow-Induced Vibrations", IAHR-IUTAM Symposium, Karlsruhe 1979, Springer Verlag 1980, paper C8, pp. 445-451. Also Delft Hydraulics publication 220. Jong, R.J. de, Jongeling, T.H.G., Kooman, D. en van der Weijde, H. (1980): "Vibration of Gates and Beams", in "Hydraulic Aspects of Coastal structures", Vol. 2, Delft University Press. pp. 25-46. Jongeling, T.H.G. en Kolkman, P.A. (1992): "Unstable Behaviour of a Floating Sector-Gate Barrier" ASME 3d Int. Symp. on Flow-Induced Vibration and Noise, Anaheim Calif. (PVP vol. 245, Bluff-Body/Fluid and Hydraulic Machine Interactions, Book G007271992. pp. 207-220. Jongeling, T.H.G. (1993): "Wave-Induced Resonance of a Flap-Gate Barrier", Proc. 2nd Conf. on Structural Dynamics: Eurodyn'93, Trondheim, Norway. Publ.Balkema Rotterdam, pp. 1149-1156. Jongeling, T.H.G. en Kolkman, P.A. (1995): "Subharmonic standing waves leading to lowfrequency resonance of a submersible flap-gate barrier", 6th Intern. Symp. on Flowinduced Vibration. Imperial College, London. Krmn, T.L. von, Wattendorf, F. (1929): "The impact on seaplane floats during landing", National Advisory Committee for Aeronautics (NACA), TN 321, Washington, USA. Kolkman, P.A. (1963): "Analysis of Vibration Measurements on an Underflow Type of Gate", Xth IAHR-Congress, London. pp. 185-191. Also Delft Hydraulics publication 33. 178

Delft Hydraulics

Kolkman, P.A. (1967): "Elastisch Gelijkvormige Modellen van Waterbouwkundige Con structies", KIVI tijdschrift "De Ingenieur" 79, nr.4, pp.W9-17. Also Delft Hydraulics publication 49. Kolkman, P.A. (1976): "Flow-induced gate vibrations; prevention of self-excitation, compu tation of dynamic gate behaviour and the use of models". Proefschrift TH-Delft. Also Delft Hydraulics publication 164. Kolkman, P.A. (1977): "Self-excited gate vibrations", 17th IAHR-congr., Baden-Baden; Invited lecture of sessionC.c, Vol.6, pp. 372-380. Also Delft Hydraulics publication 186. Kolkman, P.A. (1977): "Afmeerkrachten bij centrisch en excentrisch botsen van een schip tegen fenderpalen en bij aanvaren van een verend remmingwerk", Notitie 20 in de serie "Hydraulica bij schutsluizena" van de Vakgroep Waterbouwkunde van de Technische Universiteit Delft, afdeling Civiele Techniek, also in Kolkman 1992. Kolkman, P.A. and Vrijer, A. (1977): "Gate edge suction as a cause of self-exciting vertical vibrations", 17th IAHR-Congress, Baden-Baden, Vol. 4, paper C49. pp. 395-402. Also Delft Hydraulics publication 188. Kolkman, P.A. (1980): "Development of vibration-free gate design; learning from experience and theory". In: Naudascher, E. and Rockwell, D. (editors) "Practical Experiences with Flow-Induced Vibrations", IAHR-IUTAM Symposium, Karlsruhe 1979, Springer Verlag 1980, paper C1, pp.351-385. Also Delft Hydraulics publication 219. Kolkman, P.A. (1981): "Maximale golfdrukken volgens stromingsdrukmodel, schokgolfmodel en waterpistonmodel", Notitie SL1 in Notities van P.A. Kolkman in de periode 1969-1990, Waterloopkundig Laboratorium (1992). Kolkman, P.A. (1984): "Gate vibrations", Hoofdstuk 2 in "Developments in Hydraulic Engineering-2", editor P. Novak; Elsevier Applied Science publishers. Kolkman, P.A. (1988):"A simple scheme for calculating the added mass of hydraulic gates", J. Fluids and Structures, Vol. 2 Nr. 4, pp.339-353. Also Delft Hydraulics publication 439. Kolkman, P.A. (1992): "Notities van P.A. Kolkman", uitgegeven door het Waterloopkundig Laboratorium. Kooman, D., Korthof, R.M., Ligteringen, H. and Stans, J.C. (1980): "Wave Impact Forces, Consequences for Gate Design", in "Hydraulic Aspecs of Coastal Structures", Vol. 2, Delft University Press. pp. 47-66. Lundgren, H. (1969): "Wave shock forces: An analysis of deformation and forces in the wave and in the foundation", in Proc. of the Symposium on Research on Wave Action, Vol. 2, paper 4, Delft Hydraulics, Delft.

179

Delft Hydraulics

Miller, B.L. (1980): "Wave Slamming on Offshore Structures", report no R81, National Maritime Institute, Feltham, England. Ogilvie, T.F. (1964): "Recent progress toward the understanding and prediction of ship motions", Proc. 5th O.N.R. Symp. on Naval Hydrodynamics, Bergen, pp. 3-80. Oortmerssen, G. van (1974): "The berthing of a large tanker to a jetty", 6th Annual Offshore Technology Conf., Houston, Paper OTC 2100, pp. 665-676. Oortmerssen, G. van (1976): "The motions of a moored ship in waves", Thesis Delft University of Technology, publ. H. Veenman en zn. n.v., Wageningen. Ramkema, C. (1978): "A model law for wave impacts on coastal structures", 16th Conf. on Coastal Eng. Hamburg. Delft Hydraulics publication no. 207. Sarpkaya, T. (1978): "Fluid Forces on Oscillating Cylinders" Proceedings ASCE, Journal of Waterway, Port, Coastal and Ocean Div. WW3, paper 13941. pp.275-290. Tick, L.J. (1959): "Differential equations with frequency-dependent coefficients", Journal of Ship Research, Techn. Note, Vol. 3, No. 2, pp. 45-46. Vrijer, A. (1980): "Stability of vertically movable gates" in "Practical Experiences with FlowInduced Vibrations". IAHR/IUTAM-symposium, Karlsruhe 1979. Published in 1980, editors E. Naudascher and D. Rockwell, publ. Springer Verlag, paper C5, pp. 428-434. Also Delft Hydraulics publication 222. Witte, H.H. (1988): " Druckschlagbelastung durch Wellen in deterministischer und stochasti scher Betrachtung", in Mitteilungen Heft 102/1988, Leichtweiss Institut fr Wasser bau, Technische Universitt, Braunschweig, Deutschland.

180

Delft Hydraulics

APPENDIX I

DEDUCTION OF THE RESPONSE FUNCTION IN THE TIME DOMAIN FROM THE TRANSFER FUNCTION IN THE FREQUENCY DOMAIN

181

Delft Hydraulics

182

Delft Hydraulics

APPENDIX I DEDUCTION OF THE RESPONSE FUNCTION IN THE TIME DOMAIN FROM THE TRANSFER FUNCTION IN THE FREQUENCY DOMAIN The following is taken from Vrijburcht (1983) (Delft Hydraulics Report M1374-V). The point of departure is that in case of a harmonic oscillation of an object in water, the periodic forces that are generated in the direction of movement are known. The combination of water and object is assumed to behave in a similar way to a linear system; a twice greater amplitude results in twice greater forces. It may be shown that a harmonic movement in that case also results in a harmonic force signal. The force generated by the water in the situation described above is considered to be factorized into a component that is in-phase with the movement and a component that is outof-phase with the movement (which may also be described as in-phase with the movement velocity). The force that is in-phase is referred to as the added water mass (a) multiplied with the acceleration, therefore as if it were proportional to the acceleration of the object. The force that is out-of-phase is referred to as the added water damping (b) multiplied with the velocity, therefore as if this force is proportional to the acceleration of the object. The a and b factors (hydrodynamic coefficients) are obtained respectively by dividing the force that is in-phase by the acceleration of the object and the force that is out-of-phase by the velocity. a and b are frequency-dependent. For a harmonic movement with angular frequency it holds (in the frequency domain therefore) that: Fw (t , ) = a ( ) d2 y dy + b( ) 2 dt dt (A.1)

Fw is the force of the water. a() and b() may be determined by carrying out tests (harmonic excitation of the object in water); in certain cases these may be calculated. See Figure C3.13 in Part C, Paragraph 3.3.2. If the object performs a movement that has not been specifically described yet, it is obvious to suppose that the force exerted by the water at a moment in time t is co-determined by the history of this movement. Especially when there is wave radiation, this is unavoidable. It is assumed that only the acceleration d2y/dt2 and the velocity dy/dt at moment in time (t-) contribute to the force at moment in time t; this applies to any value of the time shift t. For a random movement it holds:

Fw (t ) = A( ) y (t )d + B( ) y (t )d
0 0

(A.2)

In this, y and y respectively represent the first and the second time coefficient of y. Remark: Often, another notation is used, i.e. t for the actual moment in time and for the moment in time at which the force element was active; this results in a change of and (t-). It is therefore possible, as an example, to deduce A.2 directly from A.1 by substituting Fw ( , t ) = Fw eit and (t ) = Y ( )eit in A.1. The determination of the inverse Fourier transformation then results in:

183

Delft Hydraulics

Fw (t ) =

y ( ) A(t )d +

y( ) B(t )d

After changing (t-) and , this results in Equation A.2 again. If the functions A() and B() were known, then it would also be possible to calculate what the forces would be in case of a harmonic oscillation. For the movement (location) the following is introduced: y = Yeit The velocity of movement of the object now is: (A.3)

dy = iY it dt For the acceleration of the object it holds: d2 y = 2Y it dt 2

(A.4)

(A.5)

After filling in A.3, A.4 and A.5, the force through the water, Fw(t), from Equation A.2, may be expressed as:
Fw (t ) = A( ). Ye
2 0 i ( t - )

d + B( ).iYei ( t - ) d
0

(A.6)

or: Fw (t ) = 2Yeit A( )e i d + iYeit B ( )e i d


0 0

(A.7)

Using Equations A.4 and A.5, this may also be written as: Fw (t ) = y A( )e i d + y B( )e i d
0 0

(A.8)

By comparing this equation with A.1, it is found that:

A( )e
0

d = a ( )

(A.9)

and:

B( )e
0

d = b( )

(A.10)

These equations are known from the theory of the Fourier transformation, provided that there is no time shift , but of the time t. Mathematically however, this makes no difference; A() and B() may also be seen as a kind of impulse-response function.

184

Delft Hydraulics

The reverse transformation from the frequency domain to the time domain now results in: 1 A( ) = 2 and: B ( ) = 1 2

a( )e b( )e

it

(A.11)

i t

(A.12)

In case a() and b() (added water mass and damping) in the entire frequency range are known, it is possible by numerically solving the Integrals A.11 and A.12, to determine A() and B(). These only need to be determined once. After filling in A.11 and A.12, it is now possible to use Equation A.2 (force due to the water at a random movement of the object) for calculations in the time domain. Vrijburcht (1983) (Delft Hydraulics Report M1374-V) provides the following equation to describe the problem of a ship that centrically collides with a fender structure:

my + mass ship

my + cy + k y + mass damping spring guiding works

Fw =0 force water

(A.13)

Equation A.2 is used for Fw again. After that, Equation A.13 may be solved numerically in time. Further references about this subject are: Tick (1959), Cummins (1962), Ogilvie (1964) and Oortmerssen (1976).

185

Delft Hydraulics

186

Delft Hydraulics

APPENDIX II

SCALE RULES AND SCALE EFFECTS FOR INVESTIGATIONS OF DYNAMIC BEHAVIOUR OF SCALE MODELS

187

Delft Hydraulics

188

Delft Hydraulics

APPENDIX II SCALE RULES AND SCALE EFFECTS FOR INVESTIGATIONS OF DYNAMIC BEHAVIOUR OF SCALE MODELS

Scale rules are the laws that a scale model needs to comply with in order to be capable of reproducing the primary flow properties and the mechanical properties on a reduced scale; scale effects are the errors that are caused due to the fact that it is not possible to reproduce secondary characteristics simultaneously. What elements are primary and what elements are secondary is a matter of conscious choice that depends on the aim of the investigation. If several properties of water or structure need to be reproduced simultaneously, these may give cause for contradictory scale rules. In that case scale model investigations are no longer possible just like that. It is however possible to set up one or more experiments in order to obtain a better understanding of physical phenomena. The scale rules for scale models concerning dynamic investigations were initially published in Kolkman (1967) (also included in Kolkman (1976)). Below, the point of departure is a model with a geometric reproduction on a length scale nL (the index refers to the magnitude to which the scale factor n relates). The requirements upon the reproduction of the flow pattern, waves and vibrations is that, at corresponding points and at corresponding moments in time, the velocity vectors in directions x, y and z are on the same scale. This is referred to as the kinematic reproduction with a velocity scale nv and a time scale nt. Naturally, nt and nv are coupled through nL.

nt =

nL nV

(B.1)

It is only possible to meet the requirements of kinematic reproduction when there is a univocal scale of the pressures and forces in fluid and structure. This is referred to as dynamic reproduction with a pressure scale np and a force scale nF. The model laws that may be deduced during investigations concerning flow and waves follow from the requirement of dynamic reproduction; it is only possible to obtain a kinematic reproduction when this requirement is met.

Model laws for flow investigations In the Navier-Stokes equation for the dynamic equilibrium of each of the water particles (below only represented with respect to the direction x), the following terms are found:
2u 2u 2u u p u u u = u + v + w v 2 + 2 + 2 x t y z y z x x

(B.2)

= density of the fluid, V = velocity vector (that is factorized again in the vectors u, v and w, in respectively the directions x, y and z), t = time and = kinematic viscosity ( is referred to as dynamic viscosity). The first term on the right indicates the part of the pressure gradient that is coupled to the local water acceleration. The second series of terms relates to the acceleration that is generated due to the fact that the particles flow toward an area where the velocity increases or 189

Delft Hydraulics

decreases (the so-called convective acceleration). The third series of terms relates to the influence of the viscosity. The Navier-Stokes equation has a similar form in direction y; for direction z there is a gravitational term in addition, that creates an equilibrium with the hydrostatic pressure. As the latter does not cause any flow, it has no influence on the scale rules. In order to obtain a univocal scale for the pressure (this only appears on the left), all terms on the right need to be reproduced according to the same scale. The local acceleration and the convective acceleration are reproduced on the same scale when the requirements of Equation B.1 are met; this therefore does not cause any problems. If there are periodic or quasi-periodic phenomena in the flow (such as turbulence), then it is possible to deduce from B.1 for the scale of the frequency ( = angular frequency, f = frequency): n n f = n = V (B.3) nL This may also be presented as a condition for the Strouhal number:
S= fL is equal for prototype and model V (B.4)

The pressure scale follows from a comparison of the left part of the equation and the convective acceleration in B.2: 2 n p = n nV (B.5) This may also be formulated as an invariance of the Euler number: Eu = p V 2 is equal for prototype and model (B.6)

The influence of the viscosity may only be represented accurately in the model when the second and third series of terms on the right side of Equation B.2 are reproduced on the same scale. This means that: n nV = v (B.7) nL This may also be formulated as a condition for the Reynolds number: Re =
VL

is equal for prototype and model

(B.8)

Conclusion 1: If the flow limitations in the model (set up as a tunnel model) are fixed, then the reproduction of the Reynolds number will probably determine the choice of the water velocity in the model. The viscosity, , is the same in prototype and in model, as it is difficult to generate flow conditions with a fluid other than water. Equation B.8 implies that the water velocity in the model needs to be much greater than in the prototype. Often, it is impossible to realize this. If it is expected that the influence of

190

Delft Hydraulics

the viscosity in the prototype is small (this is the case at hydraulic structures that are not or badly streamlined), then it is possible to choose the flow velocity in a model freely, provided that the Reynolds number does not go under a critical limit. About this limit, which depends on the shape of the structure, many references are available.
Situation of water in case of flow with a free surface of water or in case of waves In case of flow with a free surface of water, the flow is always coupled to pressure differences that cause distortion at the surface of the water level. The water level is also one of the flow limits, and the requirement of geometric reproduction applies here as well. As long as the distortion of the water level is small relative to the water depth or the dimensions of the object that is to be observed, an error in this distortion is not much of a problem. A measure for the correct reproduction is the velocity height V2/2g relative to the representative length measure L. Therefore: 2 nV = nL ng (B.9)

This may be formulated as a condition for the Froude number:


Fr = V gL is equal for prototype and model

(B.10)

This condition does not need to be met if the Froude number is small, as the water level is horizontal in that case anyhow. The precondition of the Froude scale is always contrary to the reproduction of the Reynolds number: the Reynolds number is always too small. There are means available to generate extra turbulence in the boundary layer in the model, in order to better simulate the prototype condition (extra roughness of a turbulence wire); this however is at the cost of the geometric reproduction. Conclusion 2: In a model with a free water surface, in case the flow is so strong that pressure differences generate an important distortion of the water level, the Froude number must be reproduced. This also applies to waves. It is however important that the condition that the Reynolds number in the model is big enough, is met. If the Froude number is small, both in prototype and in model, then this does not need to be reproduced. If wave investigations are carried out in the model, then the Froude number needs to be reproduced at all times. A similar condition as for the Reynolds number may also be formulated for the Weber number. The Weber number LV2/, (with = surface tension), has no influence, provided it is big enough. The influence of the surface tension is that a pressure leap is generated between air and water, that depends on the curvature of the water surface.
Reproduction in a flow model of a structure consisting of masses and springs For the deduction of scale rules, in addition to the pressure scale (Equation B.5), here the scale for the force (therefore: pressure multiplied with the surface on which it operates) is introduced: 2 2 (B.11) nF = n nV nL

191

Delft Hydraulics

Consider now the classic equation of the single mass spring system, as this is also representative of multiple systems with respect to the deduction of scale rules. This equation is: d2 y dy m 2 + c + ky = F (t ) (B.12) dt dt If the requirement3 is, that the amplitude of the movement on the length scale needs to be reproduced: n y = nL (B.13) and the force scale and the time scale are known, then it is possible to determine the scales of k, c and m on the basis of the requirement of homogenous reproduction of each of the terms on the left and the external force. For the stiffness it holds that:
2 nk = n nV nL

(B.14)

This may also be formulated as a requirement concerning the Cauchy number based on the spring stiffness: k is equal for prototype and model (B.15) Ca = V 2 L For the damping it holds that dy/dt needs to be reproduced on the velocity scale, which in that case results in: nc = n nV n 2 (B.16) L This again may be considered as a condition for the damping number:

Da =

c VL2

is equal for prototype and model

(B.17)

For the scale of d2y/dt2 it may be deduced that this is equal to the scale of V2/L. Therefore, for the scale of the mass it is found that:
nm = n n3 L

(B.18)

This mass scale is independent of the chosen velocity scale. The mass scale may also be translated to the specific weight of the structure material in the model:

n ;struct . = n ;water

(B.19)

This requirement is not always necessary; if it concerns an investigation of added water mass, damping or spring stiffness as a function of the Strouhal number, then it is possible to deviate from the correct scaling of the mass and the spring stiffness. 192

Delft Hydraulics

Conclusion 3: In case of dynamic mass spring models, the scaling of the mass is independent of the chosen velocity scale, and follows from the mass number. Therefore:
nm = n n3 L

If the velocity scale may be chosen freely (freely flow-surrounded objects, as an example flow in a tunnel), then the scale of the flow velocity and that of the stiffness are coupled through the Cauchy number.
2 nk = n nV nL

The damping number also needs to be reproduced;


2 nc = n nV nL

In practice (for the sake of safety), investigations are usually carried out in lowdamped models. As it is possible to determine the water damping in the model separately (by reducing the damping measured in flowing water by the damping measured in dry condition), it is also possible to assess the total damping in prototype in flowing water, provided the dry damping in prototype is known. In certain cases, when the vibration mechanism is known as well, it is possible to calculate the influence of a deviating damping on the vibration amplitude that is measured. In Part A, Chapter 3, reviews focus on the added water mass, water damping and water stiffness. It was found that the added water mass is proportional to L3, the flow stiffness is proportional to V2L, and the flow damping is proportional to VL2. The resulting model values correspond to the scale factors that were formulated for the corresponding mechanical properties. If the requirement that the mass number and the Cauchy number in prototype and in model are the same is met, then it automatically meets the requirement that the scale of the resonance frequency corresponds to the scale of the frequency of the excitation due to flow (the latter follows from the correspondence of the Strouhal number in prototype and in model, see Equation B.4).

Reproduction of dynamic behaviour of continuous-elastic structures in case of flow in a tunnel model The requirement concerning material distortion is that the relative distortion (deformation = L/L) is reproduced: n = nL / nL = 1
(B.20)

As the distortion now equals the material tension divided by the elasticity module, it holds that: n n = (B.21) nE (in this, is the material tension and E is the elasticity module). 193

Delft Hydraulics

Now, in case of geometric reproduction, the scale of the material tension equals the scale of the pressure in the water. Consequently, this results in the following equation:
2 nE (= n p ) = n nV

(B.22)

This may also be seen as a precondition for the Cauchy number (but this time Ca is based on the elasticity module):

Ca =

E V 2

is equal for prototype and model

(B.23)

The same precondition applies to the sliding module, G, of the material. As E and G are more or less coupled in a linear way, with a ratio that does not vary much for the various types of material, this is not represented separately here. The requirement that the damping also needs to be scaled accurately may now not be translated in the same way to a magnitude that is coupled to a flow condition. The dimensionless damping, , actually is a material constant that (except for one constant) indicates what percentage of the potential energy of the deformation of the material in case of outward spring is not returned. Therefore it must hold that:

n = 1 or:

(B.24)

is equal for prototype and model

(B.25)

For the density of the material the same applies as that which was deduced earlier at the discrete mass spring systems: n , struct . = n , fluid

(B.26)

Conclusion 4: For a model of a continuous-elastic structure, in case the velocity scale may be freely chosen, the scale of the elasticity module equals the scale of the hydrodynamic pressures, and the density and the dimensionless damping of the structure material must be the same as in the prototype. Therefore:
2 nE = n p = n nV

n , struct . = n , fluid

and:
n = 1

In order to meet the last two requirements, in case of tunnel models in which there is a free choice with respect to the flow velocity that is to be established, the model preferably is constructed from prototype material, as a result of which the flow that is generated is
194

Delft Hydraulics

according to velocity scale 1 (wind tunnel investigations airplane vibrations). The advantage of using prototype material is that the material damping is automatically correct in that case. Choosing another type of material naturally is also possible. As the requirement that the elasticity is represented to scale is met, it is possible to immediately use the elasticity that is measured in the model for the prototype values.
Reproduction of continuous-elastic structures in flow conditions with a free water surface and in waves, in which the Froude scale rules need to be met In case of flow and wave experiments on the Froude scale, the scale rule for the elasticity module is deduced from the requirement that these have to be reproduced on the scale of the flow pressure (Equation B.22). Therefore: nE = n ng nL

(B.27)

Conclusion 5: In case of reproduction of the dynamic behaviour of structures in models with water, in which the flow or the waves occur on the Froude scale, in the ideal case the elasticity module needs to be reproduced on the length scale in the elastic model. In more general terms, this is formulated as:
nE = n ng nL

while it remains valid that:


n , struct . = n , fluid

and:
n = 1 Scale rules used in reality, if flow or wave conditions are set up on the Froude scale Material with an elasticity on the length scale and a density with an equal value as that of the prototype material cannot be found for the scale magnitudes that are useful. In case of composite structures, the stiffness of the structure as a whole however appears to be linear with the plate thickness (except for the local bending stiffness of plates and solid girders), so that, at the sacrifice of the geometric reproduction, the stiffness may be corrected by selecting another plate thickness. This then results in: nd nE = nL n ng nL

(B.28)

So:
nd = n 2 n n g L nE

(B.29)

There is a practical preference for using an E that is too low and therefore a plate thickness that is too great in the model, as the model material may buckle in case of a plate thickness that is too small in relative terms. Up till now models made of synthetic material are used (Trovidur or another thermo-setting pvc), with nE = 60 relative to a steel prototype structure).
195

Delft Hydraulics

According to (B.27) this would result in a choice for the length scale of nL = 60, but in that case the models become so small that, with respect to the Reynolds number and other inaccuracies, this is not permitted. nL = 20 was used in the investigations at the visor gates of the Hagestein Weir and the sector gates of the discharge sluice in the Haringvliet (see also Examples a and c in Paragraph 6.2 of Part A), so that E is a factor 3 too low relative to the ideal scale factor. This was compensated by using a three times greater plate thickness. As the synthetic material used (Trovidur) is five and a half times lighter than steel, the three times greater plate thickness resulted in a mass of the model that was too small. The correction on this was mounted in such a way, that the stiffness was not influenced. For that reason, lead weights were added locally that were mounted and spread out, proportional to the mass shortage. With this procedure, sometimes a little extra mass needs to be added, in order to compensate that the volume of enclosed water is reduced, as a result of which the added water mass will also be too small. This correction however must not be mounted for those parts of the structure that are positioned above the water level. In case of wave impacts the stiffness of the plating is a magnitude that needs to be accurately reproduced in the model. At the scale factor that was presented in (B.29), the plating is much too stiff. In that case it is better to reduce the plate thickness, so that the bending stiffness is correct, and the stiffness of the structure as a whole may be completed using extra stiff girders. Haszpra (1979) has provided another possibility of reducing the bending stiffness of the plating in the model that is too great. If the plating in the model that is too stiff (and geometrically speaking, too thick as well), is replaced by composite plating, consisting of a number of plates similar to Venetian blinds, of which the thickness of each plate corresponds to the thickness of the prototype plating divided by the length scale, then the bending stiffness of the plating matches that of the other stiffness factors again. The objection to this is that the number of plates needs to be a whole number, which limits the choice of scales even further, and that a lot of damping is generated in the plate structure. Conclusion 6: In case of continuous-elastic models that need to meet the requirements of the Froude scale, it is pragmatic to construct the model from synthetic material, in which it is accepted that this leads to a plate thickness in the model that is too great. The scale for the thickness of the plating then becomes:
nd = n 2 n ng / nE L

It may serve as a reminder that in case of a geometric reproduction of the plate thickness, the elasticity module of the model material on the length scale should be reduced. The factor by which the real elasticity of the model material deviates from this, is compensated by correcting the thickness with this factor. If (in case of wave impact investigations) the plating stiffness also needs to be correct, the plating will be made less thick: in that case additional reinforcements need to be mounted in the model.
Scale effects during vibration investigations The scale effects with respect to the chosen scale rules first of all relate to the inaccurate reproduction of the Reynolds and Weber numbers. The influence of this may be quantified on the basis of references concerning drag terms (as a function of Re) and of calculations concerning the curvature of the water surface in case of waves and the behaviour of air bubbles in case of a reduced diameter.

196

Delft Hydraulics

In case of dynamic models it is difficult to meet the requirement of geometric reproduction; the measuring elements and the suspension should not influence the flow and in case of continuous-elastic structures, plating in the model that is too thick is a factor of interference. Elements such as rubber sealing strips that compress inward and spring outward are difficult to reproduce. In case of continuous-elastic models the error in the plating thickness also results in the fact that the bending stiffness of the plating is not indicated accurately. For this, a solution of compromise needs to be found. One point that deserves attention during the design of a continuous-elastic model is that, due to plating that is too thick, the added water mass may become too small. This deviation may often be estimated, after which a mass correction may be applied to the model. If the part of the model concerned emerges above the water during other circumstances, then this correction needs to be removed. If this concerns a pipe in flow or waves in which also water (or oil) flows, then the outer diameter on the length scale will be reduced, as a result of which the inner diameter becomes too small. The flowing fluid inside should then have such a degree of density that the mass per unit of length is correct again. A precondition that is difficult for continuous-elastic models to meet is:
n = 1

In a tunnel model in which the model is constructed from the same materials as the prototype, remains the same, as the damping is a material constant. In case of parts that cause friction, hinges or rubber seals, the damping is not automatically to scale. In case of elastic models constructed with synthetic material, the material damping is too great. Due to the fluid, scale effects may also be expected in the damping. As regards the latter, the damping will be to scale when it is due to the drag force of, for example, girders in turbulent flow, but not when the viscosity is normative (for example: in case of a plate that vibrates in its plane, the drag force depends on the boundary layer development). The damping is normative for the determination of the equilibrium amplitude in case of resonance. In general however, in case of the occurrence of resonance, the design needs to be changed in such a way, that resonance no longer occurs. Even though the damping is not exactly right, the model does give an indication whether resonance occurs or not. Preferably the model damping is lower than the damping in prototype. If no resonance occurs, the damping is much less important, as the response to nonperiodic excitation due to turbulence also includes frequencies outside the resonance frequency. The damping is of secondary importance in case of the response to impulse phenomena, as the maximum amplitude occurs shortly after the impact, so that very little energy dissipation may occur.
Scale effects during wave impact investigations As described in Part B, wave impacts are generated when the water needs to be suddenly slowed down due to the presence of the structure. The amount of movement that is lost in the process is coupled to the impulse value of the load (the magnitude of the force that is integrated over time). The impulse is correctly reproduced in a scale model, in which the Froude number is indicated accurately. The pressure distribution in space and time however, depends on the elastic properties of water, air and structure. Normative is the component that has the least stiffness. The elastic properties are not represented accurately for any of the scale models mentioned above. Especially the local stiffness of the outer plating may only be reproduced

197

Delft Hydraulics

when separate measures have been taken for this. This is difficult to combine with the reproduction of the stiffness of the entire structure and requires a separate investigation (see Conclusion 6). In order to represent the stiffness of air in a model on the Froude scale, the compressibility (and therefore the initial air pressure) should be reproduced on the length scale. This is so difficult to realize that so far it has not been done. Moreover, in case of small air bubbles the surface tension plays a role for the following two reasons: the air pressure in the bubbles deviates from the ambient pressure (as a result of which the compressibility also changes), and the magnitude and the rise velocity of the bubbles is not reproduced (as a result of which the distribution and the size of the bubbles may also deviate). Usually therefore, a global investigation of wave impacts in a continuous-elastic model suffices (response of the structure as a whole) and a detailed investigation of a stiff model, in which the wave pressures that are measured are translated to prototype values through special scale reviews.

198

Delft Hydraulics

APPENDIX III

DESCRIPTION OF A CALCULATION PROGRAM FOR THE DETERMINATION OF THE ADDED WATER MASS FOR A FLAT ROD WITH RECTANGULAR SECTION (STRIP) IN WIDE WATER IN CASE OF TRANSLATORY AND ROTATING VIBRATIONS

199

Delft Hydraulics

200

Delft Hydraulics

APPENDIX III DESCRIPTION OF A CALCULATION PROGRAM FOR THE DETERMINATION OF THE ADDED WATER MASS FOR A FLAT ROD WITH RECTANGULAR SECTION (STRIP) IN WIDE WATER IN CASE OF TRANSLATORY AND ROTATING VIBRATIONS

In Figure 1, Figure C3.3 has been included, in which the grid is represented that was used in the calculation. The calculation program that is described here, relates to a horizontally vibrating strip, of which one quarter of the space is shown. The foundations for the calculation may be found in Part C, Paragraph 3.2. The added water mass of the vibrating strip has an analytical solution, by which the calculation may be verified. At the end of this appendix it is indicated what actions are different in case a rotating strip is reviewed. In that case, calculation results are also presented that are compared with the analytical results.

Figure 1: The used mesh and definition of symbols that were used.
Translatory vibration of the strip (flat rod)

The calculation is based on L VS NS NV = Distance of the mesh points = Vibration velocity of the strip = Number of elements of the strip (= number of mesh lines connected to the strip) = Number of mesh points in vertical direction 201

Delft Hydraulics

= Number of mesh points in horizontal direction NH The potentials of the mesh points are defined as matrix: The vibration velocity of the strip is defined as column: The discharge per strip element is defined as column: The potential values at the strip are defined as column: The (potential/VS) at the strip is defined as column:

(NH, NV) VS(NS) qS(NS) S (NS) S(NS)

For NZ = 1 to Ns it holds that VS(NS) = VS (vibration velocity of strip same at all locations) For NZ = 1 to Ns it holds that for qS (discharge per strip element): qS (N Z )=VSL Initially, for all potential values it holds: (N X ,N Z )=0 The potential values at the strip initially are:

S (N Z )=0
The added mass (per unit of strip length) is calculated using:

mw =

N Z =1

NS

( N Z )L / VS

The accuracy that is required, is defined as (for the explanation see Figure C3.4):
accuracy = mwn mwn1 mwn n

(n= number of times the repeating calculation is carried out)


RA = required accuracy

Before starting the calculation procedure it holds: n mw = 0 (the counter for the amount of calculations) = 0

202

Delft Hydraulics

Now the calculation procedure starts: Repeat again and again: Previous -mw = mw (this therefore is: mwn-1) n=n+1 For NZ = 1 PROCEDURE-5 For NX = 1: For NX = 2 to NH-1: PROCEDURE-6 PROCEDURE-7 For NX = NH: For 1 > Nz Ns PROCEDURE-4 For NX = 1: For NX = 2 to NH -1: PROCEDURE-1 PROCEDURE-8 For NX = NH: For NZ = NS + 1 to NV -1: PROCEDURE-3 For NX = 1: For NX = 2 to NH-1: PROCEDURE-1 PROCEDURE-8 For NX = NH: For Nz = NV: PROCEDURE-2 For NX = 1: For NX = 2 to NH-1: PROCEDURE-10 PROCEDURE-9 For NX = NH: Calculate for NZ = 1 to NS

S (N Z ) = (1,N Z ) + 0.2 qS (N Z )
Calculate
mw =

(see Equation C3.11)

N Z =1

NS

( N Z )L / VS

Calculate
accuracy = mwn mwn1 mwn n

Until

accuracy RA (the required accuracy )


203

Delft Hydraulics

Now each result may be chosen for printing. PROCEDURE-1 through PROCEDURE-10 are presented in Part C, Paragraph 3.2.4 (see Equation C3.12 through C3.24, valid for points of category 1 through 10). As an example, PROCEDURE-1 is chosen (related to the mid area). Without over-relaxation it would hold (Equation C3.13): (NX,NZ) = {(NX-NZ) + (NX,NZ+1) + (NX+1,NZ) + (NX,NZ-1)}/4 With over-relaxation is holds that: previous- = (NX, NZ) (NX,Nz) = {(NX-1,NZ) + (NX,NZ+1) + (NX+1,NZ) + (NX,NZ-1)}/4 (NX,NZ) = previous- + over-relaxation*{(NX,NZ) previous-}
Result calculation example translatory strip

NH = 54, NV = 18 and NS = 6 over-relaxation4 = 1.5 required accuracy = 0.025 The calculation stops after 133 repeated calculations Dimensionless mw Theoretical value of this is /4 The deviation is less than 2% The calculated added length, expressed in a number of times L, is per strip element: element 1 (NZ = 1): element 2: element 3: element 4: element 5: element 6: 5.829 5.676 5.359 4.850 4.089 2.950 = 0.798692 = 0.785398

Theoretically this should result in a quarter of a circle.

A greater over-relaxation leads to a result more quickly, but it appears that if it is greater than 1.7, instabilities occur. 204

Delft Hydraulics

Figure 2: Theoretical and measured mass distribution. Figure 2 shows that the flow that surrounds a sharp edge in case of a discrete network is more difficult, resulting therefore in a greater pressure at that location.
Rotational vibration of the strip (flat rod)

In case of rotation there is no added water mass, but an added polar inertia moment Iw. The potentials at an angular velocity are calculated. It therefore holds that: VS (N Z ) = (N Z - 1 2 )L and:
Iw =
N Z =1

{( N

NS

1 )L}S ( N Z )L / 2

The other actions need no further explanation.


Result calculation example rotating strip

NH = 54, NV = 18 and NS = 6 over-relaxation = 1.5 required accuracy = 0.005 (the calculation is converging so quickly that possibly the area directly outside the strip may not be calculated sufficiently well) 205

Delft Hydraulics

The calculation stops after 37 repeated calculations Dimensionless Iw Theoretical value of this is /32 The deviation is less than 6.5% = 0.10461 = 0.09817

The deviation therefore is greater than in case of a translatory vibration, probably due to the fact that in case of rotation, the movement is greatest at the edge. This edge also generates again and again the greatest deviations of the added water mass (Figure 2). The pressure distribution across the strip elements is: element 1 (NZ = 1): element 2: element 3: element 4: element 5: element 6: 1.439 4.201 6.594 8.317 8.946 7.763

No theoretical value is known for this pressure distribution.

206

Delft Hydraulics

INDEX BY SUBJECT (PART C)

Actively excited model (Paragraph 5.3.2) Added water mass: assessment of frequency range in which this is frequency-independent (Paragraph 5.3.2) calculation with potential theory (Paragraphs 3.2.2 3.2.4) assessment using a schematized flow pattern (Paragraph 3.2.3) calculation examples (Paragraph 3.2.4) Air compression model: Linear air compression model (Paragraph 4.6.1) Non-linear air compression model (Paragraph 4.6.2) Auto-correlation (Paragraph 5.7.3) Bath plug vibration (Paragraphs 3.3.7 and 3.5.3) Calculation methods: related to vibrations (Chapter 3) related to wave impacts (Chapter 4) Calculation program for added water mass (Paragraph 3.2.4 and Appendix II) Cauchy number (Paragraph 5.2) Continuous-elastic model (Paragraphs 5.3.2, 5.4, 5.5 and 6.7 6.8) Coupled systems (Paragraphs 3.3.6 3.3.7 and 3.5.3) Critical points that may occur in dynamic scale models (Paragraph 5.4) Cross-correlation (Paragraph 5.7.3) Damping correction (Paragraph 3.5.2) Data processing: measuring system and data processing (Paragraph 5.6) statistical processing (Paragraph 5.7.2) elaborations in the frequency domain (Paragraph 5.7.4) elaborations in the time domain (Paragraph 5.7.3) Direct method for calculations in the time domain (Paragraph 3.5) Dynamic properties of scale models (Paragraph 5.3.2) Elaboration of measured data (signals) statistical elaborations (Paragraph 5.7.2) elaborations in the time domain (Paragraph 5.7.3) 207

Delft Hydraulics

elaborations in the frequency domain (Paragraph 5.7.4)

Experiences with wave impacts and vibrations in prototype (Paragraph 7.5) Extrapolation (from model to prototype) of wave impacts (Paragraph 4.8) Feedback (in case of wave load) (Paragraph 4.9) Finite Element Method (Paragraph 3.3.5) Floating gates and flap gates (Paragraphs 3.3.7 and 3.5.3) Flow-pressure model (ventilated shock) (Paragraph 4.5) Frequency-dependency (Paragraphs 3.1, 3.2.1 and 3.4.2) Frequency domain (Paragraphs 3.1 and 3.3) Froude number (Paragraph 5.2) General solution of a vibration equation (Paragraph 3.3.1) Geometrical reproduction in a scale model (Paragraph 5.3.1) Impulse, considerations on - (for the calculation of wave impact pressures) (Paragraph 4.2) Impulse-response function (Paragraphs 3.1 and 3.4.1 3.4.2) Indirect method for calculations in the time domain (Paragraph 3.4) Instable fluid oscillations (Paragraphs 3.3.7 and 3.5.3) Linear air compression model (Paragraph 4.6.1) Linear shock wave model (Paragraph 4.3) Linear system (Paragraph 3.1) Local hydraulic head (Chapter 2) Mass number (Paragraph 5.2) Matrix notation (Paragraph 3.3.3) Measuring instruments (Paragraph 5.6.2) Measuring systems and data processing (Paragraphs 5.6 5.7) Modal Analysis (Paragraphs 3.1, 3.3.4 3.3.5 and 3.4.1)

208

Delft Hydraulics

Model: see scale model Model effects (Paragraph 5.2) Mooring (Paragraphs 3.4.2 and 3.5.3) Multiple (degree of freedom) mass spring (scale) model (Paragraphs 5.3.2 and 6.6) Multiple (degree of freedom) systems (Paragraphs 3.3.3 3.3.7) Natural frequency (Paragraph 3.3.1) Natural value theory: see Modal Analysis Non-linear air compression model (Paragraph 4.6.2) Non-linear shock wave model (Paragraph 4.4) Over-relaxation (Paragraph 3.2.4 and Appendix III) Particular solution of a vibration equation (Paragraph 3.3.1) Potential theory (Paragraphs 3.2.2 and 3.2.4) Boundary conditions (Chapter 2) Pressure distribution during a wave impact (Paragraph 4.7) Prevention of (occurrence of) wave impacts (Chapter 2) Probability density function (Paragraph 5.7.2) Prototype experiences (Paragraph 7.5) Registration of measuring data (signals) (Paragraph 5.6.3) Relative damping (Paragraph 3.3.1) Response calculations: frequency domain (Paragraph 3.3) time domain (indirect method) (Paragraph 3.4) time domain (direct method) (Paragraph 3.5) (impulse) response function (Paragraph 3.4 and Appendix II) Response measurements (Paragraph 5.5) Reynolds number (Paragraph 5.2) Rigid model (Paragraphs 5.3.2 and 6.1 6.2)

209

Delft Hydraulics

Rigid wall (in case of a wave impact) (Paragraphs (4.3.1 4.3.2) Scale effects and critical points of scale models: scale effects at vibration models (Paragraph 5.2) scale effects at shock wave models (Paragraph 5.2) other model effects (Paragraph 5.2) Scale model: tunnel model (Paragraphs 5.3 and 6.1) section model (Paragraphs 5.3.1, 6.3 6.4 and 6.6) three-dimensional model (Paragraphs 5.3.1, 6.2, 6.5 and 6.7 6.8) rigid model (Paragraphs 5.3.2 and 6.1 6.2) single (degree of freedom) mass spring model (Paragraphs 5.3.2 and 6.3 6.4) multiple (degree of freedom) mass spring model (Paragraphs 5.3.2 and 6.5 6.6) actively excited model (Paragraph 5.3.2) continuous-elastic model (Paragraphs 5.3.2 and 6.7 6.8) Scale rules (Paragraphs 4.8, 5.2 and Appendix II) Section model (Paragraphs 5.3.2, 6.3 6.4 and 6.6) Sector gates (Paragraphs 3.3.7 and 3.5.3) Ship oscillation (Paragraph 3.4.2) Ship mooring at a fender structure (Paragraphs 3.4.2 and 3.5.3) Shock wave model: linear shock wave model (Paragraph 4.3) non-linear shock wave model (Paragraph 4.4) Single (degree of freedom) mass spring system: dry (Paragraph 3.3.1) in water (Paragraph 3.3.2) Single mass spring (scale) model (Paragraphs 5.3.2 and 6.3 6.4) Spectrum (Paragraph 5.7.4) Stopping criterion (Paragraph 3.2.4) Strategy of an investigation (Paragraph 5.1.2) Strip: translatory (Paragraph 3.2.4 and Appendix III) rotating (Paragraph 3.2.4 and Appendix III) Systems: single (degree of freedom) mass spring system (Paragraphs 3.3.1 3.3.2) double (degree of freedom) mass spring system (Paragraphs 3.3.3 3.3.4) 210

Delft Hydraulics

system with multiple degrees of freedom (Paragraph 3.3.5) coupled systems (structure and fluid elements) (Paragraph 3.5.3)

Three-dimensional model (Paragraphs 5.3.2, 6.2, 6.5 and 6.7 6.8) Time domain (Paragraphs 3.1 and 3.4 3.5) Transfer function (Paragraph 3.5.3 and Appendix I) Tunnel model (Paragraphs 5.3 and 6.1) Ventilated shock (flow-pressure model) (Paragraph 4.5) Verification model technique in case of artificial periodic excitation (Paragraphs 5.5 and 7.1 7.2) in case of flow conditions (Paragraphs 5.5 5.6 and 7.2) Weber number (Paragraph 5.2)

211

Delft Hydraulics

Biographies of the authors


Dr P.A. Kolkman was born in 1932 and graduated from the Delft University of Technology in 1956. He was awarded a doctorate in 1976 on the subject of vibrations in closure mechanisms. He spent his entire career at Delft Hydraulics. In addition, he was a scientific researcher at the Delft University of Technology from 1968 to 1990. Dr Kolkman has been involved in the design of almost all major Dutch navigation locks, river dams and tidal barriers and has also acted as an advisor to large hydraulic engineering projects abroad. He has also been responsible for various leading publications. In particular, he has acquired an international reputation for his work in the area of combating vibration in movable gates and hydraulic valve mechanisms. He is the co-author of a number of books published in English on saline-fresh water separation systems in sea locks and drainage relationships in man-made structures, among other subjects. At this time he is the chief editor of the International Association of Hydraulic Research (IAHR) journal. T.H.G. Jongeling was born in 1947 and completed his studies in 1976 in Applied Mechanics at the Delft University of Technology. He has worked at the HBG and IBBC-TNO and has been with Delft Hydraulics since 1977. Mr. Jongeling specialises in the dynamics of hydraulic structures and has carried out research and consultancy projects in the field of vibration and wave impact, particularly in connection with large sea defence projects in the Netherlands and overseas. He has also contributed to the design of numerous large and small hydraulic engineering projects, as well as writing a number of works in the field of structural vibration. Mr. Jongeling is now also active in the area of irrigation systems. Dienst Weg- en Waterbouwkunde (civil engineering unit of Rijkswaterstaat) Bouwdienst (construction office of Rijkswaterstaat) Delft Hydraulics 1996

212

Potrebbero piacerti anche