The auithor describes the phenomena that: occur in pipe networks when one (ot more) eIEMenigl a eis et Nk cee methods of analysing the flow behaviour are ee A suc eae uel Sa Raut act ade ae Pee cucu hn nu tog Peat amd e Noir ee ene Maumee Whe surge tanks; as well as a variety of pumps. Roce oy huntial Ree auc nse me esas aku un Pe nana) Cem Rune oe sca Pree enceRaa nec guste Sen TNO ee eRe a) Cron a aL GR Ao eu RL PR ae UR CR ued Peetu CC Further, since the theory of the analysis of ee ack pipe networks, the coverage also includes ake uence Va ae Pee een es Although in the past the analysis of unsteady flow has not been attempted by many Rn RU OU a Ae Lad PC eek cl Pee a LP uC LL Ca eR ee een a flap eS ERICA) ParagHydraulic Analysis of Unsteady Flow in Pipe Networks J. A. Reader in Gil Engineering University of Leeds© JA. Fox 1977 All rights reserved. No part of this publication may bbe reproduced or transmitted, in any form or by any means, without permission First published 1977 by ‘THE MACMILLAN PRESS LTD London and Basingstoke Associated companies in New York Dublin ‘Melbourne Johannesburg and Madras ISBN 0 333 19142 0 ‘This book is sold subject to the standard conditions of the Net Book Agreement Text set in 10/11 pt IBM Press Roman, printed by photolithography, and bound in Great Britain at The Pitman Press, BathContents Preface Notation 1 Simple waterhammer theory 1.1 Introduction 1.2. Rigid pipe—incompressible fluid theory 1.3. Sudden valve opening at the downstream end of @ pipeline 14. Slow valve closure 1.5 Distensible pipe—elastic fluid theory 1.6 Instantaneous valve closure 1.7. Separation 1.8. The calculation of the magnitude of the transient caused by complete instantaneous valve closure at the end of a simple pipeline 1.9 Pressure rise caused by instantaneous valve closure 1.10 Sudden valve closure 2 Analytic and graphical methods 2.1 Introduction 2.2. Analytic methods of solution 2.3. Stepwise valve closures at pipe period intervals 2.4 The Allievi interlocking equations 2.5 The Schnyder-Bergeron graphical method 3 Boundary conditions for use with graphical methods 3.1. Introduction 3.2. Pumps 3.3 Four quadrant pump operation 3.4 Surge tanks 3.5. Types of surge tanks 3.6 Transient analysis of surge tanks 3.7. Mass oscillation of surge tanks 3.8 Pressurised surge tanks or air vessels 3.9 Methods of integrating the surge tank equations v ix Boone wow 21 21 23 23 23 25 36 55 55 5S 60 62 63 65 66 68 70vi Contents 4 The method of characteristics 72 4.1. Introduction 72 4.2 Method of deriving the characteristic forms of the water- hammer equations 74 4.3. The characteristic forms of the waterhammer equations 77 4.4 The zone of influence and the domain of dependency B 45 The zone of quiet 19 4.6 The integration of the characteristic equations 79 4.7 Boundary conditions 81 48 The method of the regular rectangular grid 82 49 Other finite difference methods 85 5 Variable parameters in unsteady flow 87 $.1 Variation of wavespeed 87 5.2. Gas evolution 88 5.3. The magnitude of variable wavespeed and the inclusion of gas release 90 5.4 The use of the variable wavespeed equation 94 5.8 Vaporous cavitation 94 $5.6 Calculation of friction 95 5.7 The use of variable f values 96 5.8 Interpolation 97 5.9 The calculation of the free bubble content 98 5.10 Evaluation of velocities and potential heads at internal points in a pipe length 99 6 . Boundary conditions: pumps 100 6.1 Introduction 100 6.2 Pumps equipped with 2 nonreturn valve 100 6.3. The derivation of the pump’s characteristic equation 101 6.4 Dynamometer/turbine operation of a pump with forward flow 102 6.5 Pump efficiency : 105 6.6 Pump power 105 6.7 Pump start up 106 6.8 Pump run down 107 6.9 The in-line pump boundary condition 108 6.10 Suction well pumps 110 6.11 Four quadrant pump operation pat 6.12 The use of the Suter curves 118 6.13 Pump run up to steady speed of pumping 120 6.14 Pumps with by-pass valves 120 6.15 Pump stations 121 6.16 Surge suppression of transients generated by pump trip 124 6.17 Line pack and attenuation 127 6.18 Lock in 12810 Contents Other boundary conditions 7.1. Junctions 72. Joints 73° Air vessels 7.4 The motorised valve 7.8 Servocontrolled valves 7.6 Reservoirs 7.7 Bends Unsteady flow in gas networks 8.1 Introduction 8.2 Basic equations 8.3 Characteristic equations 8.4 The value of 7 8.5 Boundary conditions Impedance methods of pipeline analysis 9.1 Introduction 9.2 The analogy between electrical and hydraulic impedance 9.3 The linearisation of the waterhammer equations 9.4 The solution of the linearised waterhammer equations 9.5 The evaluation of 7 9.6 The impedance concept 9.7 Receiving and sending ends 9.8 The equation of impedance 9.9 Boundary conditions 9.10 The impedance of a network 9.11 Harmonic analysis 9.12 The forcing oscillation 9.13 The oscillating valve 9.14 A network in which resonance can be excited by forcing oscillations located at different points in the network Unsteady flow in open channels 10.1 Introduction 10.2 The equations of unsteady flow in open channels 10.3 The characteristic forms of the open channel equations 10.4 The travelling surge 10.5 The profile of a free surface flow when a travelling surge is present 10.6 The method of analysis of an unsteady free surface flow in which travelling surges are present 10.7 Other methods of analysis vii 130 130 132 134 136 142 144 146 147 147 147 149 150 154 155 155 156 157 160 162 164 165 165 167 170 172 173 174 176 177 177 178 181 192 195 195 199viii Contents 11 Global programming 11.1 Introduction 11.2. The route or link method of global programming 11.3 Pipe description 11.4 Longitudinal profiles 11.5 Upstream reservoirs 11.6 Downstream reservoirs 11.7 Pump description 11.8 Pipe longitudinal profiles at Ax intervals 11.9 Calls of procedures 11.10 Time level scanning References Bibliography Index 201 201 201 203 203 204 204 204 205 205 206 207 209 2uLPreface The reader may be interested to know how this book came to be written. The author has always found the subject of unsteady flow of great interest and throughout his career has studied it with special application. As a con- sequence most of his research effort has been in this area and he has guided many of his Ph.D. students into this topic also. In 1969 an engineer from a local Consulting Engineer's office approached him requesting information concerning surge analysis methods which could take into account variations in wavespeeds caused by free bubbles in the fluid. At that moment in time the author had already developed a computer program which could analyse surge in simple rising mains but had not in- cluded this wavespeed effect. The effect was soon incorporated into the program and it was used to analyse a main which had a history of bursting to decide what was the main cause of the bursts. At the same time, un- known to the author, one of his own ex-Ph.D, students had been employed to take measurements of the pressure history of the main. When the analy- tic results were compared with the measured results it was found that agree- ment was extremely good, the only error of significance being in the timing of the pressure peaks. The actual magnitudes of the maximum and mini- ‘mum pressures were excellently predicted. Upon seeing these results the author and the engineer from the Consult- ing Engineer's office, Bryan Smith, decided to open an office, Hydraulic Analysis Ltd, Leeds, which would routinely undertake the analysis of pro- posed or existing systems, This venture turned out quite successfully and with the passage of time the firm has been called upon to analyse more and more complex systems, ranging from simple rising mains delivering sewage to a sewage works, to undersea oil pipelines such as that of the For- ties field in the North Sea. The firm has been called in to analyse water supply networks for various authorities throughout the world, oil pipelines in the Middle East, most of the pipelines built or proposed for the North Sea, water injection schemes to improve oil delivery from underground strata, pipe networks in Condeeps and other complex networks such as those found in oil refineries and gas liquefaction plants. The firm has had to grow to handle this work, taking in a computer specialist, Andrew Keech, ixx Preface asa partner and employing more staff Throughout this period it has been necessary to develop the original pro- gram and now it has reached a considerable level of sophistication. As an academic, the author feels that the essential material in this pro- gram should be published and so he decided to write this book. It is not possible to include within the confines of one book all of the material that has gone into the program; there are many facilities which have not been included but the main material on which the program is based has been described. The author would like to warn the reader that he has not tried to write the definitive book of watethammer. It probably could not be written at present as the subject is still undergoing rapid development. Even so, this book is an idiosyncratic view of waterhammer and many people who have contributed greatly to the subject may feel slighted by the omission of their material or by the failure even to mention its existence. The author would like to apologise to such people and would plead, in advance, the limitations of space. The book is idiosyncratic in other ways, techniques of finite difference integration such as those due to Lax, Wendroff and coauthors have only been given passing mention and no mention at all has been made of what the author believes is a potential technique for the future — the finite ele- ment method. However, he has demonstrated, to his own satisfaction at least, the complete adequacy of the method of characteristics and offers this as partial justification for his limited presentation of a very large, very complex subject. Leeds, 1976 JAF.Notation Throughout this book, symbols are defined wherever they are used and these are listed below. However, variables of local interests only are not included in this list but are defined in the text. So area of flow area of valve opening at time zero (chapter 1) area of valve opening at time ¢ (chapter 1) area of pipe (chapter 1) area of valve opening pump constant in equation H = AN? + BNQ ~ CQ? (chapter 6) exit area of pump impeller (chapter 6) plan area of a suction well (chapter 6) effective valve area (chapter 9) pump constant in equation H = AN? + BNQ — CQ? (chapter 6) channel surface breadth (chapter 10) mean channel breadth (chapter 10) coefficient of discharge of a valve celerity of a small pressure wave constant in friction formula used in the surge tank analysis (chapter 3) constant in pump equation H = AN? + BNQ - CQ? (chapter 6) coefficient in the stroke equation of a servocentrolled valve (chapter 7) specific heat of gas at constant volume (chapter 8) specific heat of a gas at constant pressure (chapter 8) electrical capacitance/unit length of a transmission line (chapter 9) celerity of a small surface wave (chapter 10) Chezy C (chapter 10) coefficient of discharge of a sluice gate (chapter 10) celerity of a surge wave in an open channel (chapter 10) xixii Notation time increment _ (infinitesimal) distance increment (infinitesimal) pressure increment (infinitesimal) density increment (infinitesimal) velocily increment (infinitesimal) pipe diameter pump impeller diameter constant in the pump efficiency equation depth in a suction well internal diameter of an air vessel Young modulus of elasticity pump efficiency (chapter 3) constant in the pump efficiency equation (chapter 3) internal energy of gas/unit mass (chapter 8) constants in the characteristic forms of the unsteady gas equations as defined in text (chapter 8) g(j-i) (chapter 10) Darcy fin the Darcy formula : np (as defined in text) hoop stress in pipe wall ‘function of" and wave height when wave is travelling downstream (chapter 2) ‘function of? and wave height when wave is travelling upstream (chapter 2) constant in the pump efficiency equation in chapter 8, the force acting upon the fluid/unit fength of pipe frequency of applied head oscillation (chapter 9) Froude number (based on absolute velocity) (chapter 10) intensity of the local gravitational field throughout the text gradient of the pump’s speed ~ time rundown curve (chapter 6) potential head — the sum of local pressure head and elevation of the point above an arbitrary datum head Jost due to friction sometimes static head, sometimes head at point S, according to context head immediately upstream of a valve or nozzle. potential head change caused by momentum changes (note i= whi)a BN ARR nS Notation xiii pressure head of air in an air vessel expressed as height of the equivalent liquid column (chapter 3) atmospheric pressure head height of the base of an air vessel above the pipe centre line a8 hair above (chapter 7) head sensed by a pressure transducer controlling a servo-operated valve critical head at which a servo-operaéed valve will start to move. steady state head (chapter 9) unsteady head component (chapter 9) amplitude of pressure head wave (chapter 9) height of the reservoir surface above the spillway crest (chapter 10) the height of a wave crest above channel bed level (chapter 10) inertia of the rotating parts of a pump and motor set (chapter 6) electrical current (chapter 9) V=T (chapter 9) (see context) channel bed slope — taken positively downwards (chapter 10) frictional head loss/unit weight of fluid/unit length of channel (chapter 10) 2 constant in valve loss formula hp = KS K= a +k (chapter 1 and chapter 5) bulk modulus of liquid constant, sometimes describing local losses, 2 ie. hy = (due to bends, junctions etc) Tn chapter 1 also used: ates constant in head ~ g equation h = kq?, ie. the friction formula used in the Schnyder-Bergeron method (chapter 5) mean height of pipe roughnesses in the Colebrook-White formula pump impeller head loss coefficient (chapter 6) pump volute head loss coefficient (chapter 6) spillway constant (chapter 10) length, usually pipe length L constant in the development of the characteristic form of the differential equations of waterhammer (chapter 7) internal height of an air vessel (chapter 7) electrical inductance/unit length of a transmission line (chapter 9)xiv Notation hydraulic mean radius =4 (defined in text) area ratio “P (chapter 1) Ao running speed of a pump in rev min"! (chapter 3 and 6) polytropic index (chapter 8) pressure generated by momentum change wetted perimeter pump power in chapter 3 and chapter 6 pressure of air in an air vessel (chapter 3) atmospheric pressure pump power in chapter 6 flow flow at time r heat flow/unit area (chapter 8) steady flow rate (chapter 9) unsteady flow component (chapter 9) amplitude of flow oscillation (chapter 9) Reynolds number universal gas constant (chapter 8) hydraulic resistance/unit length (chapter 9) electrical sesistance/unit length of a transmission line (chapter 9) valve stroke at beginning of a At period (chapter 7) valve stroke at end of a Ar period (chapter 7) valve stroke required as defined by a pressure transducer (chapter 7) an integer taking the value of +1 or -I (chapters 7 and 10) pipe period ~~ (chapters 2 and 3) pipe wail thickness (defined in text) torque applied in the pump equation (chapter 3) absolute temperature (chapter 8) time velocity of impeller blade tips (chapter 6) mean flow velocity (chapter 1) velocity in pipe when ¢ > °° (chapter 1) velocity at time zero (chapter 1) velocity at time (chapter 1} air volume in an air vessel (chapter 3) volume of dissolved gasNotation xv velocity of whirl at exit from a pump impeller (chapter 6) relative velocity at exit from a pump impeller (chapter 6) absolute velocity at exit from a pump impeller (chapter 6) velocity of flow at exit from a pump impeller (Chapter 6) electrical voltage (chapter 9) velocity of a surge wave in an open channel (Chapter 10) weight density of fluid dimensionless heat parameter in the head Suter curve (chapter 6) dimensionless torque parameter in the torque Suter curve (chap- ter 6) distance along pipeline distance along pipeline elevation of the pipe centre line above datum elevation of surface in a surge tank above reservoir static surface level (chapter 3) elevation above a datum of a suction well base (Chapter 6) the elevation above datum of the centre line of the pipeline at the point of its junction with an air vessel hydraulic impedance (chapter 9) characteristic impedance (chapter 9) depth of the centroid of a channel cross section (chapter 10) real component of 7 (propagation constant) (chapter 9) constant defining nature of a channel cross section (chapter 10) Cady Jog (chapter 2) imaginary component of 7 (propagation constant) (chapter 9) pump blade angle (chapter 6) propagation constant in chapter 9 ratio of specific heat of gas (chapter 8) ratio of channel cross sectional area/surface breadth (chapter 10) distance increment (finite) time increment (finite) volume increment (with subscripts to define which volume change is intended) (chapter 1) pressure increment due to momentum change (finite) (chapter 1) potential head change: related to Apjby Apj = wAAj (chapter 1) fractional volume of free gas in liquid (chapter 5) ratio of dglag, (chapter 9) steady component of e (chapter 9) the productxvi esne sea poe < Notation unsteady component of ¢ (chapter 9) os the square root of the head ratio: (2) (chapter 2) 0 the slope of the characteristic line ay fractional valve opening: g,- (chapter 2) angle in Suter presentation of four quadrant pump characteris a(N_ 9, 6 = tan ( GB | (chapter 6 Ni | Chapter 6) constant in the characteristic formulation of the waterhammer equa: tions (chapter 4) d B oy a5 qe (Chapter 10) dynamic viscosity of fluid “ kinematic viscosity of fluid = 3.14159 mass density i ic; $22 Allievi characteristic: 37, (chapter 2) surface tension coefficient in chapter 5 viscous sheer stress in chapter 2 and chapter 10 the compound line produced by the summing of two eagres in the Schnyder — Bergeron method (chapter 2) the valve characteristic in the Schnyder — Bergeron method (chapter 2) phase angle (chapter 9) angular velocity of pump impeller (chapter 3 and chapter 6) angular velocity of the applied head oscillation (chapter 9)1 Simple waterhammer theory 1.1 Introduction The hydraulic analysis of flow in networks is usually based upon the con- sideration of steady state conditions. This is due to historic reasons; the analysis of unsteady state is an order of magnitude more difficult than that of steady state and was only possible at all if grossly simplifying as- sumptions were made. Until the relatively recent development of com- puters the only methods available were graphical in type and these could only be applied to simple networks in which the hydraulic controls were of an elementary nature and in which the number of pipes was small. Now that computers are available, a very great improvement has been made to the quality of analytic techniques that can be used and it is no longer necessary to confine the mathematical modelling of a network to that of steady state. The analysis of unsteady state can include steady state as a special case but it yields much more information than this. The behaviour of the system during starting, its run up to steady state and the transient phase that occurs after shut down can all be described with considerable accuracy. It is usually found that the conditions occurring during steady state operation are of only passing interest, what happens during the starting and shut down phases being of much greater impor- tance. The operation of complex hydraulic controls can be simulated and the only limitation upon the size and complexity of the network is the size of the computer store. Waterhammer is the name commonly used for pressure transients. The reason for the name is that when a steep pressure wave front passes through a pipe it generates a sound that resembles the noise that occurs when a pipe is struck by a hammer. In fact, all transients do not generate sound but the name has gained such a wide acceptance that there is no point in trying to change it. Wherever the word waterhammer is used in this text it should be understood to include all pressure transients even if they are not sufficiently steep fronted to cause noise. In the usual Newtonian approach to the analysis of the motion of a body it is usually assumed that a force causes an acceleration which is 12. Hydraulic analysis of unsteady flow in pipe networks simultaneously applied to all particles within the body. In fact, when a force is applied to a body those particles at the point of application of the force are immediately accelerated. The movement of these particles reta- tive to adjacent particles causes forces to be applied to them and they in turn are accelerated. The process then operates upon the next layer of ad- jacent particles and these are accelerated. Eventually all the particles in the body will be accelerated. In effect, a wave of compressive stress has passed through the body and this wave will have propagated at 2 speed that is usually large but not infinite, Most bodies are not sufficiently long in the direction of application of the force for the wave passage time to be in any way significant but the effect is always present. In the case of a long pipeline containing a fluid, the passage of a com- pressional wave through the fluid can take a significant time and the pres- sures caused by the compressional wave may be sufficiently large to burst the pipe. In suck a case an analysis which did not include the effect of such transient behaviour would be of little value. However, if the pipeline were short, and a pressure fluctuation were applied at one end of it over a time which was much greater than the time taken for the compressional wave to traverse the pipe, then an approach which assumed that all fluid particles were being simultaneously accelerated would represent a reasonably accu- rate model of the fluid’s behaviour. ‘Two ways of predicting the behaviour of a fluid colemn when under the action of a force are thus available: (1) Rigid column theory which considers the entire fluid column to be accelerating at the same value throughout its length, the wavespeed being infinitely large. (2) Elastic theory which considers any pressure change to be transmitted through the fluid column at a large, but finite wavespeed. Rigid column theory can only be used if the time of operation of the hy- draulic control is considerably greater than the time taker for a wave of pressure to pass through the fluid column. Elastic theory can always be applied and gives more accurate results but it is usually more complex in nature. 1.2. Rigid pipe—incompressible fluid theory Historically, the development of watechammer theory has followed a pat tern of increasing complexity. It started by using the solid body type analy. sis which is now called ‘rigid pipe—incompressible fluid’ theory. Later, ‘disten- sible pipe—compressible fluid’ theories were developed. The second category of theories has been the basis of most of the work performed recently and can now be considered to be in a high state of development. Rigid column theory can be of considerable value, as situations arise in which pressure transients are not of great interest but in which fluid movements are impor- tant. Rigid column theory is capable of describing such motions moderately accurately.Simple waterhammer theory 3 A simplified form of the dynamic equation First, one of the fundamental equations of waterhammer must be deve- loped. Consider flow through a pipe of length L experiencing a pressure gradient 22 which is decelerating the fluid. Note Pressures are assumed to increase in the direction of x increasing. The t velocity v at time ¢ is assumed to be the same at all points in the pipeline. i ‘The fluid mass contained in the elemental length Ax is pA Ax. The force decelerating the fluid is A 22 Ax neglecting friction. f a 4 pressure grade line * be Figure 1.1 By Newton’s second Law of Motion | op do_ AGE Ax t pd Ax S2=0 so ap, a Bet PG 20 aa This is an extremely simplified version of the Euler equation 1 Bis constant throughout the pipe length, and it is if the pipe is rigid and the fluid incompressible, the equation can be integrated to give dy Ap=~pb, (1.2) where Ap is the pressure difference over the pipe length L which it is neces- sary to produce to generate the acceleration 22 Note that if the downstream pressure exceeds the upstream pressure by the du amount Ap then will be negative, ie. a deceleration.4 Hydraulic analysis of unsteady flow in pipe networks As pressures and heads are related by the expression p=wh= pgh this result can be cast into the form Ah=-2 (12a) This solution is valid for frictionless flow but requires modification if frictional effects are to be included. 1.3 Sudden valve opening at the downstream end of a pipeline As an illustration of how the above simple theory can be used and how frictional effects can be taken into account the case of a sudden down- stream valve opening will next be examined. It suffers from the defects of all rigid pipe~incompressible fluid theories. / length & diameter d Figure 1.2 Ah; in equation 1.2a is the excess of head at the downstream end of the pipe over that at the upstream end. At the downstream end the head is atmospheric as the valve is full open (treated as zero head here) while upstream it is hs so AAj = O—hg if friction were not present, but as friction is present hs must be reduced by the amount of hg so AAj=— hg + he (3), where hg is the friction head. afl 2gd Now he aay This is the Darcy-Weisbach equation (Fanning equation in USA); the f used is one quarter that used in the USA version of the Darcy equation but is the same as that used in the Fanning equation. Therefore if local losses are also included (n, — 4fl2? _ ko? = Ldo Ahj=—{ h, ne ai (* 2ed 2gSimple waterhammer theory 5 so. or (1s) where xa tteag de Aohy— Ko? dt 2r so dv at= 21s solving by partial fractions Pak, a: | Sano 5* Ta + Vie. L ‘yeh: tvKE) = La VigisR '8 (eve cS where v is the velocity at time t. When f becomes infinite /2ghs = VK Vs where V,, denotes the asymptotic velocity as f tends to infinity 2ghs K so Y=, a standard result from normal steady state theory as would be expected. Rearranging equation 1.6 os pel Sigh k 1% Ce } 806 Hydraulic analysis of unsteady flow in pipe networks Rearranging py el2atsKI® aL — 4 E> elaahsk (1.7) Ded This can be plotted as shown in figure 1.3 The dotted line shows how the velocity curve is affected by the elasti- city of fluid and the elasticity of the pipeline wall material and demon- strates how the rigid pipe—incompressible fluid theory predicts the mean velocity changes. It is a commonplace experience that when a tap is sud- denly opened the flow is very fast for an instant then drops and rises to a steady state which is less than that of the first spurt. The high velocity (296,45 Figure 1.3 from the first spurt is approximately equal to the spouting velocity V2ghs and its rapid fall off is due to the inability of the strain energy of the fluid and pipe wall material to maintain this value for very long. A more accurate solution of this problem is described in chapter 2 1.4 Slow valve closure A further example of the use of rigid pipe—incompressible fluid theory is that of slow valve closure. Assume that the general equation for the effective valve area is: de = ao f(t) (1.8) where dy is the full open valve area and f(1) is some function of time. Effec- tive valve area means the actual area multiplied by the coefficient of dis- charge. Also assume that the Bernoulli equation can be applied to the flowSimple waterhammer theory 7 through the valve even though it is in an unsteady state. This assumption is always used and it has been experimentally justified on many occasions. Then Gt = GeV 28h, (1.9) where qe is the flow through the valve at time ¢ and fy is the head immedi- ately upstream of it at this time. Therefore, n= 2 V2ghn, where A is the area of the pipe, so by differentiation with respect to dv, _ dag Vista at 1 rg ayes aby, ae at dt A dt 4 flv? _ Ldn Wed og dt where /tg is the head in the supply reservoir. now Tig = hin = This equation is derived from equation 1.2a. Substituting for $4 and rearranging gives if hy _ nf 2ghy 4flu?) hed ee (t-te) HATO — G19) where This equation can be integrated by finite difference methods. An esti- mate of the maximum head can be made however. Multiply equation 1.10 by f(¢). This gives (fe = (1, ~~ ifLv,? d fe) oy 2) If the maximum head occurs at the time when the valve closes then 2¢= 0 and the frictional term will then have no effect. If the valve closure is of a type which generates maximum head prior to the moment of incomp- lete valve closure then v; will not be zero at the instant when te If friction is important, a finite difference integration of equation 1.10 will have to be undertaken to obtain the maximum head but if friction can be neglected the following analysis can be used to obtain it. Ifa maximum head, i the mathematical sense, occurs anywhere within the closure period (2) 2% will equal zero. Ifthe head rises throughout the valve’s closure and reaches its largest value at the instant of closure without producing a turning point then the expression 1) Sa be zero because the function f(#) must be zero when the valve is closed. So, irrespective of whether a mathematically maximum value is produced at some time within8 — Hydraulic analysis of unsteady flow in pipe networks the valve closure, or a largest value at the end of the closure is produced, setting the expression f(¢) ae to zero will define the largest value, nv 2gh, o= VA (hy tng) Bina 2 (0) ; Fa 2am lo ~ han)? = Aan (5 (FO) } Ry gn? hy a \? (. $0) Let = Imax RP ooo/tt Then Sm yE ype ft yl fy 2 4K ay Asan example, consider a valve closure in which the valve flow area reduces Rearranging gives . . t linearly with time, ie. ae = ao ( -5) whete T's the time of valve closure. Then rozi-t a _t and ag UO=-F ol geen aati) eee Then k? = ———_ =° anne Tigh “T%ehg 2ghi but ae = ug where up is the steady state velocity in the pipeline before the commencement of valve closure. Then so it is simple to evaluate Jimax/Mo from equation 1.11 Simple finite difference techniques can be used to integrate equation 1.10 to obtain the curve of fin ~ t if this should be required.Simple waterhammer theory 9 Consider the linear valve closure mentioned above in which the valve area de = do(1 ~ 1/7) and ignore friction: when = 0, hy = hs (#) ath If the time Tis split up into m increments so dhs The next step of integration can now be performed + (dhe ui teat ar Jat mm his process can be repeated until a sufficient time period has been ex- plored. As the integration scheme is of an initial value type, m must be large and so the process is best performed on a computer. The program required can be written in a very short time and the run time will be very small even when m is made large. ‘As stated already this analytic method is subject to very grave defects if the focus of interest is the magnitude of pressure transients, as it omits all elasticity effects and so only begins to approximate to reality when valve closure times are large, in which case the problem becomes trivial. The method of analysis can be useful when pressure trantsients are not of im- portance, however. Moesaat 1.5 Distensible pipe~clastic fluid theory The remainder of this book will be concerned with elastic theory. The most accurate of the methods, described later, requires the use of computers. The various theories of waterhammer described will be presented in a sequence of increasing complexity and the first three chapters will be de- voted to describing what are considered by the author to be obsolescent theories. The reader would be wise to master these obsolescent theories in the sequence in which they are presented as it is the sequence in which they were developed. He will then be able to assess the improvement in ac- curacy attainable from the use of progressively more rigorous methods.10 Hydraulic analysis of unsteady flow in pipe networks 1.6 Instantaneous valve closure ‘This section is important, as it provides the reader with an insight into the mechanism of wave propagation and reflection in pipelines. Pressure rises generated by very rapid velocity fluctuations are known as transients and have a period equal to four times the pipe length/wave celerity, ie. 4L/c. Instantaneous valve closure is a theoretical concept, as no valve can be closed in zero time but the study of instantaneous closure leads to an understanding which can be used to solve real problems. When a valve at the end of a pipeline is closed in zero time the layer of fluid immediately upstream of it is instantly brought to rest, and its impact upon the valve will cause its pressure to rise. This increase in pres- sure will cause the section of pipe containing the fluid layer to distend and the fluid in the layer to compress. The fluid layer immediately up- stream of the now stationary first layer will next be arrested a very short time later. The time delay is caused by the second layer continuing in motion for a small time while it moves forward to occupy the volume made available by the pipe distension and fluid compression of the first layer. The third layer is brought to rest in the same way as the first and second, its loss of momentum due to its impact upon the second layer causing within it a pressure rise identical to that experienced by the first and second layers. As the first and second layer cannot rebound from the closed valve their pressure cannot fall and is maintained at its initial impact value. Progressively, layer after layer of fluid is brought to rest. The situation is depicted in figure 1.4a. | reservoir distended pipe Z pressure head plot @ Figure 1.4 Eventually, the entire pipe Jength is full of fluid which is at rest but ata pressure head of hj + fs where hj is the head rise caused by the impact, i.e. an inertia head, and hg is the static head of the fluid in the upstreamSimple waterhammer theory 1 reservoir (neglecting local losses). The situation is then as depicted in figure 1.4b The process of impaction of successive layers with the small time delays involved in each layer's impact mentioned earlier is, in effect, the propaga- tion of a wave of pressure hj at a velocity c. The time taken for this wave to traverse the length of the pipe L is L/c. When the wave has traversed the pipe the entire mass of fluid in the pipe is at rest but it is also at a pressure hy + hs, This situation is unstable as the reservoir is at a lower pressure hs. The fluid therefore starts to flow out of the pipe in a direction towards the reservoir. Successive layers of fluid move towards the reservoir at the original velocity v, each layer of fluid expanding and its associated pipe Nreservie distehded pipe ath, pressure head plot wo) Figure 1.4 (continued) section contracting back to its original diameter. Figure 1.4c depicts an intermediate stage in this process. Eventually the reflected wave arrives at the valve. Figure 1.4d depicts the situation that then prevails. The flow circumstance is now exactly the same as that which existed at O except that the flow is now away from the valve instead of towards it. Again this condition is unstable. As soon as it occurs, the fluid at the valve end attempts to leave the closed valve and to move in an upstream direction. It cannot do this, so its momentum is converted into a pressure decrease. The fluid layer next to the valve is brought to rest and its pres- sure is reduced by an amount equal to the original pressure rise, ie. by hy, Successive layers are brought to rest as before but this involves a pres- sure drop as opposed to the original pressure rise. Figure 1.4e depicts an intermediate stage in this process. Eventually the entire pipe length is filled with fluid at rest but with a pressure head of fig ~ A; as shown in figure Lf. Again this situation is unstable because fluid will flow in from the reser- voir at the original velocity v. This will cause the pressure to rise back up12. Hydraulic analysis of unsteady flow in pipe networks beh pressure head plot te) pressure heod plat (a) Bopp 2h | fe) Figure 1.4 (continued)Simple waterhammer theory 13 | a reduced pipe diameter hyo pressure head plot % a Figure 1.4 (contimed) to its original pressure head / and the velocity will revert to its original value of v directed towards the valve. An intermediate stage of this process is depicted in figure 1.4g. The final consequence of this reversal of flow is shown in figure 1.4h. This situation is exactly the same as the initial circumstance so the pro- cess repeats endlessly. In fact, friction sapidly attenuates the transients that have been described so the reflected waves are sequentially reduced in magnitude. In fact, the above description has ignored the effect of friction but this is described later in this section and in section 6.17. Typically, five or six reflected waves of significant magnitude will be seen. In the description of the mechanism of wave formation given above it has been said that a wave reflects completely and negatively at a reser- voir and completely and positively at a closed end (see figures 1.4a, b and c, and figures 1.4e and f.). This means that a wave travelling over a fluid at pressure Ag with a magnitude A/ is reflected at a constant head point (a reservoir) with a magnitude ,~ Ah and at a zero velocity point (a closed end or closed valve) at a head of hs + Ah. This is an automatic consequence of the laws of conservation of energy. In a circumstance in which fluid has pi? ? velocity energy >, Nm/N but no relative strain energy [7 Nm/N, i.e. at v0 = \ cer! pipe donee reduced pe deter feo pressure head plot (9) Figure 1.4 (continued)14 Hydraulic analysis of unsteady flow in pipe networks ‘pressure head plat (ny Figure 1.4 (continued) a closed end, there is a direct conversion of kinetic energy to strain energy, i.e.a positive wave reflection and vice versa at a reservoir. (Note K is the bulk modulus of the fluid.) A major principle has been enunciated: ‘a com- plete positive reflection occurs at a closed end of a pipeline and a complete negative reflection occurs at an open end’. This suggests that partial (posi- tive or negative reflections) occur at ends which are not completely open (constant head) or completely closed (zero velocity), ie. at junctions. This will be more definitively discussed in later chapters. By careful consideration of figures 1.4a~1.4f, pressure plots at various points can be produced. At the downstream end of the pipeline a pressure—time plot can be evolved, for example see figure 1.5a. x me % A @ % z valve closure co) Figure 1.5 Ata point 1’ upstream from the valve the pressure—time diagram is as shown in figure 1.Sb, and at the reservoir end of the pipeline the pressure— time diagram is as shown in figure 1.5cSimple waterhammer theory 15 bona | {4 ) Figure 1.5 (continued) Note that although the wave shape changes greatly as the point of obser- vation of the wave moves upstream there is no attenuation of the wave magnitude. The effect of friction on the wave is, in some ways, surprising. The wave shape at the valve is as illustrated in figure 1.6. This diagram requires explanation. At point A the valve has just closed, the head hi; has been generated because the velocity v has been destroyed. At point B the velocity V has also been destroyed and an inertia head hj has been consequently generated but the wave arriving at the valve at time B was generated by stopping fluid moving at a point /’ up the pipeline. At the instant when this fluid was stopped the pressure head was greater than the original pres- sure at the downstream end by an amount 4/1'v?/(2gd) and the fluid was stopped at a time //c after the valve closure. The abrupt stopping of the flow ata point /’ upstream of the valve thus causes 4 total pressure rise of y+ 4f2'¥v*/(2gd) but this rise arrives at the valve at a time //c later. So, the a ae o Figure 1.5 (continued)16 Hydraulic analysis of unsteady flow in pipe networks 2 2 “tod pressure rise caused by stopping the flow /‘upstream takes a further !e to be transmitted to the valve at wavespeed arriving there at a time 2c after the valve closure. Considering the circumstance at the reservoir end, the liquid stops at a time L/c after the valve closure and the pressure rise is communicated to the valve after a further L/c time interval. The pressure rise is hi + 4fLv?/(2gd). Immediately following this pressure wave there will be a large negative pressure wave. This negative pressure wave will be running through stationary liquid so the pressure will drop from hy + 4fLv*|(2gd) to hj, Due to energy losses in friction the value of v will be less than the original v value so fy will be smaller than the original A, and the friction head 4fLv*/(2gd) will also be smaller than the original value of the friction head. Thus the waves attenuate, (See section 6.17 for further explanatory comments.) 1.7 Separation If a negative wave created by a reflection at the reservoir end of the pipe should attempt to reduce the pressure to a value less than vapour pressure the liquid will boil at the ambient temperature and a hole will appear within the liquid. The pressure will not be able to fall below the vapour pressure. ‘The pressure trace will then appear as illustrated in figure 1.7. Because an equivalent negative transient to the initial positive transient cannot oc- cur (as the pressure is unable to fall below the vapour pressure) the fluid moving away from the valve at a time a little greater than 2L/c cannot be brought to rest quickly. Consequently a long delay occurs whilst the in- adequate pressure difference operates to reverse the flow. The situation may then repeat until the initial transient has been so attenuated as to be unable to reduce the pressure during its negative phase to vapour pressure. Once this occurs the transient behaves in an exactly similar manner to that of any other transient, A similar phenomeon known as gas releaseSimple waterhammer theory 17 absolute zero head Figure 1.7 can simulate a very similar condition to boiling. In water, if the pressure falls below 2.4 m (8ft) absolute (approximately), air bubbles will evolve from the air dissolved in all natural water. These bubbles reduce the rate of reduction of water pressure whereas boiling prevents its reduction be- low vapour pressure absolutely. When the fluid column is at vapour pressure and a hole appears within it the phenomenon is called column separation. 18 The calculation of the magnitude of the transient caused by complete instantaneous valve closure at the end of a simple pipeline Allievi expression Consider a length of pipe Ax long through which a transient pressure of magnitude Ap; passes in time Af, reducing the velocity from v to 2er0 (valve closure case). from equation 1.2. At is the time taken for the transient to traverse the Ax length reducing the velocity by Av and is equal to Ax/c where c is the celerity of the tran- sient wAx | -Av (1.12) This is sometimes known as the Allievi expression but is also variously at- tributed to Moens, Kerteweg and Joukowsky. As the Av in the above expression is the velocity decrement occurring in time A¢ it can be replaced by v if the valve closure is total (but still instant) and then cv Any= 13) : (1.13) The remaining problem is to calculate the value of c.18 — Hydraulic analysis of unsteady flow in pipe networks Wavespeed ‘The magnitude of the wavespeed in part depends on the bulk modulus of the fluid and in part on the distensibility of the pipe. It can be calculated quite simply if the pipe distensibility is definable. A simple case will now be demonstrated and results for other cases will be quoted. Pipe fitted with expansion joints so that it can extend longitudinally with- out generation of longitudinal stress and free to distend diametrally When a pressure Ap; is applied to a pipe it will distend diametrally. The fluid within it will compress. The pipe distension plus fluid compression effects enable it to contain more fluid than it would do in its normal un- pressurised condition. This increase in volume can be calculated as follows. Volume increase due to fluid compression = AV, oA tah where K is the fluid bulk modulus and d is the pipe diameter. The hoop stress fy in the pipe wall = ond where T is the wall thickness of the pipe. The circumferential strain in the pipe wall equals the diametral strain E being Young's Modulus. z (1.14) ‘The increment in pipe radius = 0 x 5 The increment in pipe volume AVpipe = circumference x length x radial increment dil AVpipe = 1dL xn 3 ond? L 1 dpi Vine = 2, SPid so Mine = 3 ype 5 MAL The total volume now made available by pipe distension and fluid com- pression is therefore AViotat and a AViona= SP! x Fah + ap) = 47 ap 24 LAS hart api( z+ X) (15) Remembering that if the wave has not reached a section the fluid willSimple waterhammer theory 19 continue travelling at its original velocity v, then the time taken for the continuing undisturbed flow to occupy this additional volume will be Are AMeowt Ty qae so Api (1, 4 Are fPiz fh, 4 aay ea) ‘This time is the same time as that required for the wave to traverse the pipe compressing the fluid and distending the pipe. From the Allievi equation (1.13) wev Apia T= peu and equation 1.2 ‘Ap, =-eb 42 ‘Pi = —P: ar and as Au=-v aoe peo" so but from equation 1.12 (1.16) If the pipe had been infinitely rigid this result would have reduced to t fF oe20 Hydraulic analysis of unsteady flow in pipe networks so the effect of distensibility can be imagined as reducing the effective bulk modulus of the fluid from K to K ‘where 1_i,d KK TE For the case of a lined tunnel in rock: steel pipe through rock tunnel with concrete infilling between pipe and tunnel where y=f— 2D de de-d? (at) ait +a, ATE, 4dcE, | ImER where external diameter of steel liner external diameter of concrete pipe Es = Young’s modulus of steel Ee = Young’s modulus of concrete Eg = Young’s modulus of rock 1 me Poisson’s ratio for rock T = wall thickness of steel liner. For a plain tunnel in rock For a thick walled pipe aw (di -43) external diameter of pipe internal diameter of pipe 2.03067 x 109 N/m? = 42.336 x 10%Ibf/ft? for water 2.10915 x 10'' N/in? = 44.064 x 10®Ibf/ft? for ste} 9810.N/m? = 62.4 Ibf/ft® for water 32.2ft/s?Simple waterhammer theory 21 1.9 Pressure rise caused by instantaneous valve closure The pressure rise can now be calculated from Allievi’s formula wev Apj= 4pj-— 2 : R(t at (7 w\K TE, When using this formula it is vital to remember that units must be com- pletely consistent, e.g. in SI units, g must be in m/s, w in N/m®, K and E in N/m? and d and T in m. The results obtained predict very large pressure rises if a valve in a pipeline is closed instantaneously, e.g. for a typical steel pipeline of nor- mal dimensions an instantaneous valve closure will generate a pressure head rise of 125 metres of working fluid for every metre per second of velocity destroyed. As some pipelines are now working at high velocities (10 metres per second is not an extreme value) great care must be taken to make either the pipeline extremely strong or ensure that no instantaneous valve closures can occur. 1.10 Sudden valve closure The very large pressure rise created by an instantaneous valve closure can, unfortunately, be generated by valve closures which are far from instan- taneous. (This may not be the largest rise as line pack can produce even higher pressures especially in long pipelines ~ see section 6.17.) Ifa slow valve closure is thought of as occurring in a sequence of small steps of closure, each step occurring instantaneously but separated in time by a small time interval, then each step will generate a small velocity decre- ment Av associated with a small pressure rise Ap. Ap will be given by the Allievi equation and its wave form will resemble that described in figure 1.5a. Each of the steps will produce such a wave but each will start a small time interval after its predecessor. The waves so generated will superimpose upon one another and the pressure at the valve will rise. If the last closing22 Hydraulic analysis of unsteady flow in pipe networks motion of the valve is completed before the first wave’s negative reflection returns to it the sum total of the Ap values of all the waves will exactly equal that produced by an instantaneous valve closure in which the same initial velocity was destroyed. Such a valve closure is called ‘sudden’, The wave’s shape will be different from an instantaneous closure but its peak magnitude will be the same. It is produced if the closure occurs in a time less than the pipe period 2Z/c. If the valve closure is slower than this, reflected expansion waves will be returning while the later steps of vaive closure are still occurring. These Pressure decrements superimposing upon pressure increments still being generated by the continuing valve closure will cause a reduced rate of in- crease of pressure or even a decrease to occur so ensuring that the pressure rise generated by a valve closure which takes longer than 2L/c will produce a peak pressure which is Jess than that generated by a sudden closure. Pipelines are now being built which may be as long as 100 kilometres without any intermediate booster stations. The pipe period for such 2 line could be as great as 200 seconds. The closure of a valve at the end of such a pipeline in a time of 3 minutes 20 seconds might seem slow but in fact it would be fast being a sudden closure and giving rise to transients of maximum magnitude. It will be realised that it is not possible to discuss valve closure rates in terms of being fast or slow without reference to the period of the pipe (2L/c) to which the valve is fitted. Note The pipe period 2L/e must not be confused with the period of oscill- ation of the waterhammer wave 4L/c.2 Analytic and graphical methods 2.1 Introduction It is necessary to discuss the work that has been done in the past on the analysis of transients generated by slow movements of hydraulic controls. As this chapter is still concerned with providing a background to the more modern techniques of analysis that will be presented in later chapters only an outline will be provided 2.2. Analytic methods of solution Two analytical techniques for solving slow valve closures exist. Both of them require the assumption that friction in the pipeline can be ignored. The two methods are completely equivalent although this may not be obvious upon a superficial examination. ‘The assumption that friction can be ignored is extremely dangerous in the hands of an inexperienced analyst. It can lead not only to grossly wrong solutions but to unsafe solutions. Therefore, before either of the following techniques is used, the circumstance to which they are to be applied must be very carefully examined bearing in mind the above comments. 2.3. Stepwise valve closure at pipe period intervals The method is based upon the idea of considering the pressure and velo- city conditions in the pipe at every pipe period (2L/c) interval throughout the valve closure. It is necessary, of course, to know the position of the valve at the end of each of these intervals. The first step of valve closure will not have generated a negative reflection from the reservoir end of the pipe at time zero so it will be dealt with separately. ‘A small time after the first closing movement of the valve has occurred the situation in the pipeline will be as shown in figure 2.1. 2324 Hydraulic analysis of unsteady flow in pipe networks Figure 2.1 There is a wave of magnitude AJ, travelling up the pipeline at velocity c, This wave is classifed as an F wave. Waves travelling down the pipe are classified as f waves. The reason for this terminology will be explained in the next section. Initially, before any vaive closure movement has occurred, the velocity U9 exists throughout the pipe and the prevailing head is hs. Therefore Ap = CA, vy V 28s (2.1) Where ay, is the full open valve area and Cg, is the coefficient of dis- charge. Immediately after the first step of valve closure ay, ‘Av, (he aay" (22) where c is the —s Caay Denoti = v noting fe Pavy 6 Then U = Boh?'® (2.3) ‘Avs and n= (1 + ) (24) but Av =H -% (2.5)Analytic and graphical methods 25 r ) thet (2.6) In the foregoing, the assumption has been made that Bernoulli's equa- tion can be applied at the partly open valve. As stated in section 1.4 this assumption has been shown to work well even in unsteady flow conditions. Having calculated % the magnitude of the F wave: A/y is readily ob- tained from Ah; = ely ~ )/g. After a period 2L/c a further closing valve movement occurs, the B value of the valve becoming Br =Cq, ay, V28/ap However, an f wave will have been reflected from the reservoir at a time L/c after the initial step of closure and will be arriving back at the valve at the time 2,/c just as the valve makes its next closing movement. The magnitude of this F wave will equal —F because it was generated by a negative reflection of the f wave at the reservoir. The head at the valve at time 2Z/e will thus be hg f. The velocity in the pipe will be v, =v — Av. The head at the valve will rise from hs + f tohy+ f+ cl — vig 02 = Ba (ig +S + ov, ~ 02 Vig)? owl ay, = EF and f= —F Av = “fle so dy Up =~ AY, =% Hf v}= ais +s) g & Solving this quadratic gives 7) As f equals ~Ahy,, v2 can be calculated and A/tz can then be evaluated from Ahz = ele ~ Vg The head hy = fis + f+ Akg. The entire process can be repeated until the valve closure has been explored. 2.4 The Allievi interlocking equation Allievit developed the following analytic technique in 1903. It is more complex than the method given in section 2.3 and it is more elegant to the mathematically minded reader but it is no more accurate and in fact the technique described above can be manipulated so as to produce the Allievi interlocking equations quite easily.26 Hydraulic analysis of unsteady flow in pipe networks Before the Allievi interlocking equations can be presented it is neces- sary to develop the differential equations of waterhammer. These two equa- tions are differential forms of the continuity and dynamic equations and as they form the basis of all accurate analytic methods they will be devel- oped here, The continuity equation In figure 2.2 a section of pipe is shown in which a wave is travelling in the upstream direction. The mass of fluid entering the elemental Jength 5 ist time 61 is pAv6t and the mass leaving it in time 61 is le + 553) (4 +5gos) (° + & ba)oe The additional mass that can be accommodated due to fluid compression and pipe distension within the 5x length during the 8 interval is due to the increase in mean pressure that occurs over the f interval and is: @ £ #) (see equation 1.15) pAbx Por (i t The net mass inflow to the 6x length must equal the amount of mass that, the fluid compression and pipe distension can accommodate, so ap aA a0 pAvst ~ 52 5x Aust ~ pu 5” bx6t — pd 5 6x81 ~ pAvbt = pAsx Por( 2+ 4) (2.8) wave moving upstream at velocity ¢ oun Figure 2.2Analytic and graphical methods 27 neglecting second order small quantities, so A030 po Shs 0d Bt 9A B (te (2.9) but (see equation 1.16) so 5 ee oA a (2.10) but oem (: . 2) ~ oBeaye A pipe of circular cross section was assumed when developing equation 1.16 on page 197 s0 A = 5d? @d _ 2u ap ad ax” d Ox ap ine*4) but ae xd Asw(i—2)=p and w 4 = equation 2.10 becomes 8 4g £, 9h yy Oo dx , ax , av. cate Kt qe *ax~? @.1) where po is the density at the origin pressure above which pressure is measured. But & 1 +Fand as K is extremely large in comparison with any prac- tical p value p/po can be accurately approximated to unity. & oh ‘ah a) 41,.4),% 5 Sy tom (M2) K* Te) * x28 Hydraulic analysis of unsteady flow in pipe networks eae but “(e+ #8 an, hav de LY ; so at? ax tg ae Pan"? (2.12) ‘This is the continuity equation of waterhammer in differential form. The dynamic equation A force balance equation can be written for the section of pipe shown in figure 2.2. Force acting L > R = ® 5.) (44 4sx\alp sh (2455) ~ ~wAbx = pa ~(p +22 os) (44 Mx +(n+5 2 (ie Pox ~wAbx a) Q) (3) @) (5) Term (1) is the normal pressure force acting on the left end of the pipe segment. Term (2) is the normal pressure force acting on the right end of the pipe segment, Term (3) is the longitudinal component of the reaction of the mean pressure force from the pipe wall upon the fluid Term (4) is the frictional force opposing flow. Term (5) is the weight component acting along the pipe centreline opposing the flow. Note P in term (4) is the mean wetted perimeter of the pipe segment and 7 is the frictional shear stress between the fluid and the pipe wall. Ignoring second order small quantities, farce acting in the L > R direction is ~A® 5x ~ sP6x ~ wAbx Z x a& This force causes an acceleration of the fluid in the segment according to Newton's Second Law, ~A 2. gx ~2P6x ~wAbx % = ae A'S Ox ~1Pbx ~ wAbx = pA bx (2.13) «. dividing through by Adx and rearranging a a, a Pew Pp (2.14) Now 4/P =m, the hydraulic radius of the pipe, where A = cross sectional area and P= the wetted perimeter of the flow, and # can be expanded by the use of the definition of a total differential derivative, i.e.Analytic and graphical methods 29 dv__ dv, Go’ an a a2 but a a, air 2 twe)t po B+ pe tao Dividing through by w (=p) 2(h+2 42 we, Lov aro ax bw g ax gat From the Darcy-Weisbach formula (Fanning in tsa) fuel pam 2g (This implies the use of a steady-state friction formula.) Now? +2=h: the potential head dk ,2dv, Lav + fil *° ax “gax*gar* @.15a) If the pipe is of circular cross section this becomes ah vay, 10, 2foll_y L- (2.156) ax "gax' gar” ed asm=d/4, Note These two equations, continuity and dynamic, are a pair of quasi-linear hyperbolic partial differential equations and as such cannot be solved ana- lytically. Together they represent the problem of transient propagation in distensible pipes but, as they cannot be solved analytically, various simpli- fications have been made to them in an attempt to obtain an analytic solu- tion. The best known of these simplified theories is that due to Allievi and this will be presented here. Allievi decided to ignore the nonlinear terms and friction. This means that he ignored the v 2 term in the continuity x dh. equation as » 2 is of the order of + and so, in some pipelines, is ax vav 2folul , small. He ignored the 2 fae term and the a term in the dynamic equation. vdv The 25 term is ofthe order of ay 2 ignore the 2h term is to ignore friction itself and this can only be done if frictional head losses are trivial fractions of static heads. This assumption of no friction is critical; without it there is no chance of obtaining an ana- so it is usually small but to30 Hydraulic analysis of unsteady flow in pipe networks lytic solution but with it the value of the analytic solution becomes severely limited. The simplified equations that Allievi used are: av__g dh Sen eA br continuity equation (2.16) ) ih e =-8 on dynamic equation 2.17) By differentiating the first with respect to ¢ and the second with respect tox the wave equation in classic form can be obtained, i.e wv -_g Mh ax dt ~~ e? ar? and av __ th atax ~~ 8 ax? ah @h bn = e bx? (2.18) Gnce the form has been recognised it will be realised that an analytic solu- tion must be possible. The solution is usually ascribed to Riemann, h=hy ea(e+3}+s(-3) (2.19) vew-£[n(e+2)-(-2)| (2.20) (it can be found in any mathematical textbook covering the solution of partial differential equations.) ‘The reader may have noticed that in other texts the solution quoted differs from that given above, It will be found that this is due to the author adopting the convention in this book that x is measured from an upstream origin (the reservoir) and v is assumed positive if the flow is in the direction of x increasing, whereas other authors have chosen their x origin at the downstream end of the pipe (i.e. at the valve) and taken their velocity as positive if it is in a direction of x decreasing. This is mathematically in- consistent so the author has not accepted it. The symbols F and f denote ‘function of’. Clearly P(e + *) and f(t a ) must have the dimensions of head and in fact must represent contributions of head from waves; this will be more fully demonstrated in the next para- graph. Consider the Reimann head equation. It could be plotted on an x base for an instant of time f = fp. For an observer travelling upstream at wave- speed c and located at x = X at time ¢ = fo, the head seen will be hy (see figure 2.3) and hy = hy +F (' + *) , ignoring the fF (+-2)sexm for theAnalytic and graphical methods 31 Figure 2.3 moment, but the observer is travelling, so at time ¢; he will be located at position X, obeying the equation X; = X ~(f,—fo)c. ‘Now, according to the Reimann equation 0 w(t +4) hy, ce but (6 +) =F(a +X i.e. the original unchanged value of F at time f = f9at position x = X. If, to an observer travelling upstream at wavespeed, the F term does not change then the F term can only be a wave travelling upstream at velocity ¢, Similarly it is argued that the f term represents a wave travelling down- stream at speed c. Thus, like all equations, the Reimann head equation is a statement of the obvious. The head at any point on a pipeline x at time ¢ is made up of the static head Ao plus the head contributed at that point and time by a wave travelling up the pipe (F wave) plus the head contributed by any wave travelling down the pipe (f wave). From these equations it is simple to deduce the reflection circumstances at reservoirs and closed ends by mathematical means rather than by the use of engineering insight as in chapter 1. Consider a wave generated at the downstream end of a pipe travelling upstream towards a reservoir. Consider a point on it generated at time L) at the downstream end (x =). The wave magnitude will be F(0 +) As it travels upstream its shape will be unchanged (see before) so at time t and position x rebate) Der When this wave reaches the reservoir at time L/c it will still have the same magnitude and the function will be re + 3) vie. #(%). At the reservoir the é head must remain constant at Ho, so32 Hydraulic analysis of unsteady flow in pipe networks hence re) which is still equal to F(o + 4) equals -é +0) \c} J After a further period ork the f wave which has been generated will ‘2 Ly travel back to the valve arriving at x = at time 2L/c, ie. = ~> ie 7), indicating that the f wave has travelled back to the valve with unchanged magnitude. Thus the initial wave F(o + 4) will travel upstream to the reservoir with unchanged magnitude and shape at which it will be reflected completely and negatively as an f wave, which will then travel downstream to the valve with unchanged magnitude and shape arriving there 2L/e after the first F wave started out. So, in a simple pipeline, f waves arriving at the valve were generated one pipe period earlier as /* waves which were totally and negatively reflected at the reservoir Similarly at a closed end F waves are totally and positively reflected. This can be shown from the Reimann velocity equation. At a closed end the velocity must always be zero, so at such an end v and vp must always be zero Thus ie Al?) =) (221) Summarising: if time periods of a(r 2b the pipe period) are considered and any particular time is denoted by ‘T where ¢ is any appropriate integer number then: for the case of a reflection at an open end (a reservoir) fi=-Fin (2.22) and for the case of a reflection at a closed end fra (2.23) Denoting the F(t + 2) wave at a time iT by Fyand the f(t 2) wave at time iT by fj the Reimann equations can be written as Ay= lo t Pith (2.24) w= —£ (Fi) (2.25)Analytic and graphical methods 33 Considering the sequence of events at a downstream valve caused by its slow closure at pipe period intervals atr=0 Io = ho + Fo + fo (2.26) 2% = % ~£ (Fy fa) Q27) enna hy = hy + Fy + fy (2.28) % =09~£ (Ff) 2.29) atr=2T hg =hot Fy +h, (2.30) V2 = 09 -£ (Fy -f) 31) and so on, but f = 0 as there cannot be an f wave until the initial F wave hhas reflected at the reservoir and returned to the valve, Now (2.32) 2.33) (2.34) and so on, so fo =o + Fo (2.35) X= v9-£ Fo 2.36) hy = hg + Fy — Fo (2.37) 2 = v9 ~£ (Fy + Fo) (2.38) hy = ho + Fy-Fy (2.39) 2 = U9 - £@, +h) (2.40) and so on. Adding successive pairs of head equations and subtracting successive pairs of velocity equations hy + hg = 2h + F, (2.41) hay + hy = 2g + Fy - Fo (2.42) hy + hy = Iho + FF (2.43) etc34 Hydraulic analysis of unsteady flow in pipe networks and (2.44) (2.45) 2.46) etc, generally then (2.47) so (2.48) and the F and f functions have been eliminated from the problem. Nothing more can be done with these equations unless further informa- tion is supplied. If vj and 2} can be specified it is possible to solve these equations sequentially by setting i to 1, 2, 3, etc, in turn. The boundary conditions at the valve can next be introduced and this leads to a solution. Using the usual theory for defining conditions at a valve (see equation 1.9 et seq.) Site ogi (2.49) : or Cait, a, w= at or [eh (2.50) io Bp When i= 0 14 =0%0 igh, apo so i _ Cayavy fui 51) Yq ayy V ho Denote = Sai 8% by a, Taya and Mi bye? 0 nis greek letter eta is greek letter zeta (2.52) (2.53)Analytic and graphical methods 36 Substituting back into equation 2.48 ok hob ha — Iho = "mshi nat) SEoE 2 = 26 (mafia —mitd Denote Yoe Bho = then S14 Sha 2 = 2p (nadia miki) (2.54) ‘The symbols used are those suggested by Allievi in his original paper. The symbol p must not be confused with the symbol used elsewhere to denote the specific mass of a fluid. Here it is called the Allievi pipe characteristic and is equal to = Bee Fehig” Equation 2.54 represents a family of equations obtained by making i= 1,2, 3,ete, successively. They are known as the Allievi interlocking equatior When i= 1 7485 -2= 2ploto- mo) (2.55) fo will be 1 since, up to the instant of valve closure commencement, heads will be steady. no will be 1 if the valve is fully open, so when i= 1 $4 -1= 2p mo) (2.56) If the fractional valve opening 7, is known, {1 can be calculated by the usual quadratic solution. When i= 2 $3 +69 -2 = 2p(mb: ~n2b2) (2.57) $1 has just been calculated m™ and 2 must be specified, i.e. the valve closure pattern must be known, $2 can therefore be calculated. The entire valve closure period can be explored. As the solution from one step is used in the next step, the solutions interlock, hence the name of this technique. Having obtained {;$2{3, etc, heads can be readily calculated = hob t,he = hol 3, hs = hol}, ete and velocities can be similarly acquired: dy =VoM S15 V2 = VoTI2S2, Ys = Voss, etc A complete solution has thus been obtained.36 Hydraulic analysis of unsteady flow in pipe networks This method has been well worked out and techniques of applying it to pipe networks as well as to simple pipes have been evolved. These methods are well presented in books such as Hydraulic Transients by Rich and Engineering Fluid Mechanics by Jaeger. However, the author feels that as it is not possible accurately to include the effects of friction in this method and also believes that it has probably reached the end of its useful life there is little point in using space to describe such developments. 2.5 The Schnyder-Bergeron graphical method ‘The Allievi interlocking equations lead to an elegant sofution of friction- less waterhiammer but become difficult and cumbersome to use if the pipe network is not simple and if the hydraulic controls are in any way com- plex. Essentially the Schnyder”-Bergeron® graphical method solves the same fundamental equations as the Allievi method. The graphical method is much easier to generalise than the Allievi method. It does not involve the calculation of reflection and transmission coefficients if there is more than one pipe in the network and is capable of making a moderately ac- curate allowance for frictional effects in the system. It is able to deal with quite difficult boundary conditions and has been used, by highly skilled practitioners, to great effect. It suffers from a number of defects and one of these is the level of skill needed to deal with anything other than a simple situation, When considering most schemes it is necessary to in- vestigate a large number of modes of operation of the scheme. AS graphical techniques are time construing this can rarely be done. Any graphical technique consists of the following: (1) Graphically representing the boundary conditions present in the network. (2) Graphically representing the equations of waterhammer which de- scribe the conditions within the pipe as the waves traverse it. (3) Linking the events across the ends of two pipes (or more) at which a boundary condition (i.e. a hydraulic control) is present. In any waterhammer problem four variables are involved: head, velocity (or flow), position in the pipe ~x, and time ~t. To represent these four Variables on a graph an ingenious device is used which employs the fact that a transient is propagated at a speed c. In the graphical method this speed is treated as being constant and this is not true in many circum- stances. This is another defect of the graphicat method. In the early parts of this presentation friction will be ignored, but methods of introducing friction will be described later. ‘The waterhammer or eagre lines An eagre is a small wave that travels at constant velocity. In this contextAnalytic and graphical methods 37 the word is used to describe the small incremental waves that continually traverse the pipe during and after the operation of a hydraulic control. Rewriting the Reimann equations: = ho+F(e+8)47(-2) (2.18]* c wetheslea]] eo Rearranging the velocity equation gives feo-0= re +3) ~s(e-4) 2.58) Writing them both for two points at x and x’ and at times ¢ and xt ta=F(0+2)+7(-2) (2.59) hy ho =F (+2) + /(¢2) (260) and Steve =#(¢42)-1(¢-3) es) , colic) fed Stoo -2y'2) =F( + ) se x ) (2.62) Subtracting the head equations gives hag he F(t De -) x puke = F(t!) 9(r' = 2.63 efez)-e-2) ea Subtracting the velocity equations gives ¢ x x Se 2,0 (0+2)-(t-4) Pace pe -F(l + *) +7(0 x ) (2.64) Up to this point these equations are applicable to any value of x, t,x" and ¢’but if it is assumed that x, x’, ¢ and ¢’ are related as follows x=x'+ c(t—f’) then a result of great importance emerges. This relationship is that which would describe the motion of an observer * Square brackets indicate that the equation with this number was introduced earlier,38 Hydraulic analysis of unsteady flow in pipe networks travelling downstream in the direction of x increasing. This observer will be called a waverider from now on. For this waverider pte e Sti yy y's (2.65) ¢ ¢ and : x'tet—et fae) © c Thus for such a waverider travelling downstream F(t + x/a) changes but J(¢~x/a) does not ~ an idea that has been implied previously. The head relationship for such 2 waverider becomes Inge hye! = Fatt! + x'fc) + flt'-x'le) ~ F(r'+ x'le)-f(t'-x'le) = FQt~1'+x'/e)— F(t x'lo) (267) The velocity expression becomes $e dg) = FOL — 0/4 x'/e) —F(C'+ x" Ic) (2.68) Therefore fig hse = lO Pee) (2.69) It must be emphasised that this result is only valid for a waverider moving downstream. For a waverider moving upstream obeying the equation x = x'—e(t ~r') it can be shown by an identical method that c Besse hat = Usyt—Pxt') (2.70) As v= q/Awhere q is the flow and A the cross sectional area of the pipe Gg tat) (2.71) This equation describes two lines of equal but opposite slope in an h~q space. It must always be remembered that the line of positive slope implies a waverider moving up the pipe whilst the line of negative slope implies a waverider moving down the pipe. The line of positive slope is called an eagre I and the line of negative slope is called an eagre IT. Move. ment along either of these lines implies movement along the pipe at velocity cand hence movement in time also. The direction of increasing x in any pipe must be chosen so as to be in the same direction as the initial steady flow.Analytic and graphical methods 39 Boundary conditions The events that occur at the end of a pipeline are defined by the interaction of a wave with whatever hydraulic control is present. Many devices can be present at the end of a pipe, for example, a reservoir, a valve, a pump, a turbine, a surge tank, an air vessel, a dump tank, a junction. At this stage only two simple cases will be considered, i. a reservoir and a valve. (1) The reservoir is a particularly simple device as all it does it to pro- vide flow into the pipe at an (assumed) constant head if located upstream or accept flow from it at a constant head if located downstream. Its graphi- cal representation is therefore a horizontal straight line on an h ~ q plot. (2) Valve. At any instant in the valve closure 4 = Cabiay, 2h (2.72) where and dy. are the fractional valve opening and the full-open valve flow area respectively. Note § docs not have the same meaning as it did when used earlier. 2 a= (4 73) (eie,) /* ee) or h=wq? (2.74) ese where Y *iChay yo (2.75) Equation 2.74 is the equation of a family of parabolae, each member of which is defined by the value of W which in turn depends upon the cur- rent value of B. As a valve closes, 6 decreases, so y increases. B ranges from 1.0 for a fully open valve to zero for a fully closed valve and takes a posi- tive value less than unity for a partial closure. At times T, 27, 3T, etc (Tis the pipe period 2L/c) the value of 6 must be known and hence values of y for each step of closure can be calculated. Thus for every time there will be a parabola which describes all possible h values corresponding to all possible q values for the fractional valve opening then current, The graphical solution For a simple pipeline connecting a reservoir to a valve there will be three elements necessary to define the problem: the reservoir characteristic line, the eagre lines of waterhammer and the valve’s characteristic para- bolae, (Note The word characteristic is not used here to mean the charac- teristic p mentioned in connection with the Allievi interlocking equation.) Yo is the valve characteristic line for the fully open valve. The fact that this line is not a horizontal straight line through the origin shows that the full-open area of the valve is less than that of the pipe cross sectional area. In figure 2.4 the valve is shown as closing fully, in a period of four pipe periods (the subscripts to the y symbol denote how many pipe periods have elapsed since the commencement of closure). The ¥4 parabola has40 Hydraulic analysis of unsteady flow in pipe networks " OL % Gharocteristic A ‘ys the height of the reservoir static surface level above the valve Figure 2.4 degenerated into a vertical straight line because Y4 = °° as = 0 when the valve is closed. At point A, see figures 2.5, the reservoir head equals that behind the valve, so point A defines the steady state flow before valve closure commences. To solve the problem, the eagre lines must next be plotted. Tis the pipeline period (2L,/c). At time 0 the flow through the valve is defined as at Ag. The flow at B (and throughout the pipe length) is the same as that at A at time zero and the head at B is also the same as the head at A as there is no friction in the pipeline. The head and flow at B cannot alter until a wave arrives there so the point for Bp sy is the same as that for By and this in turn is coincident with Ag as shown in figure 2.5. If a waverider is started from B at time 0.57 he will travel down- stream along an eagre IJ. Plotting this line onto the diagram produces a line Bos oA oy, ie. as the waverider leaves B at time 0.57 he will arrive at time 1.07 at the valve. At this instant the valve characteristic y+ 97 comes into existence and the intersection of the valve characteristic and the eagre ‘hy, plan of pipeline Figure 2.5Analytic and graphical methods 4 IL ine defines the 4 ~ q conditions at the point A at time 1.07. Now ima- gine a waverider travelling back up the pipe at wavespeed c travelling from A to B. As he is travelling upstream he will move along an eagre I (positive slope), ie. line Ay 97 ~ By sp, and he will arrive at B at time 1.57 when the eagre I line will intersect the reservoir characteristic so defining the h ~q conditions at B at time i.57. Reversing the waverider will give an eagre IL joining B, sy to Azor; the valve characteristic ¥2,g7 comes into existence at the instant that the waverider arrives so defining the h ~q conditions at A at time 2,07. The diagram can now be completed by exactly similar methods. The diamond shape 83s > Aqor > Bast > As.or can only be produced in the absence of friction and it represents the pendulation of flow that occurs after valve closure is complete. Plotting the head at A against time produces a curve of the type illus. trated in figure 2.6. The prediction of heads at time intervals other than pipe periods To calculate the pressure heads at time periods which are fractions of a pipe period it is necessary to plot additional y curves, As an example, consider the case of the prediction of heads at times of half a pipe period (see figure 2.7). The symbol 7 will be omitted from the subscripts in the following sec- tions, ie. the subscript 1.5 should be read as 1.57. At time 0, heads and flows at A will be defined by the point 4o, the conditions at C will be the same as those at Ao until a wave reaches C at time 0.257 and at B until a time 0.57’ as before, 80, Ao, Co.asy Bo, Boas and Bos will all be coincident. Starting a waverider from B at time 0.57'will give an eagre II line Bo,s ~ Axo $0 Conditions at the valve at time 1.07 will be defined by the intersection of this eagre and the valve characteristic for the time 1.07. A waverider can be started from C at time 0.257 and this will produce an cagre II which will intersect the Yo,s characteristic, so defining the heads and flows at the valve at time 0.57, ie. at Ao.s- ‘A waverider starting off from A at time 1.07 will reach B at time 1.57, ie. at B,. and similarly one starting from A at time 0.57 will arrive at B at time 1.07, i. at By.o. Additionally a waverider starting from A at time 1.07 will arrive at the pipe’s midpoint Cat time 1.257'and one starting from B at time 1.07 will arrive at Cat 1.257 also. The intersection of the42 Hydraulic analysis of unsteady flow in pipe networks = a A Note is the midpcint of the pipe Figure 2.7 two eagres will define conditions at C at time 1.257. The rest of the dia- gram shown in figure 2.7 can then be completed. Thus by inserting the Vo.s, V1.5, Yas, family of valve characteristics the conditions at the valve at mid-period values have been obtained, and the conditions at the pipe midpoint have also been established. Slow valve opening In this case the Yo line coincides with the zero flow ordinate because the valve is initially closed. The valve is assumed to open over a 37" period in figure 2.8. Once it is full open the 3 characteristic is the characteristic for all subsequent times because the valve does not move after it has reached its full open state. At time zero the flow is zero and the head is h, The conditions at B do not change until a wave caused by the first step of valve opening arrives, at time 0.57. Starting a waverider from B at time 0.5T moving towards A gives an eagre I. This eagre intersects they. curve at Ay. Reversing the waverider’s direction produces an eagre I which intersects the reservoir characteristic at Bs. The process can be repeated to complete the diagram. Depending upon the slope of the eagre and the W5,0¢tc line two geometries can be produced as shown in figures 2.9a and 2.9b. Thus if is relatively small (ie. in the case ofa highly distensible pipe & with a low wavespeed and/or a pipe of large cross sectional area) it is pos- sible to produce transients which exceed the reservoir static head.Analytic and graphical methods 43 Slow partial valve closure The closure is assumed to occur over two pipe periods and the Ya line is the W line for the Bp 9 value of the partial opening when the valve has completed its movernent. The 2 line is thus the Ys, Ya, Ws line also. The analysis is performed as usual but the eagre lines eventually spiral in upon the intersection of the Yz line and the reservoir characteristic. Thus, even in the case of a frictionless flow, pressure transients attenuate for the case of a partial valve closure; they also do so for the case of an opening valve, See figure 2.11. If the slope of the eagre line is relatively large, the spiral shape of the eagres may be modified as shown in figure 2.10. Plotting of eagre lines The slope of the eagres is given by (2.76) tana Ag The values so calculated may be very large, i.e. if ¢ = 1000 metre/sec and A=0.5m*. Aso (a (w) Figure 2.944 Hydraulic analysis of unsteady flow in pipe networks Figure 2.12 pipe valeAnalytic and graphical methods 45 As the h~q plot is not to natural scales the value of a calculated is not of great use. The suggested method of obtaining the eagre slopes is illustrated in figure 2.12. If ¢/Ag = 200, say, draw a line from the q = 0.3 point on the abscissa, i.e. point D, to the h = 60 point on the ordinate B, i.e. BD. Complete the rectangle ABCD. Draw in the other diagonal AC. By the use of a parallel rule eagre IT lines can be drawn parallel to BD and eagre I lines can be drawn parallel to AC. ‘The method of dealing with joints Consider a pipeline in which a pipe is joined to a smaller diameter pipe asin figure 2.13. The lengths of BC and CA must be adjusted so as to be in a ratio such that the pipe periods T; and 7; for the two pipes are in a relatively simple ratio to one another, e.g. fF I this ratio is not simple, say 1.42, the graphical analysis will be com- plicated and take a very long time to perform. To illustrate the method, 7, has been assumed to be equal to 27) and arbitrary eagre slopes have been chosen. Note, the eagre slope of the up- stream pipe of larger cross section will be less than that of the downstream pipe. The closure of the valve at the downstream end of pipe II is to take 57> pipe periods so five W characteristic parabolae must be calculated and plotted. (More than this number may be necessary if the 7;/T; ratio is not a simple integer ratio.) ‘The following discussion applies to figure 2.14. The horizontal abscissa of the diagram has been moved up from the true zero head to the h = hig head value to save space and this is a usual practice when performing graphical analyses. Only the portions of lines which lie above the reser- voir line at h = hs need to be plotted. At Ao steady state is defined. Ag also defines conditions at C up until the time at which a wave will arrive there, i.e. Cos. It also defines condi- tions at B up until time 1.575, ie. B,.s- Note, the subscripts express times in multiples of 7; (also remember that + 7; = 73). ‘Thus a waverider starting from C at time 0.573 will arrive at A at time 7, — on the diagram this gives an eagre Il, ie. from Cys to Ayo. Starting from B at 0.5 7; will give an eagre II arriving at C at time 1.57, and starting a waverider from A at time 1.07; will give an eagre I arriving at Cat time 1.575. Thus the intersection of these two eagres defines con- ditions at C at time 1.57}. Having obtained C,,, two waveriders can be started, one travelling up- stream and one downstream. The downstream eagre will be an eagre II and the upstream eagre will be an eagre I. The downstream waverider will arrive at A at time 2.07 and the associated eagre will intersect the 2,0 line at ag. The upstream eagre | will intersect the reservoir characteristic at Bas. rod =1or2or3 or5 org.46 Hydraulic analysis of unsteady flow in pipe networks The subscripts insicate multiples of Tp Figure 2.14 A waverider starting from Bat time 1.57; wilt have an associated eagre ILand he will arrive at C at time 2.57. A waverider starting from A at time 2.07, will move along an eagre I and will arrive at Cat time 2.575. The intersection of these two eagres will define Cas As Bo, Bos, Bi.o and B,,s are coincident, eagres Bos > Cy,s and Bys > Cos are colinear over part of their length. ‘After the time of 1.57,the origin of any waverider starting from B will no longer be at the steady-state point but it will always be found that new starting points for B will have been established from earlier stages in the analysis, ¢.g. Bos, Bio, and By,s are coincident at the steady state point but By. will have been established before it becomes necessary to use it. From G5, Ago and Bas can be located. From Agq and Bys the point Cas can be obtained and the rest of the diagram can be completed similarly. The method of dealing with junctions At junction Cin figure 2.15 the equation of continuity must always be true, ie. N=43— a CC’ denotes conditions at the junction end of pipe 1. C? denotes conditions at the junction end of pipe 2. C? denotes conditions at the junction end of pipe 3. At the junction, neglecting local losses and kinetic energy changes hep = Neg = hep 7 i.e. the heads at the junction ends of pipes joining at the junction at any given time ¢ must be equal.Analytic and graphical methods a7 Figure 2.15 (Note that the indices 1, 2 and 3 define locations, not the first, second and third power; the subscript ¢ denotes time.) The method of analysing junctions given below is based on the method of analysing surge tanks given by Hawkins and Zienkiewicz*. Consider figure 2.15 and let the pipe periods of pipes 1, 2, and 3 be T', T? and T°. On an h ~ q graph (figure 2.16) assume that conditions at A at time ¢ 0.57", at B at time ¢~0.57? and at D at time ¢ — 0.57 are known and can be located at Atos", Bt-os1? and Dist. Through these points draw the eagre lines appropriate to waveriders traveling towards the junction C. Now, at C? the head must equal that atc, Figure 2.16 The flow into pipe 1, i.e. g; must equal g3—q> and the head at C’, hi, must equal h? and also equal A. It is therefore possible to draw a line upon the A ~q graph which is a combination of the two eagres, eagre Il and eagre Ip. Point X on eagre II and Point Y on eagre I; have the same head so the requirement that Mc* = he? is fulfilled at all points on line PQ. The abscissa of the required line must be given by qx ~ dy so if PS is set off along line PQ and PS = XY, S will define one point on the required line obtained from eagre II, and eagre I;. This process can be repeated for another pair of points, e.g. E and G and so point 7'can be obtained (RT = EG). The required line is thus ST; it is called a y line (Greek letter