Академический Документы
Профессиональный Документы
Культура Документы
By
CARY Z. WATTERS, PE
Professor of Civil and Environmental Engineering
Utah State University
Logan, Utah
Those who write books are often criticized by the practicing engineer
for lack of detailed and directly useful information or techniques,
unrealistic levels of mathematics and theoretical manipulations, and
general disregard for the direct application to engineering practice. This
book is an attempt to overcome these objections, dissipate some of the
mystique surrounding unsteady flow analysis, and produce material which
can be of direct practical use to the professional engineer. Regrettably
unsteady flow phenomena and water hammer, in particular, are rather
complex and difficult to understand and analyze. This treatment will
make the phenomena as easy as possible to grasp without seriously
restricting the range of application of the resulting analysis. Although a
certain level of sophistication is necessary, nothing is required beyond that
knowledge usually associated with the Bachelor’s degree of the last 15
years.
This book is not a comprehensive treatise on unsteady flow nor is it
written for the sophisticated analyst or practitioner. It does not cover a
number of topics which could rightfully have been included. Rather, it
addresses mainly those types of unsteady flow problems faced by the
Copyright © 1979 by Ann Arbor Science Publishers, Inc. engineer designing pipelines which convey liquids, mostly water. Within
230 Collingwood, P. 0. Box 1425, Ann Arbor, Michigan 48106 this constraint most of the topics relate to water hammer analysis.
Hopefully, this narrow scope and more detailed development will make it
Library of Congress Catalog Card No. 77-92597 more valuable to the practicing engineer.
ISBN 0-250-40228-9 The only technique of analysis used for solving water hammer
problems is the method of characteristics. It is relatively easy to apply,
Manufactured in the United States of America accurate, inexpensive and reliable; however, it does require a moderate
All Rights Reserved knowledge of FORTRAN programming. Other methods of analysis are
not addressed, because in most cases the method of characteristics holds
every advantage. Those encountering problems requiring special
techniques can turn to other reference works.
The book builds on basic hydraulic theory to give the engineer
self-confidence in solving his own unsteady flow problems. Beginning with
relatively comfortable concepts and problems, the material becomes more
comprehensive as the topic is developed. Most common types of pipeline
configurations and appurtenances which can cause water hammer
problems are discussed and examples of computer analysis are included
regularly. However, it is impossible to anticipate the variety of pipeline
problems to be encountered. Hence, the reader is led to the point of being
competent to solve his own problems, using the book and its programs as a
starting point.
Many problems can be solved on very small computers because the
storage requirements are not large for most programs. Virtually all the
programs included can be converted to BASIC, which is available on most
mini-computers. Because some level of digital computing is available to
virtually all engineering firms, it seems only reasonable to rely on these
accurate, economical, timing-saving devices.
Included are listings of many computer programs ranging from Acknowledgments
simple to rather complicated ones. While all of them have been used to
some extent to solve problems, the author accepts no responsibility for
their correctness. The results obtained from any type of computer analysis
must always be examined carefully. Too often it is a failing of the engineer The author is indebted to many for their contributions to the
to have too much faith in the results of a numerical analysis, simply publication of this book. To Drs. Victor Streeter and E. Benjamin Wylie
because the numbers appear on paper with holes on each side. who, unknown to them, first aroused in me a serious interest in unsteady
To assist the potential user who does not wish to produce the decks flow phenomena, I am most appreciative. To Dean E. J. Middlebrooks, I
himself, the author will provide the service of reproducing and mailing any owe thanks for the encouragement to begin the project. To Professor F.
or all of the programs listed in the book. This service will be provided at a M. Wood of Queens University, I am obliged for the information on water
nominal charge to cover the cost of reproducing, handling and mailing. hammer history. To those individuals and agencies, who shall remain
Interested persons should write for price quotations. unnamed, whose work was pushed aside to permit expeditious completion
The portion of the book on the steady state analysis of pipe networks of this book, I am grateful for their forebearance. To the Civil Engineering
is based on R. W. Jeppson’s “Analysis of Flow in Pipe Networks” also Department and Genevieve Fonnesbeck, I am appreciative of their
published by Ann Arbor Science. Some modifications have been made in support in preparing the draft manuscript. Many thanks are due to
some of Jeppson’s network analysis programs, particularly where Barbara South for preparing the final document. Dr. J. Paul Tullis
interfacing with a water hammer program is required. reviewed the manuscript and made several suggestions for clarification
As a college text, this work is consistent in mathematical and improvement which enhanced the accuracy and readability.
sophistication and engineering content with an advanced undergraduate
or graduate level course. It is presently used in a two-quarter course
sequence at Utah State University attended by both senior and graduate
students. For the engineer in practice, it should be sufficiently detailed in
content to provide a surmountable self-study course in unsteady flow.
In preparing material for this text, the writer was struck with the
absence of textbooks and references dedicated to the engineer initially
attempting to grasp this subject and use the technology. Regrettably, too
much of the state of the art of solving useful engineering problems resides
in proprietary isolation inside consulting engineering firms. Hopefully, as
the technology continues to mature, all engineers will contribute to the
common pool of knowledge their expertise and techniques in analyzing
and controlling unsteady flows.
Gary Z. Watters
Logan, Utah
vi
....
MW
4.1 The AH Equation ................................................
--1
4.2 ..................................................................
4.3 The a Equation ......................................................
GJ —1
Wave Speeds in Other Types of Conduits ....
4.4 ..............
Basic Differential Equations of Unsteady Flow
Solution by Method of Characteristics ......................................47
-U
5.1 Approximate Method of Characteristics ...............
5.2 Complete Method of Characteristics ....................
O'
M -q
5.3 Some Effects of Parameters on Solution Results ............
-q
viii ix
...............................
...............................
Table of Contents (Continued)
Chapter Page
Ox
Wave Transmission and Reflection at Pipe Junctions .. 153
2-2 .................... 7
VII Pumps In Pipelines 157 Pressure wave propagation in a simple pipe system....
2-3
7.1 Theory and Operation of Pumps ....................................... Pressure head vs. time at three locations along the pipe ............
7.2 The Pump in the Pipeline .................................................. 2- 4Definition sketch for unsteady flow equation deviation 11
..........
s© Ox O> O>
’e
7.3 Pump Characteristics and Power Failure Rundown... ............
7.4 Selective Pump Tripout .................................................... 3-1 13
Simple system for applying rigid water column theory
o
7.5 Pump Power Failure With Backflow and Windmilling
7.6 Selective Pump Startup ..................................................... 4- 1 38
The effect of wall thickness on C-value for p — 0.30 ..
r-
5- 1 50
VIII Problems Resulting From Unsteady Flow in Pipelines ........................ 199 The s-t plane for the simple pipe of Chapter II ................
8.1 Results of Problems............................................................ 199 5-2
The s-t plane showing characteristics for Equations 50
8.2 Causes of Problems ............................................................ 200
5-8 and 5-9 ........................................................................
IX Control Devices and Techniques .............................................. 205 5-3 52
... The characteristic grid for a typical pipe ..........................
5-4 53
9.1 Valves .................. .
Disturbance propagation on the s-t plane..........................
9.2 Air in Lines ..............
5-5
9.3 Pump Power Failure ....................................................... Computer program for approximate water hammer 59
in a simple pipe .................................................................
References ......................... 68
Appendix I ........................ 5-5
Interpolation of H and V values on a As - At grid ...........
................................................................................ 5-7
Appendix II .................................................................................. 241 Computer program for the complete method of 72
characteristics ...................................................................
Index ............................................................................................ 247
5-8
.........
Model system for showing sensitivity to system .73
parameters ........................................................................
xi
X
List of Figures (Continued) List of Figures (Continued)
Figure Page
Figure Page
5- 9 The progression of frictional effects in a simple pipe 6-12 KL and 1/KL as functions of percent open ............................................108
with sudden valve closure ......................................................... . .. .75
6-13 Linear interpolation computer code ............................................... 110
5- 10 Numerical instability in a simple pipe with sudden valve
closure ........................................................................... 78 6-14 Definition sketch for parabolic interpolation ................................. 110
5- 11 Results of water hammer tests on copper tubing (from 6-15 Parabolic interpolation computer code ........................................... 112
“Water Hammer Analysis Including Fluid Friction” by
Streeter and Lai, ASCE Hydr. Div., 88:3, Nov. 1962) ................... 79 6-16 Water hammer resulting from various gate valve closure
schemes ........................................................................................ 113
5- 12 Results of tests on the oil carrying South European Pipeline
(from “Computation of Oil Pipeline Transients” by Kaplan, 6-17 Definition sketch for the pressure reducing valve .......................... 115
Streeter and Wylie, ASCE Hydr. Div., 93:6, Nov. 1967) ............. 81
6-18 Computer program for Hardy Cross analysis of pipe
5- 13 Results of water hammer tests on buried and unburied networks ........................................................................................ 119
PVC pipe (from “Water Hammer in PVC and Reinforced
Plastic Pipe” by Watters, Jeppson and Flammer, ASCE 6-19 Linear theory program (from Jeppson [19]) ................................... 123
Hydr. Div., 102:7, July 1976) ........................................................ 82
6-20 The Newton-Raphson technique in one dimension ........................ 129
6- 1 The s-t plane for a two-pipe system with equal N’s ..................... 86
6-21 Pipe network used to illustrate the use of the
6- 2 ................................................................................................... The s-t Newton-Raphson program ........................................... 132
plane with equal At’s and large interpolations ........................................... 86
6-22 Newton-Raphson program (from Jeppson [15]) ...................................134
6- 3 The s-t plane with variable N-value and minimum
interpolation ................................................................................... 87 6-23 Three-pipe junction with valve ...................................................... 141
6- 4 .................................................... Computer program for series pipes 6-24 Water hammer program for networks ............................................ 143
88
6-25 Pipe network with a single pipe extending from the
6- 5 ............................................ The one-in, two-out three-pipe junction network ........................................................................................ 151
97
6-26 Column separation caused by sudden valve closure ...................... 154
6-6 The two-in, one-out three-pipe junction ........................................... 98
6-27 Wave transmission and reflection at a series pipe
6-7 The three-pipe junction with a constant demand
junction ......................................................................................... 155
outflow ........................................................................................... 99
6- 28 Wave transmission and reflection at a tee junction ........................ 156
6-8 The one-in, three-out four-pipe junction ........................................ 100
6-9 A model for interior major losses................................................ 101 7- 1 Schematic diagram of centrifugal pump impeller
showing velocities......................................................................158
6-10 Valve in a constant diameter pipeline ............................................ 104
xiii
6-11 Valve and reservoir at the downstream end of a pipeline .... 105
xii
List of Figures (Continued) List of Figures (Continued)
Figure Page Figure Page
7- 2 Multi-characteristics for a given pump at various 9-4 A schematic sketch of a one-way surge tank ............................... 213
speeds ............................................................................165
9-5 One-way surge tanks in a pumped pipeline ................................. .. .
7- 3 214
Typical Hp/N2 and T/N2 curves for a pump ............................... 168
7- 4 9-6 Schematic diagram of an air chamber and its
Typical parallel pump booster configuration ............................... 169 appurtenances ................................................................217
7- 5 Piecewise linear representation of Hp/N2 vs. Q/N values .... 170 9-7 Propagation of negative wave after pump power failure
with an air chamber at the pump ..................................218
7-6 Pressure wave propagation as a result of pump power
failure .............................................................. ... .........173 9-8 Propagation of negative wave after power failure with an
air chamber and a one-way surge tank ..........................218
7-7 Computer program for source pump rundown as a result
of power failure .............................................................174 9-9 Column separation prevention using an air chamber and
one-way surge tanks ......................................................223
7-8 Pressure extremes resulting from power failure in a booster
pump configuration .......................................................182 9-10 Computer program for analyzing the effects of air
chambers and one-way surge tanks ............................................ 224
7-9 Computer program for booster pump rundown as a result
of power failure ........................................................................... 183
9- ....................................................................................................... 3 xv
Open-end surge tank in a hydroelectric installation xiv .................................... 213
List of Tables LIST OF COMPUTER PROGRAMS
4- 1 Moduli of elasticity and Poisson’s ratio for common pipe I Approximate water hammer program for a simple pipe ................... 59
materials ........................................................................................
Complete characteristics method program for a simple pipe ... 72
5- 1 The effects of friction on maximum water hammer pressures
at the valve for sudden closure (N = 6) ......................................... 74 Water hammer in a series pipe .........................................................88
5-2 Effect of N on pressure head at the valve for rapid and Hardy Cross network analysis - simple version .............................119
slow velocity change (V = 5 fps) ............................................... 76 ........................................................................ 1
'J'X
6- 1 Loss coefficients for a gate valve .................................................. 107 Elementary linear theory - Darcy-Weisbach version .....................
...........................
6-2 1/KL for even increments of percent open ............................................. 109 Newton-Raphson network analysis program ..................................134
Program for booster pump rundown from power failure ............... 183
10 Surge valve in a simple pipe - approximate method
of characteristics ............................................................................ 209
xvi xvii
1
I. Introduction
1
2 UNSTEADY FLOW IN PIPELINES INTRODUCTION 3
solution by arithmetic, graphical or algebraic means. Nonlinear terms water hammer pressure. According to Rouse and Ince [3], Nicolai
were neglected, friction was included by lumping or approximating, or it Joukowsky in 1898 was clearly the first to show that the pressure rise in a
was left out altogether. Matching of boundary conditions at pumps and water line was related to the change in flow velocity, the wave speed and
turbines was, at best, difficult and understood by relatively few engineers. the fluid density. However, Wood [2] states that in a less well known but
Today, modem analysis techniques, including numerical methods of equally important study, J. P. Frizell in 1897 conducted an analysis of the
solving partial differential equations, has brought within reach of most effect of water hammer pressures on speed regulation of a hydroelectric
engineers the capability of solving accurately a wide range of water plant turbine in Ogden, Utah. Apparently, without knowledge of
hammer problems. Although digital computers are needed, they need not European work, he developed his own wave speed and pressure intensity
be large. Undoubtedly in the near future certain simple types of unsteady equations for sudden valve closure. He also noted the effect of branched
flow problems will even be solved on programmable “pocket” calculators. lines and wave reflection including the relationship between gate closure
Accordingly, it is the purpose of this book to make available to the time and wave period.
engineering profession the means of employing both rigid water column At virtually the same time, Nicolai Joukowsky in Moscow published a
theory and elastic theory in solving problems related to the design of report of his analytical and experimental studies of water hammer as it
pipeline systems. Sophisticated and obscure points and developments will related to the Moscow municipal water system. Joukowsky was
not be included. Methods of analysis other than recent computer-oriented well-acquainted with previous work. He derived equations for wave speed
methods will not be addressed. The emphasis will be placed on providing a and pressure increase and considered the problems of pressure wave
readable self-study book with which the engineer can instruct himself on propagation into smaller pipes, wave reflection from open pipes, the
unsteady flow analysis with enough applications and computer programs relationship between gate closure time and wave period, effects of air
included to give him a start in building his own library of programs. chambers, and the use of spring-controlled surge valves. Certainly, in
Because the emphasis in this work is on water hammer analysis, the retrospect, Frizell and Joukowsky would have to share the title of fathers
history of unsteady flow without elastic effects will not be included. Rather of water hammer analysis.
a brief history of water hammer analysis will be presented to give the The next giant to appear in the field of water hammer analysis was
reader a perspective on the evolution of this type of analysis over the last Lorenzo Allievi. According to Rouse et al. [3], in 1913, this Italian
100 years. hydraulician created a mathematical and graphical treatment of water
hammer problems which was the foundation for further developments in
1.2 History of Water Hammer Analysis
the field for the next 50 years. His contribution is too monumental to
review in any detail.
While it is difficult to establish the beginnings of unsteady flow
The first part of the 20th Century was devoted to applying the work of
analysis in pipelines in general, it certainly dates back to early in the 19th
Joukowsky and Allievi to water hammer problems. By a quirk of history,
Century. However, water hammer analysis history is more readily
Frizell’s significant contribution was largely ignored. Much of the work
documented. Some of the earliest work, according to Wood [2], was
seemed directed to the problems associated with hydraulic turbines in
when Wilhelm Weber in the 1850’s measured the effects of pipe wall
hydroelectric plants. Most of the books available today [4, 5, 6, 7] are
elasticity on wave propagation speed. He also developed the continuity
and fluid dynamic equations which were the basis for later analytical largely related to this application.
studies. Wood [2] also states that in 1875 Marey published the results of In the 1930’s, friction was included in the analysis of water hammer
his careful laboratory work which proved wave speed was constant for a problems and the First Symposium on Water Hammer was held in
given situation and depended on pipe elasticity. In 1878, Korteweg Chicago in 1933. Topics covered included high-head penstocks,
considered both pipe and fluid elasticity in developing an equation for compound pipes, surge tanks, centrifugal pump installations with air
wave speed and his equation is essentially that used today. chambers and surge relief valves.
Again from Wood we learn that Jules Michaud first dealt In 1937, the Second Water Hammer Symposium was held in New
experimentally with water hammer in 1878 by using air chambers and York with presentations by both American and European engineers. The
pressure relief valves in pipelines to reduce the effects of sudden gate or leaders in the field were in attendance as papers were presented on air
valve closures. In 1883, Grameka published an analysis showing the chambers, surge valves, water hammer in centrifugal pump lines and
effects of friction but he was unable to solve the equations. effects of friction on turbine governing.
It is less difficult to identify the beginnings of water hammer analysis During this period graphical techniques of analysis thrived under the
wherein fluid and pipe elasticity are important in the computation of work of Allievi, Angus, Bergeron, Schnyder, Wood, Knapp, Paynter, and
Rich. In later years moves were made to more accurately incorporate
4 UNSTEADY FLOW IN PIPELINES
5
6 UNSTEADY FLOW IN PIPELINES FUNDAMENTAL CONCEPTS 7
EL-HGL
EL=HGL
----------- x
V EZ3 t s 2 L/o
1=5/2
t= 1/2 L/o |
•V
Figure 2-1. Steady state flow situation for simple water hammer.
o
The increase in pressure at the valve results in a swelling of the pipe
°I
1
and an increase in the density of the liquid. The amount of pipe stretching
and liquid volume decrease depends on the pipe material and size and the I *= L/° I 3>G ~ .A
liquid elasticity. Generally, for common pipe materials and liquids, the
percentage change is less than 0.5 percent. The deformation has been
Bl'
greatly exaggerated in Figure 2-2 for purposes of illustration. V-0 3
0
The pressure increase propagates upstream at a wave speed of a,
which is determined by the elastic properties of the system and liquid and [7= 7/2 L/o |
the system geometry. The wave speed will remain constant so long as the EL=HGL
above remain constant. Traveling at a speed a, the wave will reach the 0 J
reservoir in a time L/a. At this time the velocity in the pipe is everywhere
I ______
1I
zero, the pressure head is everywhere H + AH, the pipe is stretched and | t- 3/2 L/o~|
the fluid is compressed.
Under these conditions the liquid in the pipe is under a condition of t:4L/0
non-equilibrium because the pressure head in the reservoir is only H. As a -V ' V=O
result, flow begins to occur toward the reservoir as the distended pipe J
ejects liquid in that direction. The reverse velocity is equal in magnitude to J
the initial steady velocity (as a result of neglecting friction) and the source
of liquid for the reverse flow is the liquid previously stored in the stretched
pipe walls as compressed liquid.
This process continues and at time 2 L/a, the pressure has returned
to normal (but with reverse flow occurring) throughout the pipe. However,
there is no source of liquid at the valve to supply the upstream flow hence
the pressure head drops an additional AH to force the reverse velocity to Figure 2-2. Pressure wave propagation In a simple pipe system.
zero. This drop in pressure causes the pipe to shrink and the liquid to
expand.
8 UNSTEADY FLOW IN PIPELINES FUNDAMENTAL CONCEPTS 9
At time 3 L/a this effect has propagated to the reservoir and the
velocity of flow is everywhere zero. However, the pipe pressure head is AH
below that of the reservoir. Consequently, the pipe sucks in liquid from
the reservoir creating a velocity of flow equal to and in the same direction
as the original steady flow. While this is occurring the pressure in the pipe
is also returning to its original value.
Head
After time 4 L/a this wave has reached the valve and at this instant
the flow is identical to its original steady state configuration. This elapsed
1
„ _______________
time constitutes one wave period. As time goes on, this cycle of events will
continue without abatement (in the absence of friction).
Some fundamental concepts can be gained from examining more
Jlo
closely what occurs in this system. For example, it is clear that the time .L. TIME
CD
oir
a
parameter which best describes the sequence of events in a meaningful
fashion is not time alone but the ratio L/a. It is informative to plot the Pressure head vs. time at the valve.
ce
pressure head at various points in the pipeline as a function of time as
shown in Figure 2-3. Note the pressue head as the valve fluctuates between
H ± AH whereas the pressure head at other locations also experiences
periods of time when pressure head is H.
_l|o
_l|o
TIME
(XI
to
oil-
closely. Suppose that instead of closing the valve suddenly, we were to
close it in 10 steps, each increasing the pressure head at the valve by b) Pressure head vs. time at the midpoint.
AH/10. A further requirement would be that the complete closure of the
valve would be accomplished before 2 L/a seconds had elapsed. It is clear
that the pressure head at the valve would still build up to the full AH value
because “relief’ from the reservoir could not arrive before 2 L/a seconds.
The point to be made is that a valve need not be closed suddenly to create
the maximum water hammer pressure. Indeed, any closure time less than
the time necessary for relief to return from a reservoir (a larger pipe may
suffice) will result in full water hammer pressures. In fact, as we will see
later, because of the nature of the way a valve shuts off flow in a pipeline
by creating large head losses, it may be necessary to close the valve in a
time much greater than 2 L/a to prevent high pressures from occurring.
CXI
I'
TIME
oV
CJI
air
a|r
air
r
Unsteady flow in piping systems is a common occurrence. Indeed, c) Pressure head vs. time just inside the pipe at the reservoir.
steady flow is so rare that one might question the advisability of devoting
so much time to a study of its behavior. Virtually, all hydraulic design is
based on steady flow analysis and to a significant degree, the unsteadiness
Figure 2-3. Pressure head vs. time at three locations along the pipe.
10 UNSTEADY FLOW IN PIPELINES FUNDAMENTAL CONCEPTS 11
Substituting the force components and mass from Figure 2-4 into this _j_3p__ 3z _ V2 = dV .............................................................. (2-5)
equation results in .................................................................................................
pAA - [ p + A s') AA - W sin 0 - TAs7rd = — 4^- ....................... (2-2) y 3s 3s D 2g g dt
\ 3s] g at
After some manipulation, we end up with the one-dimensional Euler
equation Recognizing that z is a function only of s and represents the elevation
above some datum of the pipe centerline, we can change the partial
derivative to a total derivative. Finally, the equation has the form
1 dp 3z 1 dv ....................................................
(2-3) . 1 _ X X X = 1X
_4T dt " 7 3s " ds ’ D 2g g dt ..................................................................’
3s 3s yd
Expanding the particle diameter to the size of the pipe cross-section Familiarization with the application of the equation will be gained by
and introducing the average velocity V gives a more useful equation examining several different flow situations.
HI. Rigid Water Column Theory
The unsteady flow equation can be used to solve a wide range of
pipeline problems which fall within the domain of rigid water column
theory. We will begin with some of the simple problems and proceed to
more comprehensive ones.
o
I
1
i
VALVE
+
—
©
L
Figure 3-1. Simple system for applying rigid water column theory.
13
................ ...
In a horizontal constant-diameter pipe, the integration is made quite easy Example 3-1
because (dz/ds) = 0 and V is a function of time only. We also assume the
A horizontal pipe 24 inches in diameter and 10,000 feet long leaves a
f-value in unsteady flow is the same as for a steady flow at a velocity equal reservoir 100 feet below the surface and terminates in a valve. The steady
to the instantaneous value. The result is state friction factor is 0.018 and it is assumed to remain constant during
the acceleration process.
If the valve opens suddenly, calculate how long it will take for the
fi^v2 =—— ............................................. (3-1) velocity to reach 99 percent of its final values. Neglect minor losses.
2gD g dt
Solution
Because the pressure head p t/y = constant = Ho and because p2 /y — 0
LV 2.65 x 10,000 x Vo
for t > 0, the equation is reduced to (3.6) <„ « 2.6S » - 322x,00 ■
TV2
Solving for V0>hf = f^^-
dV ...................................................... (3-2)
era | f <
0.018 x 10,000 xV2
dt 100
= ----------- 27644 ---------- >V0 = 8.46 fps
t99 = 70 sec.
dV ........................................ (3-3)
_ JL. v2
2gD
The following graph illustrates how the velocity approaches its steady
The integration gives the following equation for the time necessary to state value with time.
accelerate the flow to a given velocity V.
...................... (3-4)
y^°-+v
= Vo, the steady state velocity, the
Recognizing that V2g H0D/fL
equation for t becomes
.v
V +V ......................................... (3-5)
0
log
vo-v
Ho
It is important to note that as steady flow is approached, V V o and as a
consequence t -*•<». Of course this answer is unacceptable so we propose
that when V = 0.99 Vo, we have essentially steady flow. With this 10 20 30 40 50 60 70
interpretation, TIME
3.2 Pressure Caused by Valve Closure in a Horizontal Pipe where each of the variables on the right hand side would be evaluated at
time t.
Valve closure can cause some analysis problems beyond those of That there is indeed a limit of applicability to this approach can be
instantaneous valve openings. This possibility is apparent when one seen with Equation 3-7. As faster and faster valve closure times are used,
considers the rapid valve closure which caused the elastic water hammer dV/dt becomes quite large and, in the limit, goes to infinity. According to
problem discussed in an earlier chapter. The difficulty occurring in this Equation 3-7, in the limit p2 /y 00 also. The point at which rigid water
problem is precipitated by the fact that the pressure just upstream of the column theory fails to give acceptable results and a move to elastic theory
valve is no longer zero, but is determined by loss characteristics of the flow is necessary is hard to establish, because it depends on the individual
through the valve. problem and the accuracy in analysis required.
Figure 3-1 can still be used to represent the problem. At t = 0 the
velocity is Vo and the EL - HGL is approximately a straight line between Example 3-2
the reservoir surface and the pipe outlet (neglect minor losses) under
steady flow conditions. Water flows from one reservoir to another through the pipe at a
The differential equation representing this problem is the same as velocity of 10 fps. The shutdown plan calls for a valve closure scheme
Equation 3-1 which will cause the velocity to decrease linearly to zero in 100 seconds.
The valve is located at the center of a 6440-ft long pipeline.
L dV Estimate the maximum and minimum pressures which will occur in
Ho- (3-
g dt 7) the system, locate them and give the time at which they will occur.
Unfortunately, there are two dependent variables so we need another
equation.
The second equation devolves from an energy equation written across
the valve
)2O' ?
STEADY STATE _ > V7
(3-8)
where KL is the valve loss coefficient. Substituting this equation into PIPE MIDPOINT
Equation 3-7 gives
VALVE
3-9)
o
I
V(t+At) = V(t) 2g D g dt
+
18 UNSTEADY FLOW IN PIPELINES RIGID WATER COLUMN THEORY 19
Given that the velocity will decrease linearly with time, Under steady flow conditions, p3/y = 90 ft. At the instant the valve
begins to move, p3/y drops to 80 ft.
■10
- 0.10 ft/sec2
II
100 At the instant of valve closure, p3/y = 70 ft.
To solve the problem, we will consider it in two sections. In summary,
Upstream section
(p3/y)max = 90 ft at steady flow just before valve begins to close
(p3 /ylmin = 70 ft at t = 100 sec
Both sections
+ L dv =0
dt
er
=
(p/y)max HO ft at upstream side of valve at t = 100 sec
a
3220 (-0.10) (p/y)min = 70 ft at downstream side of valve at t = 100 sec
OX)
8 3.3 Unsteady Flow in Series Pipes
= 100 + 100 (0.10) - V2
3220' Engineers are generally confronted with piping systems which are
y® = 110-^-V2 more complex than single constant-diameter pipes. This section illustrates
2gD how to reduce a series pipe to a dynamically equivalent single pipe so that
the previous analysis techniques for single pipes can be used.
The concept of equivalent pipes is familiar to engineers. It is possible
Because we are looking for extreme values of pressure, it is clear that to replace any minor loss or any given pipe with another pipe of any
(P2/y)max = 110 ft when V = 0 which is at the instant of valve closure. convenient diameter. The only concern is that both the actual system and
The minimum pressure occurs at steady flow where p 2 /y = 90 ft. An the equivalent system have the same frictional losses at the given flow rate.
instant after the valve begins to close, p2/y jumps to 100 ft. In applying the equivalent pipe idea to unsteady flow problems, the
In summary, concept of equivalence must be extended to include dynamic behavior as
well as friction. That is, the inertial effects of the actual and equivalent
(p2 /y)max = 110 ft at t = 100 sec systems must be similar.
Using the three-part series pipe below as a general model, we will
(p2/y)min = 90 ft just before valve starts to close determine the relationships necessary to size the equivalent single
diameter pipe shown.
Downstream section
P4 P3
cr
fL
Q
V2 +
+
I
i
I
2g i
D
P3 3220
(-0.10)
3220
Equivalent Pipe
The first requirement is that the frictional head loss between A and B
be the same in each case. The equation is
20 UNSTEADY FLOW IN PIPELINES RIGID WATER COLUMN THEORY 21
= Hr + Hr + Hr
f
'eq 'l *2
we conclude3
3-10
K>
where K is a collection of numerical values including pi and g. Because the In more general form,
discharge in each section of the series pipe is the same at any instant, an
expression for friction loss equivalent follows from the above.
I—
fxL, f3L (3-12)
tl
w
|a
+
D,s
a
5
a
sq
In more general terms for any number of pipes in series, where N is the number of pipes in series. With Equations 3-11 and 3-12
and an arbitrarily picked f-value for the equivalent pipe, it is possible to
z
solve for the length and diameter of the equivalent pipe. Once the
equivalent pipe configuration is found, Equation 3-1 can be used to solve
for pressures.
where N is the number of pipes in series. 3.4 Unsteady Flow in Parallel Pipes
Now considering the dynamic behavior of the flow in the series pipe,
we begin by writing the unsteady flow equations for each section. The development of an equivalent pipe for a parallel pipe system is
similar to that previously done for series pipes. Again, we will use three
L, dV, I dQ pipes and generalize the results to any number of pipes.
- Hf = —1 =
f
i g dt A. dt
h _2 L
2 1 dQ
JI
IC
g dt 2 dt +
on
. 11___________________ 3 ___________________ 1 „
M1 1D
J-
g dt
w
u ______________________________________________________
Adding the three equations together,
CD
Equivalent Pipe
dQ
dt Following the criterion of frictional head loss equivalence,
Now, let us write the unsteady flow equation for the equivalent pipe.
_ PB _ PJ _ Leg d^eq _ Leq J dQ
7 7 feq g dt " g A£q "dt" This relationship leads to the equations
Noting that pa = p^ and Pjj = pg, and that friction loss equivalence
gives K Q2 = K 3-13)
a
<s
tn
22 UNSTEADY FLOW IN PIPELINES RIGID WATER COLUMN THEORY 23
where
Q,+Q2+Q,=Qeq
4-
Substituting Equation 3-13 into the above expression for continuity gives
Dividing out dQeq,
% % % ro/
£ rf’
eq L^eq £ rf
eq T^eqi
Q+ eq
5
- Deq , L
fi i. eq
L
D 5
n
J2L2J
II
O'
4-
eq eq
Example 3*3
Now, addressing the dynamic behavior of the parallel system, we
write a dynamic equation for each pipe A three-unit pumped storage facility is operating in the generating
mode. During emergency shut-down, the wicket gates on the turbines are
dx 1 ±
closed in such a manner that the velocities in the penstocks at the turbines
t"1 OQ I t"*
Aj dt
dt decrease linearly from 60 fps to zero in 30 seconds.
1 dQ
dV
2 2 1 v2
Compute the maximum pressure head at the wicket gates during
dt g A, dt shut-down. Assume f-values are the same for all pipes.
dVa
)—I
J__a
— - —-H dt A, dt
EL 800
bO
7 7 *3
t-1 era
dt
H-
er
a
UNITS
Because of friction loss equivalence, the left hand side of the above L = 2000'
0=8'
T
equations are equal, giving
L dQ L dQ ■ L dQ L dQ n
7
Solution
Aj dt Aj dt A3 dt Agq dt
Writing the equation of continuity in differential form, First, the three parallel pipes are replaced by a single equivalent pipe.
............ ..
24 UNSTEADY FLOW IN PIPELINES RIGID WATER COLUMN THEORY 25
Cu
2000 555 f'L dV ...................................
eq =
II
(3-12)
OO |f
= 7.64 2gD dt
2
242 11.542
Q
eq where
QU
= 0.00296 f' = f + KL ................................................................ (3-18)
(3-11)
11.54s and
Solving these two equations simultaneously gives
dV ................. (3-19)
HJ
OU
D = 13.7 ft and Lpn = 1437 ft
II
dt
I
eq eq
Computing the velocity in the equivalent pipe, It is important not to use the traditional equivalent length method to
represent the minor loss. This technique adds length to the pipe and the
subsequent increase in liquid mass will distort the true dynamic behavior
Now, that we have a single constant diameter pipe, we can use the
of the system.
unsteady flow equation to solve for the maximum pressure.
If the pipe system is complex, the previous techniques must be
fLV2 L dV applied to the individual components of the system. The second method
2g D at discussed earlier is recommended with the technique of computing an f1 to
oo
distribute the minor loss along the entire pipe. After this has been
fLV2 (1437^ 0-61.3 accomplished, the analysis proceeds as before.
400- 2g D y32.2 J 30
fLV2
400+91.2-
2gD
The above equation for p2 /y is good durmg the deceleration process. It is
clear from this equation that the highest pressure will result when V = 0,
hence,
For situations in which the velocity changes suddenly and the pipeline
is relatively long, the elastic properties of the pipe and liquid enter into the
analysis. In Chapter II we saw how a pipeline behaves under the action of
a sudden closed valve. The suddenly closed valve caused an increase in
pressure head AH to occur, which propagated at a speed a. It remains now
to develop means to calculate AH and a and broaden the range of
applications from that of the simple example in Chapter II.
The previously derived and integrated unsteady flow equations
cannot be used because they have not included elastic effects. We will
employ the impulse-momentum equation and the conservation of mass
principle to develop an appropriate set of equations for an impulsive
change in velocity.
>V-AV
>
1
Unsteady Case
28 UNSTEADY FLOW IN PIPELINES ELASTIC THEORY 29
this case it is possible to use a translating coordinate system to transform The side shear force caused by friction will be neglected because its
the unsteady flow into a “steady” flow. size is limited by a very small dL. Also, because we are considering only
If we move our reference system to the left at a speed a we have, for all relatively rigid pipe (steel, concrete, etc.), the pipe bulge will be very small
appearances, a steady flow. and F3 will also be negligible.
Application of Equation 4-2 gives
Fi -F2 = QP(V-AV + a-V-a) = Qp(-AV)
where Qp = (V + a) AP
If the pressure at (1) were p0 then the pressure at (2) would be p0 +
Ap.
Po A’(P0 +Ap)(A + 6A) = (V+a)Ap(-AV)
Steady Flow
Expanding this equation and recognizing that Ap = yAH and dA is very
From basic fluid mechanics we have available the one-dimensional small compared to Ho, AH, A and y, we can neglect the small terms with
impulse-momentum equation the result
SFext = (SQPV)out.(SQpV).n ..............................................(4.J) -AH? A = (V+a)Ap(-AV)
where Q is the discharge, p is the liquid density and Z Fexf is the sum of In slightly different form, this equation can be written
the external forces acting. The momentum correction factor for
nonuniform velocity profiles has been assigned the value of 1.0. AH = -2-AV(V + a)
Considering only the component of this vector equation parallel to the
pipe and noting that momentum enters and leaves the section of pipe dL or
long at only one section each, we can write AH = aAV
(”«■), = :I
' ''-'i ............................................ <«>
In most cases involving rigid pipes (even PVC with a wave speed of
To apply the impulse-momentum equation we must specify a control about 1200 fps), the value of V/a is less than 0.01. Accordingly, Equation
volume and take into account all forces acting on the fluid in the control 4-3 is generally used (and is always used in this text) as
volume at a particular instant and at that same instant evaluate the
momentum fluxes into and out of the control volume. We will choose a
control volume coinciding with the inside of the pipe walls over the length AH = — AV ................................................ ......... (4-4)
g
dL and including the flow cross-section at each end of the pipe section dL
long. This control volume, the fluid in it and the external forces acting are
shown below. It is clear from Equation 4-4 that AH depends on a and cannot be
determined until a value of a is established.
SLA (4-7)
8V = - Ap
where dV is the change in volume of the liquid in the pipe section dL long
as the result of a pressure change of Ap.
Because the increased pressure stretches the pipe, there is more room
made available to store the net mass inflow of liquid. When the pipe
stretches circumferentially it may also stretch longitudinally so both
time = t time = t + dt
contributions to change in pipe volume should be evaluated.
Before proceeding, it is important to recognize that there is an
It is clear that dL and dt are related via the wave speed by interconnection or relationship between pipe wall strains in two
perpendicular directions. If a material is strained in one direction an
dL = adt
amount £t then a strain E2 will occur in the perpendicular direction
Net Mass Inflow (provided the material is free to strain without developing stress in that
direction) such that E2 = p£j where p is Poisson’s ratio. If there is a
During the time period dt an amount of liquid has accumulated in the restriction to strain in either direction caused either by restraint or applied
section of pipe given by the amount stress, the relationship is more complicated. In any case, Timoshenko [11]
gives, for thin-walled pipes,
8M = Mass accumulated = VApdt - (V - AV) (p + 6 p)(A+6A)6t
+
Expanding parentheses and neglecting small terms gives - 1 ( A
---- —
c
E or gj - -g- (Oj -po2) ........................ (4-8a)
8M = ApAVdt 1-p
or writing in terms of wave speed and dL, ej+pej
E or e2 (o2 - poj ....................... (4-8b)
n
1-M
2 E
8M = ApAV—............................................. (4-5)
a where and £j are the stress and strain, respectively, in the direction
This amount of extra liquid is accumulated in section dL by being parallel to the pipe axis and o2 and s^are the values in the circumferential
compressed slightly and by stretching the pipe slightly to provide storage direction. E is the modulus of elasticity of the pipe wall material.
room. In the case of water hammer waves, there is generally a resident stress
and strain already occurring in the pipe before wave passage. Hence, we
Change In Liquid Volume will write the above equations in incremental form
Because the pressure has increased during the passage of the wave,
Ae, + pAgj
the volume of the liquid in the section will compress slightly to a higher Ao, = E or Aej = (AtJj -pAcr2) (4-9a)
density. The equation describing this relationship is that defining the bulk
modulus of elasticity which can be found in any text on fluid mechanics.
Aea + p Aej
K=- (4-6) A<J2 E or Mo.) .............. (4-9b)
dV- IV
V The change in volume caused by circumferential stretching is
where K = the bulk modulus of elasticity of the liquid and p, V are the
6Vc =
o
pressure and volume, respectively. Recognizing that dp ~ Ap (K is
relatively constant over a wide range of pressure), Equation 4-6 becomes where 7r6D = 7rDAe2
32 UNSTEADY FLOW IN PIPELINES ELASTIC THEORY 33
Combining the two equations gives Because a buried pipeline might be expected to be restrained from
strain by soil friction and anchor blocks, we will pursue case (b) restraint
5V = y 7rD28LAe2 .............................................................. (4-10) to develop an equation for wave speed.
Case (b) restraint — For this restraint, Ae 2 = 0 and Equation 4-9a
The change in volume caused by longitudinal stretching is becomes
PAe2
5VL = j D2 SLACj ............................................................... (4-11) A
^i = T77 E =
1 [J.
Adding the two equations together gives the total volume change
5v=
and Equation 4-13 becomes
y D2 8L + Ae2j............................................................... (4-12)
Change in circumferential stress in the pipe wall caused by Ap is ApD Ae2
2e 1-
K>
Substituting this equation into Equation 4-12 gives
ApD ............................ (4-14)
A<72
2e
m
where e is the pipe wall thickness Now considering conservation of mass, we already have Equation 4-5
So Equation 4-9b becomes expressing the amount of mass which has accumulated in the dL pipe
ApD _ Aea + ^Aei section in dt seconds. We can write a different expression for the mass
4-13 change in the dL pipe section after wave passage. The mass change in the
tn
2e 1-u2 section is
Unfortunately, the longitudinal pipe restraint condition determines
8M = (p + 8p) (A8L + 8 V) - p A8 L
Aoj. For example, if the pipe were anchored at some point and free to
stretch longitudinally (much like a long slender pressure vessel), the Equating this expression with Equation 4-5, expanding and dropping
longitudinal stress would be small terms gives
under static conditions. However, the dynamic conditions of a water To arrange this equation in more useable form, note that for a mass of a
hammer situation will cause the pipe to stretch axially in a dynamic given substance, an increase in pressure causes a decrease in volume and
fashion wherein the hardware inertia is important. That is to say, any an increase in density.
valves, fittings, etc. in addition to the weight of the pipe itself must be
displaced by the pressure changes. The pipe may even be partially pV = constant
restrained by supports. In order to determine the value of Ao lt we would V8p + pS’V- = 0
have to solve a rather complex coupled set of equations relating the fluid 5V
O
n
dynamics to the hardware dynamics.
I!
V
Rather than attempt this task, it is suggested that the dynamics of the Substituting Equation 4-6 into this equation gives
pipe be ignored and the above equation for Ao t be used. Because this type
of restraint is rare and because restraint does not have an excessive impact
to
II
Q.
Q.
on wave speeds in typical pipelines, we will not be greatly concerned with
precisely fixing this type of restraint. Replacing Ap with yAH in the above equation, substituting it and
On the other hand, if the pipe were rigidly anchored to prevent axial Equation 4-14 into Equation 4-15
strain then Aoj = pAo2 because A^ = 0. If, however, the pipe had
_______ i
functioning expansion joints throughout its length, then Aoj = 0 and Ac t = Av
7AH a (4-16)
is of no interest. Following Streeter and Wylie [1], we will identify the
above as case (a), (b), and (c) restraints, respectively. In a practical sense, Combining this equation with Equation 4-4 gives
the pipe restraint lies between these values somewhere.
34 UNSTEADY FLOW IN PIPELINES ELASTIC THEORY 35
Table 4-1. Moduli of elasticity and Poisson’s ratio for common pipe
1 materials.
tn
Steel 106 psi 30 x 0.30
W td M td W M
or in a more conventional form for wave speed,
Ductile Cast Iron 106 psi 24 x 0.28
Copper 106 psi 16 x 0.30
[K/p]K Brass 106 psi 15 x 0.34
» ll ll II ll II
(case b) x 10® 0.33
O
tn
(4-17) Aluminum
PVC psi M « 0.45
X
Fiberglass reinforced plastic = 4.0 10s psi = 0.27- 0.30
WWW
(FRP) = 1.3 x 10® = 0.20 0.24
« 3.4 psi - 0.30
3=
Asbestos Cement x 10®
»»
Concrete 57,000 0.30
It is now possible for us to compute wave speed and pressure increase
t
psi
in simple situations where Equation 4-4 can be used. x 10®
Streeter and Wylie [1] have shown that the equation for wave speed where f'c = 28 day strength.
psi
can be more conveniently expressed as
Example 4-1
EL-HGL
where
C = 5/4-p for case (a) restraint ............................................... (4-19a)
200
C = 1-p2 for case (b) restraint ................................................ (4-19b)
-------J- - - - e-a_ A
C = 1.0 for case (c) restraint................................................... (4-19c) Flow in the 24-inch pipeline above occurs at a velocity of 6 fps. The
pipeline is fabricated of steel and has a wall thickness of 0.25 inches,
Recall that this set of equations applies only to thin-walled pipes where a) Calculate the wave speed for all three cases of restraint.
D/e is generally greater than about 40 (see Art. 4.3).
To assist in calculating wave speeds in pipes constructed of common
materials, the following table of E-values and p-values is included. The Case (a) 4720 _______ = 3413 fps
value of K for water can generally be taken as approximately 300,000 psi. _ /i + 3 x 1Q5
- — (5/4-0.30)
In the limit the pipe can become completely rigid without causing the V 3 x 107 -25
wave speed to become infinite. This limiting value is obtained by passing E
to oo in Equation 4-18. With the nominal value of K = 300,000 psi, the
resulting wave speed is approximately 4720 fps. This number has no _______ 4720
practical value in design because it is far too high to serve as even an Case (b) a
approximate wave speed for preliminary design. With even a limited Vl +0.96 (1 -0.302)
amount of experience, the designer can make far better estimates for wave 4720
speed in the pipe he is working with. Case (c) a = 3371 fps
x/1 +0.96 (1.0)
36 UNSTEADY FLOW IN PIPELINES ELASTIC THEORY 37
In a practical sense, the differences are negligible. 4.3 Wave Speeds in Other Types of Conduits
b) Find the head increase resulting from sudden valve closure for all
three cases of restraint. The simplest case of thin-walled pipes has been used previously to
derive equations for wave speed. It is obvious that many hydraulic
3413 conduits are constructed of thick-walled pipe and often using two or more
Case (a) = 636 ft
Ox
X
materials (reinforced concrete). Also tunnels may be carved in rock, lined
with steel and back-filled with concrete. It is necessary to be able to
Case (b) AH = 344° x 6 = 642 ft calculate wave speeds in all these cases.
32.2
A concise summary of the calculation of the wave speeds for these
3371 cases is given by Halliwell [12]. The most obvious extension of the previous
Case (c) AH = 32 2 x 6 = 628 ft example of thin-walled pipes is to thick-walled pipes. In a thick-walled
The variation in head increase among the three cases is about 2 pipe, the wall thickness is so great that stress varies noticeably between the
percent. inner and outer surfaces and this affects the expression for wave speed. An
c) Compute the axial and circumferential pipe wall stresses before analysis reveals that we may continue to use the same basic form for the
and after valve closure for all three cases of restraint. wave speed equation, but we must find a different value for the C in
Equation 4-18.
consider the uncertainties of pipe restraint and its effect on wave speed.
Figure 4-1 shows that uncertainty with respect to the type of restraint oc-
curring can cause differences of about 10 percent between C-values at the
two extremes of restraint. If we accept a similar error in deciding whether
to use thin-walled or thick-walled formulas, then a D/e value of 20 is an
appropriate dividing line. If, however, we decide to remove as much
uncertainty as possible, then the thick-walled formulas should always be
80
used. The additional computation required is negligible. In a practical
sense though, because of the relative size of terms in the denominator of
Equation 4-18, using thick-walled formulas beyond D/e values of 40
generally makes no significant improvement in the value of the wave speed
70
except in cases where softer pipes such as PVC are used. It should also be
noted that using the thin-walled formulas leads to higher (more
conservative) wave speeds. To see the effect, consider the following
Example 4-2
50
reservoirs. The inside diameter of the pipe is 9.522 inches and the wall
thickness is 0.239 inches.
Compute the C-values and wave speeds using both thin and
D/e
thick-walled formulas and compare results.
40
Wave Speed-fps
30
Restraint Thin-walled Thick-walled Error (D/e = 40)
Case (a) 4023 3999 0.6%
Case (b) 4047 4021 0.6%
20
Circular Tunnels
10
lim (D/e) C = — x —
ee = 2(1 + p)
e -> oo
A reinforced concrete pipe, 30 inches inside diameter, is prestressed In this example we will replace the cement lining (prestressed in
using 3/8-inch diameter wrapping wire placed 1.25 inches o.c. The pipe is compression) .by an equivalent area of steel.
42 UNSTEADY FLOW IN PIPELINES ELASTIC THEORY 43
II
used in the standard analysis, further discussion of the analysis of water
Now the thickness of the equivalent steel pipe is hammer with entrained air will be deferred to Chapter VI.
1—
N 1 co
dV .............................................. (4-24)
CPIQ?
dt
When free air occurs in a substantial portion of a pipeline, either as
small bubbles or in larger discrete lumps, the wave speed in the pipe is
decreased. As a consequence, the pressure extremes and the wave As before, the pipe wall friction will be difficult to evaluate directly.
propagation patterns are altered. Because we will be working with circular cylindrical pipes, we use the
The wave speed of the air-liquid mixture is computed as previously equation
done for a homogeneous liquid, but with the use of an average density for
the mixture. This approach implicitly assumes the mixture is evenly = i pfvivi
distributed throughout a significant portion of the pipe. The elasticity of
the liquid-air mixture is dramatically affected by a small amount of for relating wall shear to the Darcy-Weisbach friction factor. The peculiar
entrained air so the elasticity of both substances must be included. form of the velocity representation is introduced in the above equation to
Application of the momentum equation and conservation of mass point out how the proper shear force direction is preserved when the
leads to the following equation for wave speed from Tullis, Streeter and velocity reverses. If we had used V2 instead, then the negative velocity sign
Wylie [13], occurring on flow reversal would be lost in the squaring process.
^KjAve Substituting this equation into Equation 4-24 gives
(4-23)
Ko D £K dV +J_ ........................... (4-25)
Cfl | N
V|V| =
Q-J co
OIO
+
1 + -g- — C + (void fraction) dt p
OQ
olo
+
a
where the subscripts Z and a refer to properties of liquid and air, It is appropriate to point out at this time that both V and p are functions
respectively, and pave is the average density of the mixture. The value of of time t and location s along the pipe. The term 3z/3's is the slope of the
Ka is dependent on the thermodynamic process, e.g., for an isothermal pipe and can be written as the total derivative dz/ds. Equation 4-25 is an
process, Ka equals the absolute pressure. The void fraction is the volume equation with two dependent variables V (s,t) and p (s,t) hence, we need a
of air per unit volume of mixture at a given pressure. second equation relating the same dependent variables.
Unfortunately, the wave speed depends on the pressure because the To obtain this equation the conservation of mass is applied to a
void fraction is a function of pressure. Hence, an accurate analysis must control volume coinciding with the interior of the pipe and of length ds.
ELASTIC THEORY 45
44 UNSTEADY FLOW IN PIPELINES
dp 1 dA 9V
!{*)■« ........................ (4-28)
•to b-
+
dt A dt 9s
because ds, fixed to the pipe walls, varies only with time.
From the previous section,
d d
P - P
d¥ dp_
V p
Conservation of mass gives so
Q.
■—'I
=_L 22. .............................................................. (4-29)
AV- pAV + ^d(pAV)dsj =A (pAds) dt K dt
(4-26)
' g7^AV)ds = |r(PAds)
To develop a useful expression for dA/dt in terms of pressure, the
elastic pipe deformations must be considered. For stretching of the
At this point we employ a rather unusual form of the control volume cross-sectional area, Equation 4-10 is used to give
concept in that we permit the ends of the control volume to move
longitudinally with the pipe as it stretches. This device is employed dA = VZTTD2
because the pipe stretching affects the storage volume available and the
connection between pipe elasticity and the volume available for the fluid is
dA = -| (do2 - p do j)
identical to that developed in the previous section.
Expanding the parentheses of Equation 4-26 gives In evaluating the stresses we will again use case (b) restraint
- (pA ds + pV ds + AV-^- ds'j
1 9s 9s 9s I " dp £
a>[a>
Q-J|CU
(4-30)
+
+
<
|*t)
................................................ (4-27)
Considering longitudinal expansion,
Recognizing that d(ds) = de j ds
dp
75 I nJ
Q.I +-
at
and
9A = dA ± A (ds) = 0 ........................................................................... (4-31)
+V
CO CO
St dt ds dt
co
. ..................... ............................... ............................... . ......
D dp
4-
>M
1 dp + (1-M2) =0
K dt eE dt
+(1 -^2) D
"Ol’O
pj
K
'•-*
eE
dz f .............................. (5-1)
S V|V| = 0
ds 2D
47
..........................
48 UNSTEADY FLOW IN PIPELINES SOLUTION BY METHOD OF CHARACTERISTICS 49
1 3p
CM
5-2) Choosing A = -a as the other scale factor permits writing Equation
O.
>1 w
cd
4-
p at 5-4 as
4-
Cuj
and 5-2 may be replaced by some linear combination of themselves. Using dt ap dt ds
A as a linear scale factor, one combination is
Remember Equation 5-6 is good only when
+ 1 12 VIVI = 0-...(5-3)
4-
<<
II
4-
p 3s P at
o
>
>
4-
+x ...(5-4)
QjICL’
£
4-
\p at The result of these manipulations is that we have replaced two partial
<z>
then
dV £ dH ds .............. (5-8)
ID VM '0 only when
4-
4-
and dt a dt dt a
nJ -•-*
32 is to be replaced by 1
►*
+
>
8s
while
<<IQ.
II
— |Q.
then £ dH f
V |V| = 0 .............. (5-9)
4-
I
a dt 2D
Equating the expression for (ds/dt) in each case, we find
The fact that the relation between s and t in Equation 5-8 must satisfy
(ds/dt) = a, has caused the equation (ds/dt) = a to be called the
characteristic of Equation 5-8. A similar line of reasoning suggests that
or .................................................................................... (5-5) (ds/dt) = -a is the characteristic of Equation 5-9.
Because Equations 5-8 and 5-9 apply only along the appropriate
So both equations are satisfied if A = ±a. Arbitrarily picking A = +a characteristics, it is customary to view the situation graphically. This is
permits us to rewrite Equation 5-4 as done by drawing the s-t coordinate system (known also as the s-t plane),
determining the plot of the characteristics on this plane and thereby
deciding how to solve Equations 5-8 and 5-9.
In drawing the s-t coordinate axes, the origin of the s-axis (abscissa)
Dividing through by wave speed gives is placed at the upstream pipe end and extends in the positive direction
along the pipe for a distance L equal to the pipe length. Figure 5-1
1 dz f .................................. (5-6) illustrates how the s-t plane can be used to verify the occurrences that we
V |V| = 0
4-
CTQ
dV £ dH -3— = 0 where -a
dt a dt
Multiplying by (a/g) dt gives
dH = — dV
g
/-
along the C* characteristic. This equation is identical to the AH — (a/g)
Q
X x
x
X
X
AV equation derived earlier. Basically it means the change in H between
X
two points on the C’ characteristic can be computed from the change in
O
s=L velocity between the same two points. This result is the same conclusion
reached in Chapter II in Figures 2-2 and 2-3. The C" characteristic
Physical Configuration s-t Plane equation also gives
ds = - adt
Figure 5.1. The s-t plane for the simple pipe of Chapter II. which illustrates how long it takes information to travel between the two
points ds apart.
Now that some physical grasp of the meaning of Equations 5-8 and
5-9 is in hand, it seems appropriate to formalize the ideas more carefully.
In order to generate values of H and V at various selected locations in the
pipe as functions of time we must supply initial conditions along the s-axis
in the s-t plane and boundary conditions for all times at s = 0 and s = L.
Then, a solution for values of H and V can march upward on the s-t plane.
To see how the marching process occurs, examine Figure 5-2, which
is the s-t plane for some unknown problem. We know that at any point on
the s-t plane, say point P, the value of H and V is unique (i.e., the H and V
C values are independent of which characteristic they were approached
from). We also know that if we construct the C + and C" characteristics
through this point we have two ordinary differential equations which apply
-T- t along their respective characteristics. The ordinary differential equations
U
...................................
Le
►s 2D
ps
o
o
CO
I
CO-
CL
CO
Le Ri
As
(A
t>
CL
4->
CL
O
2D
In the above equations, tp-0 in general is At, and when these equations are
A disturbance is caused by shutting the valve. The disturbance
multiplied by At they become
propagates upstream at speed a making the track shown on the s-t plane.
The negatively sloped dashed line is the negative or C' characteristic along
which Equation 5-9 applies. If friction is negligible, Equation 5-9 gives C+: (Vp-V^) 12)
bQ
+
52 UNSTEADY FLOW IN PIPELINES SOLUTION BY METHOD OF CHARACTERISTICS 53
and This procedure described for point P 2 can be continued ujitil the
values of H and V at points P2 through Pjq are calculated. The values of H
C-: (Vp-VR1)- f (Hp.HR1) + ^ vw IVR1I = 0 .................................... » and V at PpPjq+j must be calculated with the assistance of boundary
conditions. At the left hand end of the pipe (near s = 0) we have only the
C' equation relating conditions between s2 and sv The boundary
Each of the characteristic equations can be integrated to show that
condition at s2 (s = 0) must provide another equation in HPj and/or V Pj
As = a At ............................................................................. (5-14) to solve simultaneously with the C'equation. The same situation and
requirements exist at the downstream end of the pipe with the C+
Once we have committed ourselves to a finite difference numerical characteristic.
solution we must divide our pipe into a discrete number of sections and Once the boundary conditions are established, then velocities and
proceed accordingly. If we decide to divide the pipe into N sections, then total heads at all the grid points at t = At can be calculated. These values
each section will be As = L/N long. Now that As is fixed, At can be are now used to write new equations to solve for values of V and H at the
calculated. It is now possible to construct a grid of characteristics as next time step where t = 2At. This procedure is repeated until the solution
shown in Figure 5-3. has progressed the required amount of time. The result is a set of H and V
Grid points along the s-axis represent points spaced As apart along values at discrete time intervals for N + 1 locations along the pipe.
the pipe and the values of V and H at these points on the s-axis represent Finally, it is possible to generalize from our development some
initial conditions. Initial conditions are generally some steady state flow concepts on disturbance propagation. Because of our previous discussion
situation in the pipe. Picking a point P 2 in the pipe on the s-t plane, concluding that disturbances propagate along characteristic lines, we can
writing Equations 5-12 and 5-13 for the known Vj^ VRJ, HJ^, HRJ, f say that “zones of unawareness” can be identified on the s-t plane which
and D, it is clear the equations each have two unknowns. The two cannot be affected by disturbances occurring at a particular location and
equations can be solved easily for Hp2 and Vp . Note that the known time (see point S on Figure 5-4). Conversely, if a location along the pipe is
velocities are used to represent the friction term. To do otherwise would chosen and a time specified (point P), then no disturbances outside the
render the equations difficult to solve and would not noticeably improve portion of the s-t plane subtended by the C+ and C" characteristics
the solution except when friction losses are very large. In the exceptional through the point can influence values of V and H at the point.
case, an iterative procedure would have to be developed wherein friction
losses would depend on a velocity averaged over the time interval. Setting Up the Numerical Solution
t
I \ \/./ . . ,
u
Zone of Zone of Zone of
Unoworeness Unoworeness Unoworeness
of the pipe were determined by using boundary conditions. It remains now The equation for Hp would be (from Equation 5-12)
to set up a solution procedure so that computer techniques can be used.
Looking first at solving Equations 5-12 and 5-13 for interior grid
intersections, we solve these two equations for Vp and Hp.
N+l
» fcro
4-
for any value of VpN+1 (including zero).
fat Constant speed pump boundary conditions — This boundary
2D 1)1 .................................. (5-15) condition offers the added complexity of a boundary condition containing
both H and V. This boundary condition equation must be solved
H
simultaneously with the C" Equation 5-13 to get equations for Hp, and
Ri)
Vp, •
4-
To keep the result uncomplicated at this time, we will assume the
head Hp put out by the pump can be adequately represented by a
16) tic equation
Q*
3
..................................
cd
45
= A'Q2 +B'Q+C'
CL
The boundary conditions at each end of the pipe are comprised of
externally imposed conditions of velocity and/or pressure head. To assist where Q is the discharge through the pump. The Q’s in this equation can
in understanding how to handle boundary conditions, it is best to examine be replaced by VP, x A and the head increase Hp can be added to the
a few common ones. supply sump H-value to give Hp, . The modified pump equation can now
Reservoir boundary condition — Where a pipe exits from a reservoir, be written:
the H-value remains constant for all time (neglecting velocity head). This
fact is symbolized in equation form as = A Vp + B Vp + C .................................................... (5-21)
............................................................................. (5-17) where C includes the H-value of the sump. It is important to point out at
this time that if this curve is to be concave downward and always slope
This equation is solved simultaneously with the C' Equation 5-13 to downward with increasing Q (generally it should do this), then A < 0, B <
yield an expression for velocity 0 and C > 0. These signs will be used subsequently when computerizing
this procedure.
When Equation 5-21 is solved simultaneously with the C" Equation
............................ (5-18) 5-13, the elimination of Hp gives
aa2 2D O..(5-23)
Tc seconds. The equation for Vp would be
=V
t1 ’ Tj ’ 0<t<T Using the formula for solving quadratic equations, V P, can be found.
o1 c Back substitution into Equation 5-21 will give the value of Hp .
N+l ....................................(5-19)
As a sidelight, it is interesting to note that if a check valve were
=0 t>T installed downstream of the pump, we would have to check the sign of
N+l c
56 UNSTEADY FLOW IN PIPELINES SOLUTION BY METHOD OF CHARACTERISTICS 57
VPj. If VPj were negative, we would change to VPj =0 and calculate collection of known terms by a single symbol. In this case, looking first at
Hpj from Equation 5-13, not Equation 5-21. the C‘ characteristic, from Equation 5-25,
Computerizing the Numerical Procedure = c, +C2H .......................................... (5-28)
The problem solving approach we have used is relatively easily where
computerized. If a pipe of length L is divided into N sections, the node
= V„ - £ f At V2|V2|
points between sections can be numbered sequentially from 1 to N + 1
ns
2D
beginning at the upstream end. Keeping in mind the parallel between the
subscripts in our equations and the indices of subscripted variables in C2 = g/a
computer programs, we rewrite Equations 5-12 and 5-13 as
Of course, in the computer Ct and C2 are just numbers
CD
</)
p
c
calculated using known values from the previous time step.
f At ........... (5-24) Combining Equation 5-28 with Equation 5-21 to eliminate
ns kra
4-
+
2D
and = A Vp + B Vp + C
The equations for interior values of Vp and Hp (Equations 5-15 and 5-16)
=0
are now written
- I KH
+
p kra
............................. -c3 ±
-.-^(VHIVH l-VwIVw()] .............................................. (5-27)
II
-4
for 2 i N. For the boundary conditions previously posed, the resulting C4
or in another form
Equations 5-17, 5-18, 5-19 and 5-20 must be similarly modified where
required. However, when certain boundary conditions require simulta-
neous solution of two equations to obtain H and V values, the algebra
............................... (5-30)
must be performed by the engineer and prepared for the computer. An
CD tO
appropriate illustration of this technique (which will be used many times
in succeeding sections) is the boundary condition for a constant speed It remains only to determine which of the “±” signs to use.
pump which led to Equation 5-22. Because this decision will have to be made in several other instances,
Constant speed pump (computerized) — For purposes of algebraic it is appropriate to discuss it in detail now. We begin by determining the
manipulation it is convenient to simplify the equations by representing a signs (where possible) of the C-terms. From Equation 5-28
..
Cj = unknown sign
C2 = (+) 2
3
C PROGRAM NO. 1
C APPROXIMATE WATER HAMMER PROGRAM FOR A SIMPLE PIPE
From the previous section on the pump curve, 4
c
5
6 C — RESERVOIR AT UPSTREAM END —
< cq O
II II II
7 C — VELOCITY DECREASES LINEARLY WITH TIME TO ZERO AT DOWNSTREAM END —
8 C
9 C IOUT = INDEX WHICH GIVES AN OUTPUT EVERY IOUT-TH TIME STEP
10 C NPARTS = NUMBER OF PARTS INTO WHICH PIPE IS DIVIDED
11 C L = PIPE LENGTH-FT D » PIPE DIAM-IN F » DARCY-WEISBACH F-VALUE
12 C VZERO = VELOCITY-FPS A » WAVE SPEED-FPS HZERO » RESERVOIR ELEV-FT
From the definition equations for C3 and C , 13 C
C
ELEVUP » ELEVATION OF UPSTREAM END OF PIPE-FT
ELEVDN = ELEVATION OF DOWNSTREAM END OF PIPE-FT
14
15 C TMAX = MAXIMUM REAL TIME OF SIMULATION-SEC
16 C TCLOSE « TIME REQUIRED FOR VALVE CLOSURE-SEC
o
17 C
18 DIMENSION X(20),V(20),H(20),HLOW(20),HHIGH(20),HEAD(20),
19 $VNEW(20),HNEW(20),PIPEZ(20)
20 REAL L
(+) + (unknown)L , . A 21
NAMELIST /SPECS/ IOUT,NPARTS,L,D,F,A,VZERO,HZERO,ELEVUP,ELEVDN,
------- = (unknown) 22
STMAX,TCLOSE
READ(5,SPECS)
23
A
II
PI = 3.141592
24
WTT-L/A
25
DELL=L/NPARTS
We conclude C3 is always positive and we are not sure of the sign of C4. 26
T=0 .
27
DELT=DELL/A
Equation 5-30 can be written in terms of signs as 28
C=32.2/A
29 INDEX«TMAX/DELT + 1
30 DELEL’(ELEVDN-ELEVUP)/NPARTS
31 NODES»NPARTS+1
32 WRITE(6,200)
(sign Vp ) = (+) 33 WRITE(6,201) IOUT,NPARTS,L,A,D,F,VZERO,HZERO,ELEVUP,ELEVDN,
r
i 34 $WTT,TCLOSE,TMAX,DELT
35 AK=12»*F*DELT/(2•*D)
36 DELHF=12.*F* DELL*VZERO** 2/(64.4*0)
37 DO 300 1 = 1,NODES
38
At the beginning of the unsteady flow process, V p t is positive, hence, we 39
V(I)=VZERO
B(I}=HZERO-(1-1) *DELHF
should be able to get some positive values from this equation. This can 40
41
HLOW(I)«H(I)
HHIGH(I)=H(I)
only be true if [ ] = (+). Because we would like to also permit negative 42
43
X(I)=(1-1)*DELL/L
PIPEZ(I)=ELEVUP+(1-1)*DELEL
velocities to occur occasionally, then [ ] = (-) must be possible. Because 44 HEAD(I)«H(I)-PIPEZ(I)
45 300 CONTINUE
the square root is always (+), the only way that [ ] can be both positive 46 WRITE(6,202)
47 WRITE(6,204) T,(X(I) ,HEAD(I) , H( I) ,V(I) ,1 = 1 ,NODES)
and negative would be that if the (+) sign in the (±) were used. This line of 48 DO 99 11=1,INDEX
49
reasoning leads us to the final form of the equation 50 C
T=T+DELT
** COMPUTE II AND V AT INTERIOR NODES **
51 DO 20 1=2,NPARTS
52 VNEW(I)=0.5*(V(I-l)+V(1+1)+C*(H(1-1)-H(1+1))-AK*(V(I-1)*ABS(V(I-1)
S)+V(1+1)*ABS(V(1+1))))
4C, 53
54 20 HNEW(I) =0.5* (II (1-1) +H(I + 1) + (V( I-1)-V(l+1) )/C-AK* (V(I-l) *ABS (V(I-l)
..................... (5-31) 55 $)-V(I+l)*ABS(V(I+1)))/C)
U> K>
56
57 C — THIS BOUNDARY CONDITION IS FOR A CONSTANT-HEAD RESERVOIR —
58 KNEW(1)=HZERO
and 59 VNEW(1}=V(2)+C*(HNEW(1)-H(2))-AK*V(2)*ABS(V(2))
60 C ** COMPUTE H AND V AT DOWNSTREAM END **
61 C — THIS BOUNDARY CONDITION IS FOR LINEARLY DECREASING VELOCITY —
..................... (5-32) 62 IF(T.GT.TCLOSE) GO TO 30
63 VNEW(NODES)=VZERO*(1,-T/TCLOSE)
64 GO TO 31
65 30 VNEW(NODES)=0.0
Again, if we were to install a check valve to prevent backflow 66 31 HNEW (NODES) =11 (NPARTS) + (V (NPARTS) - VNEW (NODES) - AK*V ( NPARTS) *
SABS(V(NPARTS)))/C
through the pump, we would have to test V and, if negative, set VPj = 67
68 DO 50 1 = 1 ,NODES
IF(HNEW(I).LT.HLOW(I)) BLOW(I)=HNEW(I)
0. We could then go ahead and calculate Hpj from Equation 5-32. 69
70 IF(HNEW(I).GT.HHIGH(I)) HHIGH(I)=HNEW(I)
However, we could not use Equation 5-21 to compute Hp , because the 71
50 -HEAD(I)=HNEW(I)-PIPEZ(I)
IF(MOD(II,IOUT} .EQ.0) WRITE(6,204) T, (X (I) ,HEAD(I) ,HNEW(I) ,VNEW(I)
72
check valve has isolated the pump from the pipeline. 73
S,1=1,NODES)
IF(T.GT.TMAX) GO TO 400
74
DO 40 1=1,NODES
75
V(I)=VNEW(I)
The Basic Computer Program 76
40 H(I)=HNEW(I)
77
99 CONTINUE
78
79
A basic computer program which can be modified to suit a number of Figure 5-5. Computer program for approximate water hammer in a
different situations is presented in Figure 5-5. The program has simple pipe.
60 UNSTEADY FLOW IN PIPELINES SOLUTION BY METHOD OF CHARACTERISTICS 61
Example 5-1
4
a
89: >S ,I3//31X,’L =■,E7.1,' ET'/31X,'A -',E7.1 , ' EPS'/
90: S31X,'D =',F7.2,' IN'/31X,'F ,F6.4//
91: S27X,'VZERO -',E7.2,' FPS'/27X,'HZERO -',F7.1.' FT'/
92: 326X,’ELEVUP »’,F7.1,' FT'/26X,'ELEVDN =',F7.1,‘ FT’//
93: S29X,'L/A »’,F7.3,1 SEC*//26X,’TCLOSE «',F7.2,’ SEC*/
94 : 528Xf'TMAX =’,F7.2,' SEC'/28XDELT »',F7.3,' SEC')
95: 202 FORMAT(lHl/’ PRESSURE HEADS, H-VALUES AND VELOCITIES AS FUNCTIONS
96: 3 OF TIHE'/2X,60(’-’))
97: 204 FORMAT(//18X,3(5X. ' X HEAD-ET H-FT V-FPS')/' TIME ='.F7.3,
98: S’ SEC',3(5X,’ --------------------------------------------------- ---------------- *)/(18X,3(5X,F5.3,2F7»0,
99 : SF8.2)))
100: 205 FORMAT(1H1/18X,27('*')/18X* TABLE OF EXTREME VALUES **/18X,27(“
101: S')//13X,’X MAX HEAD MIN HEAD MAX H MIN H'/11X , 5( ' - *) ,2X , 8 (
102: S'-') ,2X,8(*-') ,2X ,6 c-') ,2X,6 (’-*))
103: 206 FORMAT (1IX , F5.3,2X , F7.0,3X , F7.0,3X , F6.0 , 2X , F6.0)
104: END
NAMELIST input because of the ease of reading the data cards that have
been punched. The input parameters are listed in the program under the
NAMELIST /SPECS/ statement and each parameter is defined in The 30-inch steel pipe has a wave speed of 4060 fps and an initial
COMMENT statements at the beginning of the source program. velocity of 5.0 fps. The valve is closed in a manner which causes the
In this basic version of the program, the user must make his own velocity to flow at the valve to decrease linearly with time to a value of zero
steady state hydraulic computations to determine discharges, friction in 1.0 seconds. There is initially negligible head loss in the valve.
factors and EL-HGL location. These values of steady state conditions are Find the maximum pressure head in the system which would occur in
then input into the program to establish the required initial conditions. the first 4.0 seconds after closure begins.
In summary, the program reads in the basic information, generates Solution — If the pipe is divided into 5 sections and if the steady state
steady state H and V values at the grid intersection points (nodes) along hydraulics have been established, it remains only to set up the data for the
the pipe and then begins the unsteady flow calculations. The interior grid computer.
intersection points are first calculated using Equations 5-26 and 5-27. The
upstream and downstream boundary conditions are used to get values of
Hp and Vp at each end of the pipe. The whole process begins again using
the just-computed values of Hp and Vp as the known values. The process
continues to “loop” until the time has reached TMAX. Before execution is DATA CARDS
terminated, “envelope” values of H are printed for each node. These
values are the highest and lowest values of H and pressure head computed
SSPECS IOUT»1,NPARTS-5,L»5000.,D»30.,F«.020,A=4060.,VZERO»5.00,
at each point during the analysis. HZERO®715.5,ELEVUP®100,,ELEVDN»50•,TMAX«4.0,TCLOSE*1«0
SEND
To simulate transient problems with other boundary conditions, it is
necessary only to change those parts of the program listed under upstream
and downstream boundary conditions. Occasionally the input to the The computer output is listed below. A quick scan of the Table of
program must be altered to include new parameters, however, the basic Extreme Values shows the maximum pressure head was 1292 ft at the
program architecture remains the same. valve.
To illustrate the use of the program, we will consider an elementary
example.
62 UNSTEADY FLOW IN PIPELINES
*********************************
* WATER HAMMER IN A SIMPLE PIPE *
*********************************
INPUT DATA
IOUT = 1
NPARTS = 5
—
L 5000.0 FT
A= 4060.0 FPS
D= 30.00 IN
F
- .0200
***************************
* TABLE OF EXTREME VALUES *
***************************
>
4-
>
II
ex
In solving the complete equations, we can proceed in a manner
similar to that for the approximate method. However, we will use the The final set of equations which compare with Equations 5-8 and 5-9 are,
complete equations of Euler and Continuity 4-33 and 4-34.
after replacing p with y(H - z),
The Complete Equations
+. ffH. V — V |V| = 0 only if = V + a • • (5-33)
Again the multiplier A is used to combine the two partial differential dt a ds
4-
P fa
equations. Multiplying A by Equation 4-33 and adding the result to
Equation 4-34 gives
^<| Q.
to)
CU
CO ko
>”
|V| a2 dt a dt a ds
o
OD
4-
4-
ds
CO
c/3
ex
To carry forth the same procedure as was used previously, we must break The ordinary differential equations are quite similar to those for the
dV/dt and dp/dt down into component parts. The result is approximate case. However, the important distinction is that the
characteristic lines on the s-t plane, which were of constant slope before,
CO | CO
>«
CX CX
oo | N
Xg IV | a2 are now curved and their slope is a function of V(s,t). This result
<x|a;
4-
4-
+
w fa
<73 fa
^l<
4-
4-
S’fe’
o
II
1
4-
P dt ds P dt 1 P dt
and Rj, but pass through the t = constant line at points L and R
Rewriting the restriction equations for ds/dt, somewhere in between.
ds v , a2 ds . J_, In this case the finite difference approximation to Equations 5-33 and
dF V+yandir=X + V 5-34 becomes
Equating the two expressions for ds/dt and solving, we get again
Vp-V -£V f ..........
T
X=±a a L ds 5-35)
4-
At At 2D
Our characteristic equations are, in this case,
68 UNSTEADY FLOW IN PIPELINES SOLUTION BY METHOD OF CHARACTERISTICS 69
where
t
Ax
At
Solving for VL and HL gives
V
P L=
+
<
n
and
< Le Ri
L
b R1
As As
Substituting the value of (Ax/At) = a + VL gives
------------- ►
Figure 5-6. Interpolation of H and V values on a As - At grid.
............................................. (5-37)
and
H
The difficulty, of course, is that the values VL> HL, VR, HR are not C+7HHL -Hc)(a + VL) .................................... (5-38)
known. However, the values of VLe, HLe, VR, , HRj , Vc and Hj. are
known. The unknown values of H and V at points L and R can be
estimated by interpolation. In this case, we will use linear interpolation A similar analysis for the C' characteristic gives
and the sketch below illustrates the relationships.
n.
X
and
...................................
H
R’Hc + £<HR|-Hc,(a’V«) ...................................................................................... ,M01
Because (At/As) (VRJ - V^) is of the order V/(a+V) which is very small
compared to 1, it is a good approximation to neglect the second terms in
the denominators of Equations 5-37 and 5-39. The result is
.............................................. (5-41)
and
H
Ax L-HC
............................................. (5-42)
As
70 UNSTEADY FLOW IN PIPELINES SOLUTION BY METHOD OF CHARACTERISTICS 71
The simultaneous solution of Equations 5-35 and 5-36 for Vp and Hp where max |a+V| is the maximum expected absolute value of the sum of
gives the wave speed and flow velocity. If locations of points R and L fall outside
the grid points Le and Rj, numerical stability problems and accuracy
problems begin to develop. (See Art. 5-3.) These problems are associated
(VL + VR) + f (HL - HR) +f At (VL - VR) sin 6
with the previous discussion on “zones of unawareness,” i.e., where points
R and L fall outside the grid points, information from the zone of
unawareness is influencing the values of Hp and Vp. This is improper
numerical procedure, because the disturbances have not had time to reach
point P. As a result, this occurrence must be guarded against to prevent
inaccurate results.
The value of As is still determined by the number of sections into which we Effect of Friction
have chosen to divide the pipe. Because our interpolation procedure
implies that the points R and L are between points Rj and Le, we must The effect of friction will be examined by looking at a simple pipe and
choose At so small as to guarantee this always occurs. The above equations seeing how various amounts of friction effect the pressure extremes in the
suggest that pipe. The model system used is shown in Figure 5-8.
The results of three runs with different velocities which cause
At < ------ — ........ ......................................................... (5-45) different frictional losses are shown in Table 5-1. It appears that one can
max |a + VI
72 UNSTEADY FLOW IN PIPELINES SOLUTION BY METHOD OF CHARACTERISTICS 73
73 : GO TO 31
74: 30 VNEW(NODES)=0.0
75: 31 VL=V (NODES)+DTDS*A* (V (NPARTS)-V~(NODES) ) ’
C PROGRAM NO. 2 76: HL»H(NODES)+DTDS*(H(NPARTS)-H(NODES))*(A+VL)
2:
77: HNEW(NODES)«H(NPARTS)+(V(NPARTS)-VNEW(NODES)-AK*V(NPARTS)*
3: C COMPLETE CHARACTERISTICS METHOD PROGRAM FOR A SIMPLE PIPE
78: SABS (V (NPARTS) ) )/C+DELT*VL*SINE ___________________________________________
79 : DO 50 1=1,NODES
c
80: IF(HNEW(I).LT.HLOW(I)) HLOW(I)=HNEW(I)
6: C — RESERVOIR AT UPSTREAM END —
81: IF(HNEW(I).GT.HHIGH(I)) HHIGH (I)-HNEW(I)
7: C — VELOCITY DECREASES LINEARLY WITH TIME TO ZERO AT DOWNSTREAM END —
82: 50 HEAD(I)»HNEW(I)-PIPEZ(I)
8; C
83: IF(MOD(I I,IOUT).EQ.0) WRITE(6,20 4) T,(X(I) , HEAD(I) ,HNEW(I) ,VNEW(I)
9 : C IOUT - INDEX WHICH GIVES AN OUTPUT EVERY IOUT-TH TIME STEP
84 : S,1=1,NODES)
10: C NPARTS » NUMBER OF PARTS INTO WHICH PIPE IS DIVIDED
8 5: IF(T.GT.TMAX) GO TO 400
11: C L = PIPE LENGTH-FT D = PIPE DIAM-IN F - DARCY-WEISBACH F-VALUE
86 : DO 40 1=1,NODES
12: C VZERO = VELOCITY-FPS A - WAVE SPEED-FPS HZERO ~ RESERVOIR ELEV-FT
87: V(I)=VNEW(I)
13: C ELEVUP » ELEVATION OF UPSTREAM END OP PlPE-FT 88 : 40 H(I)=HNEW(I)
14 : C ELEVDN » ELEVATION OF DOWNSTREAM END OF PIPE-FT 89: 99 CONTINUE
IS; C THAX • MAXIMUM REAL TIME OF SIMULATION-SEC 90 : 400 WRITE(6,205)
16: C TCLOSE = TIME REQUIRED FOR VALVE CLOSURE-SEC 91: DO 401 1 = 1 ,NODES
17: C 92: HEADMX-HHIGH(I)-PIPEZ(I)
18: DIMENSION X(20),V(28),H(20),HLOW(20),HHIGH(20),HEAD(20), 93: HEADMN=HLOW(I)-PIPEZ(I)
19 : SVNEW(20) ,HNEW(20) , PIPEZ(20) 94 : 401 WRITE(6,206) X(I),HEADMX,HEADMN,HHIGH(I),BLOW(I)
20: REAL L 95:
21: NAMELIST /SPECS/ IOUT,NPARTS,L,D,F,A,VZERO,HZ ERO,ELEVUP,ELEVDN, 96: 200 FORMAT(1H1,19X,33(’*')/20X,'* WATER HAMMER IN A SIMPLE PIPF *’/
22: STMAX,TCLOSE 97: S20X,33('*'))
23 : READ(5,SPECS) 201 FORMAT(///29X,1 INPUT DATA‘/29X,10('- 1)//28X,’IOUT =’ ,I3/26X,1NPART
98:
24: PI = 3.141592 $S =',I3//31X,1L =’,F7.1,' FT'/31X,'A =’,F7.1,' FPS1/
99:
25: WTT»L/A S31X,'D =’,F7.2,' IN’/31X,'P =',F6.4//
100 :
26: T=0 . S27X, ‘VZERO -'^7.2,’ FPS '/27X , ‘ HZERO =’,F7.1,' FT'/
101:
27: C=32.2/A S26X,'ELEVUP =‘,F7.1,’ FT‘/26XELEVDN =',F7♦1,’ FT1//
102:
28: DELEL«(ELEVDN-ELEVUP)/NPARTS $29X,'L/A »',F7.3,' SEC'//26X,'TCLOSE =’,F7.2,' SEC'/
103:
29: DELL°L/NPARTS _________________ 104: S28X,’TMAX «‘,F7.2, 1 SEC'/28X,'DELT =',F7.3,‘ SEC*)
30 : 202 FORMATCIHI/' PRESSURE HEADS, H-VALUES AND VELOCITIES AS FUNCTIONS
DELT-DELL/(A+VZERO) 105:
31: $ OF TIME'/2X,60(*-'))
DTDS’DELT/DELL 106 :
32: 204 FORMAT(//18X,3(5X,' X HEAD-FT H-FT V-FPS')/' TIME =',F7.3,
SINEaDELEL/DELL _______________ 107:
33: S' SEC',3(5X,' --------------------------------------------------- ---------------- 1 )/(18X.3 (5X,F5.3,2F7.0,
INDEX-TMAX/DELT + 1 108: SF8.2)))
34 :
NODES-NPARTS+1 109 : 205 FORMAT(1H1/18X,27('*’)/18X* TABLE OF EXTREME VALUES *'/18X,27('*
35:
WRITE(6,2O0) 110: S’)//13X,’X MAX HEAD MIN HEAD MAX H MIN H'/l1X,5('-') , 2X , 8 (
36:
WRITE(6,201) IOUT,NPARTS,L,A,D,F,VZERO,HZ ERO,ELEVUP,ELEVDN, 111: $'-'),2X,8(*-'),2X,6(*-*),2X,6(’-’))
37:
SWTT,TCLOSE,TMAX,DELT 112: 206 FORMAT(11X,F5.3,2X,F7.0,3X,F7.0,3X,F6.0,2X,F6.0)
38:
AK = 12.*F*DE LT/(2.*D) 113: END
39 :
DELHF=12.*F*DELL*VZERO** 2/(64.4*D) 114:
40 :
41: DO 300 I-l,NODES
4 2: V(I)=VZERO
43: H(I)-HZERO-(1-1)*DELHF Figure 5-7. Continued.
44: HLOW(I)»H(I)
45: HH IGH(I)= H(I)
46: X (I) = ( I-1)*DELL/L
47: PIPEZ(I)=ELEVUP+(I-l)*DELEL
48: HEAD(I)=H(I)-PIPEZ(I)
49 : 30e CONTINUE
WRITE|6,2«2)
v
50 :
51: WRITE(6,20 4) T, (X(I), HEAD(I) ,H ( I) ,V(I) , 1 = 1 , NODES) EJ=lHGL_
52: DO 99 11=1,INDEX
53: T=T+DELT
54 : C ** COMPUTE H AND V AT INTERIOR NODES **
55: DO 20 I=2,NPARTS _____________________________________________________________________
56 : VL=V(I)+DTDS*A*(V(I-l)-V(I))
57 : VR = V(I)+DTDS*A *(V(I + 1)-V(I))
58 : HL=H(I)+DTDS*(H(I-l)-H(I))*(A+VL)
2000
59 : HR=H(I)+DTDS*(H(1+1)-H(I))*(A-VR)
60: VNEW(I)=0.5*(VL+VR+C*(HL-HP)+C*DELT*(VL-VR)*SINE-AK*(VL*ABS(VL)+
o
61: SVR*ABS(VR)})
62: 20 HNEW(I)=0.5*(HL+HR+(VL-VR)/C+DELT*SINE*(VL+VR)-(AK/C)*(VL*ABS(VL)
T
63: S-VR* ABS-(VR-L-) ) ______________________________________________________________________
64: C ** COMPUTE H AND V AT UPSTREAM END **
65: C — THIS BOUNDARY CONDITION IS FOR A CONSTANT-HEAD RESERVOIR —
80001'24"
66: HNEW(1)=HZERQ _________ _________________________________________________ Datum
67: VR-V(l)+DTDS*A*(V(2)-V(l))
68: HR=H(1)+DTDS*(H(2)-H(1))*(A-VR)
69: VNEW(1)=VR+C*(HNEW(1)-HR)-AX *VR*ABS(VR)-C*DELT*VR*SINE
a = 4000fps
70 : C “ COMPUTE H AND V AT DOWNSTREAM END **
71: C — THIS BOUNDARY CONDITION IS FOR LINEARLY DECREASING VELOCITY —
72 : IF(T.GT.TCLOSE) GO TO 30
VNEW(NODES)=VZERO*(1.-T/TCLOSE) Figure 5-8. Model system for showing sensitivity to system parameters.
Figure 5-7. Computer program for the complete method of
characteristics.
74 UNSTEADY FLOW IN PIPELINES SOLUTION BY METHOD OF CHARACTERISTICS 76
hf
TT AH= —AV AH+h,+H
f0
H
min Ho-AH+2hf
Velocity n
max g
(fps) (ft)
(ft) (ft) (ft) (ft) (ft)
Ml.
5.0 31 2616 621 2621 1412 1410
10.0 124 3221 1242 3242 882 882
1500
estimate maximum pressures occurring at the valve for sudden valve 1 = 4.67sec
closure if the following formula is used:
max
« H + hf + AH ..................................................... (5-46) 1000'
It also appears from Table 5-1 that the minimum pressure occurring at the
valve can be estimated with the formula
500
• H0 + 2hf-AH ............................................................(5-47)
The user should keep in mind these are rough approximations which
O
apply well to this case, but should be used with caution if applied to other
situations. .33 .67 1.0
o
To help develop a grasp of how frictional effects develop in a pipe, we x/L
will use the situation in Figure 5-8 with a 10 fps velocity and a suddenly
Figure 5-9. The progression of frictional effects in a simple pipe with
closing valve. Refer to Figure 5-9 which shows how the EL-HGL moves as
sudden valve closure.
the pressure wave propagates through the system. Basically the increase in
head AH caused by sudden valve closure propagates upstream increasing
the steady state head at each point by an amount AH. It might seem that
after a time L/a, the EL-HGL would be a line parallel to the steady-state the maximum head value at the valve can be approximated with Equation
EL-HGL, but positioned AH above it. Actually, a chain of events occurs 5-46. Pursuing this line of reasoning for another 2 L/a seconds would lead
beginning at the end of the first time step At, which causes the pressure to Equation 5-47. Keep in mind, the purpose of this example is to give the
head at each point in the pipe to continue to edge upward after the first engineer a physical feel for what is occurring; it is not intended as a
large AH occurs. This happens, because while V = 0 in the pipe section As substitute method of analysis. It also serves to illustrate the value of a
long, there is a pressure difference (Ah = (hf/N)) occurring (sloping computer program which handles all of the bookkeeping necessary to
EL-HGL). As a result, the system is not in equilibrium and a small include friction.
downstream velocity occurs to try and bring the system into equilibrium.
Each end of the As-long portion of the EL-HGL moves about one-half of Effect of the Size of N
the vertical distance to horizontal. As the pressure wave progresses to the
reservoir and returns to the valve the EL-HGL has had the time to move It would seem reasonable to expect that the greater the number of
upward an amount hf after the initial AH move. This process explains why pieces into which the pipe is divided, the greater the accuracy of the
76 UNSTEADY FLOW IN PIPELINES SOLUTION BY METHOD OF CHARACTERISTICS 77
results. Because As = L/N, it would seem that the larger N, the better the To demonstrate these problems, the simple pipe system in Figure 5-8
accuracy. It is surprising, then, to discover that the choice of N has very was used. The value of As was fixed, the valve was to be closed suddenly
little effect on the numerical solution. For example, for sudden valve and the value of At was calculated so that points L and Le coincided
closure with an initial velocity of 5 fps in the example of Figure 5-8, the (minimum interpolation). Higher values of At would cause L to move
maximum and minimum values of H differ less than 10 ft for N = 3 and N outside of Le and possibly R to move outside Ri and numerical instability
— 18. (See Table 5-2.) For slower velocity change the difference is should occur. Substantially lower values of At should cause points L and R
negligible. to move towards C (see Figure 5-6) thereby leading to interpolation errors.
There are two points to be made here. First, except in the case of In the demonstration, the reference value of At0 for minimum
rapidly varying velocity, there is little to be gained by using a larger interpolation was computed as above and a run was completed to use as a
N-value than necessary. Second, for a given run time, the number of grid standard of comparison. Then runs with At’s of 0.90 At 0, 1.02 At0 and
points vary as N* and so does the cost of computer analysis. However, care 1.10 At0 were made to see what happened. The results are shown on
should be taken not to make N so small that the resulting At would be too Figure 5-10. Note that for At = 0.90 At0, the interpolation error has
great to give good representation of results for time-varying boundary caused rounding of the sharp corners on the H vs t plot. For At = 1.02
conditions such as valve movements. At0, the results are not too different from the base run. However, when At
is increased to 1.10 At0, the solution begins to run wild. This result is
Effect of Sloping Pipes termed numerical instability and it is clear that the admonition in
Equation 5-45 should be followed.
Pipelines ranging in slope over ± 25% were tested to determine the On the other hand, when At is smaller than required, the result is not
effect of slope on the results. For both extremes of slope, the results were
instability, but inaccuracy. It appears then that the At parameter should
virtually identical. While the slope of the pipe should not be ignored (it is
be controlled so that interpolation is minimized, but care is taken to
necessary to include it to obtain pressure head anyway), the engineer
remain within the Le-Ri grid points.
should be comfortable “smoothing" out the pipe profile to include only a
manageable number of constant-slope pipes. By this means, it is possible
to use the series pipe program to obtain good approximations of pressure
head along the line and still exercise control over the As and At Experimental Verification
parameters.
The literature has several examples of comparisons between
Numerical Instability and Inaccuracy computed and measured pressures in systems. The work of Streeter and
Lai [14] illustrate comparisons for a simple pipe system. They employed
Earlier it was stated that At should be kept small enough to insure 0.5-inch and 0.375-inch copper tubing approximately 300 ft long in their
that points L and R in Figure 5-6 remained between grid points Le and Ri experiments. The results in Figure 5-11 represent simple pipes and series
for all situations. If not, numerical instability was presumed to occur as pipes and confirm decisively the accuracy with which the characteristics
well as inaccuracy in the computational results. method can predict pressure variation with time for turbulent pipe flow.
For laminar flow situations the work of Kaplan, Streeter and Wylie
[15] illustrates the accuracy of the method of characteristics. Figure 5-12
Table 5-2. Effect of N on pressure head at the valve for rapid and slow illustrates the comparison between theory and experiments conducted on
velocity change (V = 5 fps). the 465 mile long 34-inch diameter South European Pipeline. These
experiments substantiate the validity of the application of the method of
characteristics to laminar flow problems.
To illustrate the effect of burying the pipe on water hammer
Sudden Valve Closure Valve Closure in 4 L/a
calculations, the work of Watters, Jeppson and Flammer [16] is used.
These experiments on PVC pipe also verify that the theory works well on
max
relatively flexible pipe. Figure 5-13 shows that under unburied conditions
2611 for velocities up to 10 fps, the predicted and measured pressures were
1417 2288 1969
essentially identical. Further, burying the pipe, which might be suspected
2616 1412 2290 1969
of giving support to the pipe to make it act “stiffer,” was discovered to
18 2619 1409 2292 1969
have no effect on the wave speed.
78 UNSTEADY FLOW IN PIPELINES SOLUTION BY METHOD OF CHARACTERISTICS 79
1000
20
i
Figure 5-10. Numerical instability In a simple pipe with sudden valve closure.
J -------- 1 ___ i ____ i ____ i ____ L
1
t , sec
TIME- seconds
10
5
At = 1.02 of Normal
At = 1.10 of Normal
At = 0.90 of Normal
Normal At Value
in
o
0001
1500
3000
2500
2000
500
IN TIAL HEAD800
AT R
600
PRESSURE.
PSI
400
H , ft
200
t, sec
1000 III ~ I ~
NO 4 ----------- EXPERIMENTAL RESULS
800
600
PRESSURE. PSI
X 400
200
100
0
0 25 50 75 100 !25 ISO 175 200
0 0.2 0.4 0.6 0.8 LO 1.3 1.4 TIME - SECONDS
I , SEC.
Figure 5-12. Results of tests on the oil carrying South European Pipe-
line (from “Computation of Oil Pipeline Transients” by
Figure 5-11. Continued Kaplan, Streeter and Wylie, ASCE Hydr. Div., 93:6, Nov.
1967).
82 UNSTEADY FLOW IN PIPELINES
200
A natural extension of the analysis of single pipelines is to consider
180 more complex pipe systems. In actuality, practical situations almost
always involve systems more complicated than simple pipes. We have at
160
hand most of the analysis techniques necessary to handle complex systems
MEASURED PEAK PRESSURE NCREMENT
so we will begin with the simplest which are the series pipe problems.
AT DOWNSTREAM TRANSDUCER (PSI)
140
120
6.1 Series Pipes
IOO
In series pipe systems, each pipe carries the same steady flow
80
discharge, but has its own values of velocity, diameter, wave speed, etc.
60 This does not exclude the single size pipe with breaks in slope which can
be treated as a series pipe; however, the analysis is the same in either case.
40
20
The Internal Boundary Conditions
83
84 UNSTEADY FLOW IN PIPELINES COMPLEX PIPE SYSTEMS 85
case, we will assume that the minor head loss occurring at the junction is
negligible. The equation along the C+ characteristic is The values of the velocities are calculated from Equations 6-2 and 6-4 so
they will not be repeated here.
Computation of At
-2- L At VT sin 0. - H In the previous analysis we have assumed that At is the same in each
a< i 1 a. 6-1
PH
...
pipe. This is rarely true, so it is important to look carefully at this aspect of
the problem. The previous formula for computing At with the complete
or in short form method of characteristics is rewritten here for convenience.
............................................................. (6-2)
HJ
At = .................................................................... (6-8)
N(V + a)
For the downstream pipe the equation along the C' characteristic is
Because each pipe in a series system generally has different lengths, wave
speeds and velocities, it is obvious that strict application of this formula
(even using different N’s in each pipe) will lead to differing values of At in
................................ (6-3) each pipe. As an example, refer to the two-pipe system shown in Figure
6-1. In this example, equal N’s were used to compute At in each pipe. The
result at the internal boundary is that characteristic lines from the two
or in short form pipes do not intersect the internal boundary at the same point in time on
the s-t plane. Because it is also unlikely that we can find two N’s which will
= C, + C, H ....................................................... 6-4) give identical At’s in the two pipes, we must try another strategy.
Figure 6-2 displays another approach based on our knowledge of how
It is clear from Equations 6-2 and 6-4 that we have four unknowns to interpolate when characteristic lines do not intersect grid points. We
and only two equations. The remaining two equations are obtained from can reduce the larger At in one pipe down to the value of At in the
continuity and energy principles. remaining pipe, thereby overcoming our previous problem of matching
In applying the equation of continuity, the point P 2 in pipe 1 and the the At’s at the internal boundary. However, this is done at the risk
point Pi in pipe 2 are assumed to be so close together that there is no room of introducing serious interpolation inaccuracy in the one pipe.
to store any mass. As a result, continuity gives A technique exists to reduce this interpolation error without loosing
internal boundary compatibility. This is done by increasing N in the pipe
VRi Aj = Vp^ A2 .................................................................. (6-5) whose At must be decreased. Figure 6-3 illustrates the effect of this
In this case, when applying the energy equation and neglecting any approach. In practice, N is increased until one of the characteristic
small head loss caused by the contraction (or expansion), intersection points L or R move outside the grid point. The next smaller N
is then used to divide the pipe. The algorithm used to find this N-value is
N=4
Figure 6-1. The »-t plane for a two-plpe ayatem with equal N’a. PIPE NO. I PIPE NO. 2
'-—-
TRY N = 6 N=4
Figure 6-3. The a-t plane with variable N-value and minimum
interpolation.
The basic computer program for series pipes is shown in Figure 6-4.
The program is primarily an extension of the program presented earlier
for the complete method of characteristics. As before, the input
parameters are all defined with COMMENT cards and the user must
make all the steady state hydraulic computations by hand. A more
complicated version of this program is given in Appendix I which
performs its own steady state hydraulic computations.
It should be noted that double subscripting of variables now occurs
N=4 N=4 because quantities vary from pipe to pipe. A word of caution in regard to
( SAME AT ) selecting an N-value is advised. If a relatively short pipe occurs in the
system, it will be the “controlling” pipe with the minimum At. If there are
Figure 6-2. The a-t plane with equal At’a and large interpoladona.
other considerably longer pipes in the system, it is possible to end up with
the long pipe having a very large number of sections. Making a computer
run with a very short TMAX value and a small N-value will permit you to
see what is happening for a few At’s and give you the opportunity to make
appropriate adjustments.
This program is set up to handle an upstream reservoir with a gate
valve closure at the downstream end where the pipe exits into another
reservoir. The gate valve, whose loss coefficients are input via a DATA
statement, can be closed at two different rates.
88 UNSTEADY FLOW IN PIPELINES COMPLEX PIPE SYSTEMS 89
15 c**** ****** •• *• • ** **
2: C PROGRAM NO. 3
3: C WATER HAMMER IN SERIES PIPES
4: C (PROGRAM DIMENSIONED FOR UP TO 8 SERIES PIPES) LA»L(I)/A(I)
83:
84 : 15 WRITE(6,201) I ,0 ( I) , L( I) , A <I) , P IPEZ (I) , F (I) , VZERO (I) ,
6: C 85: SDELTT(I),NPAR(I),SINE(I),LA
7: C — RESERVOIR AT UPSTREAM END — 86: C
8: C — RESERVOIR AT DOWNSTREAM END WITH TWO-STAGE CLOSING VALVE IN LINE --
87: C ** SET UP CONDITIONS FOR STEADY STATE (T=fl) **
9: c 88 : C
10 s c PIPES ARE NUMBERED CONSECUTIVELY BEGINNING WITH PIPE 1 AT UPSTREAM END 89 : H(1,1)-HZERO
11: c — PIPE DATA ARE STACKED IN SEQUENCE BEGINNING WITH PIPE 1 — 90 : V(1,1)=VZERO(1)
12: c IOUT = INDEX WHICH GIVES AN OUTPUT EVERY IOUT-TH TIME STEP 91: DO 20 1=1,NPIPES
13 : c NPARTS = NUMBER OF PARTS INTO WHICH PIPE IS DIVIDED 92: DELL=L(I)/NPAR(I)
14 : c HDOWN = DOWNSTREAM RESERVOIR ELEV-FT 93: DELHF’l2.*F(I)* DELL*VZERO(I)**2/(64.4*D(I))
15: c HZERO » UPSTREAM RESERVOIR ELEVATION-FT 94 : AK(I)=12.*F(I)* DELT/(2.*D(I) )
16 : c NPIPES = NUMBER OF SERIES PIPES 95: K=NPAR(I)+1
17: c TMAX = MAXIMUM REAL TIME OF SIMULATION-SEC 96: X(I,1)=0.
18 : c TCI = TIME AT WHICH STAGE 1 OF VALVE CLOSURE SCHEME IS COMPLETED-SEC 97: HEAD(I,1)=H(I,1)-PIPEZ (I)
19: c TC2 « TIME AT WHICH VALVE IS COMPLETELY CLOSED-SEC 98 : RATIO(I)“DELT/DELL
20 : c PCI = PERCENT VALVE IS OPEN AT END OF STAGE 1 99 : DO 21 J = 2,K
21: c QZERO = STEADY STATE DISCHARGE-GPM 100 : X(I,J)=(J-l)*DELL/L(I)
22: c ZEND = ELEV OF DOWNSTREAM END OF LAST PIPE-FT 101: H (I, J)(1,1)-DELHF*(J-l)
23 : c HATM = ATMOSPHERIC PRESSURE HEAD-FT 102: HEAD(I,J)=H(I,J)-(PIPEZ(I)+X(I,J)*L(I)*SINE(I))
24 : c L() = PIPE LENGTH-FT D() = PIPE DIAM-IN F() * DARCY FRICTION FACTOR 103: 21 V(I,J)*VZERO(I)
25: c A() » WAVE SPEED-FPS PIPEZf) » PIPE ELEVATION AT UPSTREAM END-FT 104 : IF(I.EQ.NPIPES) GO TO 20
26 : c KI() = TABLE VALUES OF 1.0/KL 105: H (1+1,1)=H(I,K)
27: c PCT () = VALUES OF PERCENT OPEN FOR THE ABOVE KI()-VALUES 106: V(1+1,1)=VZERO(1+1)
28 : c 107 : 20 CONTINUE
29: DIMENSION X(8,30),V(8,30),H(8,30),HLOW(8,30),HHIGH(8,30), 108: DO 23 1*1,NPIPES
30: $HEAD(8,30),VNEW(8,30),HNEW(8,30) 109: K=NPAR(I)+1
31: DIMENSION L(8),D(8),A(8),PIPEZ(8),F(8),VZERO(8),NPAR(8), 110 : DO 23 J=1,K
32: $C(8),AK(8),SINE (8),AREA(8),DELTT(8),RATIO(8) BLOW(I,J)“H(I,J)
111:
33: DIMENSION KI (11),PCT(11} 23 !IH IGH (I, J) *H {I, J)
112:
34 : REAL L,LA,KI,KLI 113: PHMAX“-100.
35: LOGICAL PAIL 114: PHMIN=100000 .
36 : DATA (KI (I),1*1,ll)/0,,.0167,,0313,.0556,.100,,1787,.3333,.625, 115: C
37: $1.25,2.50,5.27/ 116: C ** WRITE OUT STEADY STATE CONDITIONS **
38 : DATA (PCT(I) ,I»1,11)/0.,10.,20.,3A.,40.,50.,60.,70.,80.,90., 100./ 117: C
39: NAMELIST /SPECS/ IOUT,NPARTS,HDOWN,HZERO,NPIPES,TMAX,TC1,TC2,PCI, 118 : WRITE(6,204)
40: SOZERO,ZEND,HATM 119: WRITE(6,205) T
41: c
READ(5,SPECS) 120: DO 22 1=1,NPIPES
42: K=NPAR(I)+1
READ(5,10 0) (DC I) ,L(I) ,F(I} ,A(I) ,PIPEZ(I) ,1=1,NPIPES) 121 :
43 :
c 122: WRITE(6,206) I,(X(I,J),HEAD(I,J),H(I,J),V(I,J),J=1,K)
44: 123: 22 CONTINUE
45: PI=3.141593
124: C
46: T»0.
125: C --------------------------------------------------------------------------------------------------------------------------------------------------------
47: FAIL*.FALSE.
126: C BEGIN TRANSIENT ANALYSIS
48: DO 11 1*1,NPIPES
127: c
49: AREA(I)».25*PI‘D(I)*D(I)/144.
128: ----------------------------------------------------------------------------------------------------------------------------------------------------------
50: C (I) => 32.2/A ( I)
IF(I.NE.NPIPES) SINE(I)»(PIPEZ(1+1)-PIPEZ(I))/L(I) 129 : C DO 99 11=1,INDEX
51: 130:
IF(I.EO.NPIPES) SINE(I)*(ZEND-PIPEZ(I))/L(I) T=T+DELT
52:
VZERO(I)=QZERO/(449.‘AREA(I)) 131: C
53: 132:
11 DELTT(I)*L(I)/(NPARTS*(VZERO(I)+A(I))) C ** COMPUTE H AND V AT INTERIOR NODES **
54 : 133 : C
55: c 134 :• DO 30 1=1,NPIPES
56 : c “ COMPUTE MINIMUM DELTA T ** 135:
c K’NPAR(I)
57: 136: DO 30 J=2,K
58 : DELT’DELTT(1) 137 : RA=RATIO(I) * A(I)
59 : KMIN»1
138: VMINUS’V(I,J)-RA*(V(I,J)-V(I,J-1))
60: DO 12 1=2,NPIPES 139: VPLUS=V(I,J)-RA*(V(I,J)-V(I,J+l))
61: I F(DELTT(I) .GT.DELT) GO TO 12 140: HMINUS’H(I,J)-RA*(H(I,J)-H(I,J-l))
62: DELT’DELTT(I) 141: HPLUS=H(I,J)-RA*(H(I,J)-H(I,J+l))
63: KMIN’I 142: VNEW(I,J)=0.5 * (VMINUS+VPLUS+C(I)*(HMINUS-HPLUS)+C(I) *DELT*SINE(I)
64 : 12 CONTINUE 143: $*(VMINUS-VPLUS)-AK(I)*(VMINUS*ABS(VMINUS)+VPLUS*ABS(VPLUS)))
65: DO 13 1*1,NPIPES 144: KNEW(I,J)=0.5*(HMINUS+HPLUS+(VMINUS-VPLUS)/C(I)+DELT*SINE(I)*
66: IF(I.EQ.KMIN) GO TO 14
145: $(VMINUS +VPLUS)-(AK(I)/C(I))*(VMINUS *ABS(VMINUS)-VPLUS*ABS(VPLUS)))
67: ANPAR=L{I)/(DELT*(VZERO(I)+A(I))) 146: 30 CONTINUE
68 : NPAR(I)’ANPAR 147: C
69 : TEST’NPAR(I) 148: C ** COMPUTE H AND V AT INTERIOR JUNCTIONS **
70: IF(ABS(TEST+1-ANPAR).LT..0001) NPAR(I)»NPAR(I)+1 149 : C
71: GO TO 13 KK»NPIPES-1
150:
72: 14 NPAR(I)’NPARTS DO 40 1=1,KK
151:
73 : 13 CONTINUE 152: K=NPAR(I)+1
74 : INDEX*TMAX/DELT + 1 153: VMINUS"V (I, K) - RATIO (I) *A (I) * (V ( I , K) -V (I, K- 1) )
75: c
154: HMINUSMI (I,K)-RATIO(I)*A(I)*(H(I,K)-H(I,K-1))
76: c ** WRITE OUT INPUT DATA **
155: CCC-VMINUS+C(I)»HMINUS+C(I)»VMINUS*DELT'SINE(I)-AK(I)‘VMIHUS*
77: c
156: SABS(VMINUS)
78 : WRITE(6,200) 157: VPLUS=V(I+1,1)-BAT 10(1+1)*A(I +1) » (V (1+1,1)-V (1 +1,2))
79 ; WRITE (6,201) LOUT ,NPARTS ,NPIPES ,Q2ERO,HDOWN ,¥.1 ERO, 7.END, HATM 158: IIPLUS-H (1 + 1 ,1) - RATIO(1 + 1) »A(I+1)*(H(1+1,1)-H(1+1,2))
80 : STMAX,DELT,TC1,PCI,TC2 CC=VPLUS-C(1+1)*HPLUS-C(1+1)*OELT*VPLUS*SINE(1+1)—AK(I+1)*VPLUS*
159:
81: WRITE(6,202) SABS(VPLUS)
82: 160:
DO 15 1=1,NPIPES HNEW(I,K) =■ (—CC*AREA(1+1)+CCC*AREA(I) ) / (C (1 + 1) ‘AREA (1 + 1) tC ( I) •
161:
Figure 6*4. Computer program for series pipes. Figure 6-4. Continued.
90 UNSTEADY FLOW IN PIPELINES COMPLEX PIPE SYSTEMS 91
Example 6-1
PIPE DIAM-IN LENGTH-FT WAVE VEL-FPS PIPEZ-FT F-VALUE VEL-FPS DELT-SEC PARTS SINE L/A-SEC
SEC-VALVE IS CLOSED
PCI = 100.00 PERCENT OPEN
.75 SEC
.111 SEC
.00 SEC
FT
GPM
FT
FT
32.0
.0
1.00
881.6
935.7
* WATER HAMMER IN A SERIES PIPE *
1
3
HZF.RO = 1000.0 FT
INPUT DATA
HATM =
TMAX =
ZEND =
TC2 =
DELT =
TCI =
IOUT =
NPAPTS =
HDOWN =
QZEP.O =
NPIPES =
1,000' 1,000'
V 2.5 fps f = .O2O VQ= 10
a = 3OOOf fps
ps
Configuration No. 1
DATA
$ SPECS IOUT=1,NPARTS”3,HD0WN»935,7,HZERO=1000,,NPIPES-2,TMAX-4.0,
TCl“0.,TC2*1.0,PCl»100.,QZERO»881.6,ZEND»0.,HATM-32.
SEND
12. 1000. .020 3000. 0.0 I
6. 1000. .020 3000. 0.0
PRESSURE HEADS, H-VALUES AND VELOCITIES AS FUNCTIONS OF TIME
PIPE 1 .000 1000. 1000. 2.50 .333 999. 999. 2.50 .667 999. 999. 2.50
1.000 998. 998. 2.50
PIPE 2 .000 998. 998. 00* IdT .333 977. 977. 10.00 .667 957. 957. 10.00
1.000 936. 936. 00-0T
XI
I
I
I
I
TIME = .111 SEC
94 UNSTEADY FLOW IN PIPELINES
PIPE 2 • 000 998. 998. 10.00 333 977 977 10.00 667 957 957 10.00
1.000 936. 936. 10.00
PIPE 1 .000 1000. 1000. 2.50 333 666 666 2.50 .667 999. 999. 2.50
1.000 998 . 998. 2.50
PIPE 2 • 0 00 998. 998. 333 977 977 10.00 .667 957. 957. 10.00
1.000 937. 937. 66*6
PIPE 1 .000 1000 . 1000. 2.50 333 666 666 2.59 667 999 666 2.50
1.000 998. 998. 2.50
PIPE 2 •000 998. 998. 9.99 333 978 978 9.99 667 959 959 9.97
1.000 941. 941. 9.94
PIPE 1 .00 0 1000. *0001 2.50 333 666 999 2.50 667 666 666 2.50
1.000 998. *866 2.50
PIPE 2 .000 998. 998. 9.98 333 980 980 9.97 667 962 962 9.94
1.000 947 . 947. 9.88
4.0
four-pipe junctions.
Three-Pipe Junctions
The model three-pipe junction is shown below in Figure 6-5, with the
flow directions indicated by vectors whose directions are dictated by the
steady state conditions. That is, the signs of terms in the equations to be
written and the characteristic lines to be followed will be determined by
the steady flow configuration (as has been done in all previous cases).
3.0
For the pipe junction with one in-flowing pipe and two out-flowing
pipes, the following equations can be written to solve for the six
unknowns.
(
Pipe 2-C = C, + C, II (6-11)
TIME sec
2.0
Pipe 3-C = C, +C< H (6-12)
CM
-
m
Continuity: Vp A, = Vp A2+Vp A3 (6-13)
Q- =0.333 SEC IN EACH PIPE
Energy: (6-14)
VALVE CLOSED
PIPE 3
0001
o
1500
2000
500
-500
A
3A3VA IV QV3H 3dnSS3dd
-
______________________ i
PIPE 2
Solving this linear set of equations gives Most three-pipe junctions fit into one of the above categories. However, it
is possible in pipe networks that a constant demand might be specified at
C] Aj ■ C3 A2 - Cs A3 a pipe junction.
(6-15)
C2AI +C4A2 +C6A3 As an example of the three-pipe junction with a constant demand, see
Figure 6-7. Equations 6-10, 6-11, and 6-12 still apply as does Equation
Back substitution into Equations 6-10, 6-11, and 6-12 will give the values 6-14. However, Equation 6-13 must be modified to the form
of the velocities.
If we have, instead, a three-pipe junction with two pipes in-flowing Vp A, = VPj A2 + Vp^ A3 + Q ..................................................... (6-20)
and one out-flowing as shown in Figure 6-6, an analysis similar to the The impact on the solution of the equations is shown below in the equation
above would give the following equations for the unknown velocities and for Hp.
H-values.
CjAj +C3A2 -CSA3 CjAj - C3A2 ~C5A3 -Q
(6-16)
C2AJ + C4A2+C6A3 C2AJ +C4A2 + C6A3 .................. (6-21)
= C, -CH -
Examination of the form of each of the equations for Hp for the
17)
= C3 -C4H various types of junctions reveals a consistency in the form which would
have permitted the engineer to write down the equations for Hp by
CM
CM
(6-18)
inspection.
-
19)
f
rQ
PIPE 3 PIPE 3
O
PIPE I
Q.
PIPE PIPE 2
Four-Pipe Junctions The results of the analysis give the following equation for Hp.
C,A, ■ C3A2 -C5A3 -C7A4 -Q > i(6.28)
The four-pipe junctions are handled using the same techniques as the
three-pipe junctions. We will examine only one possible combination and C2AI + C4A2 +C6A3 +C8A4
conclude that the others can be found by inspection.
Figure 6-8 shows the one-in, three-out four-pipe junction with the As before, back substitution into Equations 6-22, 6-23, 6-24, and 6-25
possibility of a constant demand discharge Q. The equations along the gives values for the velocities.
characteristics for this situation are As we will see subsequently when pipe networks are discussed,
computer programs can be written which investigate the pipes at each
Pipe 1-C. = C. -C ........................................... (6- junction and automatically classify them according to number and
22) junction configuration. The general approach used to perform this
Pipe 2- = C, +C,FI .......................................... classification employs the pattern we have observed in the previous
CM
M
C
= C. + C.H 23)
...
Pipe 3- 6.3 Interior Mqjor Losses
co
co
hi
C": = C„ +C„H (6-
...
CL,
Occasionally a device occurs in the interior of a pipeline which causes
4k
24)
The remaining equations are
Pipe 4-C': a significant loss in the system, either by design (a pressure reducing valve)
(6- or by necessity (a constriction, a meter, a partially closed valve, etc.). In
Continuity: V A, = V A2 + Vp A3 +V A4 + Q ............................... (6-26)
25) any case, if the loss is significant compared to other frictional losses in the
system, it must be included in the analysis.
Energy: Hp = H2 p = H ..................................... Possible approaches to approximating this loss would be to distribute
r
1 r
3
3 r
d (6-27)
i
it uniformly as though the pipe were rougher, lump it into the boundary
•"o
A
CM
KJ
............................................ (6-
Continuity gives 30)
an equation.
Continuity: V
PjAi Vp A2 ........................................... (6-31)
PIPE 2
l-ol
i
Figure 6-8. The one-ln, three-out four-pipe Junction.
PIPE I
The application of the energy equation across the loss assumes that The solution with all three constants must be recalculated and an
the loss would be the same as for steady flow at the instantaneous unsteady alternate equation for Vp2 must be derived.
velocity.
6.4 Real Valves
(6-32) Of all the unsteady flow situations the engineer will face, the largest
r*
single cause will be from valve movement. By virtue of constricting the
2g flow, the closing valve creates increasing head loss in the pipe system
Combining these equations together by substituting Equations 6-29, which causes the flow to decelerate. Different types of valves create head
6-30, 6-31 into 6-32 gives the following equation for Vp . loss in different manners which are determined by the closing procedure
as well as the mechanical design of the valve.
In a traditional sense, for steady flow hydraulics, the head loss
= 0 ............. (6-33) through a valve is characterized by the equation
instances the engineer may have information which gives the head loss of
his valve at only two or three openings. It is important that he can use this
C= information to synthesize a continuous variation of KL with valve position
and make an analysis with some confidence that his results will be
reasonably accurate.
so that Equation 6-33 becomes
In all cases the instantaneous head loss across a valve will be assumed
+ BV„ -C = to be unaffected by the unsteadiness of the flow, i.e., it will depend only on
valve stem position and valve characteristics as determined by laboratory
testing. With this in mind, we will take the approach that if valve stem
Solving,
movement as a function of time is known, and if KL is known as a
function of valve stem position, then a hydraulic head loss equation can be
written across the valve and used as a boundary condition in our method
of characteristics solution technique. Once we see how this approach can
Because C2 and C4 are always positive, B is always positive; therefore, the
provide solutions to water hammer problems, we will examine valves and
(+) sign in front of the radical must be retained. The final equation is
their head loss characteristics in more detail.
pq]cM
V 2
B
If a valve in the interior of a pipeline is programmed to close or open
as some function of time, water hammer pressures are created. The
The other velocity and the H-values may be found by back substitution solution procedure follows closely the approach used for series pipes
into Equations 6-29, 6-30, and 6-31. except the heads on each side of the valve are unequal (as with major
One additional problem is created by this type of loss. Should V P2 internal head losses). Figure 6-10 illustrates the typical situation of an
become negative, the energy equation (6-32) is likely to be invalid. Under internal boundary condition at the valve.
reverse flow, Again we have four unknowns and we need four equations. The
V2
V
P, equations are:
Energy (reverse flow): Hp +K. ------ -- HD
rev 2g
Pipe 1-C+: = C, -C.H ....................................... (6-35)
COMPLEX PIPE SYSTEMS 105
104 UNSTEADY FLOW IN PIPELINES
4CK
r
rev 6-41)
i Note that the C+ and C' characteristics are determined by the steady flow
situation and need not be redefined as the result of a temporary flow
reversal. However, the reverse-flow head loss characteristics of valves are
generally different than the normal flow characteristics. It is also likely
PIPE I PIPE 2
that they will be difficult to find, so it may be necessary to use the
forward-flow loss characteristics as a best estimate.
Figure 6-10. Valve In a conitant diameter pipeline.
Valve at Downstream End of a Pipe Next to a Reservoir
Pipe 2-C : = Ci +C2H (6-
36) One other case of valve location in a pipeline is common enough to
Continuity: warrant an analysis. In this case the valve is at the downstream end of a
(6- pipeline. It is possible that there could be some confusion as to the nature
37) of the boundary condition (it is no longer an interior boundary condition),
Energy: = H„ +K
2g hence, Figure 6-11 is provided to define the situation. There are now only
(6- two unknowns at Pi so only two equations are required. They are:
Because these equations are similar to Equations 6-27, 6-28, 6-29,38)and
6-30, we can write the appropriate form of Equation 6-31 directly. = C. - C. H (6-
42)
= 0 .................. (6-39)
+
2g
r
(6-
43)
To solve the quadratic equation, we let Arranging the simultaneous solution of these two equations in
quadratic form,
B = 2g
C = 2g
The solution is
4CK
Q
C
(6-40)
>
II
where KL0 is the loss Coefficient when the valve is completely open. This
K>
dimensionless form of loss coefficient has the advantage of varying
+
between 0 and 1.0 and is preferred by some engineers. Because head loss
Defining data will occur in both forms, and because it is easy to convert from one
form to another, this work will use the more common KL form.
B= Several attempts have been made to present the water hammer
pressures generated by valve closure in graphical form. Wood and Jones
[17] briefly comment on these methods and present their own more
C= comprehensive graphs. As they point out, it is impossible to cover all the
effects of friction and system configuration in simple graphical form. This
permits us to write the quadratic equation solution as conclusion is valid and it is for this reason that we will not devote time to
understanding the available graphs, but rather work on computerizing the
procedure in a manner which will permit us to utilize any valve in any pipe
........................ (6-44) configuration.
1+ We will begin by synthesizing KL as a function of position for the
situation where there is a minimum of loss coefficient information. In this
or for reverse flow, example, consider a gate valve whose loss coefficients are known for only
four positions (see Vennard [18] and Table 6-1 below). Additional data on
loss coefficients for other types of valves is given in Appendix II. The next
....................... (6-45) step is to make a plot of KL and 1/KL as a function of percent-open by
rev
plotting the values from Table 6-1 and sketching in a smooth curve. The
result is shown in Figure 6-12.
It is now clear that the impact of valve movement on pressures and It is clear that a problem is encountered when sketching KL near the
velocities in the system can be calculated if we can determine the valve closure. To avoid this problem, we will work only with 1/KL which
particular value of KL for a given valve closing (or opening) scheme. We can be sketched in with more confidence. It should be noted also (see
will now examine the technique and resources available to accomplish this Equations 6-40 and 6-44) that 1/KL *S the term appearing in our analysis.
objective. Next, we construct a table of 1/KL values for even increments of
percent-open for the valve by reading the graph just constructed.
The resulting table is given as Table 6-2.
Finding KL Now, if the valve closing schedule is known, i.e., valve stem position
(percent open) is known as a function of time, we need only to read Table
The procedure for solving the transient problem was described 6-2 to find 1/KL an^ so^ve the appropriate equation at the given time step.
previously. It remains now only to find KL as a function of time. If there To accomplish this, the computer must be programmed to perform table
are tests available for the proposed valve which detail KL as a function of interpolation. This is done by fitting a curve (or a straight line) between
several stem positions, then there is little to do except develop an
interpolation scheme between data points in the KL VS. percentage-open
table. If there are only two or three data points available, then more work
is involved in establishing the KL VS. percentage-open table. Table 6-1. Losa coefficient! for a gate valve.
Head loss characteristics of valves are generally expressed one of two
ways. One of these ways is the loss coefficient KL used before. The second
method is by use of the function T where T is defined as % Open IK
1.1 0.18
II
04
5 0.04
108 UNSTEADY FLOW IN PIPELINES COMPLEX PIPE SYSTEMS 109
100 5.27
90 2.50
35
80 1.25
o
70 0.625
c
X
<
1“
L2a 60 0.333
I
th
(Vennard " 4 Ed.) 50 0.179
30
40 0.100
30 0.0556
20 0.0313
25
10 0.0167
0 0.0
time, finding the value of 1/KL at 72.4 percent open, we would calculate it
20 *
from the following equation and the data in Table 6-2.
I / Ku
72,4-70.00
1/K (1.25 -0.625)+ 0.625 ............................... (6-47)
II
80.0 - 70.0
ui
-j
10
The information on 1/KL for various percent-opens is stored in two
one-dimensional arrays by means of DATA statement shown below.
DATA (KI(I),1=1,11)/0.,.0167,.0313,.0556,.100,.1787,.333,.625,
$1.25,2.50,5.27/
DATA (PCT(I),1=1,11)/0.,10.,20.,30.,40.,50.,60.,70.,80.,90.,100./
% VALVE IS OPEN Because of its simplicity, the linear interpolation procedure should be
used whenever possible. However, for functions which are sharply curved
Figure 6-12. and 1/K^ a* function! of percent open. or for table points which are too far apart, higher order interpolation
should be used.
two data points in the table and interpolating. In our problems, straight
line interpolation is often adequate and parabolic interpolation (using a Parabolic Interpolation
second degree equation) is broad enough to cover any case.
The purpose of parabolic interpolation is to obtain more accurate
Linear Interpolation interpolation than possible with the linear method. While this result is
generally achieved for relatively smoothly varying functions, those with
Because linear, or straight line, interpolation is easiest to understand, sharp curvatures or points of inflection can result in poor interpolations.
we will cover it first. If the valve closing schedule required, at a particular
xxv UNSIXUADY FIJUW IN nrtjLINEB COMPLEX PIPE SYSTEMS 111
DO 32 1=1,11 B= (6-50)
ITEST=(OPEN-PCT(I))*.10 2 Ax
IF(ITEST.EQ.0) GO TO 33
32 CONTINUE Recognizing that the two coordinate systems are related by the
33 FACT=(OPEN-PCT(I))*.10 equations
KLI=KI(I)+FACT*(KI(I+1)-KI(I))
fix ) = fix)
+
Figure 6-13. Linear interpolation computer code.
c
II
X
The technique is to fit a parabola through the three data points which the parabolic equation can be transformed from local coordinates to the
bracket the range over which interpolation is required (see Figure 6-14). original coordinate system.
Once the parabolic equation has been found, the interpolated value is
calculated by direct substitution. The value x in Figure 6-14 can be fix) = f(xn) + { f(xn+1) + f(xn4) - 2f (Xn) }
thought of as the percent-open of the valve and the value fix) can be
thought of as the 1/KL value. A displaced local coordinate system is ...................... (6-51)
located at xn and the parabolic equation is written in terms of this local
+
coordinate system.
While this equation appears bulky, the computer programming is straight
+
CQ
+
A
A
_ f<xn+l ) + f(Xn-l) - 2f (X„) fix) = fix,) + { f(x3) + f(Xl) ■ 2 f (x2) )
T73 ................. ........................................... (6-49)
Vi co
M O = Data points into computer
TRUE CURVE shown below in Figure 6-15. While the code could be incorporated into
fix) subrputine, it is so brief that this is not necessary.
PARABOLA
Effect of Valve Closure Schedules on
Water Hammer Pressure
f(x)
f(Xn) The use of real valves in an unsteady flow situation has a more
substantial impact on pressures than our previous technique of artificially
f (Xfi-i) varying the velocity at the valve. This effect is more pronounced with gate
valves where the valve must be almost closed before it generates enough
head loss to slow down the flow noticeably. The result is that, for simple
pipe-reservoir systems, the linear valve closure time must be substantially
X
x
n-i x
nx x
n+i greater than 2 L/a in order to reduce the water hammer pressure
appreciably below the value computed for sudden valve closure. To
Figure 6-14. Definition sketch for parabolic interpolation. illustrate this effect an example problem was worked with three closure
schedules.
112 UNSTEADY FLOW IN PIPELINES COMPLEX PIPE SYSTEMS 113
IF(OPEN.LT.10.) GO TO 9000
DO 9001 1=2,11
ITEST=(OPEN-PCT(I))*.10
IF(ITEST.EQ.0) GO TO 9002
9001 CONTINUE
9002 FACT=(OPEN-PCT(I))*.10
KLI=KI(I)+.5*FACT*FACT*(KI(I+1)+KI(I-1)-2.*KI(I))+.5*FACT*
$(KI(1+1)-XI(1-1))
GO TO 9004
I
9000 FACT=OPEN*.10 FOR SUDDEN VALVE CLOSURE
KLI=KI(1)+.5*FACT*FACT*(KI(3)+KI(1)-2.*KI(2))-.5*FACT*
$(KI(3)+3.*KI(1)-4.*KI (2))
9004 CONTINUE 700
Figure 6*15. Parabolic Interpolation computer code.
Example 6-2
A pipe-reservoir system has a gate valve at the downstream end with a LINEAR VALVE CLOSURE
steady state head of 300 feet at the valve. For sudden valve closure, an 600
increase in head AH = 431 feet is realized. CLOSE TO 10% IN I SEC
The pressure head at the valve for the following three closure CLOSE TO 0% IN 5SEC
schedules is shown to compare with sudden closure.
CLOSE TO 5% IN I SEC
HEAD AT VALVE-ft
a) Linear closure in 6 seconds CLOSE TO 0% IN 5 SEC
b) Close linearly to 10%-open in 1 second; close linearly to 0%-open
in the next 5 seconds.
c) Close linearly to 5%-open in 1 second; close linearly to 0%-open
500
in the next 5 seconds.
The results of this analysis are shown in Figure 6-16. It is clear that the
last 5 percent or less of valve closure is critical in this case. As an
approximation here; it appears from case (a) that the last 2 percent of
valve closure should be accomplished in a time greater than 2 L/a if any
significant reduction in pressure is to be realized.
10
st
O
r
actuated and undamped so that it responds instantaneously to changes in
<
>
5
o
r
flow conditions. Because the purpose of the PRV is to maintain (within
limits) a specified pressure on the downstream side, the pressure there
TIME - sec
remains fixed as long as the upstream pressure is greater. Figure 6-16. Water hammer resulting from various gate valve closure
schemes.
114 UNSTEADY FLOW IN PIPELINES COMPLEX PIPE SYSTEMS 115
\
pressure setting plus head loss through the valve, then the specified
downstream pressure cannot be held and the valve acts as a major interior
\°+
loss in the pipe. If the pressure situation is altered so greatly as to cause
\
\
back flow in the line to begin, the PRV generally acts as a check valve and 2^—
+
prevents back flow.
It is clear that the PRV can operate in any one of three modes. The
PRV 4-
6
\ PIPE 2
equations describing these situations follow. We assume first that the
PRV is operating in its intended manner. If solution for velocity or
pressure indicates it is operating in a different mode, then we must switch Figure 6-17. Definition sketch for the pressure reducing valve.
to the appropriate set of equations to solve.
For normal operation, the equations are derived as though the PRV
then the flow is still positive, but the pressure on the downstream side has
were a major interior loss. Figure 6-17 defines the flow variables. The
equations along the characteristic lines are dropped below the specified value. We must now resort to Equations 6-29,
6-30, 6-31, and 6-34 to find the unknown velocities and H-values.
IfHp becomes larger than Hpt (orVP2 becomes negative), then the
Pipe 1 - C. = -c (6- velocity must be set to zero and Equations 6-53 and 6-54 used to find the
C» 53) respective H-values.
Pipe 2 • C':
= c3 (6- Slower Acting Pressure Reducing or
54) Pressure Sustaining Valves
Continuity and energy give
Continuity: Most valves of this type are operated by a pilot system which senses
the pressure on the downstream side of the valve and actuates a
(6- diaphragm to change the valve setting and maintain the required
Energy: = H„ +K 55) downstream pressure. Because this pilot operation requires a flow of
liquid to move through the pilot system, there is a finite response time to
sudden and large pressure changes which should be of concern.
(6-2 =
These equations are altered for normal PRV operation in that Hp Most of these types of valves were designed to respond to slower
HpRv = constant. This means that for normal operation 56) varying pressures than those normally found in water hammer situations.
Further, the time-varying response of these valves is unknown and varies
(6-57) from valve to valve as well as with the magnitude of the water hammer
pressure increase. In most instances, these problems preclude any
and sophisticated analysis and the system is assumed to respond instanta-
neously. The approach in the previous section is then used to find
velocities and H-values.
ii
.................................................................... (6-58)
n
l 2
►o
V
only to relatively simple systems. However, we will discuss all three of the
K>
116 UNSTEADY FLOW IN PIPELINES COMPLEX PIPE SYSTEMS 117
most popular methods, the Hardy Cross method, the linear theory method To produce the proper sign on AQ, the denominator is kept negative and
and the Newton-Raphson approach. the terms in the summation in the numerator are positive or negative,
depending on whether one moves with or against the flow while proceeding
Hardy Cross Method clockwise around the loop.
Once the AQ is computed for each loop, it is added (or subtracted)
Because of its simplicity of application, its easily understood theory
from the flow rates in each member of the loop to get a better estimate of
and its amenability to hand calculation, the Hardy Cross method has
the true flow rate. Because the decomposition of (Qj + AQ) n with the
enjoyed (and still enjoys) considerable popularity among practicing
binomial theorem was not exact and because pipes which are common to
engineers.
The basic approach is to first estimate flow rates in all the pipes in a more than one loop have multiple AQ corrections, the calculated AQ’s will
network so that continuity is satisfied at each junction (node). Of course it not be correct. Therefore, the process is iterative and must be continued
is unlikely that the EL-HGL is continuous throughout the network until the error is acceptably small (or no convergence to a solution occurs).
because the original estimates of the flow rates are always erroneous to Although this numerical method is not so sophisticated as the other
some degree. Therefore, this method assumes that there can be found a methods, the results are just as valid, provided convergence is obtained.
unique flow rate adjustment which can be applied to each loop in the Actually, a more careful investigation would reveal that Hardy Cross
network which will cause the EL-HGL to be continuous around each loop. analysis is a decoupled Newton-Raphson analysis.
In hydraulic terms this continuity is expressed as To complete the presentation on Hardy Cross analysis, a simple
program is listed and an example is presented. The main purpose is to
illustrate the simplicity of the approach for uncomplicated networks.
Z Mil
around each loop where “i” is the pipe number and N is the number of
pipes in the loop. Assuming that the head loss can be written in the form Hardy Cross Example 6-3
N i cfs
t=i
2?/(5) / 17J
Because “n” is generally a non-integer, Equation 6-60 is generally O/
I—I
'55
^22e^§
expanded by the binomial theorem to yield an equation for AQ. Retaining
-
only the first two terms of the binomial expansion, we get the following
equation for AQ.
a
(3)
ZZ
2OOQ-I8-.0165
*’ 3
M
t6l
O
,
AQ i=,i 1 1
......................................... . ................. (6-61) \
N 1.78 cfs
-nS K. Qn_1
i=l 11
118 UNSTEADY FLOW IN PIPELINES COMPLEX PIPE SYSTEMS 119
17: WRITE(6,93) (NO <I) ,D(I) ,AL(I) ,E(I) ,Q(I) ,NP(I) ,NS(I) ,I = 1,N)
18: IT-0
with it. 19:
20:
DO 3 1 = 1, N
AD=D(I)/12.
4. Construct a table for each loop as an orderly means of computing the 21: 3 AK(I)»F(I)*AL(I)/(64.4‘(AD**5.)*.7854*.7854)
22: 4 DO 2 1 = 1, KK
AQ-value for each loop. 23: SN(I)=0.0
24 : 2 SD( I) =0.0
Enter the first estimate for Q in the second column of the table for
C/i
DATA this. In the linear theory the nonlinear equations for the sum of the head
losses around each loop are linearized. Then the complete set of linear
10 4 0.01
1 18. 1600. .0175 2.45 1 equations (the continuity equations are already linear) are solved.
2 15. 2000 . .018 1.00 1 2
3 18. 2400 . .0175 -2.00 To understand how the procedure works, we need to look at the
4 12.
5 12.
1800.
1900 .
.019
.019 .50
.78 2
2 3
equations involved. For each loop in the network the following equation is
6 10.
7 15.
1300.
1700.
.019
.018
-.83
.20
3
3 4
valid.
8 18. 2000 . .0165 -1,50 3
9 24. 1200. .016 2.31 4 N
10 15. 1800. .018 .08 4
MZ
h = s K Qn = 0
L
i i=l
Computer Output
where N is the number of pipes in the loop. To linearize this equation the
Qjn is decomposed into two pieces so that this equation becomes
THE PIPE NETWORK IS COMPOSED OF THE FOLLOWING
zwy
PIPE NO. DIAM. -IN . LENGTH-FT DARCY F O-CFS PRIMARY LOOP SECOND. LOOP S (KiQ.n'1)Qi = =0
O'
1 18.0 1600.0 .0175 2.45 1 0
2 15.0 2000.0 .0180 1.00 1 2
i= 1
3 18.0 2400.0 .0175 -2.00 1 0
4 12.0 1800.0 .0190 .78 2 0
5 12.0 1900.0 .0190 .50 2 3
6 10.0 1300.0 .0190 -.83 3 0 Of course Kj is now a function of Qj so we are still involved in an iterative
7 15.0 1700.0 .0180 .20 3 4
8 18.0 2000.0 .0165 -1.50 3 0
process.
9 24.0
15.0
1200.0 .0160 2.31 4
4
0 As we solve the set of linear plus linearized equations, the estimate of
10 1880.0 .0180 .08 0
Kj’is revised after each solution. After several iterations the values of Qj
and Kj should converge to their final values. The mathematical form of
the iteration equation is
THE RESULTS OF THE NETWORK ANALYSIS ARE
Z M II
o
« =0
o’
O'
PIPE NO. K 0-CFS HEAD LOSS'
1 .092818 2.20 .45
3 .139228 -2.25 .70 where j is the iteration number. For example, if we are making the 8th
4 .860911 .61 .32 iteration (j=8) then we would calculate Kj r’s from the 7th iteration.
5 .908740 -.27 .07 Experience with the linear theory has shown that the numerical
6 1.547161 -.24 .09
solution tends to oscillate around the final values. To damp out this
7 .252408 .65 .11
numerical oscillation, the iteration equation is altered to include the last
8 .109393 -.91 .89
two iterations for Qj in computing Kj.
9 .015104 2.46 .09
+
o
Linear Theory .0) = 0 ..................... (6-62)
O
NJ
J
The linear theory method is a technique for solving a set of network
equations, some of which are nonlinear, for the unknown flow rates in the
pipes. The equations are generated by writing continuity equations for When starting an analysis, only the direction of flow (not the amount)
flow into and out of each junction and by specifying that the algebraic sum has to be specified. This is a substantial savings in effort over the Hardy
of the head losses around each loop is zero. Solving a set of nonlinear Cross approach. For the first iteration, Kj(>) is assumed to be equal to Kj.
equations is an iterative process and there are many techniques for doing For the second iteration, Kj^) will equal KjfQjUljn-l. Thereafter,
Equation 6-62 will be used for each loop.
122 UNSTEADY FLOW IN PIPELINES COMPLEX PIPE SYSTEMS 123
The previous example used for the Hardy Cross analysis will now be C'
c PROGRAM NO. 5
solved using the linear theory. The linear theory program used is shown in c
c
ELEMENTARY LINEAR THEORY - DARCY-WEISBACH VERSION
PREPARED BY G. WATTERS FROM PROGRAM BY JEPPSON
Figure 6-19. On the following page, the data cards and the computer C1
c
output from the analysis are shown.
This program is more powerful than the previously listed Hardy Cross c
program in that it calculates its own Darcy-Weisbach friction factor from c
DIMENSION A(50,51)
IF A( , ) CHANGES DIMENSIONS, STATEMENT 8888 AND 9999 ALSO MUST BE CHANGED
the input pipe roughness and liquid viscosity. The simultaneous equation c
REAL L,KP
solver is appended to be used as a subroutine to solve the set of linearized INTEGER XFLOWS
LOGICAL CFS,GPM,CMS
equations. DATA DI/4IHIN) ,4H(FT) ,4H (M) , 4 H (CM)/LE/4 H (FT) , 4 H ( FT) , 4 H (M) ,
$4H (M)/HLS/4H(FT),4H(FT),4H (M),4H (M)/VIS/5HFT-SQ,5HPT-SQ,5H M-SQ
$,5HCM-SQ/VIS 2/4H/SEC/
DATA DJSCHE/5H(GPM)/DISCHC/5H(CFS)/DISCHM/5H(CMS)/
Newton-Raphson Method NAMELIST /SPECS/ NPI PF.S , NJ UNCT, NLOOPS , ITERAT,NUNIT,ACC,VISCOS,
22 : $XFLOWS,CFS,GPM,CMS
23: C
The Newton-Raphson technique has the same conceptual basis as the 24 :
25:
NPIPES«0
30 CONTINUE
Hardy Cross method. The flow rates in each pipe are assumed which 26 :
27:
GPM=.FALSE.
CFS=.FALSE.
satisfy continuity and these flow rates are corrected so that the sum of the 28 :
29:
CMS’.FALSE.
DELQl’0,05
head losses around each loop is equal to zero. In the Hardy Cross method 30:
31 :
IF(NPIPES.EO.fl) GO TO 330
DO 777 I»1,NJUNCT
the flow rates in each pipe are corrected after each AQ computation. In the 32: JZ=NN(I)
33 : NN ( I)=0
Newton-Raphson method the equations containing AQ are written for 34 : OJ(I)=0.
35: DO 777 J-l ,JZ
each loop, then this nonlinear set of equations is solved for the final value 36 : JN (I,J)“0
37: 777 CONTINUE
of AQ in each loop. When the solution is complete, only then are the initial 38 : DO 778 I’l ,NLOOPS
39: JZ’LL(I)
flow rates in each pipe adjusted to their final value. 40: LL (I) =0
41 : DO 778 J»1 ,JZ
The method gets its name from the technique used to solve the 42: LP (J, I)=0
43: 778 CONTINUE
nonlinear set of equations. The Newton-Raphson technique is a frequently 44: DO 779 I=1,NPIPES
45: JC (I)=0
used, powerful method of numerical analysis. In operation, it adjusts 46 : NPIPE(I)=0
47: JUP(I)=0
successive approximations to the solution by computing the way the 48: JDN(I)=0
solution is moving with respect to each variable and then, based on that 49 :
50 :
D ( I)=0 .
L(l)=0.
computation, calculates new trial values for the unknowns. 51:
52:
E(I)=0.
KP(I)=0 .
The method is best understood by examining it in one-dimension 53: Q( I)“0.
54 : EXPP( I) =0 .
(solution of one equation). The equation to be solved is written 55:
56 :
AR(I)«0.
ARL(I)=0.
57 : 779 CONTINUE
58 : 8888 DO 780 1=1,50
F (x) = 0 59 : DO 780 J«1,51
68 : 780 A (I,J)=0.
61: C
and we must find the value (or values) of x which satisfy the equation 62:
63:
C tlPIPES
C NJUNCT
= NO. OF PIPES
» NO. OF JUNCTIONS
(make F = 0). Our first guess will be x(1) and when we put this number in 64 :
65:
C NLOOPS
C ITERAT
= NO. OF LOOPS
= MAX. NO. OF TRIALS ALLOWED IN SEARCH FOR SOLUTION
the equation, F (x(O) 0 will result. The question is, what is the next 66 :
67:
C NUNIT
C
« B - D AND E IN INCHES, L IN FEET, VISCOS IN FT-SQ/SEC
= 1 - D AND E IN FEET, L IN FEET, VISCOS IN FT-SO/SEC
logical guess for x? To understand the Newton Raphson technique, refer 68 :
69:
C
C
= 2 - D AND E IN METERS, L IN METEPS, VISCOS IN M-SO/SEC
= 3 - D AND E IN CENTIMETERS, L IN METERS, VISCOS IN CM-! ■SO/SEC
to Figure 6-20. The solid line represents F (x) and the true solution is the 70 :
71 :
C ACC
C VISCOS
= ITERATION ACCURACY TO WHICH DISCHARGE IN EACH PIPE IS (
= KINEMATIC VISCOSITY OF LIQUID IN UNITS OF 'NUNIT'
COMPUTED
x-value where F (x) crosses the x-axis. In this technique the slope of F (x) is 72:
73:
C XFLOWS
C CFS
= NO. OF EXTERNAL FLOWS AT JUNCTIONS
= LOGICAL. IF .TRUE,, EXTERNAL FLOWS ARE IN CFS
evaluated at x^), projected to the x-axis and that value of x is used as x<2). 74 :
75 :
C GPM
C CMS
= LOGICAL. IF .TRUE,, EXTERNAL FLOWS ARE IN GPM
8
LOGICAL. IF .TRUE., EXTERNAL FLOWS ARE 111 CU-M/SEC
The cyclic process continues until the value of |x0) - x0’^| is less than 76: C
77 • •c
some specified tolerance. 78 : C READ IN GENERAL NETWORK DESCRIPTION
330 READ(5,SPECS,END®99)
In mathematical form, the previous statements mean C -------------------------------------------------
C
■c
N.7 = NJUNCT
4)
dF = FQ - 0 NP»NPIPES
tJL’NLOOPS
dx x(j-l) - x« NPP’NPIPES+l
169: JU=JUP(1)
86: NJ1-NJUNCT-1 170: JD-JON(I)
87: IF(NUNIT,GE.2) DISOUT-DISCHM 171 : NN (JU)=NN(JU) +1
88: IF (NUNIT.LE.l.AND.GPM) DISOUT-DISCHE 172: NN(JD)=NN(JD)+1
89: IF (NUNIT.LE.l.AND.CFS) DISOUT-DISCHC 173: JJ-NN(JU)
90 t C 174: KK=NN(JD)
91: C WRITE OUT PAGE HEADING 175: JH (JU,JJ)-I
92: WRITE(6,200) 176: JN (JD,KK)= -I
93: 200 FORMAT(1H1,llX,27('*')/12X,'* INPUT DATA FOR NETWORKS *'/12X, 177: 300 CONTINUE
94 : $27('*')//) 178: c
95: C 179:
96: C WRITE OUT GENERAL NETWORK CONFIGURATION
97: WRITE(6,203) NPIPES,NJUNCT,NLOOPS,XFLOWS,ITERAT,ACCtDISOUT,NUN IT, 180: c READ IN PIPES COMPRISING EACH LOOP. EACH PIPE NUMBER IS GIVEN A SIGN.
98: SVISCOS/VIS(NUNIT+1),VIS2 181 : c WHEN MOVING CLOCKWISE AROUND THE LOOP, IF YOU ARE MOVING IN THE
99: 203 FORMAT(10X,'NO. OF PIPES =',I3/6X,'NO. OF JUNCTIONS =',13/ 182: c ASSUMED FLOW DIRECTION, THE SIGN IS (+). IF YOU MOVE AGAINST THE FLOW,
100 : $10X,'NO, OF LOOPS =',13/IX,'NO. OP EXTERNAL FLOWS -',13// 183: c THE SIGN IS (-).
101: $4X,'MAXIMUM ITERATIONS I3/4XITERATION ACCURACY - *,F6.3,IX,A5 184 : c JZ = NUMBER OF PIPES IN LOOP
102: S//17X,'NUNIT I2//13X,'VISCOSITY = ' , F9.7,1X,A5,A4//) 185: c LOOPS SHOULD BE CONSIDFRED IN NUMERICAL SEOUENCE
103; C 186: c
104 : C NPIPE() » ARRAY OF PIPE NUMBERS 107: DO 35 1=1,NLOOPS
185: C D() - PIPE DIAMETER - UNITS ACCORDING TO 'NUNIT1 1(18: READ(5,102) JZ,(LP(J,I),J = 1,JZ)
106: C L() = PIPE LENGTH - UNITS ACCORDING TO 'NUNIT' 189: 35 [,L(I)=JZ
107: C E() -PIPE ROUGHNESS - UNITS ACCORDING TO 'NUNIT’ 190: 102 FORMAT(8I10)
108: C JUPO = JUNCTION NO. AT UPSTREAM END 191 :
109: C JDNO - JUNCTION NO. AT DOWNSTREAM END 192: c
110: C 193: c WRITE OUT PIPES IN EACH LOOP
194: WRITE(6,206)
111: C ------------------------------------------------------------------------------------------------------------------------------------------------------------- c
195: 206 FORMAT(// 11X,'*** PIPES IN LOOPS * *’//10X,'LOOP 1 ' ,8X, 'PIPES'/
112: C READ IN DATA ON PIPES
113: 196: 510X, ,2X.23('-'))
READ(5,100) (NPIPE (I) ,D(I) ,L(I) ,E(I) ,J UP (I),JDN(I) ,1-1,NPIPES)
114: 106 FORMAT(110,3F10.0,2110) 197: DO 36 1=1,NLOOPS
115: C------------------------------------------------------------------------------------------------------------------------------------------------------------- c 198 : LLL=LL(I)
116: c 199 : WRITE(6,207) I,(LP(J,I),J=1,LLL)
117: C WRITE OUT PIPE INPUT INFORMATION 200 : 36 CONTINUE
118: WRITE(6,201) DI(NUNIT+1),LE(NUNIT+1),DI(NUNIT+1) 20] : 207 FORMAT(10X,13,2X,15(1X,I 3) )
119: ' WRITE(6,202) (NPIPE(I),D(I),L(I),E(I),JUP(I),JDN(I),1-1,NPIPES) 202 : DO 5 1=1,NP
203 : IF(NUNIT .GT. 1) GO TO 66
120: 201 FORMAT(2X,'PI PE DIAM LENGTH ROUGHNESS J-UP J-DWN’/
20 4 : KP(I)*« 0009517* I, (I)/D (I) **4.87
121: $8X,A4,3X,A4 ,5X,A4/2X, — ------- ',6('-'),2X,9('-'),2X,4(’-'),2X, 205: GO TO 5
122: 55()) 206 : 66 KP (I)=.00212 *L(I)/D(I)* * 4.87
123: 202 FORMAT(2X,13,2X,F5.1,IX,F7.0,IX,FB .6,4X , 13,4X , 13)
207: 5 CONTINUE
124: C
208 : ELOG=9.35*ALOG10(2.71828103)
125: DO 48 1=1,NP
209 : NCT=0
126: 48 F.(I)-E(I)/D(I) 210: c
127: IF(NUNIT-l) 40,41,42 211: c CALCULATE ELEMENTS OF COEFFICIENT MATRIX
128: 40 DO 43 1 = 1,Np 212: c
129: 43 D(I)-D(I)/12. 213: 20 CONTINUE
130: 41 G2-64.4 214: DO 6 1 = 1 , NJ1
131: GO TO 50 215: DO 7 J = 1 , N P
132: 42 IF(NUNIT .EQ. 2) GO TO 45 216: 7 A(I,J)=0.
133: DO 46 I»1,N P 217: NNJ-NN(I)
134: 46 D (I) = .01*D(I) 218: DO 8 J=1,NNJ
135: VISCOS=.0601*VISCOS 219: IJ=JN(I,J)
136: 45 G2-19.62 220: IF(IJ .GT. 0) GO TO 9
137: 50 CONTINUE 221: IIJ=ABS(IJ)
138: DO 70 1=1,NP 222: A (I, IIJ)—1.0
139: AR (I)=0.7854*D(I)**2 223 : GO TO 8
140: 70 ARL(I)=L(I)/(G2*D(I)*AR(I)**2) 224 : 9 A(I,IJ)=1.0
141 : C 225: 8 CONTINUE
142: c ............. ............. ............. - --------------- -------------------------------------------------------------------------------- - --------------------- C 226 : IF(ABS(QJ{I)).LT.0.0001) GO TO 10
143: c READ IN EXTERNAL FLOWS AT JUNCTIONS 227: A(I,NPP)-QJ(I)
144: C I = JUNCTION NUMBER
228 : GO TO 6
145: C QJ () « EXTERNAL FLOW IN UNITS GIVEN BY LOGICALS 'GPM', 'CFS', 'CMS'
229: 10 A(I,NPP)=0.
146: C IF FLOW IS INTO NETWORK, SIGN OF QJ () IS (+) - IF OUT OF NETWORK, SIGN IS (-)
230 : 6 CONTINUE
147: READ(5,101) (I,QJ(I),11 = 1,X FLOWS) 231 : DO 11 I=NJ,NP
148: 101 FORMAT(4(IIP,F10.0))
C---------------------------------------------------------- -------------------------------------------------------------------------- - ---------------------- c 232 : DO 22 J«1,NP
149: 23 3 : 22 A(I,J)=0.
C
150: 234 : II-I-NJ1
C WRITE OUT EXTERNAL FLOWS AT JUNCTIONS
151': 235 : NNJ-LL(II)
WRITE(6,204)
152: 236: DO 12 J-l.NNJ
204 FORMAT(///l1X,'*** EXTERNAL FLOWS ***'//10X,'JUNCTION INFLOW OU
153: 237 : I J = LP(J,11)
STFLOW*/10X,8() ,3X,6( ' - ’) ,2X,7 ('-'))
154: 238 : IIJ=ABS(IJ)
DO 771 I=1,NJUNCT
155: 2 39: IF(IJ .LT. 0) GO TO 13
QIU-0 .
156: 240 : A ( 1, 11J)=KP(11J)
QOUT-0,
157: 241 : GO TO 12
IF(ABS(QJ(I)).LT.0.001) GO TO 771
158: 242 : 13 A (1,11J)=-KP(11J)
IF(QJ(I).LT.0 , ) OOUT-QJ(I)
159: 24 3 : 12 CONTINUE
IF(QJ(I).GT.0.) QIN - QJ(I)
160: 24 4 : 11 A(I,NPP)=0.
IF(GPM) QJ(I)»QJ(I)/449.
161: 24 5 : V(1)=4 .
WRITE(6,205) I,QIN,QOUT
162: 246 : c
771 CONTINUE
163: 247 : c SOLVE SIMULTANEOUS EQUATIONS
IF (GPM) ACC-ACC/449.
164 : 248 : c
205 FORMAT(12X,13,4X,F8,1,IX,F8,1)
165: 249 : 9999 CALL GJR(A,51,50,NP,NPP,JC,V)
C
166: 2 50 : IF(NCT.EQ.0) GO TO 499
C DETERMINE SIGNS OF PIPE NUMRERS AT EACH JUNCTION
167: 251: DO 49 I=1,NPIPES
168: DO 300 1 = 1,NPIPES 252 : BB=A(I,NPP)
DATA
Computer Output
•**•********•**•»*»»»*»•*»*»***
* FINAL NETWORK CONFIGURATION *
ITERATIONS ■ 3
IL.
10 15.00 1800. 107.23 .02
IL.
NO. OF PIPES = 10
NO. OF JUNCTIONS = 7
NO. OF LOOPS = 4
NO. OF EXTERNAL FLOWS = 7
MAXIMUM ITERATIONS - 20
ITERATION ACCURACY * 1.000 (GPM)
NUNIT ’ 0
o.
X
4 12.0 1800 . .010200 2 4 dF<J-D
5 12.0 1900 . .010200 4 3
6 10.0 1300 . .009000 5 4 dx
7 15.0 1700 . .010200 5 6
8 18.0 2000. .009000 3 6
9 24.0 1200 . .010200 5 7
10 15.0 1800. .010200 7 6 Because the functional form of F is known, the derivative can be found
and its numerical value determined by substitution of the appropriate
x-value.
130 UNSTEADY FLOW IN PIPELINES COMPLEX PIPE SYSTEMS 131
Of course it is possible that there is more than one solution to the approximation for x; and the inverse computed. However, this is a very
equation (F(x) crosses the x-axis more than once) and it is possible to find large computational task for large numbers of simultaneous equations,
the wrong one. Also it is possible to start searching at an x-value which hence, we employ a slightly different approach when working with
leads you further from the solution rather than closer. Fortunately, the hydraulic networks.
network equations generally do not exhibit this behavior and solutions for In hydraulic networks,
AQ are unique.
The Newton-Raphson technique in two or more dimensions (two or
MZ
more equations with two or more unknowns) proceeds in the same manner KQ" = 0
ll
except each function Fj can depend on a large number of xj^’s. As a result =1
the iteration equation for finding the next trial values for the x^’s is not so
simple as Equation 6-63. However, because the Qj’s are unknown, we must guess a value Q Oj in
We now have a set of equations each pipe and search for AQ’s which will correct the Q0’s to the proper
F
value. The loop equations now are of the form
1 (xI,x2,x3, ...,x ,...x^) =
O
F2 (Xl, XJ , x3
a
r
*
(6-67)
II
o
= 0, k= 1,K
because any pipe in a given loop may be a member of other loops and their
II
o
In finding the next trial value for the x^’s, we use the equations AQ’s must be included. Then, in general, we may write
dFk0-D Fk (AQj.AQ,, ........... AQK) = 0, k=l,K ................................... (6-68)
dx7
If we now represent AQ® - APO'!) as dQ® then Equation 6-65 can
be used to show
4
-
where k represents the loop number. There are K of these equations, one This equation is now solved for dQ® and we can write
for each loop.
It is most convenient to express this set of equations in matrix form AQ® = AQ®1}
4
-
after each iteration. When dQ becomes small enough, an acceptable
{ F®®} + [J®1^ x® -x®1*} = {0} ..............................................(6-65) solution has been obtained.
It should be noted here that this method only requires the solution to
where [J] is a K x K matrix of (dFCHVdxfc) known as the Jacobian. a set of equations equal in number to the number of loops. The linear
Converting this to the same form as Equation 6-63 gives theory must solve a set of equations equal in number to the number of
unknown flow rates. Consequently, the Newton-Raphson technique
requires substantially less storage space and computer time for solving the
[j(J-D]{xw} -[J®1)] = -{ F®1}) set of equations.
or
Newton-Raphson Example 6-5
■ [J®1} 1 _1{ F®1}) .................................. (6-66)
An example of the Newton-Raphson approach has been chosen which
In one dimension, the inverse of the Jacobian is l/(dF/dx). Because all the can also be used to illustrate the transient program. A sketch of the
F
k can be differentiated, the Jacobian can be evaluated at each new network is shown on Figure 6-21. Because this program will handle
132 UNSTEADY FLOW IN PIPELINES COMPLEX PIPE SYSTEMS 133
Boundary Conditions
Figure 6-22. Newton-Rapluon program (from Jeppson [15]). Figure 6-22. Continued.
136 UNSTEADY FLOW IN PIPELINES COMPLEX PIPE SYSTEMS 137
Computer Output
3 27 : 40 3 FORMAT(3X, I3,3X,F6.2f3X,F8.«’,3X,F7.4,3X,F9.2,3X,F8.2,5X,F10.2,3X,
3 28 : $Flfl.2) NO. OF ITERATIONS = 4
329: 4033 FORMAT(3X,I 3,3X,P6.2f3X,Ffl.0,3X,F6.0,4X,F9.2,3X,F8.2,5X,Fl 0.2,3X,
3 30 : SF10.2) PIPE DIAM-IN LENGTH-FT ROUGHNESS FLOW-CFS HEAD LOSS-FT ELHGL UP-FT ELHGL DN-FT
331: IF(GPM) QI (I)-QI (D/449 .
332: 9 CONTINUE
333 : WRITE(6,404) 1 15.00 800. .0200 8.59 9.75 1480.00 1470.25
3 34 : 404 FORMAT(////lOXNODE ELHGL-FT ELGRND-FT P-PSI'/10X,' ---------------------------------------- ',2X, 2 12.00 1000 . .0200 3.38 5.75 14 7 0.2’5 1464.50
335 : SB ( ),2X,9(),2X,5('-')) 3 12*.00 600. .0200 5.21 8.22 1470.25 1462.03
3 36 : DO 23 1=1,NODES 4 12.00 1200 . .0200 2.21 2.96 1462.03 1459.07
337: P° ( ELHGL(I)-ELG(I))/2.31 5 10.00 800. .0200 2.33 5.43 1464.50 1459.07
338 : 23 WRITERS, 405 ) I,ELHGL(I) ,ELG(I) ,P 6 8.00 1200. .0200 .95 4.14 1468.64 1464.50
339 : 40 5 FORMAT(10X,13,2X,F7.1,4X,F7,1,2X,F6»1) 7 8.00 1500 . .0200 .24 .33 1459.39 1459.07
340 : IF (.NOT. TRANS) GO TO 98 8 8.00 1800 . .0200 .78 4.21 1459.07 1454.85
341: C 9 10.00 1000 . .0200 3.41 14.55 1400.02 1468.64
34 2 : C ” WRITE DATA FOR TRANSIENT ANALYSIS ON PILE (9) •* 10 8.00 400 . .0200 2.46 9.24 1468.64 1459.39
343: C 8.00 800. . 0200 4.54 1459.39 1454.85
11 1.22
344 : WRITE(9) NPIPES,NPUMP,NODES
345 : DO 511 1-1,NODES
346 : 511 WRITE(9) ELHGL(I),ELG(I)
347 : DO 510 I-l/NPIPES
348: IF(QI(I).GT.0.) GO TO 510
NODE ELHGL-FT ELGRND-FT P-PSI
349 : JU“JDN(I)
350 : JD-JUP(I)
351: JUP(I)»JU 1 1480.0 1460.0 8.7
352 : JDN(I)«JD 2 1470.2 1300.0 73.7
353 : QI(I)=-QI(I) 3 1462.0 1290.0 74.5
354 : 510 WRITE(9) PIPE (I) ,D(I) ,L (I) ,CHW(I) ,QI (I),JUP(I) rJDN(I) 4 1464.5 1310.0 66.9
355: DO 520 1=1,NODES 5 1459.1 1280.0 77.5
356: DO 523 J-1,4 6 1400.0 1380.0 8.7
357 : 523 LOP(I,J)-fl 7 1468.6 1320.0 64.3
358 : J-l 8 1459.4 1310.0 64.7
359: DO 521 111 “1,NPIPES 9 1454.9 84.4
1260.0
360: II-PIPE(111)
361: IF(JUP(II).EQ.I) GO TO 522
36 2 : IF(JDN(11)•EQ.I) GO TO 524
36 3 : GO TO 521
DATA FOR TRANSIENT ANALYSIS HAS BEEN WRITTEN ON FILE (9).
364: 524 LOP(IrJ)«-II
365: GO TO 525
366 : 522 LOP(I,J)-II
367: 525 J-J+l * NETWORK INPUT DATA *
368: 521 CONTINUE
369: WRITE(9) (LOP(I,J),J=1,4)
370: 520 CONTINUE
371: DO 512 I>1,NPUMP NUMBER OF PIPES = 11
37 2: 512 WRITEC9) LLP(I),A(I),B(I),HO(I) NUMBER OF NODES a 9
373: REWIND 9 NUMBER OF LOOPS a 4
374 : WRITE(6,540)
NUMBER OF PUMPS a 1
375 : 540 FORMAT(///lOX, 1 DATA FOR TRANSIENT ANALYSIS HAS BEEN WRITTEN ON FIL
NO. PSEUDO LOOPS 1
376: SE (9).')
377 : C MAX. ITERATIONS = 50
378 : 99 STOP
379: END
ITERATION ACCURACY = .010 CFS
DATA
“ PIPE DATA **
$ INDATA NPIPES-11,NLOOPS»4,MAX=50,NPUMP»1,NSLOOP«1,ACC».01,GPM»P,TRANS=T,
NODES-9,NELEV-1,ELELEV-1480., PIPE DIAM-IN LENCTH-FT FLOW-CFS ROUGHNESS COEFF. NODE UP NODE DN
SEND
1 15. 800. .020 8. 1 2 1460. 1 300. 1 15.00 800 . 8.00 ,0200 1 2
2 12. 1000. .020 3. 2 4 1310. 2 12.00 1000. 3.00 .0200 2 4
3 12. 600. .020 5. 2 3 1 290 . 3 12.00 600. 5.00 .0200 2 3
4 12. 1200 . .020 2. 3 5 1280. 4 12.00 1200 . 2.00 .0200 3 5
5 10. 800 . .020 2. 4 5 5 10.00 800 . 2.00 .0200 4 5
68 . 1200. .020 1. 7 4 6 8.00 1200 . 1.00 .0200 7 4
7 8. 1500. .020 1. 8 5 7 8.00 1500 . 1.00 .0200 8 5
8 8. 1800 . .020 1. 5 9 8 8.00 1800. 1.00 .0200 5 9
9 10. 1000. .020 4. 6 7 1380. 1320 . 9 10.00 1000. 4.00 .0200 6 7
10 8. 400 . .020 3. 7 8 1310. 10 8.00 400. 3.00 .0200 7 8
11 8. 800 . .020 1. 8 9 1260 * 11 8.00 800. 1.00 .0200 8 9
1 4 2 5 -4 -3
2 4 -6 10 7 -5
3 3 -7 11 -8
4 4 9 6 -2 -1
9 -.50 -.30 90.
4 -80.
140 UNSTEADY FLOW IN PIPELINES COMPLEX PIPE SYSTEMS 141
** PIPES IN LOOPS **
LOOP PIPES
1 2 5 -4 -3
2 -6 10 7 -5
3 -7 11 -8
4 9 6 -2 -1
** PUMP INFORMATION «*
LINE PUMP
IS IN
o
T
A B
HO
9 -.50 -.30 90.0
PIPE I
** PSEUDO LOOP INFORMATION **
LOOP DIFF.
4
, IN RESERVOIR ELEV.-FT
-80.00
We will investigate the analysis of a valve or other major loss .......... (6-77)
CTQ
occurring near a junction. To minimize bulky equations, we will use a
one-in, two-out, three-pipe junction. It will be clear from the analysis how
to extend the analysis, by inspection, to the other types of pipe junctions. Inserting Equations 6-77, 6-72, 6-73 into Equation 6-74 results in the
The problem to be solved is defined by the junction sketch in Figure following equation for Hp.
6-23. The equations to solve are shown below.
UQ
Pipe 2 C = c3 + c.
................................... (6-
+
................... (6-78)
Pipe 3 C’ : C' + C, H 72) KLC2
This equation must be solved iteratively by trying successive values of Hp
Continuity: Vp Aj + V.................................
P Ag + Q ..................(6-73)
(6-74)
until F(Hp) = 0 is satisfied within acceptable tolerances. The iterative
II
FtHpO’1))
H FW’1)) (6-79)
*■0
o.
2; C* PROGRAM NO. 7 *
C* WATER HAMMER IN PIPE NETWORKS *
where F = (dF/dHp) is obtained by differentiating Equation 6-71. 3:
C* PROGRAM INTERFACES WITH NEWTON-RAPHSON PROGRAM NO. 6 *
C* (PROGRAM DIMENSIONED FOR UP TO 20 PIPES, 15 MODES, AND 5 SOUPCE PUMPS) *
6 :
7: c
8 :
(J
9 : C A() = WAVE SPEED-FPS
1- 10: C HATH = ATMOSPHERIC PRESSURE HEAD-FT
O
bJ
11: C TMAX = MAXIMUM REAL TIME OF SIMULATION-SEC
C NPARTS “ NUMBER OF PARTS INTO WHICH PIPE IS DIVIDED
12:
C IOUT ■ INDEX WHICH GIVES AN OUTPUT EVERY IOUT-TH TIME STEP
13:
C N3CQ “NUMBER OF NODES WHERE THE DISCHARGE IS SUDDENLY SET EQUAL TO ZERO
14: C NODEQO ■ ARRAY OF NODE NUMBERS WHERE DISCHARGE IS SUDDENLY ZEROED
15:
-C4A, (6-80) 16:
17:
C NODOUT • LOGICAL. IF TRUE, INFORMATION ON HEADS,
C ALLOUT = LOGICAC. IF TRUE, INFORMATION ON HEADS,
H-VALUES, ISOUTPUT AT NODES
H-VALUES, ANDVELOCITIES
C IS OUTPUT FOR INTERIOR SECTIONS OF PIPES.
18: C TCI » TIME (SEC) WHEN FIRST STAGE OF VALVE CLOSUPE IS COMPLETED
19: C PCI - PERCENT OPEN OF VALVE AT END OF STAGE ONE OF VALVE CLOSURE
20 : C TC2 = TIME (SEC) WHEN SECOND STAGE OF VALVE CLOSURE FINISHES CLOSING VALVE
21: C PVALVE = NUMBER OF PIPE WITH A TWO-STAGE VALVE PLACED AT THE DOWNSTREAM END
The iterative procedure is generally started by using the value of Hp from 22:
23:
C NSHUT => NUMBER OF PIPES WITH SUDDENLY CLOSING VALVES IN THEM
C StIUTO = ARRAY OF NUMBERS OF PIPES WITH SUDDFNLY CLOSING VALVES IN THEM.
the previous time step as the first estimate. Programming precautions are 24:
25 :
C IF VALVE IS AT DOWNSTREAM END, PIPE NUMBER SHOULD HAVE A MINUS SIGN.
C IF VALVE IS AT UPSTREAM END, PIPE NUMBER IS POSITIVE.
also taken to avoid looping in the iterative solution. 26:
27:
C NSAVE = NUMBER OF SETS OF H VS T DATA
C SAVE() = ARRAY OF PIPE NUMBERS WITH H
TO BE SAVED (4 WITH 500 PTS MAX)
VS T DATA SAVED FOR DOWNSTREAM END
28 : C IOUTSA » SAME AS IOUT BUT APPLIES TO THE H VS T DATA
29: C
Network Water Hammer Examplea 6-6 30 :
31:
DIMENSION X(20,30), V(20 ,30) ,H(20,30) ,BLOW(20,39),HHIGH (20,30),
SHEAD(20,30) ,VNEW(20,30) , IINEW (20,30)
32: DIMENSION L(20),D(20),A(20),PIPEZ(20),F(20),VZEPO(20),NPAR(20),
33: $C (20) ,AK(20) ,SINE (20) ,AREA(20),DELTT(20) ,RATIO(20) ,JUP(20) ,
The computer program shown in Figure 6-24 can handle a variety of 34:
35:
SSHUT(20)
DIMENSION NODE?(15,4),ELHGL(15),ELG(15),NTYPE(15),SAVE(15),
boundary conditions. The examples of data input given illustrate the 36:
37:
S0DEMI15),HNODE(15),NODFO(15)
DIMENSION PLINE(5),AZ(5),BZ(5),CZ(5),HRES(5)
interfacing between the steady state and unsteady versions of the program. 38:
39:
DIMENSION SAVEH(4,500),TIND(5),HIND(5)
DIMENSION CODD(4),CEVEN(4)
The examples are relatively simple ones which demonstrate many of the 48:
41:
DIMENSION KLI(11),PCT(11)
INTEGER PLINE,PVALVE,SAVE,SHUT
common boundary conditions. However, the user may need to tailor a 42:
43:
REAL L,LA,KI,KLI
LOGICAL FA IL,NODOUT,ALLOUT,VALVE
program for his own particular network, incorporating the proper special 44 :
45:
DATA FAIL,UODOUT,ALLOUT,VALVE/4 *.FALSE./
DATA NBCO,PVALVE,NSHUT,NSAVE,IOUTSA/5*0/
boundary conditions at junctions where they occur. 46:
47:
DATA (KLI(I)r1-1,11)/.0,.0167,.0313,.0556,.10,.1787,.3333,.625,
$1.250,2.58,5,27/
48: DATA (PCT(I),I“1,11)/0.,10.,2O.,30.,40.,50.,60.,70. ,8P. ,90.,100./
The program can generate unsteady flows by a) permitting the 49: NAMELIST /SPECS/ IOUT,NPARTS,THAX,HATH,NHCO,NODEO,NODOUT,ALLOUT,
50: STCl,PCI,TC2,PVALVE,NSHUT,SHUT,NSAVE,SAVE,IOUTSA
demand at any or all nodes to be suddenly set to zero, b) permitting valves 51: PI-3.141593
52: T=0 .
at the upstream or downstream end of any line (or several lines) to be 53: C
54 : C ** READ IN TRANSIENT DATA AS WELL AS DATA FROM STEADY STATE ANALYSIS FROM (9)
closed suddenly, or c) permitting two-stage valve rate closure of a gate 55 : C
56: READ(9) NPIPES,NPUNP,NODES
valve at the downstream end of any one line. 57: C
58 : READ(5,SPECS)
Output can be either at junctions only or at all nodes; it can be at 59 : READ(5,100) (A(I),I»1,NPIPES)
60: 100 FORMAT(8 Fl 0,0)
every time step or only at intervals of several steps. Pressure head 61: C
DO 4 1=1,NODES
62:
maximum and minimum values are output at the end and it is possible to 63: 4 PEAD(9) ELI1GL (I) ,ELG(I)
DO 2 11 = 1 ,NPIPES
64 :
save the complete H vs. time history at four selected junctions in the 65:
66:
READ(9) I,D(I),L(I),F(I),VZERO(I),JU, JD
JUP(I)=JU
network. 67:
68 :
PI PEZ(I)=ELG(JU)
SINE (I) = (ELG (JD)-ELG (JU) )/L (I)
The first three sets of data shown generate water hammer situations 69 :
70:
AREA(I)=.25*PI *D(I) *D(I)/]44.
VZERO(I)=VZERO(I)/AREA(I)
C (I)=32,2/A (I)
in the network shown previously for the steady state Newton-Raphson 71:
72: 2 DELTT(I)=L(I)/(NPARTS•(VZERO(I)+A(I)))
DO 3 1=1,NODES
analysis. Because the purpose here is to illustrate data deck setup, the 73:
74 : 3 READ(9) (NODEP(I,J),J=1,4)
IF(NPUMP.EQ,0) GO TO 6
analysis results are not given. 75:
76:
DO 5 1 = 1,NPUMP
5 READ(9) PLINE(I) ,AZ(I) ,BZ (I),CZ(I)
In case (a), the demands at nodes 4 and 9 are set equal to zero. The 77 :
78:
6 CONTINUE
REWIND 9
output is printed only for the junction values every tenth time step. The H 79:
30 :
C
C
81:
vs. time data for every second time step is saved for the downstream ends 82:
C *• COMPUTE MINIMUM DCLTA T **
C
83:
of pipes 5, 8, 9, and 11. 84 :
DELT=DELTT(1)
K.MIN’l
85: DO 12 I=2,NPIPES
86:
87: IF(DELTT(I).GT,DELT) GO TO 12
88 : DELT=DELTT(I)
89: KHIN-I
90s 12 CONTINUE 174: GO TO 24
91: DO 13 1-1.NPIPES 175: 25 O’AREA(I)'VZERO(I)
92 : IF(I.EQ.KHIN) GO TO 14 176: ORES 'J)=ELHGL(JU)
93: ANPAR-L(I)/(DELT*(VZERO(I)+A(I))) 177: H (1,1)«H( I,1)+AZ(J)*O*Q+BZ(J)*Q+CZ(J)
94 : HPAR(I)"ANPAR 178: 24 DELL=L(I)/NPAR(I)
95; TEST-NPAR(I) 179: □ ELUF«12.*F(I)* DELL*VZERO(I)* * 2/(64,4‘D(I))
96: IF(ABS(TEST+l-ANPAR).LT..0001) NPAR(I)-NPAR(I)+1 180: AK(i;=12.*F(I)‘DELT/(2.*D(I))
97: GO TO 13 181: K=NPAR(I)+1
98: 14 NPAR(I)"NPARTS 182: xd,i)"0.
99 : 13 CONTINUE 183: V( 1,1)»VZERO(I)
100: INDEX«THAX/DELT + 1 184: BLOW;I,1)«H(I,1)
101: C 185 :
IIHIGB (1,1) "If ( 1,1)
102: C * * IDENTIFYING NODE TYPE AND EXTERNAL DEMAND “ 186:
187: HEADd,1)=H(I,1)-PI PEZ (I )
103: C
188: RATIO(I)"DELT/DELL
104 : DO 50 1 = 1 ,NODES
105: QDEM(I)"0. 189: DO 21 J"2,K
106 : HNODE(I)"ELHGL(I)-ELG(I) 190 : Xd,C)«(J-l) *DELL/L(I)
107: DO 51 J"1 ,4 191 : H(I»C)»H(I,I)-DELHF*(J-l)
108 : JJ"NODEP(I»J) 192: HLOWiI,J)=H (I ,J)
109 : IF(JJ.EQ.0) GO TO 52 193: HHIGH(I,J)"H(I,J)
110: JJJ-ARS(JJ) 194: HEADI I ,J) =11 ( I, J) - (PIPEZ (I) + X (I , J) *L( I) ‘SINE (I ) )
111: 51 ODEM (I)-ODEM(I)-(ABS(JJ)/J J)* AREA(JJJ) *VZERO(JJJ) 195: 21 V(I,J)"VZERO(I)
112: 52 IF(J.EQ.2) GO TO 59 196 : 20 CONTINUE
113: GO TO 5fl 197: PIIMAXs-100.
114: 59 IF(NODEP(I,1).GT.fl) GO TO 53 198: PHMIN»1 fl0003 .
115: NTYPE(I)«97 199: C
116: GO TO 50 280 : C •• WRITE OUT STEADY STATE CONDITIONS **
117: 53 QDEH(I)=fl. 2.11: C
118: NTYPE(I)=98 202 : IF(.NOT, ALLOUT) GO TO 26
119: 50 CONTINUE 203:
120: IF(HPUMP.E0.0) GO TO 56 WRITE(6,204)
204 : WRITE(6,205) T
121 : DO 55 1-1,NPUHP 205 :
122: II=PLIKE(I) DO 22 1-1.NPIPES
206 :
123: JU-JUP(11) 207 : K"NPAR(I)+1
124: 55 NTYPE(JU)"99 208: WRITE(6,206) I, (X(I,J) .HEAD(I.J) ,H(I,J) ,V(I,J) ,J»1,K )
125: 56 CONTINUE 209 : 22 CONTINUE
126: IF (NBCQ.EQ.fl) GO TO 58 210 : GO TO 29
127: DO 57 1=1, MBCQ 211: 26 IF(.NOT. NODOUT) GO TO 29
128: II»NODEQ(I) WRITE(6,230)
212:
129 : 57 ODEMdI)-0. 213: WRITE(6.231) T
130: 58 CONTINUE 214 : WRITE(6,232) (I,HNODE(I),ELHGL(I),I«1,NODES)
131: C 215: 29 CONTINUE
132: C 216: 1F(NSAVE.EO.9) GO TO 31
133: C ** WRITE OUT INPUT DATA “ 217 : DO 32 I = 1,NSAVF.
134 : C 218: II-SAVEd)
135: WRITF(6,200) 219:
136: WRITE (6 , 201) IOUT,NPARTS,KPIPES. IIATM.TMAX , DF.LT K=NPAR(II)+1
220 : 32 SAVEH(1,1)=H (11 , K)
137: WRITE(6.220) 221 :
138: IF (NRCQ.EQ.0) GO TO 19 I HD" 2
222 : 31 CONTINUE
139: WRITE(6,221) (NODEQ(I),I -1,NI3CQ) 2 23 :
140: 19 CONTINUE C
2 24 :
141: IF(PVALVE.EQ.0) GO TO 9 2 25 :
142: WRITE(6,212) PVALVE.TC1.PC1.TC2 cBFGIN TRANSIENT ANALYSIS
226:
143: 9 CONTINUE c - ---
227 :
144: IF (NSH'JT.EQ.0 ) GO TO 199 c
228 :
145: DO 198 1=1.NSHUT DO 99 II r 1 = 1 .INDEX
14 6: IF(SHUT(I).LT.0) WRITE{6,222) SHUT{I) 229 :
230: T"T+DELT
147: IF (SHUT(I) ,GT.0) WRITE(6,223) SfiUT(I) IF(T.GT.TMAX) GO TO 400
231 :
148: 198 CONTINUE c
232 :
149: 199 CONTINUE C " COMPUTE H AND V AT INTERIOR NODES **
233 :
150 : WRITE(6,202) C
234 :
151: DO 15 I-l.NPIPES 235 : DO 30 1=1,NPIPES
152: LA-L(I)/A(I) K-NPA’d)
236 :
153: 15 WRITE(6,203) I,D( I) ,L(I) ,A( I) ,PIPEZ (I) ,F(I) .VZEROd) , DO 30 J«2,K
154: 9DELTT(I),HPAR(I).SINE(I), LA 2 37 :
2 38 : RA»RATIO(I)*A(I)
155: WRITE(6,2022) VMINUS"V(I,J)-RA*(V(I.J)-V(I,J-1))
2 39 ;
156: DO 16 1=1,NODES 240 : VPLUS"V(I,J)-RA* (V (I,J)-V(I,J + 1))
VNEWd ,J) =0.5* (VMINUS+VPLUS+C (I) * (HMINUS-HPLUS) +C (I) *DFLT*S I NE (I)
157: 16 WRITE (6,2033) I, FLHGL(I) ,ELG(I) , (NODEP (I ,.7) , J = 1 ,4) , ODEM (I) 241: HMINUS"!t (I, J ) -RA * (1| (I,J)-H ( I , J-l))
S' (VMIHUS-VPLUS)-AK(I)•(VMINUS*ABS(VMINUS)+VPLUS*A3S(VPLUS)))
158: IF(NPUMP.FO,0) GO TO 18. 24 2 : HPLUS=H(I,J)-RA* (ll|I,J)-ll|I,J+l) )
KNEW(I,J)=0.5* (HMINUS+HPLUS+(VMINUS-VPLUS)/C(I)+DELT*SINE(I) *
159 : WRITE(6.2044) 243: S (VHIHUS+VPLUS) - ( AK (I)/C (1) ) *'(VMI NUS * ABS (VMI NUS) -VP LUS* ABS (VPLUS) ) )
160 : DO 17 I=1,NPUMP 244 : 30 CONTINUE
161: 17 WP.ITE ( 6,2055) PLINE (I) , AZ (I) , RZ (I) , CZ (I) 24 5 : C
162: 18 CONTINUE 246 : C ** COMPUTE H AND V AT JUNCTIONS **
163: C 247 : C
164 : C “ SET UP CONDITIONS FOR STEADY STATE (T=0) ** 248 : DO 4 0 111 = 1.NODES
165: C 249 ; IF(NTYPE(111).EQ.97) GO TO 81
166: DO 20 I=1,NPIPES 250 : IF (HTYPE(111).EQ.98) GO TO 30
167 : JU’JUP(I) 251: IF (NTYPE(111) .EQ.99) GO TO 80
168: I! (1,1) "RLHGL (JU) 252 : DO 41 J«1,4
169: IF (NPUNP.EQ.0) GO TO 24 253 : CODD(J)=fl.
170: DO 23 J=1,NPUMP 254 : 41 CEVEN>J)=0,
171: I1»PLINE(J) 255 : C
172: IF(II.EO.I) GO TO 25 256 : C • STANDARD JUNCTIONS *
173: 23 CONTINUE 257 : C
258 : DO 42 J=1,4
Figure 6-24. Continued. 259: 11=NODF.P|III ,J)
260: I-IAOS(II)
261: IF (11.EQ.0) GO TO 45
262: IF(II.LT.fl) GO TO 43 3 47: DO 801 N-l,NSHUT
263 : VPLUS-V(I,1)-RATIO(I)*A(I)*(V(I,1)-V(I,2)} 3 48 : IF (SHUT(N).NE.11) GO TO 801
264 : HPLUS-H(I ,’l) - RATIO(I) * A (I) * (H (1,1) -U (1,2) ) 349 : VNEW(I,l)-0.
2 65 : CODD(J)--AREA (I)*(VPLUS-C(I)‘HPLUS-C(I) *DELT*VPLUS*SI MJ(I)- 350 : HNEW(I,1)--CC/C(I)
266: SAX(I)*VPLUS*ABS(VPLUS)) 351: GO TO 40
267 : CEVEN(J)-C(I)*AREA(I) 352 : 801 CONTINUE
268 : IF(NSHUT.EQ.0) GO TO 42 353 : 800 CONTINUE
2 69 : DO 430 N-l,NSHUT 354 : HNEW(1,1)»ELHGL(111)
270 : IF (SHUT(N).NE.I) GO TO 430 3 55 : VNEW (1,1)aCC+C(I)*HNEW(1,1)
271: HNEW(I,1)-CODD(J)/CEVEN(J) 356: GO TO 40
272 : VN EW (1,1) = 0 , 357: C
273 : CODD(J)=0. 358 : C * SOURCE PUMP JUNCTIONS
274 : CEVEN(J)=0. 359 : C
275 : GO TO 42 360 : 85 DO 86 JJ-1,NPUMP
276 : 43O CONTINUE 361 : II-PLINE(JJ)
277 : GO TO 42 362 : IF(I.EQ.II) GO TO 87
278 : 43 K»NPAR(I)+1
363 : 86 CONTINUE
279 : VM INUS-V(I,X)-RATIO( I) * A ( I) * (V (I , K) -V (I , X-1) ) 364 : STOP
280: HMINUS-H(I,X)-RATIO(I)*A(I)*(!I(I,K)-II(I,K-1))
365 : 87 CC-H PLUS-VPLUS/C (I) +VPLUS*S INE (I) *DELT+ (1.0/C (I) ) *AX (I) * VPLUS*
281: CODD(J)-AREA(I) * (VMI N'.IS+C (I) *HM ItlUS+C (I) *VMINUS*DELT*SINE(I)-
366 : SABS(VPLUS)
282 : SAX(I)*VMINUS*ABS(VMINUS) )
283 : CEVEN (J)-C(I) *ARF.A(I) 367 : C3-(BZ(JJ) *AREA(I)-1.0/C (I))/(AZ(JJ) *AREA(I)**2)
284 : IF(NSHUT.EQ.0) GO TO 440 368 : C4-(CZ (JJ)+HRES(JJ)-CC)/(A2(JJ)‘AREA(I)* *2)
369 : C5=4;*C4/(C3*C3)
285 : DO 441 N-l/NSUUT IF(C5.GT.0.) GO TO 88
370 :
286 : IF(SHUT(N).NE.11) GO TO 441
371: VNEW(I,1)-0.5*C3*(-1.+SQRT(1,-C5))
287 : X-NPAR(I)+1
372 : GO TO 89
288 : HNEW(IrK)’CODD CT)/CEVEN(J) 373 : 88 VNEW(I,l)-0.
289 : VNEW(I,K)=0. 374 : 89 HNEW(I,1)-CC+VNEW(I,1)/C(I)
290 : CODD(J)-0 . 375: ELHGLdI I) =HNEW(I ,1)
291: CEVEN(J) =0. 376 : HNODE(111)-ELHGL(111)-ELG(111)
292 : GO TO 4 40 377: GO TO 40
293 : 441 CONTINUE 378 : C
294: 440 IF (I.NE.PVALVE) GO TO 42 379 : C * DOWNSTREAM END OF SINGLE PIPES *
295 : C1-CODD(J)/AREA(I) 380 : C
296 : CODD(J)=0. 381: 81 I-1ABS(NODFP(111,1))
297 : C2-CU)
298 : 382 : K-NPAR(I)+1
CEVEN(J)“0. 383 : VMINUS-V(I,X)-RATIO(I)*A(I) * (V(I,X)-V(1,X-1) )
299 : VALVE-.TRUE, 3 84 : HMINUS-H{I,K)-RATIO(I)*A(I)*(H(I,K)-H(I,K-l))
300 : 42 CONTINUE 385 : CC-VMINUS+Cd) ‘HMINUS+C (I) * VMINUS * DELT* SINE (I) -AK (I) *VMINUS*
301: 45 CONTINUE SABS(VMINUS)
386 :
302 : ANUM-0. 387 : IF (NSHUT.EQ.0) GO TO 810
303 : DENOM-R. 388 : II-NODEP(111,1)
304 : DO 46 J = 1,4 389 : DO 809 N-l,'NSHUT
305 : ANUM-ANUM+CODD(J) 390 : IF(SHUT(N).NE.11) GO TO 809
306 : 46 DENOM-DENOM+CEVEN(J) 391 : VNEW(I,K)-0.
307 : IF(DENOH.LE..IE-10) DENOM-.1E-10 392: HNEWd,K)-CC/C(I)
308 : IF(VALVE) GO TO 140 393 : GO TO 40
309 : 449 HPP-(ANUH-QDEM(111))/DENOM 394: 809 CONTINUE
310 : 49’ HNODE(111)“HPP-ELG(111) 395 : 810 CONTINUE
311 : ELHGL(111)-HPP 396 : VHEW(I,X)-V2ERO(I)
312 : IF(DENOM.LE..1E-10) GO TO 40 397 : HNEW(I,K)=(CC-VNEW(I,X))/C(I)
313: DO 47 J-1,4 398 : GO TO 40
314 : II-NODEP(111,J) 399 : C
315: I-IABS(II) 400 : C * STAGED VALVE CLOSURE AT JUNCTION *
316: IFdI.EQ.0) GO TO 40 401: C
317: IF(I I.LT.0) GO TO 48
318: 402: 140 IF(T.GE.TC2) GO TO 150
IF(NSHUT.EQ. 0) GO TO 450 403 : IF(T.GT.TCl) GO TO 149
319: DO 4*51 N-l,NSHUT
320 : 404 : OPEN-100.-(100.-PCI)* T/TC1
IF(SHUT(N).NE.I) GO TO 451 4 05 : GO TO 148
321: GO TO 47
322 : 406 : 149 OPEN-PCI *(TC2-T)/(TC2-TC1)
451 CONTINUE 407 : 148 DO 145 KK-1,11
323 : 450 HNEW(I,1)=HPP
324 : 408 : ITEST-(OPEN-PCT(XX))*.10
VNEW (1,1) a (-CODD (J)/AREA( I) )+C (I) *1IPP 409 : IF(ITEST.EQ.0) GO TO 146
3 25 : GO TO 47
326 : 410 : 145 CONTINUE
48 K-NPAR(I)+1 411: 146 FACT-(OPEN-PCT(XX))*.10
327 : IF(NSHUT.EQ.0) GO TO 460
328 : 412: XI-KLI (XX) + FACT*(X LI (XX+1)-XLI (XX) )
DO 461 N-l,NSHUT 413: CC=?-2.2*AREA (PVALVE) *XI/C2
3 29 : IF (SHUT(N) .NE.11) GO TO 461
330 : 414: C8-DENOM
GO TO 47 415: C9-ANUM-QDEM(111)-CC
331: 461 CONTINUE
332 : 416: CCC-2.*C2*C2/(32.2 * XI)
460 IF(I.EQ.PVALVE) GO TO 47 4 17: X-NPAR{PVALVF)+1
333 : HNEW(I,X)-HPP
334 : 418: H P-H ( PVALVE, K)
VNEW(I,X)=(CODD(J)/AREA(I))-C(I)*HPP 419: DO 142 XXX-1,20
3 35 : 47 CONTINUE
336 : 4 20: FHP-CC*SQRT(1,-CCC*(HP-C1/C2))-C8*HP+C9
GO TO 40 4 21:
337 : c FPRIME®(0.5*CC/SORT(1,-CCC*(RP-C1/C2)))*(-CCC)-C8
4 22 : HPP-HP-FHP/FPRIME
338 : C * RESERVOIR JUNCTIONS *
339 : c 4 23 : IF (ABS(HPP-HP).LT.0.5) GO TO 143
4 24 : HP-HPP
340 : 80 I-1ABS(NODEP(111,1))
4 25 : 142 CONTINUE
341: VPLUS-V(I,1)-RATIO(I) * A (I) * (V(1,1)-V(1, 2) ) 4 26: WRITE(6,160)
342: HPLUS»H(I,1)-RATIO(I)*A(I)*(H(I,1)-H(I,2)) 4 27 : 160 FORMAT (///5XNEWTON-P.APHSON PROCFDURE AT THE VALVE HAS FAILED")
343 : IF(NTYPE(111).EQ.99) GO TO 85 4 28 : STOF
344 : CC-VPLUS-C(I)•HPLUS-C(I)* DELT*VPLUS*SINE(I)-AX(I)*VPLUS*ABS(VPLUS) 429 : 143 VALVE-.FALSE.
3 45 : IF (NSilUT.EQ.0) GO TO 800
430 : VNEW(PVALVE, K)-(32.2 * XI/C2)*(-1.+ SQRT(1,-CCC*(PPP-C1/C2)) )
3 46 : II-NODEP(111,1) 431: HNEW(PVALVE,X)-HPP+VNEW(PVALVE,X)* * 2/(64.4 * X I)
4 32 : GO TO 49
Figure 6-24. Continued.
Figure 6*24. Continued.
148 UNSTEADY FLOW IN PIPELINES COMPLEX PIPE SYSTEMS 149
Caw (a)
In case (b), valves are suddenly shut at the downstream ends of pipes
Figure 6-25. Pipe network with a single pipe extending from the network.
5, 8, and 11. The H vs. time data is saved for each time step at the
downstream ends of pipes 1, 9, 3, and 10.
4 cfs
remaining distance in a total of 2.0 sec. from the beginning of valve
movement. Output for all internal nodes is printed every tenth time step
and H vs. time data is saved for the downstream ends of pipes 5 and 6 for
every second time step.
Caae(c)
Case (d) that the homogeneous liquid column has been separated means that our
analysis techniques to date can no longer be applied directly.
Steady State Analysis Actually whenever the pressure at any point in the pipeline drops
below the pressure exerted on the liquid at the pipeline source, the
$ INDATA NPIPES*10,NLOOPS=3,MAX*50,NPUMP = 1,NSLOOP=1,ACC =,01,GPM=F,TRANSIT, saturation pressure may be reached and the dissolved gases may begin to
SEND
NODES»9»HELEV»1,ELELEV=1480.,
come out of solution. This is one of the reasons for placing air release
1 15 . 800 .
1000.
.020
.020
8.
3.
1
4
2 1460. 1300
1310
valves at pipeline summits. The amount of gases coming out of solution
2 12 2
3 12 600 . .020 5. 2 3 1290 depends on how low the pressure drops and how much gas there is in
4 12 1200 . .020 2. 3 5 1280
5 10 800 . .020 2. 4 5 solution. Certainly, if the pressure drops to the liquid vapor pressure,
6 8. 1 200. .020 1. 7 4
7 8. 1500 . .020 1. 8 5 large cavities filled with vapor and gas will form.
8 8. 1800. .020 1. 5 9 1260
9 10 1000 . .020 4. 6 7 1380. 1320
A simple example of column separation can be found downstream of
1
10 8.
4 2 5
400 .
-4 -3
.020 3. 7 8 1310
a suddenly closed valve. Upon sudden valve closure, the pressure head just
2
3
4
4
-6
9
10 7
6 -2
-5
-1
downstream of the valve attempts to drop the AH value required to bring
9
3 -80.
50 -.30 90. the liquid column to rest. If the required head drop causes the pressure to
reach the liquid vapor pressure, no further drop can occur and the liquid
Water Hammer Analysis column separates at the valve (see Figure 6-26). Analysis of the flow
situation would require some technique for representing cavity formation
SSPECS IOUT=1P,NPARTS=3zTMAX® 4.0 , HATM*34., and changes in wave speed.
ALLOUT=T,
NSHUT-1, SHUT(l)*-8,
NSAVE=4,SAVE*4,5,7,8, 11*0, IOUTSA»1,
SEND
Analysis With Column Separation and Released Air
3300. 3300 . 3300. 3300. 3303. 3300. 3300. 3300
3300. 3300 . 3300 .
Tullis, Streeter, and Wylie [13] thoroughly discuss the effects of air
release at low pressures, as well as column separation at vapor pressure.
They suggest that volumes of released air can be either uniformly
distributed throughout the flow or concentrated in pockets. In the former
6.8 Column Separation and Released Air case, Equation 4-23 can be used to find the reduced wave speed. Because
the change in wave speed leads to large interpolations with the rectangular
It is common knowledge that excessively high pressures in pipes grid method of solution, they suggest using the method where the
resulting from water hammer can cause damage. It is also generally characteristic lines are followed as closely as possible to minimize
known that low pressures, in certain circumstances, can cause collapse of interpolation errors and instability.
pipes with thin walls or those with high external loading. What is not so If it is undesirable to use this approach, the regular wave speed is
commonly known or understood is the phenomenon referred to as column used, the air is collected at discrete points and internal boundary
separation. conditions are formed at each bubble. The size of the bubbles are
monitored and their thermodynamic behavior is included.
Cause of Column Separation and Released Air Because analysis with released air and vapor is rather difficult to
accomplish and still is not completely understood, no further discussion
When water hammer causes the pressure at points in a pipeline to will be included here. However, the effects of released air are not
approach the vapor pressure of the liquid, gases in solution begin to come noticeable until the pipeline pressure drops below the saturation pressure
out of solution and dramatically affect the flow behavior. If the drop in of the dissolved air. Unless this low pressure is anticipated in the system to
pressure is severe enough to cause the local pressure to reach the vapor be analyzed, the engineer need not be concerned with its effects.
pressure, then the liquid boils, i.e., cavitates, vaporizes, forming large
pockets of undissolved gases and liquid vapor. This phenomenon is 6.9 Wave Transmission and Reflection at Pipe Junctions
referred to as column separation.
A consequence of this occurrence is the rather substantial change in In many instances it is desirable to be able to estimate portions of
the wave speed because of entrained gases and vapor bubbles affecting pressure waves reflected and transmitted at pipe junctions. We already
the liquid elasticity. This effect was discussed in Chapter IV. Also the fact know that at reservoirs, none of the positive (or negative) pressure wave is
I COMPLEX PIPE SYSTEMS 155
154 UNSTEADY FLOW IN PIPELINES
ol
HV
—T
EL-HGL <
-------- J.
PI
CF 1 , PIPE 2 PIPE I
PIPE 2
o
x HV
BEFORE AFTER
A MIXTURE OF
r
WATER AND Figure 6-27. Wave transmission and reflection at a series pipe Junction.
WATER VAPOR
AT THE VAPOR
PRESSURE
Example 6-7
Column separation caused by sudden valve closure. As an example, consider a 24-inch pipe reducing to a 6-inch pipe. An
a-value of 3300 fps and a velocity of 1.0 fps occur in the 24-inch pipe and
Figure 6-26. an a-valve of 3700 fps and a velocity of 16 fps occur in the 6-inch pipe.
The AH-value for sudden flow stoppage in the 6-inch pipe would be 1838
transmitted into the reservoir. We will look first at series pipe junctions ft.
and then at tee junctions, assuming no head loss occurs at the junctions. According to Equation 6-81, only 194 feet of this increase would be
passed on into the 24-inch pipe. In many respects, the large pipe acts as a
Series Pipe Junctions
reservoir.
The equations of momentum and continuity are applied to a pressure
Tee Junctions
head increase of AH approaching a junction. After the wave reaches the
junction, AHi passes through (is transmitted) and (AH-AH,) is reflected. The situation for tee junctions is shown in Figure 6-28. Using the
Figure 6-27 shows the configuration of the EL-HGL before and after the
same analysis technique as before leads to the following equations
occurrence. The results of the analysis show that
2a. a, A,
2ajA2 AH. = AH2 = ------------------- ------------------- AH .................(6-83)
AH. AH ............................................... (6-81) a2a3A! +3J33A, H-a^Aj
+ a, A2
or for similar a-values in all pipes,
where A is the pipe cross-sectional area. For equal a-values,
2A,
2A, AH.1 = AH2 =——-
2
2
— ...........................................................(6-84)
2A + A,
AHL ....................................................... (6-82)
= AFA7 AH
156 UNSTEADY FLOW IN PIPELINES
EL-HGL
Basic Theory
157
158 UNSTEADY FLOW IN PIPELINES PUMPS IN PIPELINES 159
........................................................ (7-5)
noted that forces on the fluid other than those caused by the pump 7-7)
CS
+
)
Mr N
K)| <
+
N
+
hardware are either negligible (shear and weight) or have zero moment
+
arm (pressure at inlet and outlet). Consequently, the torque exerted on the
fluid by the pump is With some adjustments, depending on the pump and manufacturer, this
Hp-value is the value shown on the pump characteristic curve.
T = Qp(r2V2t -rtVlt) ..................................................................... (7-2)
Pump Power and Efficiency
It is now possible to develop an expression for the head produced by
the pump. Using the conventional fluid power equation The power applied to the pump shaft is generally given in horsepower
with the symbol BHP (brake horsepower). Because of various losses only a
fraction (70 percent to 90 percent) of this power is added to the fluid in the
Power = QyH 7-3)
pipeline. The overall pump efficiency is given as
where Qy is the weight flow rate and Hp is the energy added to each pound
of fluid, and equating it to the mechanical power equation QyH (7-8)
e
X
Power = Teo £
550 BHP
.......................................................................... (7-4)
160 UNSTEADY FLOW IN PIPELINES PUMPS IN PIPELINES 161
and this value is found from laboratory tests for a number of different impossible. Therefore, we must choose the most important requirement
operating conditions. and make adjustments for the effects of other non-similarities. Because
Losses between the pump shaft power and the final power added to smooth flow through the pump is most important, Equation 7-9 is used as
the pipeline fluid are the result of many causes. There are mechanical the primary similarity criteria in pump similitude. This means that viscous
losses caused by friction in bearings, seals and packings. There is a loss effects are not properly accounted for and some adjustments must be
resulting from pumping of dead water behind the impeller (disk friction). made to test results to take into account unequal Reynolds numbers.
There are hydraulic losses in the pump intake, the impeller and the Generally, this is done with small adjustments to the efficiency. Pumps
diffusing section at the outlet. There is a loss because of leakage backward satisfying Equation 7-9 are referred to as homologous pumps.
from the high pressure to the low pressure side of the impeller through the Equation 7-9 doesn’t directly show the relationships between head
wear-ring gaps. All of these causes extract energy and reduce the pump increase and power requirements between model and prototype. However,
efficiency. However, the engineer has little control over these factors, appropriate equations can be deduced using the basic theoretical
except that he can consider a range of different pumps to maximize relationships. Equation 7-6 shows that head increase is proportional to
efficiency for his particular application. pump speed, size and flow velocity, or in equation form,
Similarity Laws
a
The principles of hydraulic similitude find important application in
both constant-speed pumping situations and in water hammer situations Replacing Q with ND3 from Equation 7-9 and collecting terms gives
caused by pump power failure. The similarity principles permit us to
predict heads, discharges and torques over a wide range of speeds when
using only the basic information given at one speed. ............................................................ (7-11)
For those familiar with conventional modeling parameters for - constant
N2D2
hydraulic problems, i.e., Reynolds number, Froude number, etc., it is
interesting to recognize that in the case of pumps, kinematic similarity If the same arguments are applied to Equation 7-2 for torque,
takes precedence over dynamic similarity. In kinematic similarity the
velocity fields in the model and prototype should be identical (scaled T.QpD
images of one another). Because pumps are run at a speed determined by
the motor and not affected by the flow, then we must select flows which Again using Equation 7-9 gives
give identical ratios of fluid flow velocity to impeller speed. Flow velocities
are proportional to Q/D2 and impeller velocities are proportional to ND
where D is a typical pump dimension (generally the impeller diameter)
3
(7-12
-
and N is the rotational speed in revolutions per minute (rpm). Kinematic n = constant
pN2D5
similarity requires these ratios to be constant, which gives the following
equation Using the relationship BHP Too gives an appropriate relationship for
Q power
=■ constant .................................................................... V(7-9)
--- 3-.....................................................................................................................................
ND ’
-------- = constant (7-13)
pN3D5
Because viscous effects would seem to be important in pumps, it
would be desirable to have Reynolds number equivalence in model and These relationships can be used to predict changes in discharge, head
prototype. In terms of flow rate, the equation is and power requirements for pumps of different size (impellers trimmed)
-3— = constant ..................................................................... (7-10) or pumps running at different speeds. The accuracy of the prediction
DP decreases as the changes in speed and size become more different from the
test conditions. The application we are most interested in for water
where v is the kinematic viscosity of the fluid. hammer situations is when power to a pump fails and the pump begins to
Requiring both Equations 7-9 and 7-10 to be satisfied in a decelerate. In this case, the decelerating pump is a prototype of itself
model-prototype situation would be so restrictive as to be essentially
162 UNSTEADY FLOW IN PIPELINES PUMPS IN PIPELINES 163
(model scale — 1:1) at successively changing speeds. Under these Specific Speed
conditions we will use the similarity rules as the best estimate of pump
performance at any speed. This iterative process of pump selection can be accomplished several
ways. One approach is to calculate the specific speed of the pumping
Types of Pumps situation. The specific speed is calculated from the equation
We will consider three basic types of pumps—centrifugal, mixed-flow N _ N (rpm) VQ (gpm) ...........................................................(744)
(turbine), and axial-flow (propeller) pumps. The centrifugal pump is S
' h3'4(ft)
primarily designed to increase pressure substantially for relatively small
flow rates. The flow through the impeller is in the radial direction (normal Entering a chart such as Figure 17-16 in Daugherty and Franzini [20], one
to the pump shaft) and energy is added primarily through centrifugal can determine the type of pump required (or whether any type will work).
action. As an example of the value of the approach, assume that your specific
At the other extreme, the axial-flow pump is designed to pass speed calculation gives a number twice as large as the optimum value for
relatively large flow rates at a relatively small head increase. Because the the type of pump you prefer. This result suggests that if you use four
pump impellers resemble propellers, the fluid passes over the vanes pumps in parallel, the Ns-value will be just right. You can now search for
parallel to the pump axis and energy is added through increasing velocity pumps with one-fourth the line capacity with some assurance of finding
and pressure. ones that have good operating efficiency.
The mixed-flow pump is a combination of the two previous types and
it provides moderate pressure increases for moderate flow rates. The Cavitation
mixed-flow pump generally finds greatest application in restricted space
situations such as wells where multistaging can provide the required The possibility of cavitation is particularly important in pump
pressure increase without increasing the pump size to objectionable levels. selection. While the pump vendor is quite capable of assisting in this
Multistaging is the placing of pump impellers in series (generally on process, it is good practice for the engineer to make his own evaluation.
the same shaft) so that pressure undergoes a sequence of incremental The occurrence of cavitation is determined in the testing laboratory.
increases to the required value. This approach is generally used when When the discharge and pump location are such that the pressure at some
space requirements are limited (as discussed above) or when efficiency point in the pump drops to the vapor pressure of the liquid, cavitation
requirements are better served. begins. A further deterioration in the conditions will lead to massive
Paralleling of pumps is generally accomplished by simply placing two cavitation and result in sharp drops in efficiency, possible vibration
or more pumps side by side and connecting them to the pipeline via a problems and, if continued for long periods, failure of the pump
manifold. Parallel pumps are used when flow rate requirements are so hardware.
high that no single pump can efficiently or economically accomplish the Information which can be used to design pump installations to
joh. prevent cavitation is given by pump manufacturers in several ways. The
There are a variety of other pumps not discussed here and many other
simplest is “suction lift.” This value is supposed to be the maximum
considerations in pump selection, depending on the problem situation.
height the pump may be situated above the sump to prevent cavitation
The Pump Handbook by Karassik, Krutzsch, Fraser and Messina [21] is a
under normal conditions. Of course if one has non-normal conditions, it is
good compilation of the various aspects of pumps and pump selection.
not clear what to do.
Another technique (popular with turbine and propeller pump
manufacturers) is to provide a curve of required submerged vs. pump
discharge. This graph will show what distance the impeller must be
7.2 The Pump In the Pipeline submerged to prevent cavitation (or air intake) at sea level for water at
normal temperatures. Adjustments must be made for non-standard
The pumps under consideration here fit into pipeline systems as conditions.
reactive elements. That is, the head increase and flow rate they produce The most fundamental measure of requirements to prevent cavitation
depend on the pipeline system in which they are placed. A large is the net positive suction head (NPSH). An equation for NPSH is
component of pump selection involves the trial-and-adjust process of
finding a pump compatible with the given pipeline system. NPSH = — - - Zs - h L (7-
yy.S entr.
15)
164 UNSTEADY FLOW IN PIPELINES PUMPS IN PIPELINES 165
CL
em
between the sump and the pump impeller. A curve of NPSH vs. discharge
is given by some manufacturers. With a value of NPSH, Equation 7-15
can be used to provide information on Zg for non-standard values of p0
and pv.
Once the pumping configuration has been established, the pumps
selected, and the pipeline sized and laid out, it is time to evaluate the
effects of water hammer. The next sections will be devoted to the
calculation of the pressures and velocities resulting from various
conditions of pump power failure and startup,
As the pump slows down after power failure, its head vs. discharge
and torque vs. discharge characteristics change. It is customary to assume
that as the pump speed changes, the pump characteristics at any speed
can be found using the similarity relationships for homologous pumps.
While we are interested in the change in pump torque as rundown occurs,
we will concentrate at this time on the head vs. discharge characteristic. Figure 7-2. Multi-characteristics for a given pump at various speeds.
For various speeds, the pump characteristics could be shown as on
Figure 7-2 where each curve is calculated from the original curve at No The curves at speeds Nn N2, etc. can be generated by choosing data
using Equations 7-9 and 7-11. It should be noted that in our applications, points along the base curve at No and using the equations
D is the same at all speeds so we can modify the previous equations to the
form
ZZ
Qi = % (7-19)
—o
olz
constant ......................................................(7-16)
II
points all fall on a parabola from Co passing through the origin. If the Setting Up the Equations
same procedure is repeated for points C_ 1( C_2, C-3, etc. on the original
curve, then the complete set of pump characteristics can be drawn. Along each of the parabolas passing through C_ 2, C_2, Co, C+i, etc.,
As the speed of the pump changes, a line is traced on the in Figure 7-2, the values of Q/N and Hp/N2 is a constant. It is, therefore,
characteristic diagram of Figure 7-2 which depicts the head and discharge possible to represent the entire pump characteristic data of Figure 7-2 on a
generated as rundown progresses. Of course the rundown progression is single curve. The same reasoning applies to the torque data T/N2. Typical
determined by the rotational inertia of the rotating parts and the torque examples of these two curves are shown in Figure 7-3. To demonstrate how
exerted by the liquid on the impeller as well as by the pipeline hydraulics. the data from Figure 7-2 can be condensed so markedly, consider finding
Two general types of situations are depicted on Figure 7-2. Situation a given head Hp for a given discharge and speed. If the speed is known,
1 typically occurs when the static lift is high and the pipeline is short. The compute Q/N, read the curve in Figure 7-3(a) for a value of Hp/N2, then
inertia of the water in the pipeline is relatively small and the gravitational compute Hp. In order to find the changed N-value, we must use the torque
force is assisting in the slow-down process. As a consequence, the information and the rotational inertia of the system.
discharge through the pump drops to zero rather quickly while a positive In a more explicit manner, if N is known, then Figure 7-3(a)
head across the pump still exists. Then flow backwards through the pump represents Hp = f(Q) in a manner similar to constant speed pumps. We
occurs unless a check valve is installed in the line. could again generate a parabolic curve to approximate the true function
In the second situation, the pipeline is generally relatively long and a Hp = f(Q) and proceed as before. This time, however, we will take a more
large portion of the head generated by the pump is used to overcome flexible approach.
frictional losses in the line. When power fails in this configuration, the Let us pose the problem of a set of parallel identical pumps in a
large inertia of the fluid in the pipeline dies out slowly, mostly by viscous pipeline. Schematically, they are shown in Figure 7-4. There are five
action; the pump tends to decelerate more slowly and the head drops to unknowns in this system (once the new speed has been computed). The
zero before the flow rate does. At this point the flow either continues equations are as follows:
through the pump, doing work on the impeller in a “windmilling”
situation, or a bypass line around the pump opens and flow bypasses the Suction side C . Vp = Cj • C2 (7-
pump with relatively little loss in energy. Discharge side C': Vpr = C3 + C4 21)
It should be apparent now that, in general, there are four possible d
flow configurations through a pump. Continuity: Vp A = Vp A,r b r u (7-
1. Flow is in the forward direction through the pump and the pump s d 22)
is rotating forward.
Flow in the reverse direction occurs while the pump is still Energy: Hp + H (7-
r
rotating forward. s 23)
Flow in the reverse direction occurs when the pump is rotating Pump: H = f(Q) = f' (V ) or f" (V )................................ ... . . .(7-
backwards. 25) (7-
Finally, flow in the forward direction occurs while the pump is P r
a s 24)
rotating backwards. We have five equations and five unknowns and it remains only to
The possible occurrence of any of these pump flow and rotation solve them. Recall from the section on constant speed pumps, we selected
configurations depends on the existence of bypasses, check valves and three data points and found a quadratic equation for 7-25, then solved for
regular line valves. In general, only the data on configuration no. 1 is all the unknowns. Because this approach was rather restrictive, what we
available from manufacturers. If the pump is expected to operate in any of will do here is construct a table of Q/N vs. Hp/N2 data and use linear
the other three modes, then special tests must be run to obtain the desired interpolation to set up an equation for Hp = f(Q). Care must be taken to
characteristics. divide the pipeline discharge equally between all pumps.
To develop familiarity with the problem solving approach, we will
assume first that there is a check valve in the discharge line and, if the Finding the Change In Speed
pump is a booster pump, we will assume there is a forward bypass around
the pump which opens to bypass the pump in the event the head drops to The previous discussion was based on the premise that the new speed
zero under forward flow conditions. Under these conditions we will be was known at the time the new head and velocity values were to be
operating in the first configuration only. computed. This speed is found by calculating the decelerating torque and
computing the speed change over a At time interval.
168 UNSTEADY FLOW IN PIPELINES PUMPS IN PIPELINES 169
BYPASS
I
CL CM
I2 ■3
-I
vnc-urs VMUVC
H
II
where I is the rotational inertia of the system.
To find the change in speed we must integrate Equation 7-26.
f dN = f dt
J T 2TTI J
We don’t know how T varies with time at this point, so we choose to keep
At small and let T be constant at its known value at the previous time level.
Incorporating this approach, the new rotational speed becomes
By this means the new speed can be determined explicitly using a value of
torque at a known speed obtained by table interpolation from the T/N 2 vs.
Q/N table. It is now possible to proceed with the calculation for the heads
and velocities.
Hp/N2 vs. Q/N will be a sequence of straight line segments (see Figure we can now combine Equation 7-29 easily with the other appropriate
7-5). The equation of the straight line over the anticipated range of Q/N is equations. The result of such a combination gives
z
st
(7-28) ................................ (7-30)
QA. ' QB N C7AdNst
z
where Nst is the number of stages for each of the pumps. Condensing this DU
equation to the form where Npu is the number of pumps in parallel. If Vps > 0, then Equations
7-21 through 7-5 can be used to find the remaining unknowns.
In addition, we must check Hp. If Hp < 0, then we set Hp = 0 and
olz
........................................................(7-29) HPS = Hpd and compute the unknown velocity values from the following
t-
equations:
c c + c c
.................................................... (7-31)
A7”
c+c -
2 4A
S
........................................................... (7-32
The Equations 7-21 and 7-22 can be used to find Hps and HPd.
If, however, Vpj is negative, then Vps = VPd. — 0 and Equations
7-21 and 7-22 are used to compute HPS and Hpd . Also we must check
Q/N and if we are outside the interval A-B, a new set of coefficients C7 and
Ca must be computed and the solution process repeated.
Example 7-1
Solation
DATA
TOTAL HEAD IN FEET PER STAGE
Hi
r;
BRAKE HORSEPOWER
100
75
N.RS.H.— FEET
50
25
10 20 30 40 50
CAPACITY IOO GALLONS PER MINUTE
♦ K»7 Dia.
Q********************************************************** *********************
1:
c* PROGRAM NO. 8 *
2:
3: C* PROGRAM FOR PUMP RUNDOWN FROM POWER FAILURE * 62: C COMPUTE STEADY STATE DISCHARGE
Q********************************************************** ********************* 63: OTRY=QN(4)
64: COF.F-1 2 . * 1 4 4 . * 1 4 4 . * 16 . *NPUMPS* NPUMPS/(6 4.4 * 4 49 • * 4 4 9 • * PI * Pl)
5 : c DO 320 NZ-1,20
6: C WATER HAMMER PROGRAM FOR PUMP RUNDOWN RESULTING FROM POWER FAILURE 65: DO 300 1=1,5
7: C SYSTEM CONFIGURATION HAS PUMPS AT UPSTREAM END AND RESERVOIR AT DOWNSTREAM END 66: IF(OTRY.GT.QN(I).AND.QTRY.LE.QN(1+1)) GO TO 310
8: C PROGRAM WILL HANDLE ANY NO. OF PARALLEL PUMPS-ALL FAIL SIMULTANEOUSLY 67: 300 CONTINUE
9: PROGRAM WILL HANDLE UP TO 8 SERIES PIPES- MUST BE AT LEAST TWO SERIES PIPES 68: WRITE(6,250)
<JO<JOUOOOUO<JOOOOOO<JOO
10 : 69: STOP
11: NPIPES = NUMBER OF PIPES IN SERIES - PIPE NO. 1 IS JUST DOWNSTREAM FROM PUMP 70: 310 HN»HNSQ(I)+(QTRY-QN(I))*(HNSQ(1+1)-HNSQ(I))/(QN(1 + 1)-QN(I))
12: IOUT = PARAMETER WHICH WRITES OUT EVERY IOUT-TH COMPUTATION 71: SUM=0,
13: NPARTS = MINIMUM NUMBER OF PARTS INTO WHICH PIPES ARE DIVIDED 7 2: DO 301 J-l,NPIPES
14: TMAX=HAX. REAL TIME OF RUN-SEC HATM=ATMOSPHERIC PRESSURE HEAD (POSITIVE)-FT 73: 301 SUM=SUM+F(J)*L(J)*QTRY*QTRY/D(J)**5
15: HRES=ELEVATION OF WATER SURFACE IN RESERVOIR 74: PUNCT=HSUHP-HRES+NSTAGE*HN-COEF*SUM
16 : ZEND«ELEVATION OF DOWNSTREAM END OF LAST PIPE AT RESERVOIR - FT 75 : FPRIME-NSTAGE*(HNSQ(1+1)-HNSQ(I))/(ON(1+1)-QN(I))-2.*COEF*SUM/QTRY
17 ; QACC=ACCURACY OF DISCHARGE ITERATION FOR STEADY STATE COMPUTATION - GPM. 76: QNEXT-QTRY-FUNCT/FPRIMF
PIPE»PIPE NUMBER 77: IF(ABS(QNEXT-QTRY).LT.QACC) GO TO 350
18: QTRY-QNEXT
19: AO-WAVE SPEED-FPS L()=LENGTH-FT D()«DIAM)-IN FO-FRICTION FACTOR(DARCY) 78:
79 : 320 CONTINUE
20 : PIPES()-UPSTREAM ELEVATION OF PIPE - FT HEADH()-PRESSURE HEAD IN PIPE - FT WRITE(6,251)
21: 80 :
NPUMPS-NUMBER OF PUMPS IN PARALLEL HSUMP=ELEVATION OF WATER SURFACE IN SUMP 81: STOP
22:
RPM=STEADY STATE PUMP SPEFD-RPM NSTAGE-NUMBER OF STAGES FOR PUMP 82: 350 QLINE=QNEXT*NPUMPS
23 :
24 : 83 : Q-ONEXT
WRSO-MOMENT OF INERTIA OF PUMP AND MOTOR IN LBM-FT SQUARED
25 : QN () ■'DISCHARGE FROM PUMP CHARACTERISTIC CURVE - GPM 84 : HPUMP=NSTAGE*HN+HSUMP
26 : HNSQO-HEAD PER- STAGE FROM PUMP CHARACTERISTIC CURVE - FT 85: C0EF=4 .* 1 4 4 ./( 4 49 ,*PI)
TNSQO- BHP PER STAGE FROM PUMP CHARACTERISTIC CURVE 86: DO 2 1=1,NPIPES
27:
87: VZE RO(I)=QLINE*COEF/D(I) * * 2
28: AREA(I)=0.25*PI*D(I)*D(I)/144.
29 : DIMENSION X(8,30) ,V(8,30) ,H(8,30),HLOW(8,30),HHIGH(8,30), 88 :
5-JCAD (8,30) ,VNEW(B,30) ,HNEW(8,30) 89 : C (I)=32.2/A(I)
3 0:
DIMENSION L(8),D(8),A(8) ,PIPES (8) ,F(8) ,VZERO(8) ,NPAR(8) , 90 : DELTT(I)=L(I)/(NPARTS*(VZERO(I)+A(I)))
31:
5C(8) ,AK(8),SINE(8) ,AREA(8) ,DELTT(8) ,RATIO(8) 91: IFfl.EQ.NPIPES) SINE(I)=(ZEND-PIPEZ(I)»/L(I)
32: IF(I.EQ.NPIPES) GO TO 2
33: DIMENSION QN(6),HNSQ(6),TNSQ(6) 92:
LOGICAL FAIL,SHUT,ZEROH 93: SINE(I)=(PIPEZ(1+1)-PIPE2(I))/L(I)
34 : 2 CONTINUE
35 : INTEGER PIPE 94 :
C
36: REAL LA,L 95:
C ** COMPUTE MINIMUM DELTA T **
37: NAMELIST /SPECS/ NPIPES,IOUT,NPARTS,TMAX,HATM,HRES,ZEND,QACC 96:
C
38 : NAMELIST /PUMPS/ NPUMPS,RPM,WRSQ,NSTAGE,HSUMP,ON,TNSQ,HNSQ 97: DELT=DELTT(1)
39 : C 98: KMIN=1
40: READ(5,SPECS) 99: DO 12 1=2,NPIPES
41: READ(5,200) (PIPE,D(I),L(I),F(I),A(I),PIPEZ(I),1=1,NPIPES) 100: IF(DELTT(I).GT,DELT) GO TO 12
42: 200 FORMAT(I10,5F10,0) 101: DELT-DELTT(I)
43: READ(5,PUMPS) 102: KMIN»I
44: C 103: 12 CONTINUE
45 : WRITE(6,202) 104 : DO 13 1=1,NPIPES
46 : WRITE(6,101) NPIPES,NPARTS,HRES,HATM,ZEND,TMAX,IOUT,QACC 105: IF(I»EQ,KHIN) GO TO 14
4 7: WRITE(6,102) 106 : ANPAR=L(I)/(DELT*(VZERO(I)+A(I)))
DO 3 1 = 1,NPIPES 107: NPAR(I)=ANPAR
4 8:
LA-L(I)/A(I) 108: TEST-NPAR(I)
49 :
3 WRITE(6,103) I,D(I) ,L(I) ,F(I) ,A (I) , PIPES (I) , LA 109: IF(ABS(TEST+l-ANPAR).LT..0001) NPAR(I)-NPAR(I)+1
53: GO TO 13
51: WRITE(6,104) 110:
NPUMPS,NSTAGE,RPM,WRSO,HSUMP 111: 14 NPAR(I)-NPARTS
52: WRITE(6,105)
112: 13 CONTINUE
53: WRITE(6,I06)
(QN(I),HNSQ(I),TNSQ(I),1-1,6) 113: C
54 : WRITE(6,107) C CONVERT TO 0/N, H/NSQ, T/NSQ
55: C 114:
FAIL-.FALSE. COEF-60.* 550./(2.*PI)
56: 115:
DO I 1=1,6
57: SHUT-.FALSE. 116: QN (I)-QN(I)/( 4 49 .*RPM)
58: ZEROH-.FALSE. 117: TNSQ(I)=NSTAGE*COEF*TNSQ(I)/RPM** 3
59 : PINSRT-WRSQ/32.2 118: 1 HNSQ(I)=HNSQ(I)/RPM**2
60 : PI-3.141592 119:
61: C 120 :
121:
122: C 182: C
123: C 183: C BEGIN TRANSIENT ANALYSIS
124 : C ♦* SET UP CONDITIONS FOR STEADY STATE (T-0) ** 184: C ---------------------------------------------------------------- --- -----------------------------------------------------------------------------------
125: C 185: C
126: H(1,1)-HPUMP 186: DO 99 II«1,INDEX
127: V(1,1)»VZERO(1) 187: T-T+DELT
128: DO 20 I»l,NPIPES 188: C
129: DELL“L(I)/NPAR(I) 189: C ** COMPUTE H AND V AT INTERIOR NODES **
130: DELHF=12.*F(I)*DELL*VZERO(I)**2/(64.4*D(I)) 190: C
131: AK(I)»12.*F(I)* DELT/(2.* D(I)) 191: DO 30 1 = 1,NPIPES
132: K=NPAR(I)+1 192: K«NPAR(I)
133: X(I,l)-0, 193: DO 30 J = 2,K
134: HEAD(I,1)"H(I,1)-PIPEZ(I) 194 : RA«RATIO(I)*A(I)
135: RATIO(I)"DELT/DELL 195: VMINUS“V(I,J)-RA*(V(I,J)-V(I,J-l))
136: DO 21 J«2,K 196: VPLUS=V(I,J)-RA*(V(I,J)-V(I,J +1))
137: X (I,J)»(J-1)*DELL/L(I) 197: HMINUS»H(I,J)-RA*(H (I,J)-H(I,J-l))
138: H (I, J)=H(1,1)-DELHF*(J-l) 198: HPLUS“H(I,J)-RA*(H(I,J)-H(I,J + l))
139: HEAD(I,J)«H(I, J)-(PIPEZ(I)+X(I,J)*L(I)*SINE(I) ) 199: VNEW(I,J)=0.5*(VMINUS+VPLUS+C(I)*(HMINUS-HPLUS)+C(I)*DELT*SINE(I)
140: 21 V(I,J)«VZERO(I) 200:
141: IF(I.EQ.NPIRES) GO TO 20 $*(VMINUS-VPLUS)-AK(I)*(VMINUS*ABS(VMINUS)+VPLUS*ABS(VPLUS)))
201: HNEW (I,J)=0.5*(HMINUS+HPLUS+(VMINUS-VPLUS)/C (I)+DELT*SINE(I)*
142: H(I+1,1)-H(I,K) 202 :
143: V(1+1,1)»VZERO(1+1) $(VMINUS+VPLUS)-(AK(I)/C(I))*(VMINUS*ABS(VMINUS)-VPLUS*ABS(VPLUS)))
203:
144 : 20 CONTINUE 30 CONTINUE
204 :
145: DO 23 1-1,NPIPES C
205:
146: K=NPAR(I)+1 C ** COMPUTE H AND V AT INTERIOR JUNCTIONS **
206:
147: DO 23 J«1,K C
207:
148: HLOW(I,J)«H(I,J) KK»NPIPES-1
208:
149: 23 HHIGH(I,J)®H(I,J) DO 40 I»1,KK
209:
150: PHMAX--100, K=NPAR(I)+1
210:
151: PHMIN-100000. 211: VMINUS’V(IfK)-RATIO(I)*A(I)*(V(I,K)-V(I,K-1))
152: DO 11 I«l,NPIPES HMINUS-H(I,K)-PATIO(I)*A(I)*(H(I,K)-H(I,K-l))
212:
153: K«NPAR(I)+1 213: CCC“VMINUS+C(I)*HMINUS + C(I)*VMINUS*DELT*SINE(I)-AK(I)*VMINUS*
154 : DO 11 J-1,K 214: SABS(VMINUS)
155: IF(HEAD(I,J).LT.PHMAX) GO TO 16 215: VPLUS»V(1+1,1)-RATIO(1+1)*A(I+1)*(V(I+1,1)-V(1+1,2))
156: IPMAX-I 216: HPLUS = H(1 + 1,1)-RATIO(1+1)* A (1+1)* (U (1+1,1)-H(I +1 , 2))
157: XMAX-X(I,J) 217: CC=VPLUS-C(1+1)*HPLUS-C(1 +1)* DELT*VPLUS*SINE(1 +1)-AK(1+1)*VpLUS*
158: PHMAX=HEAD(I,J) 218: SABS(VPLUS)
159: TTMAX’T 219: HNEW(I,K)=(-CC*AREA(1+1)+CCC*AREA(I))/(C(1+1)*AREA(1+1)+C(I)*
160: GO TO 17 220: SAREA(I))
161: 16 IF(HEAD(I,J).GT.PHMIN) GO TO 17 221: HNEW(1+1,1)“HNEW(I,K)
162: IPMIN-I 222 : VNEW(I,K)=CCC-C(I)*HNEW(I,K)
163: XHIN»X(I,J) 223: VNEW(I+1,1)=CC+C(1+1)*HNEW(1+1,1)
164: PHMIN-HEAD(I,J) 224: 40 CONTINUE
165: TTMIN»T 225: C
166: 17 CONTINUE 226: C ** COMPUTE H AND V AT DOWNSTREAM END **
167: 11 CONTINUE 2 27 ; C —THIS BOUNDARY CONDITION IS FOR A CONSTANT HEAD RESERVOIR—
168: C 228 : C
169: C ♦♦ WRITE OUT STEADY STATE CONDITIONS ** 229 : NV=NPAR(NPIPES)+1
170: C 230: VMINUS’V(NPIPES,NV)-RATIO(NPIPES)*A(NPIPES)*(V(NPIPES,NV)-
171: T“0.0 231: SV(NPIPES,NV-1))
172: INDEX»TMAX/DELT+1 232 : HMINUS=H(NPIPES,NV)-RATIO(NPIPES)* A(NPIPES)*(H(NPIPES,NV)-
173: WRITE(6,204) 233:
174: WRITE(6,205) T SH(NPIPES,NV-1))
234 : CC«VMINUS+C(NPIPES)*HMINUS+C(NPIPES)*VMINUS*DELT*SINE(NPIPES)-
175: DO 22 1-1,NPIPES 235:
176: K»NPAR(I)+1 SAK(NPIPES)*VMINUS*ABS(VMINUS)
236 : HNEW(NPIPES,NV)-HRES
177: WRITE(6,206) I,(X(I,J),HEAD(I,J),H(I,J),V(I,J),J = 1,K)
237: VNEW (NPIPES,NV)=CC-C(NPIPES)*HNEW(NPIPES,NV)
178: 22 CONTINUE 238:
179: WRITE(6,212) RPM,Q C
239: C BOUNDARY CONDITIONS AT THE PUMP
180: Q-Q/449. 240 :
181: C IF(SHUT.OR.ZEROH) GO TO 42
241:
QQ=Q/RPM
DO 550 1-1,5
302 : K=NPAR(I)+1
242: IF(OQ.GT.pN(I).AND.QQ.LE.ON(1+1)) GO TO 551 303 : DO 61 J«1,K
243 : 550 CONTINUE 304 : IF(HNEW(I,J).LT.HLOWfl,J) ) HLOW(I,J)=HNEW(I,J)
244: WRITE(6,260) 30 5 : IF(HNEW(I,J).GT.HHIGH(I,J)) HHIGH(I,J)= HNEW(I,J)
245 : STOP 306 : HEAD(I,J)=HNEW(I,J)-(PIPEZ(I)+X(I,J)*L(I)‘SINE(I))
246 : 551 TN2=TNSp(I)+(OO-QN(I) ) * (TNSO (1 +1) -TNSQ (I) ) / (ON (I +1) -pH (I) ) 307: IF(HEAD(I,J).LT.PHMAX) GO TO 66
247: RPM’RPM-30. *TN2*DELT/(PI*PINERT)*RPM*RPM 308: IPMAX’I
248 : 42 VPLUS’V(1,1)-RATIO(1)*A(1)*(V(1,1)-V(1,2)) 309 : XMAX=X(I,J)
249: HPLUS3 fJ (1,1)-RATIO (1) *A(1) * (H (1 , 1)-H(1,2) ) 310 : PHMAX-HEAD(I,J)
250 : C1«VPLUS-C(1)*HPLUS-C(1)*OELT*VPLUS*SINE(1)-AK(1)*VPLUS*A3S(VPLUS) 311: TTMAX’T
251: IF(SHUT) GO TO 43 312: GO TO 67
252: IP(ZEROH) GO TO 518 313: 66 IF(HEAD(I,J).GT.PHMIN) GO TO 67
253 : DO 500 NZX1,20 314 : IPMIN«I
254 : HA=HNSO(I) 315: XMIN«X(I,J)
255 : HD»HNSO(I+1) 316: PHMIN=HEAD(I , J)
256: QA»pN(I) 317: TTMIN’T
257: QF3 = QN (1 + 1) 318: 67 CONTINUE
258 : C5=NSTAGE*RPM*RPM 319: IF(HEAD(J,J).GT.-HATM) GO TO 62
259 : C7“lHA-HB)/(QA-QB) 320 : WRTTE(6,214) T,I,X(I,J)
260 : C8’HB-QB*C7 321: FAIL’.TRUE.
261: C6=HSUMP+C5*C8+C5*C7*C1*AREA(1)/(NPUMPS*RPM) 322 : 62 CONTINUE
262 : C9“C5*C7*C(1)*AREA(1)/(NPUMPS*RPM) 323: 61 CONTINUE
263 : HPP=C6/(1.-C9) 324 : C
264 : VPPxCl+HPP*Cd) 325: C “ WRITE OUT H,V,AND HEAD VALUES AND TEST FOR TMAX **
265 : QPP=VPP*AREA(1)/(NPUMPS*RPM) 326: C
266 : DO 510 J»1,5 327 : 70 IF(FAIL) GO TO 71
267 : IF(QPP.GT.pN(J).AND,QPP.LE.QN(J+l)) GO TO 520 328 : IF(MOD(II.IOUT).NE.0) GO TO 72
268 : 510 CONTINUE 329 : 71 WRITE(6,205) T
269 : IF(OPP.GT.0.) WRITE(6,280) 330 : DO 73 I=1,NPIPES
270 : IF(QPP.GT.0.) GO TO 518 331: K»NPAR(I)*1
271: 519 J»I-1 332 : WRITE(6,206) I,(X(I,J),HEAD(I,J),HNEW(I,J),VNEW(I,J),J=1,K)
272 : IF(J.EQ.0) GO TO 43 333 : 73 CONTINUE
273 : GO TO 521 334 : IF (SHUT.OR.ZEROH) GO TO 74
274 : 520 IF(J.EQ.I) GO TO 530 335 : OQQa449.*Q
275: 521 I=J 336 : WRITE(6,212) RPM,00Q
276: 500 CONTINUE 337 : 74 IF(FAIL) GO TO 400
277 : WRITE(6,270) 338 : 72 CONTINUE
278 : STOP 339 : C
279 : 530 HNF,W(1,1)=HPP 340 : C ** PREPARE FOR NEXT TIME STEP COMPUTATION **
280 : VNEW(1,1)= VPP 341; C
281: O=QPP*RPH 342: DO 80 I=1,NPIPES
282 : GO TO 540 343: K»NPAR(I)+1
283 : 518 ZEROH-.TRUE. 34 4 : DO 80 J»1,K
284 : SHUT=>FALSE. 345: V (I,J)=VNEW(I,J)
285: HNEW(1,1)-HSUMP 346: 80 H(I,J)»HNEW(I,J)
286: VNEW(1,1)=C1+HSUMP*C(1) 347: IP(T.GT.TMAX) GO TO 400
287 : IF(VNEW(1,1).LE.0.) GO TO 43 348: 99 CONTINUE
208 : Q»VNEW(1,1)‘AREA(1)/NPUMPS 349 : C
289 : GO TO 540 350: - ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------
290 : 43 VNEW(1,1)= 0.0 351: C END OF TRANSIENT ANALYSIS
291: 0=0. 352: C-
292 : SHUT®.TRUE. 353: C
293: ZEROH®,FALSE. 354 : c
294 : IINEW (1 ,1) ® (VNEW (1 ,1) -Cl)/C (1) 355 : c ** WRITE OUT MAXIMUM AND MINIMUM HEADS AND H-VALUES **
295 : IF(KNEW(1,1).LT.HSUMP) GO TO 518 356: c
296 : 540 CONTINUE 357: 400 CONTINUE
297 : C 358 : WRITE(6,207)
298 : C 359 : DO 401 IX1,NPIPES
299: C ** LOCATE MAXIMUM AND MINIMUM HEADS AND H-VALUES ** 360: WRITE(6,208) I
300 : C 361: K’NPAR(I)+1
301: DO 61 I»1,NPIPES
Example 7-2
2: C* PROGRAM NO. 9 *
3: C* PROGRAM FOR BOOSTER PUMP RUNDOWN FROM POWER FAILURE *
4:
5 : c
Figure 7-8. Pressure extremes resulting from power failure in a booster pomp configuration.
6: C WATER HAMMER PROGRAM FOR PUMP'RUNDOWN RESULTING FROM POWER FAILURE
7: C BOOSTER PUMP IS AT INTERNAL BC WITH RESERVOIRS AT UPSTREAM AND DOWNSTREAM ENDS
8: C PROGRAM WILL HANDLE ANY NO. OF PARALLEL PUMPS-ALL FAIL SIMULTANEOUSLY
9: C PROGRAM WILL HANDLE UP TO 8 SERIES PIPES- MUST BE AT LEAST TWO SERIES PIPES
10: C
11: C NPIPES = NUMBER OF PIPES IN SERIES - PIPE NO. 1 IS AT UPSTREAM END
12: C IPUMP = NUMBER OF THE PIPE WHICH HAS BOOSTER PUMP AT DOWNSTREAM END
13: C IOUT ' PARAMETER WHICH WRITES OUT EVERY IOUT-TH COMPUTATION
14: C NPARTS « MINIMUM NUMBER OF PARTS INTO WHICH PIPES ARE DIVIDED
15: C TMAX-MAX. REAL TIME OF RUN-SEC HATM'ATMOSPHERIC PRESSURE HEAD (POSIT IVE)-FT
16: C A()=WAVE SPEED-FPS L()'LENGTH-FT D()XDIAH)-IN F()'FRICTION FACTOR(DARCY)
17: C PIPEZ{)'UPSTREAM ELEVATION OF PIPE - FT HEAD()=PRESSURE HEAD IN PIPE - FT
18: C ZEND'ELEVATION OF DOWNSTREAM END OF LAST PIPE AT RESERVOIR - FT
19: C HRESUP = ELEVATION OF WATER SURFACE IN UPSTREAM RESERVOIR
20: C HRESDN = ELEVATION OF WATER SURFACE IN DOWNSTREAM RESERVOIR
21: C QACC=ACCURACY OF DISCHARGE ITERATION FOR STEADY STATE COMPUTATION - GPM
22: C
23 : C NPUMPS'NUMBER OF PUMPS IN PARALLEL
24 : C RPM'STEADY STATE PUMP SPEED-RPH NSTAGE'NUMBER OF STAGES FOR PUMP
25 : C WRSQ'MOMENT OF INERTIA OF PUMP AND MOTOR IN LBM-FT SQUARED
26: C ON()'DISCHARGE FROM PUMP CHARACTERISTIC CURVE - GPM
27: C HNSQO'HEAD PER STAGE FROM PUMP CHARACTERISTIC CUPVE - FT
28 : C TNSQ()» BHP PER STAGE FROM PUMP CHARACTERISTIC CURVE
29: C
30 : DIMENSION X(8,30),V(8,30),H(8,30) ,11 LOW (8,30) , HHIGH (8,3fl) ,
31: $HEAD(8,30),VNEW(8,30),HNEW(8,30)
32: DIMENSION L (8) ,D(8) ,A(8) ,PIPEZ(8) ,F(8) ,VZERO(8) ,NPAR(8),
33: $C(8),AK(8), SINE(8),AREA(8),DELTT(8),RATIO(8)
34 : DIMENSION QN(6),HNSQ(6),TNSQ(6)
35: LOGICAL FAIL,SHUT,BYPASS
36: INTEGER PIPE
37: REAL LA,L
38: NAMELIST /SPECS/ NPIPES,IOUT,NPARTS,TMAX,HATH,HRESUP,HRESDN,ZEND,
39: SQACC,IPUMP
40 : NAMELIST /PUMPS/ HPUMPS,RPM,WRSQ,NSTAGE,QN,TNSO,HNSQ
41: C
42: READ(5,SPECS)
43: READ(5,200) (PIPE,D(I),L(I),F(I),A(I),PIPEZ(I),1=1,NPIPES)
44 : 200 FORMAT (11 0,5 Fl 0.0 )
45 : READ(5,PUMPS)
46: C
47: WRITE(6,202)
48: WRITE(6,101) NPIPES,NPARTS,I PUMP,HRESUP,HRESDN, ,HATH,ZEND,TMAX,
49: $IOUT.,QACC
50: WRITE(6,102)
51: DO 3 1=1,NPIPES
52: LA-L(I)/A(I)
53 : 3 WRITE(6,103) I,D(I),L(I),F(I),A(I),PIPEZ(I),LA
54 : WRITE(6,104)
55: WRITE(6,105) NPUMPS,NSTAGE,RPM,WRSO
56: WRITE(6,106)
57: WRITE(6,107) (QN(I), HNSO(I),TNSQ(I),1=1,6)
58: C
59 : FAIL'.FALSE.
60: SHUT'.FALSE.
61: BYPASS'.FALSE.
62 : PINERT’WRSO/32.2
63: PI'3.141592
o
o O
ro O
C Figure 7-9. Computer program for booster pump rundown as a result of
O
power failure.
184 UNSTEADY FLOW IN PIPELINES PUMPS IN PIPELINES 185
190: DO 99 11«11INDEX
191: T=T-DELT
253: IF(QPD.GT.QN(33).AND.QPD.LE.QN(33+1)) GO TO 48
192: C 47 CONTINUE
254:
193: C ** COMPUTE H AND V AT INTERIOR NODES ** IF(OPD,GT,0.) WRITE(6,280)
255;
194 : C IF(QPD.GT.0.) GO TO 50 *
256:
195: DO 30 1*1,NPIPES
257: 33=3-1
19’6: K = NPAR(I) 259: IF(33.E0.K) GO TO 43
197: DO 30 3 = 2,K 259: GO TO 49
198 : P.A=RATIO (I) * A (I) 48 IF(3.EQ.33) GO TO 50
260 :
199: VMINUo=V(I,3)-RA*(V(I,3)-V(I,3-1)) 261: 49 3*33
200 : VPLUS«V(1,3)—RA*fV(I,3 -V(I,3+1)) 262: 46 CONTINUE
201: HMIWJS’H (1,3) -RA’ (H (I, J) -Fl (1,3-1) ) 263: WRITE(6,270)
20 2: HPLUS’H(I,3)-RA*(H(I,3’-H(I,3+1)) 264: STOP
203: VNE«(I,3) =0.5* (VMINUS+VPLUS+C(I) * ( HMINUS-HPLUS) +C (I) * D^LT*!5I NE (1) 265 : 50 CONTINUE
204 : S*(VMINUS-VPLUS)-AK(I)*(VMINUS*ABS(VMINUS»+VPLUS*ABS(V?LUS)]) 266: Q*QPD*RPM
205 : KNEW(I,3)=0,5*(HMINUS+HPLUS+(VMINUS-VPLUS)/C(I)+DELT*SINE(I)* 267 . HNEW(I + 1,1) *HPD
206: $(VMINUS+VPLUS)-(AK (I) /C (I) ) *(VMINUS*A8S(’MINUS)-VPLUS*ABS(VPLUS3)) 268: VNEW(I+1,1)=VPD
207 : 30 CONTINUE 269 : VNEW(I, K)= AREA(1 + 1)*VPD/AREA(I)
208 : C HNEW(I,K)*(C1-VNEW(1 +1 , 1))/C(I)
270:
209 : C ’* COMPUTE H ANC V AT INTERIOR 3UNCTIONS ** HTEST*HNEW(1+1,1)-HNEW(I, K)
271:
210: C
272: HPUMP’HTEST
211: KK=NPIPES-1
273 IF(HTEST.GE.O.) GO TO 40
212 : DO 40 1*1,KK 274 : 4 2 VNEWtI + l,1)*(C1*C(1 + 1)+C(I)*C3)/(C(I)+C(1+1)*AREA(I+ 1)/AREA(I ) )
213: K*NPAR(I)+1 275 ; VNEW(I,K) = ARF.A(I+1) *VNEW(1+1,1)/AREA(I)
214: VMINUS’V(I,K)-RATIO 11) * A (I) * (V (I, K) - V (I, X-l) 276 : KNEW(1+1,1)=(VNEW(I+1,1)-C3)/C(1+1)
215 : HMINUS*H(I,K)-RATIO!I)"A (I) * (H(I,K)-H(I,X-l) 277: HNEW(I,K)=HNEW(I+1,1)
216: C1*VMINUS+C(I)*HMINUS+C(I)*VMINUS*DELT’SINE(2)-AK(I)*VHIKU5* 278: IF(VNEW(1 +1,1)♦LT,0 .) GO TO 43
217: $ABS(VMINUS) 279:
218 : VPLCS«V(I+1,1)-RATIO(I-1)*A(1+1)*(V(I+1,1)-/-I+1,2)) BYPASS*.TRUE.
280: SHUT*.FALSE.
219: HPLUS-iKI + l, 1) -RATIO (1*1) *A(I + 1)*(H(I + L,_)-4'I + 1,2))
261: HTEST»HNEW(1+1,1)-KNEW(I,K)
220 : C3»VPLUS-C (1+1) ‘HPLUS-C (1 + 1) *DELT*VPLUS*SINE ’ I +1) -AK (’-1) *7PLUS ‘
221: 282: IF(HTEST,LE.0.) GO TO 40
$ABS(VPLUS)
222: IF(I.EO.IPUMP) GO TO 41 263: 43 VNEW(I,K)=0.
264: VNEW(1 + 1,1)=0.0
223: HNEW(I,K) = (-C3*AREA(1 + 1]+C1 *AREA(I) )/(C(Z+l•’AREA(1 + 11~C(Ii*
265: Q=0.
224 : $AREA(I))
286: SHUT*.TRUE,
225: HNEl\(I+l,l)=HNF,W(I,K)
267 : BYPASS*.FALSE.
226 : VNEW(I,K)=C1-C(I)*HNEW(I,K)
268 : HNEW(I,K)*Cl/C(I)
227 : VNEW(1+1,1)-C3+C(1+1)*FNEW(I+1,1)
269 : HNEW(1 + 1,1)*-C3/C(I +1)
228 : GO TO 40
290: HTEST*HNEW(I+1,1)-HNEW(I,K)
2 29: C
291: IF(HTEST,GE.0.) GO TO 40
230 : C COMPUTE CONDITIONS AT BOOSTER PUMP
292 : GO TO 42
231: C 293 : 40 CONTINUE
232 : 41 IF(SHUT) GO TO 43 294 :
233 : IF(BYPASS) GO TO 42 295 : C ** COMPUTE H AND V AT UPSTREAM END **
234 : Q0=Q/RPM 296 : C —THIS BOUNDARY CONDITION IS FOR A CONSTANT HEAD RFSERV0IR--
235: DO .4 1 3=1,5 297 : C
236 : IF(QQ.GT.QN(3).AND.QQ.LE,QN(3+1)) GO TC 45 298 : VPLUS*V(1,1)-RATIO(1)*A(1)*(V(1,1)-V(1 , 2) )
2 37 : 44 CONTINUE 299: HPLUS = H(1,1)-RATIO(1)*A(1)*(H (1,1)-H(1,2))
238 : WRITE(6,260) 300 :
239 : 4 5 TN2=7NSQ(3)+(QQ-QN(3) 1 *(TNSQ(3+1)-TNSO(3))/ i CN (3 + 1)-02(3)) C3=VPLUS-C(1)*HPLUS-C(1)* DELT*VPLUS*SINE(1)-AK(1) *VPLUS*ABS(VPLUS)
301: HNEW(1,1)“HRESUP
240 : PPM=FPM-30.*TN2*DELT/(PI*PINERT)*RPM*PP« 302 : VNEW(1,1)=C3+C(1)*HRESUP
241: DO 46 NZ»1,20 303 :
242: HA=HNSQ(3) 304 : C *’ COMPUTE H .AND V AT DOWNSTREAM END **
243 : HB=HNSO(3+1) 305 : 2 —THIS BOUNDARY CONDITION IS FOR A CONSTANT HEAD RESERVOIR—
2 4 4 : CA=0N(3) 306: C
245: QE=QN(3+1) 307 : NV*NPAR(NPIPES)+1
246 : C7= (flA-HB) / (CA-QB) 308 : VMINUS=V(NPIPES,NV)-RATIO(NPIPES)* A(NPIPES)*(V(NPIFES,NV)-
247 : C8=HB-QB*C7 309 : $V(NPIPES,NV-1))
248 : 310: HMINUS=H(NPIPES,NV)-RATIO(NPIPES)*A(NPIPES)*(H(NPIPES,NV)-
VPD*(Cl/C (I) +NSTACE*RPM*RPM*C8+C3/C (I + l’i'il TIT + IUMFA (TXH / 311:
249 : $ (C ( I)‘AREA ( I) )-RPM*NSTAGE*C7*AREA (1+1)/NatjMps’ ■ (I )/ $H(NPIPES,NV-1))
312: Cl *VMINUS + C(NPIPES) *HMINUS + C(NPIPES) *VMINUS*DELT*S7ME(NPIPES)-
2 50 : HPD=(VPD-C3)/C(I + D .........
2 51: QPD=V?D*AREA(1+1)/(NFUMPS*RPM) 313: $AK(NPIPES)*VMINUS*ABS(VMINUS)
252; 00 47 33*1,5 314 :
HNEW(NPIPES,NV)=HRESDN
315 :
VNEW(NPIPES,NV)*C1-C(NPIPES)*HNEW(NPIPES,NV)
z
7.4 Selective Pump Tripout
Ci-
Q-
>
Combining these equations with Equations 7-21 through 7-24 gives
Occasionally it is necessary to find the pressure resulting from the
the following equation for Vp .
shutdown of one or more of a set of parallel pumps. The effect is generally
of less significance than power failure to all the pumps. We will take the
ZI
z
N
-Nfa
z
Pu \ pu
approach of modifying Equations 7-21 through 7-35 to this application fa
+
o
and assume each individual pump discharge line has a check valve. N0NstC7/
■J
'fa
z
z
Setting Up the Equations PU fa
T3
A-sCiJ N NstC,
Referring again to Figure 7-4 for a typical booster pump application,
we see that Equations 7-21, 7-22, 7-23 and 7-24 remain unchanged. ........................................................ (7-37)
Equation 7-25 must be replaced because all the pumps no longer operate
in harmony. If V pd >, 0, then we must check tripped-out pumps to see if check valves
If Nfa is the number of pumps failing, then the new speed of Nfa of should be closed. To accomplish this we must compute Qfa from the
the pumps is given in Equation 7-27 as following equation
N = N(t) - T(t)At
27rl
•Q
Each of the tripped out pumps generates head according to Equation 7-29
H = N2N?t ............................ (7-33) If Qfa < 0, then the computation for must be redone with the
>
OH
T5
p st
tripped-out pump check valves closed.
o |o
where Qfa is the discharge through each of the decelerating pumps. The
II .
pumps still operating under power generate head according to the
(7-39)
equation
T5
Hp = No2 .................................. (7-34)
Calculation of the remaining heads and discharge values proceeds as
before. The individual pump check valves, once closed, should never be
Of course the head increase across the pump configuration must permitted to reopen on the failing pumps. However, on running pumps
equal the head increase in each pump. The flows through the individual check valves have to reopen.
pumps adjust to satisfy this necessity. Equating 7-33 and 7-34 gives the
relationship between the discharges in the tripped out and constant speed 7.5 Pomp Power Failure With Backflow and Windmilling
pumps.
In certain pumping installations, neither check valves nor bypass
valves are employed. Backflow, if it occurs, is either permitted to continue
Z,
operating characteristics are not normally available from the pump of q/n, h/n2 and r/n2 become undefined, though finite. Some technique
manufacturers. must be used to circumvent this problem. The technique most commonly
Under these conditions the engineer is generally faced with the employed for small n is to redefine the dimensionless relationships so that
options of paying for extensive laboratory tests to determine the n no longer appears in the denominator.
non-standard characteristics, configuring the system so neither backflow For example, use n/q as the abscissa for the plots with small n. To
nor windmilling occurs, adopting known non-standard characteristics replace the ordinate h/n2, combine it with (q/n)2 to give h/q2. The same
from similar pumps to his application, or omitting this type of analysis approach suggests using T/q2 for torque. This adjustment is standard
altogether. We will assume here that the necessary non-standard practice and permits use of the similarity relationships over the full range
characterises are available and set up the analysis accordingly. of n for both positive and negative directions of rotation. Examples of
these types of characteristics are taken from Miyashiro [22] and shown in
Complete Pump Characteristics Figure 7-10. It only remains to illustrate how the various characteristics
are handled in setting up and solving the equations for a typical system.
If head, discharge and power information are available for both
standard and non-standard configurations of flow direction and rotation, Setting Up the Equations
then the complete set of pump characteristics is said to exist. Because we
have had data on head and power for both forward and reverse flow and We will set up the equations for a typical source pumping plant with a
forward and backward pump rotation, four sets of data seem to be mainline discharge control valve, no check valve and no bypasses. The
required. However, forward flow with backward pump rotation is so rare, sketch of the system is shown in Figure 7-11. There are NpU identical
it is generally not included. Hence, we have three sets of characteristics to pumps in parallel and we will assume power fails simultaneously to all.
manage. The C‘ characteristic equation in the pipeline gives us one equation
In Art. 7.3 we employed the basic data taken directly from pump and continuity gives a second.
characteristic diagrams to include in our analysis. Because it is much
more common to use normalized pump data, we will define the Vp - Cj + C2 ........................................................... (7-
normalized parameters based on values taken at the operating condition 41)
of maximum efficiency. The equations would be
<
CL, .
O'
(7-
o2|z
........ (7-40a) pu
42)
S3
II
II
-r
2g
........ (7-40d)
In this case, because backflow will occur, reverse flow KLRCV information
ev
should be obtained for the valve.
The resulting dimensionless characteristic diagrams would all lie within Writing the energy equation between the sump and the discharge line
limited ranges of for positive flow,
= Hd ................................................................................................
(7-44)
S3
Another problem must be addressed in this application which was where Hs is the H-value of the sump. Hydraulic losses in the manifold
not important earlier. Now that backflow through the pump can occur, have been ignored here for simplicity and because their impact on
pump rotation can reverse. When n passes through zero speed, the values significant water hammer pressures is small.
194 UNSTEADY FLOW IN PIPELINES PUMPS IN PIPELINES 195
•2.0
PUMPS
*1.5
//////// //////77
N
-C C
'0.5
7777777777777777777777/
•1.0 L~
-l.o -0.5 0 *0.5 ‘l.o *l.s Figure 7-11. Pumping plant
configuration for illustrating effects of
~n-
backflow.
(a) Characteristics for forward rotation with forward and reverse flow. To complete the set of equations, the pump characteristics must be
included. Again, assume flow is in the forward direction
Hp = f(Q)
Assuming the pump characteristics are going to be represented by
piecewise continuous linear equations, we can represent the above
equation in the form
IP
P IP
st
~4
O
+
p
o
h= > 2 ........................................ (7-45b)
LQ
st
p
where Nst, C7, Ct are defined as in Equation 7-29.
Recognizing that Hp = h-HpQ and Q = q-Q0, Equation 7-45 can be
written
C7Q ..........................................
(7-46a)
a
O
o
(b) Characteristics for forward and reverse rotation with reverse flow.
Once we have selected the proper interval in the table of pump where
characteristics and evaluated the coefficients in Equation 7-46, we can
proceed with solving the equations.
(7-55)
I
It should be noted that if q = (-), then the following set of equations
I
would have to be solved in conjunction with Equations 7-41 and 7-42. Rev
and
Hd = Hp-K .......................................................... (7-47)
Rev 2g
r
......................................
10
Hs - HP = H d ....................................................................... (7-48) Rev
For positive flow direction (using Equation 7-45a), If, however, q/n is less than zero or greater than 1.0, then the
following equations for C9 and C10 must be used in conjunction with
Equation 7-54.
................................ (7-49)
NstC7Apn
where
QoNP ..............................................
............................ 7-50)
II
Rev NsCaA
r
and 2g
QoN
2
PU
and
............................ (7-51)
10
CL,
r
I
2g
Q0Npu
c
1
QoNpu ...........................................................
II
interpolation schemes for the pump characteristics you are operating in.
2g Q N 2v 12 pump, but the type of motor used to drive the pump. Because of these
o pu
uncertainties, we will analyze pump startup by assuming that the pump
If, however, the flow is in the reverse direction, the following equation attains its normal operating speed instantaneously and make provision for
for Vp must be used. bringing several pumps on line at given multiples of At. This approach
should give more severe pressures than a more precise analysis and should
....................... (7-54) result in a conservative design.
tn
o
.........
High Pressures
<Z
o
n
z
z
Unsteady flow in a pipeline system can result in objectionably high or
i
.........................
ii
N0NstC7A low pressures. Excessively high pressures can result in damage to pumps,
CL
>
Q.
ru itself. However, it should be kept in mind that “failure” can refer only to
The remaining variables are calculated from Equations 7-21 through 7-24 inability to meet a given performance standard, hence, no physical
and 7-59 and 7-60. damage must occur to constitute a failure. For example, it may be
Because in pump startup there is no reason to expect back flow to required that under all conditions, the pressure in a pipeline must remain
occur, no provision is generally necessary. It only remains to devise a greater than atmospheric pressure to prevent air entering the line through
program for bringing the pumps on line and write some input code for the vacuum breakers. If analysis showed the pressure to drop below
computer program which will accommodate the proposed startup scheme. atmospheric under any conditions, the pipe has “failed.”
Low Pressure*
199
............
200 UNSTEADY FLOW IN PIPELINES PROBLEMS RESULTING FROM UNSTEADY FLOW 201
Vibrations is generally controlled by a dashpot or pilot system and this type of valve
functions much like an in-line surge relief valve.
A type of unsteady flow problem which can cause substantial
problems in pipelines is that of vibration. Severe vibration occurs when Filling Empty Lines
some periodic unsteady flow phenomenon excites some component of the
Filling empty lines, particularly in pumped systems, can give
pipeline hardware at a frequency close to its natural frequency of
velocities well in excess of expected steady state velocities. At the low head
vibration. Under these conditions, large stresses and strains (as well as
generally occurring during the filling process, the pump is operating at a
noise) can occur which may ultimately cause system failure.
point on its characteristic curve where large flow rates occur. If the line is
Because this type of analysis relates to periodic motion and implies
terminated by any device acting as a flow obstruction, e.g., a
some knowledge of the vibration of the hardware, it is generally analyzed
partially-closed valve or a closed valve with an air-vacuum valve upstream,
by techniques other than method of characteristics. Therefore, this type of
then a serious water hammer situation can develop (see Figure 8-1).
unsteady flow will not be considered in this work and the reader is referred
The air rushing out of the pipe ahead of the water will leave the pipe
to Streeter and Wylie [1] for further information.
more easily than the water behind it. As a result, when the last air leaves
and the water tries to escape, severe decreases in velocity generally occur
8.2 Carnes of Problems
causing large water hammer pressures.
Another situation is depicted in Figure 8-2 where the discharge
Valve Movement
column of a pump is initially empty and the pipeline full of water is held
by a closed check valve. High pressures will occur when the rushing water
Probably the commonest and most well-known cause of unsteady flow
fills the column and slams shut the vacuum valve.
problems is valve movement. Either opening or closing the valve causes
pressure waves to propagate through the system. The severity of the
pressure waves depends on the type of valve movement and the system
hydraulics and elastic properties, as well as the type of valve. It was shown
earlier that under certain circumstances, the last few percent of valve
movement during the closing process had an overpowering effect on the
maximum pressure generated.
Proper analysis of the impact of valve movement on the high and low
pressures in a system depends strongly on the head loss characteristics of
the valve. Charts and diagrams are available to estimate the maximum
pressures caused by valve closure, but it is far more reassuring to be able
to calculate the effects of valve closure in any specific situation. Following
the analysis shown in Article 4.4 will provide the means.
Check Valves
Figure 8-3. Propagation of negative wave resulting from pump power fidhue.
Objectionable pressures generally occur in pipelines because of
changes in velocity imposed on the system as a result of valve movement,
pump power failure, column separation, etc. Because the change in
pressure is directly proportional to the change in velocity, avoiding sudden
velocity changes generally prevents serious water hammer pressures. Most
control devices and techniques are designed to function in a particular
application to satisfy this generalization.
9.1 Valves
205
206 UNSTEADY FLOW IN PIPELINES CONTROL DEVICES AND TECHNIQUES 207
** *
1: C
2: C PROGFAM NO. 1(1
3: C SURGE VALVE IN A SIMPLE PIPE - APPROXIMATE METHOD OF CHARACTERISTICS
4: C****************************************************************************
C
6: C — PIPE EXTENDS BETWEEN TWO RESERVOIRS --
7: C — SURGE VALVE AND GATE VALVE ARE LOCATED AT DOWNSTREAM END OF PIPE —
8: C — GATE VALVE CLOSES LINEARLY WITH TIME —
9: C
10: C IOUT = INDEX WHICH GIVES AN OUTPUT EVERY IOUT-TH TIME STEP
1,000 11: C NPARTS = NUMBER OF PARTS INTO WHICH PIPE IS DIVIDED
12: C L ’ PIPE LENGTH-FT D = PIPE DIAM-IN F = DARCY-WEIS3ACH FRICTION FACTOR
--------- SURGE VALV E OPENS 100% 13: C QZERO = DISCHARGE-GPM A = WAVE SPEED-FPS HZERO = RESERVOIR ELEV-FT
E OPENS 50% 14 : C ELEVUP = ELEVATION OF UPSTREAM END OF PIPE-FT
15: C ELEVDN = ELEVATION OF DOWNSTREAM END OF PIPE-FT
--------- SU DDEN VAI VE CLOSl IRE - NO URGE 16: C TMAX - MAXIMUM REAL TIME OF SIMULATION-SEC
900 VAL\ E 17: C TCLOSE = TIME REQUIRED FOR VALVF. CLOSURE-SEC
18: C POPEN = PRESSURE AT WHICH SURGE VALVE OPENS - PSI
—<— -- ' 19: C DSURGE = DIAMETER OF SURGE VALVE - IN
20 : C KLSUR = WIDE OPEN SURGE VALVE LOSS COEFFICIENT
21: C OPENI = PERCENT OF WIDE OPEN WHICH SURGE VALVE INITIALLY OPENS
22: C TOPEN « TIME SURGE VALVE IS OPEN - SEC CLOSES LINEARLY WITH TIME
23 :
800 24:
C
DIMENSION X(20),V(20),H(20),BLOW(20),HHIGH(20),HEAD(20),
25: $VNEW(20),KNEW(20),PIPEZ(20)
26 : DIMENSION KI (ID ,PCT(11) ,KLISUR(11)
27: REAL L,KI,KLl,KLSUR,KLISUR,KLIS
28 : LOGICAL SURGF,CLOSE
29;
700 ,'z 1
30:
DATA (KI(I), 1 = 1, 11)/0 .,.0167,.0313,.0556,.100,.1787,.333,.625,
$1.25,2.50,5.27/
1 31:
z 1 DATA (PCT(I),1=1,11)/0.,10.,20.,30,,40,,50.,60.,70.,80.,90.,100./
,* 1 32 : DATA (KLISUR(I),1-1,11)/0.,.0004,.0016,.081,.194,.342,.504,
1
33 : $.664 , .819 , .931 ,1.00/
1 34 :
1 NAMELIST /SPECS/ IOUT,NPARTS,L,D,F,A,0ZERO,HZERO,ELEVUP,ELEVDN,
1 35: $TMAX,TCLOSE,POPEN,DSURGE,KLSUR,OPENI,TOPEN
600 36: C
ELEVATION , feet
37 : READ(5,SPECS)
/ t 38 : C
• 1
39 : PI=3.141592
1 t i
/ ! 40 : WTT=L/A
1 / : 41: DELL=L/NPARTS
500 1 t
42:
J T=0 .
n 1 43: SURGE=.FALSE.
400 STEADY
1
/ALVE ] XX
V I
1
1
1
1
45:
46 :
47 :
CLOSE=.FALSE.
DELT=DELL/A
C=32.2/A
INDEX=TMAX/D£LT + 1
STATE I IE AO, AT ' 1 48:
DELEL=(ELEVDN-ELEVUP)/NPARTS
v/ \ 49:
SINE=DELEL/DELL
1 50:
1 51 :
NODES=NPARTS+1
1 52 :
VZERO=144 « * 4 .*QZERO/(449.*PI*D*D)
1
\■
t ’ H0PEN = 2.31 *POPEN+ELEVDN
*» 1 53:
AS=»25*PI*DSURGE**2/144.
300 54 :
AA=.25 * PI *D** 2/144.
55:
—\ 56 :
DO 1 1=1,11
1 KLISUR(I)=KLISUR(I)*(1./KLSUR)
57 :
\t 58:
C
C ** WRITE OUT INPUT DATA **
\i 59 :
200 C
60:
WRITE(6,200)
V1 61:
WRITE(6,201) IOUT,NPARTS,L,A,D,F,OZERO,VZERO,HZ ERO,ELEVUP,ELEVDN,
62:
SURG i VALVE SURGE VALVE 63 : $WTT,TCLOSE,TMAX,DELT
C PEN CLC SED 64 : WRITE(6,211) DSURGE,KLSUR,POPEN,OPEN I,TOPEN
65: AK=12•*F*DELT/(2.*D)
100 66 : DELHF = 12.*F* DELL*VZERO** 2/(64.4 *D)
67: C
68 : CC** SET UP CONDITIONS FOR T = 0. **
69 : DO 300 1=1,NODES
70 : V(I)=VZERO
71: H ( I)=HZERO-(I-1)*DELHF
72: HLOW(I)=H(I)
73: HHIGH(I)=H(I)
74: X (I) = (I-1)*DELL/L
O
C
10 12 75:
76:
PIPEZ(I)=ELEVUP+(I-1)*DELEL
HEAD(I)-H(I)-PIPEZ(I)
14 77: 300 CONTINUE
78 • HDOWN=H(NODES)-VZERO**2/(64.4*KI(11)
TIME-seconds
Figure 9-1. Pressure head at the gate valve showing the effects of surge Figure 9-2. Computer program for a surge valve at the downstream end
valves. of a simple pipe with a line valve closing.
210 UNSTEADY FLOW IN PIPELINES CONTROL DEVICES AND TECHNIQUES 211
79: C DO 38 111-1,20
80: C ** WRITE OUT STEADY STATE CONDITIONS “ 159: FUNCT-CC-C*HTRY-A1*SQRT(HTRY-HDOWN)-A2*SQRT(HTRY-ELEVDN)
81: C 160: PRIME’-C-.5*A1/SQRT(HTRY-HDOWN)-.5*A2/SQRT(HTRY-ELEVDN)
82: WRITE(6,202) 161: HNEXT’HTRY-FUNCT/PRIME
83 : WRITE(6Z204) T,(X(I), HEAD(I),H(I),V(I),1-1,NODES) 162: IF(ABS(HTRY-HNEXT).LT.1.0) GO TO 388
84 : C 163:
HTRY-HNEXT
85: C ------ ----- --------- ------- ------ --------- ------ --------- ------ ----------- ----------------------- -------- ------- -------- ------- 164:
C BEGIN TRANSIENT ANALYSIS 38 CONTINUE
86: 165:
87: C ----------- -------- ------ ------ ----------- ------- ------ _ __ ___ ___ __ _____ „ ______ ____ _______________ _______ ____ 166: WRITE(6,208)
C 167: STOP
88:
DO 99 11 = 1,INDEX 168 : 388 HNEW(NODES)-HNEXT
89: T-T+DELT 169: VNEW(NODES)»CC-C*HNEW(NODES)
90: C 170: VV-SQRT(64.4*KLI*(HNEW(NODES)-HDOWN))
91: C “ COMPUTE H AND V AT INTERIOR NODES “ 171: QSURGE- (VNEW (NODES) -W) *AA
92: C 172: 39 CONTINUE
93 : DO 20 I«2,NPARTS 173: C
94 : VNEW(I)-0.5*(V(1-1)+V(I+1)+C*(H(1-1)-H(I+1))-AK*<V(1-1)*ABS(V(1-1) -C ** COMPUTE MAXIMUM AND MINIMUM H-VALUES **
$)+VQ+1)*ABS(V(1+1)))) 17 4:
95: 175:
20 HNEW(I)-0.5*(H(I-l)+H(1+1)+(V(I-l)-V(I+1))/C-AK*(V(I-l)*ABS(V(I-l) C
96: 176:
97: $)-V(1+1}*ABS(V(1+1)))/C) DO 50 1-1,NODES
C 177: IF(HNEW(I).LT.HLOW(I)) HLOW(I)-HNEW(I)
98: IF(HNEW(I).GT.HHIGH(I) ) HHIGH(I)’HNEW(I)
C “ COMPUTE H AND V AT UPSTREAM END ** 178:
99: C — THIS BOUNDARY CONDITION IS FOR A CONSTANT-HEAD RESERVOIR — 179: 50 HEAD(I)-HNEW(I)-PIPEZ(I)
100 : C 180: C
101: HNEW(l)»HZERO 181: C ** WRITE OUT H, V, AND HEAD VALUES AND TEST FOR THAX **
102: VNEW(1)-V(2)+C*(HNEW(1)-H(2))-AK*V(2)*ABS(V(2)) 182: C
103 : C 183:
104: C ** COMPUTE H AND V AT DOWNSTREAM END “ IF(MOD(II,IOUT),NE.0) GO TO 60
C -THIS BOUNDARY CONDITION FOR GATE VALVE WITH SURGE VALVE JUST UPSTREAM- 184 :
105: WRITE(6,204) T, (X(I) ,HEAD(I) ,HNEW(I) ,VNEW(I) ,I-1,NODES)
C 185:
106 : 186: IF(CLOSE) GO TO 59
QSS’QSURGE*449.
107: CC-C*H(NPARTS)-AK*V(NPARTS)*ABS(V(NPARTS))+V(NPARTS)
IF(T.GT.TCLO$E) GO TO 31 187: IF(SURGE) WRITE(6,209) QSS,OPENS,KLIS
108: 188: 59 I F (T. LT .TCLOSE) WRITE(6,210) W,OPEN,KLI
C ** COMPUTE LOSS COEFFICIENT FOR GATE VALVE **
109: 189: 60 IF(T.GT.TMAX) GO TO 400
OPEN’100.*(TCLOSE-T)/TCLOSE
110: 190: C
DO 32 1=1,11
111: ITEST-(OPEN-PCT(I))*.10 191: C ** PREPARE FOR NEXT TIME STEP COMPUTATION **
112: IF(ITEST»EQ.0) GO TO 33 192: C
113: 32 CONTINUE 193: DO 40 1-1,NODES
114 : 33 FACT-(OPEN-PCT(I))*.10 194: V(I)-VNEW(I)
115: KLI-KI (I)+FACT* (KI (I + D-KI (I) ) 195: 40 H(I)-HNEW(I)
116: 31 CONTINUE 196: 99 CONTINUE
117: IF(.NOT.SURGE.OR.CLOSE) GO TO 310 197: C
118: C ** CHECK TO SEE IF SURGE VALVE CLOSED AND COMPUTE LOSS COEFFICIENT IF NOT ** 198: C ** WRITE OUT MAXIMUM AND MINIMUM HEADS AND H-VALUES **
119 : 35 IF(T.GE.(TSTART+TOPEN)) CLOSE’.TRUE. 199: C
120: 200 :
IF(CLOSE) GO TO 310 400 CONTINUE
121: 201:
OPENS’OPENI*(TSTART+TOPEN-T)/TOPEN WRITE(6,205)
122: 202: DO 401 1-1,NODES
123: DO 320 1-1,10
203: HEADMX-HHIGH(I)-PIPEZ(I)
124: ITEST’(OPENS-PCT(I))*.10
204 : HEADMN-HLOW(I)-PIPEZ(I)
125: IF(ITEST.EQ.0) GO TO 330
205: 401 WRITE (6,206) X (I) , HEADMX , HEADMN , IIHIGH (I) , BLOW (I)
126: 320 CONTINUE
206:
127: 330 FACT-(OPENS-PCT(I))*.10 C
207 :
128: KLIS’KLISUR(I)+FACT*(KLISURl1+1)-KLISUR(I)) c ••••••• FORMAT STATEMENTS *******
208: C
129: IF(T.LT.TCLOSE) GO TO 37 209: 200 FORMAT (1111 ,19X , 33 ('*')/20X * WATER HAMMER IN A SIMPLE PIPE •’/
130: IF(T.GE.TCLOSE) GO TO 312 210: S20X,'* WITH A SURGE VALVE *'/
131: 310 IF(T.GE«TCLOSE) GO TO 34 211: $20X , 33 ( ' * ’))
132: C ** SOLVE FOR HNEW AND VNEW WITH GATE VALVE OPEN AND SURGE VALVE CLOSED ** 212: 201 FORMAT(///29X,'INPUT DATA’/29X,10('-')//?.8X, *IOUT =',I3/26X,’NPART
133 : 311 C5 = 64.4 *KLI/C 213: $S -',I3//31X,‘L =',F8.1,' FT'/31X,'A =’,F7.1,' FPS'/
134 : C6-64.4*KLI* (CC/C--HDOWN)
214: $31X,'D =’,F7,2,’ IN'/31X,‘F =',F6.4//27X,'QZERO =',F7.1,' GPM'/
135: VNEW(NODES)’.5*C5*(-1.+SQRT(1.+ 4.*C6/C5* * 2))
215: $27X,'VZERO »■,F7,2,' FPS'/27X,'HZERO -', F7 .1 ,' FT'/
136: HNEW(NODES)’(CC-VNEW(NODES))/C
216: S26X,'ELEVUP ,F7.1,' FT'/26X,'ELEVDN »' ,F7.1 , ' FT'//
137: 36 IF(HNEW(NODES),GT.HOPEN) SURGE’.TRUE.
IF(SURGE) TSTART’T 217 : S29X,'L/A =',F7.3,' SEC'//26X,'TCLOSE -',F7.2,' SEC'/
138: 218: S28X,'TMAX - ’ , F7.2 , ' SEC/28X, 'DELT - ' , F7.3 , ' 'SEC' )
139: IF(SORGE) GO TO 35
W’VNEW(NODES) 219: 202 FORMAT(1H1/' PRESSURE HEADS, H-VALUES AND VELOCITIES AS FUNCTIONS
140: 220 :
141:
GO TO 39 $ OF TIME'/2X,60(’-’))
34 VNEW(NODES)’0 . 221: 204 FORMAT(//18X,3(5X,* X HEAD-FT H-FT V-FPS’)/' TIME =',F7.3,
142: HNEW(NODES)“CC/C 222 : S' SEC’,3(5X, ’--------------------------------------------------- ---------------- d/dBX^tSX.FSd.ZF?.?,
143: IF(CLOSE) GO TO 39 223 : SF8.2)))
144 : GO TO 36 224 : 205 FORMAT(1H1/18X,27(’*')/18X ,'* TABLE OF EXTREME VALUES *’/18X,27('*
145: C ** SOLVE FOR VNEW AND HNEW WHEN SURGE VALVE OPEN AND GATE VALVE CLOSED ** 2 2 5‘: $’)//13X,’X MAX HEAD MIN HEAD MAX H MIN H’/11X,5('- 1 ),2X , 8 (
146: 312 C5«64•4 *AS *AS/(AA*AA*C*(1,+1«/KLIS)) 226: $'-'),2X,8('-'),2X,6l‘-'),2X,6('-')}
147 : C6»C5*C*(CC/C-ELEVDN) 227: 206 FORMAT(11X,F5.3,2X,F7.0,3X,F7.R,3X,F6.0,2X,F6.01
148: VNEW(NODES) = .5*C5*(-1.+SQRT(1.+4.*C6/C5** 2)) 228: 208 FORMAT(///l0X,'NEWTON-RAPHSON FOR SOLUTION FOR SURGE AND LINE VALV
i 4 9 s HNEW(NODES)’(CC-VNEW(NODES))/C 229 : SES OPEN HAS FAILED')
150 : QSURGE’VNEW(NODES)*AA 230 : 209 FORMAT(6X,'SURGE VALVE DISCHARGE ,F7.1,' GPM *,8X,'SURGE VALVE IS
151: GO TO 39
231: $',F7.2,’ PERCENT OPEN WITH KLIS »',F9.6)
152: C ** SOLVE FOR VNEW AND HNEW IF BOTH SURGE VALVE AND GATE VALVE ARE OPEN **
232 : 210 FORMAT(6XLINE VALVE VELOCITY - ' ,F6.2,' FPS WHEN VALVE IS’,
153: 37 HTRY’H(NODES)
233: SF6.2,' PERCENT OPEN AND KLI ,F9.6)
154: Al’SQRT(64•4 *KLI)
234 : 211 FORMAT(//16X,1 SURGE VALVE DIAM =',F7.3,' IN'/I2X,'WIDE-OPEN LOSS C
155: A2-SQRT(64•4 *AS*AS/(AA*AA*(1.+ 1,/KLIS)))
235: SOEFF -',F6.2/8X,1 SURGE VALVE OPENS WHEN P ’',F5.1,' PSI'/
156: 236: $5X,'SURGE VALVE INITIAL OPENING -',F5.1,’ PERCENT'/
157: 237 : $2X,'TIME OF CLOSURE OF SURGE VALVE «',F5.1,’ SEC')
158: 238 : END
!
In pumped pipelines it may be necessary to provide a discharge ■
If the line is mostly filled with liquid with only relatively small pockets
of air created by a shut-down, caution must still be used in startup. The
best approach is to use the technique suggested above for filling empty Figure 9-3. Open-end surge tank In a hydroelectric installation.
lines and not resume normal operation until all the vacuum valves and air
release valves have closed.
Surge Tanks
o
C
A
Upstream C . = C,-CH (9-1)
Continuity: 2
+
o
II
2gC O
>
(9-7)
Energy: Qs = C0An V2g(Hs + AB ’Hp) ........................... ^-4) c2 AB + c4
where An is the cross-sectional area of the pipe connecting the tank to the 2gC02
pipeline and Hg is the instantaneous value of the elevation of the water [Ci Ag - C3 AA + (C2AB + C4 AA)(HJ + ZAB)J
O'
C2AB + c4
>
surface in the surge tank above the pipe. The Hp’s are not subscripted
because the values at A and B are the same.
Note that the short-tube orifice equation is used to find an equation
relating discharge from the tank into the pipeline. The Co values range
216 UNSTEADY FLOW IN PIPELINES CONTROL DEVICES AND TECHNIQUES 217
Air Chambers
]
An open-end surge tank placed at the discharge side of a pump Air-on Level
E
7
Compressor l
chamber is sealed and a layer of compressed air overlays the liquid inside Chamber Discharge Nozzle
(see Figure 9-6).'After pump power failure, liquid is drawn from the T // / / Z7
chamber into the pipeline and the volume of air within the chamber
expands, causing the pressure to drop. The rate at which the chamber
pressure drops is dependent on the initial air volume, the thermodynamic
process which the air undergoes, and the rate at which liquid is being Pipeline
drawn out of the chamber. With the chamber supplying liquid to the pipe
at the ever-decreasing pressures, the liquid in the pipeline is slowly
decelerated without causing unnecessarily low pressures. By this means Figure 9-6. Schematic diagram of an air chamber and its appurtenances.
column separation can often be entirely avoided. A typical situation is
shown in Figure 9-7, which illustrates how gently the air chamber can
bring the system to rest. Compare this result with Figure 8-3. Air Chamber Analysis
The air chamber must be large enough to supply the needs of the
pipeline without reaching the point where air enters the line (emptying). The air chamber contribution to the upstream boundary condition
Also the initial air volume must be large enough so that the pressure drop must be incorporated with that of the pumps. We will assume there are
rate is not too high. Too small initial air volume will cause the pressure to NpU pumps in parallel and the power fails to all simultaneously.
drop much like the pumps alone would cause, thereby introducing little The equations for this situation are:
assistance in preventing column separation. In addition, to damp out high
pressure surges, the flow path back into the air chamber is made more Downstream C : .(9-9)
difficult so that large frictional losses can be generated and strong
damping achieved. Ultimately, all this must be accomplished with the Continuity: N UQ + Q C “ V p A (9-10)
smallest and least expensive configuration possible.
Figure 9-7 also shows that an air chamber is not always adequate to Energy (pump): (9-11)
completely prevent column separation. Low pressure can occur at local
a
summits in the pipeline and the air chamber often cannot help enough at Energy (chamber): Q = V2g (HT - Hp) (9-12)
these locations. In these cases a one-way surge tank is provided at the
summit to “drape” the EL-HGL above the pipe as shown in Figure 9-8. Pump: (9-13)
218 UNSTEADY FLOW IN PIPELINES CONTROL DEVICES AND TECHNIQUES 219
where H/p is the H-value in the chamber, An is the area of the discharge
nozzle, and Qc is the discharge from the chamber.
We have five equations and six unknowns, so another equation is
required. This equation is obtained from the thermodynamic process
which the air undergoes. The equation describing a polytropic process is
_E_ =
yV
where p0 and y0 are the initial absolute pressure and specific weight of the
air, p and y are these values at a later time and rj is the polytropic
exponent, generally taken as rj = 1.2. Using Hatm as the atmospheric
pressure head, this equation can be written as
-3
If the initial air volume is CT0 and the volume at a later time is CT, the
equation can be written in the form
Figure 9-7. Propagation of negative wave after pump power failure with Z
P ' ^atm 9-14)
-r
+
an air chamber at the pump.
N
atm
where the air volume Op can be calculated m the same manner used to
keep track of the level in the one-way surge tank
It now remains only to solve the six equations and six unknowns.
In the most general case we will assume there is still flow through the
pumps and there is flow from the air chamber. Knowing C-p we can
calculate H-p and reduce the equations and unknowns to five. Solving the
five equations for Qc gives the following expression for Qc
.............................. (9-16)
where
NNst
H (CjC.A+NN C,) 4 C6
(9-25)
z
pu
NNstAC2C7
z
pu where
(9-18) 2g Ci2 A (9-26)
AC.
After calculating Qc, Q is calculated from the following equation
and
Q= .................. (9-19)
2gC0‘ A C6 = 2gC2A
z
If Q> 0, then the solution is acceptable and the remaining unknowns may The value of Cj depends somewhat on the nozzle configuration, but in
be calculated. general discharge coefficients for re-entrant type short tubes are used.
If, however, Q < 0, then the check valves to the pumps must be closed Therefore, recommended values of Cj range from 0.80 to 0.50.
and Q = 0. In this case, Qc must be redetermined using the following
Sizing an Air Chamber
equation
Pressure variation with time at the head of the pipeline where the air
chamber is situated depends primarily on the initial air volume in the
chamber when power failure occurs. The smaller the initial air volume,
where the more rapidly the pressure drops.
The first step in the sizing procedure is to try successive values of
2gC02A 2
initial air volume until the computer analysis shows that the minimum
n ................................................... (9-21)
pressures occurring along the pipeline are within acceptable limits. This
C2A initial air volume establishes the upper emergency level. If power fails
when the air volume is smaller than this, pressures lower than desirable
and will occur in the system.
Because pressures fluctuate during normal operation there must be
= 2g Co2 A.2 (9-22) some space reserved in the air chamber for this fluctuation. Evans and
Crawford [26] recommend allowing 25 percent of the initial air volume at
the upper emergency level.
Once the pump check valves are closed, they are never reopened. Now because it is possible for power to fail when the initial air volume
Flow out of the air chamber will continue until flow reversal occurs. With is at this greater volume, we must be sure the tank is of adequate size to
flow into the air chamber a different set of equations must be solved.
prevent its emptying under downsurge. To accomplish this, we make one
For flow into the air chamber, the pumps are no longer involved and
more analysis using this larger initial air volume (lower emergency level).
the following equations are used in conjunction with Equation 9-9.
The maximum air volume during this analysis establishes the minimum
total volume of the air chamber. It is also recommended an additional 10
Continuity: VpA = Qc ..............................................................(9-23) percent or so of volume be added to act as a safety factor and perhaps
prevent vortexing and air entrainment at the connection.
Energy (chamber): Qc = - CjAn \/2g(Hp - HT) .........................(9-24)
The configuration shown in Figure 9-6 is the most common type of air
chamber design for small chambers. When requirements dictate large
where Cj is the discharge coefficient for flow into the chamber.
chambers, e.g., 6 feet in diameter and 40 feet long, the chambers are
Solving these equations simultaneously for Qc gives the following
equation. generally placed horizontally.
222 UNSTEADY FLOW IN PIPELINES CONTROL DEVICES AND TECHNIQUES 223
Figure 9-10. Computer program for analyzing the effects of air Figure 9-10. Continued.
chambers and one-way surge tanks.
226 UNSTEADY FLOW IN PIPELINES CONTROL DEVICES AND TECHNIQUES 227
166: DO 13 1=1,NPIPES
249 C
167: IF(I.EQ.KMIN) CO T9 14
252 DO 99 11 = 1 ,INDEX
168: ANPAR=L(I)/(DELT* (VZF.ROf 1) +A {I) ) )
251 T=T+EELT
169: NPAR ( I) =ANPAR
252 C
170: TEST=NPAR(I)
253 C ** COMPUTE H AMO V AT INTERIOR NODES “
171: IF (ARS(TFST+l-ANPAR).LT,.0001) MPAP(I)=NPAR(I)+1
254 C
172: GO TO 13
173: 14 NPAR(I|=NPARTS 255 DO 33 1=1,NPIPES
174 : 13 CONTINUE 256 K-NFAR(I)
175: C 257 DO 30 J - 2,K
176: C CONVERT TO O/N, II/NSD, T/NSQ 258 PA=PATIO(I)*A(I)
259 VMINUS=V(I,J)-RA*(V(I,J)-V(I,J-1))
177: COEF = 60.* 550./(2.*P11
170: DO 1 1=1,6 262 VPLUS’V (I, J) -PA* (V{ I , J)-V(I ,J + 1) )
179: QN{I)=ON(I)/(449.*PPM) 261 HMINUS = H (I ,J)-RA* (11 (I ,J) -H (I, T- I) )
2 62 HPLUS’H ( I , J )-RA* (II ( I , J )-H (I, J + l) )
180 : TN3Q(i)=NSTAGE*COEF*TNSD(I)/PPM**3
263 VNEW(I,I).5*(VMINUS + VPLUS + C(I) *(HMINIJS-PPLUS)+0(I)*DELT*SIME(I)
181 : I !INSQ(I I - UN SO{I)/RDM**2
182: C 264 $* (VM INUS-VP LUS) -AK(I)* (VMlftUS*ARS (VM INUS) + VPLUS * ADS (VP LUS) ) )
183: C 265 HNEW (I, J) = 0.5* ( HMIMUS + HPLUS + (VM I NUS-VP LUS ) /C (I) +DELT*SINE (I) *
184: C * SET UP CONDITIONS FOR STEADY STATE (T=8) ** 2 6 6 $ (VMINUS + VPLUS) -(AK(I)/C (I))* (VM I NUS* A8S (VM I'JUS) -VPLUS*ABS(VPLOS) ) )
165: C 267 30 CONTINUE
186 : U (1,1)»HPU4P 263 C
187: V(1 ,1)=VZERO(1) 269 C ** COMPUTE H AND V AT INTERIOR JUNCTIONS **
18B : DO 20 7=1,NPIPES 270 C
189 : DELL=L'I)/NPAR(I) 271 KK=NPIPES-1
190 : DELUF«1?..*F(I) *DELL*VZERO(I) * * 2/( 64.4 * D (I) ) 272 DO 4 0 I = 1, K K
191: AK: I) = ’2.*F(I)* DELT/(2.*D(I)) 273 K=NPAR(I)+1
192: K=HPAR<I)+1 274 VMINUS = V(1,K)- RAT IO(1)*A(I)*(V(I,K)-V(I,K-1))
193: X(I,1)=0. 275 liMINUS=H (I, K) - RATIO (I) * A (I) * (11 (I, K) -H ( I , K-l) )
194 : HEAD(1,1)= H(1,1)-P I PEZ (I) 276 C3=VVINUS+C (I) * IIHINUS + C ( I) *VMINUS*DE LT*SINE (I) - AK (I) ‘VMINUS*
195: RATIO (I) = DF.LT/DELL 277 SABS(VMINUS)
196: DO 21 .7 = 2, K 278 VPLUS=V(1+1,1)-RATIO(1 + 1)*A(1 + 1)*(V(1+1,1)-V(1 + 1 , 2) )
197: X(I,J)=(J-1)*DELL/L(I) 279 HPLUS=H(1+1 ,1) -RATIO(1+1) *A (1 + 1) * (I! ( 1+1, 1) -H (1 + 1 , 2) )
198: 1) (I, J) =11 (1,1) -DELilF* (J-l) 280 OVPLUS-C ( 1 + 1) *(IPLI'S-C (1 + 1) * DELT* VPLUS*S INK (1 + 1) -AK (1 + 1) * VP LUS*
199: 1IEAD(I,J)=H(I,J) - (PIPEZ (I) +X(I,J) *L(I) ‘SINE (I) ) 281 SABS(VPLUS)
206 : 21 V(I,J)=VZERO(I) 2 8 2 I F (NSURGE.EQ.0) GO TO 44
201 : IF I I.EQ.NPIPES) GO TO 23 233 C
20 2 : H(1+1,1)= H(I,K) 284 C — ONE-WAY SURGE TANKS —
20 3 : V(1+1,1)-VZERO(1+1) 285 C
204 : 23 CONTINUE 2B6 DO 41 J = 1 , NSURGE
20 5 : DO 23 1=1,NPIPES 237 IF (IAAS( ISURGE(J)) .EQ.1) GO TO 45
206: K=NPAR(I)+1 288 41 CONTINUE
207 : DO 23 .7 = 1,K 289 GO TO 44
2P8 : BLOW(I,J)=H(I,J) 290 45 I F (I SURGE(J) ,LT.3) GO TO 46
209 : 23 iill IGH ( I ,J)=H (I , J) 291 ii PP= (-C1 * A°EA (1 +1 ) +C3 * AREA ( I) ) / (C (I+ 1) *AREA (1 +1) +C (I) * AREA (I) )
210 : PHNAX=-100. 292 IF (HPP.LT.(PIPEZ(1 + 1)+HS(J))) GO TO 46
211 : PHMIN=100000. 293 GO TO 44
212: DO 11 1=1,NPIPES 294 46 ISURGE(J)=-IABS(ISURGE(J))
213: K=NPAR(I)+] 295 C5 = 64 .4*CS (J) *CS (J) *AC (J) *AC (J) / (C (I + D ‘AREA ( 1 + 1) +C (I) ‘AREA (I) )
214: DO 11 J=1,K 296 C6 = C5*(Cl *AREA (1 +1)-C3*AREA( I) + (C(I+ 1)*AREA(I +1)+C (I)*AREA(I))*
215: IF(HEAD(I,J) .LT.PBMAX) GOTO 16 297 $(HS(J)+ PIPEZ (1 + 1) ) )
216: IPMAX=I 298 IF(C6.GT.0.) GO TO 47
217: XMAX=X(I,J) 299 QS(J)=0.
21B: PHMAX=HEAD(I,J} 303 ISURGE(J)=IABS( ISURGE(J))
219: TTMAX=? 301 GO TO 44
220: GO TO 17 33'2 47 QS(J)=0.5*C5*(-1,+SQRT(1.+4.*C6/(C5*C5)))
221 : 16 IF (HEAD(I,J) .GT.PHMIN) GO TO 17 303 HNEW(1+1,1) =BS(J)+PIPEZ ( 1 + 1) -OS (J) **2/(64 .4* (CS(J) *AC (J) ) “21
222: IPHIN=I 334 HNEW(I, K)=HNEW(1+1,1)
223: XMIN=X• I ,J) US (J)«HS(J)-QS(J)*DELT/AS(J)
305
224 : PHMIN=HEAD(I,J) GO TO 48
225: TTMIN’T 306
226 : 17 CONTINUE 307 C
227 : 11 CONTINUE 308 44 HNEW(I,K)=(-Cl‘AREA(1+1)+C3*AREA(I))/(C(I+l)*AREA(I+l)+C(I)*
228: C 309 SAREA(I))
229 : C ** WRITE OUT STEADY STATE CONDITIONS ** 310 HNEW(1+1,1)=HNEW(I,K)
2 30 : C 311 40 VNF.Wf I, K)=C3-C(I)*HNEW(I,K)
231 : T = 0.0 312 VNEW (1+1,1)»Cl+C(1 + 1) *IINEW( 1 + 1 ,1)
232 : INDEX=TMAX/DELT+1 313 40 CONTINUE
2 33 : WRITE(6,204) 314 C
234 : WRITE(6,213) DELT 315 C ** COMPUTE H AND V AT DOWNSTREAM END **
235 : WRITE(6,205) T 316 C —TiilS BOUNDARY CONDITION IS FOR A CONSTANT BEAD RESERVOIR —
236 : DO 22 1 = 1 ,NPIPES 317 C
237 : K«N?AR(I)+1 310 NV-NPAR(NPIPES) +1
2 38 : WRITE(6 , 206 ) I, (X(I,J) ,HEAD(I,J) ,H(I,J) ,V(I,J) ,J=1,KY 319 VMINUS=V(NPIPES,NV)-RATIO(NPIPES)*A(NPIPES)*(V(NPIPES,NV)-
239 : 22 CONTINUE 320 $V(NPIPES,NV-1))
240 : HPUHP=H(1,1)-HSUMP 321 HMINUS=H(NPIPES,NV)- RAT IO(NPIPES)*A(NPIPES)*(H(UP IPES,NV)-
241 : WRITE(6,212) RPM,Q,HPUMP 322 $H(NPIPES,NV-I))
242: Q=Q/449. 323 CC=VMINUS+C(NPIPES)*HMINUS+C(NPIPES)*VMINUS*DELT*SINE(NPIPES)-
24 3 : HT = H (1,1) 324 SAK(NPIPES)*VMINUS*ABS(VMINUS)
24 4 : HTZERO=HT 325 KNEW(NPIPES,NV)= HRES
24 5 : C 326 VNEW(NPIPES,NV)*CC-C(NPIPES)‘KNEW(NPIPES,NV)
246 : c- 327 C
247 : C BEGIN TRANSIENT ANALYSIS 328 C ** BOUNDARY CONDITIONS AT THE PUMP **
248 : C- 329 C
330 IF(SHUT) GO TO 42
b
complexity of the column separation phenomena, computer modeling is
3
Department of Civil Engineering, Queens University, Kingston, Ontario.
not reliable at this time and experience with these devices is necessary to
select the proper size valve and adjust it. 3. Rouse, H., and S. Ince. 1963. History of Hydraulics. Dover Publications,
New York.
Closure
4. Rich, G. 1951. Hydraulic Transients. McGraw-Hill Book Co.
There are a great many techniques and devices available which can be
economically employed to prevent serious problems from occurring as a Bergeron, L. 1961. Waterhammer in Hydraulics and Wave Surges in
CJI
result of unsteady flow. It is incumbent upon the engineer, however, to Electricity, (translated under the sponsorship of the ASME), John Wiley
make the appropriate analysis and, in cooperation with the equipment & Sons, Inc., New York.
sales engineer, select the proper control system. Indiscriminate 6. Parmakian, J. 1963. Water-Hammer Analysis. Dover Publications, Inc.,
installation of the various control devices currently available from vendors New York.
can be a very costly approach to design which does not necessarily improve
the safety of the pipeline system. Moreover, this practice is not 7. Jaeger, C. 1977. Fluid Transients in Hydroelectric Engineering Practice.
engineering in the true sense of the profession. A thorough analysis by a Blackie & Sons, Ltd., Glasgow and London.
knowledgeable engineer, well-versed in transient analysis techniques, is a
desirable and economical design procedure in any pipeline system. 8. Tullis, J. P. 1970. Control of Flow in Closed Conduits. Proc, of the
Institute held at Colorado State University, Colorado State University
Press, Fort Collins.
Wylie, E. B., and V. L. Streeter. 1978. Fluid Transients. McGraw-Hill
235
236 UNSTEADY FLOW IN PIPELINES APPENDIX I 237
253 :
169 : X(I,l)-0,
254 : NV’NPAR(NPIPES)+1
170: HEAD(I,1)=H(I,1)-PIPEZ(I)
2 55: VMINUS’V (HP IPES,NV) -RATIO{NPIPES) * A(NPIPES) * (V (NPI PES , MV) -
171: RATIO{I)’DELT/DELL
172: DO 21 J»2,K 256 : $V(NPIPFS,NV-1))
173 : *DELL/L(I) 2 57: HMINUS=H(H?IPES,NV) -RATIO (N PI PES) * A (NP IPES) *(H(N?IPF.S,NV)-
H (I,J)aH (1,1)-DELHF*(J-l) 258 : 5H(HPIPES,NV-1))
174 :
175: HEAD(I,J)«H(I,J)-(PIPEZ(I)+X(I,J)*L(I)*SINE(I)) 255 :
CC=VMINUS+C(NPIPES)*HM INUS + C(NDI°FS)*VMI HUS* DELT*SINE(NPIPES)-
21 V(I,J)“VZERO (I) 2 6 0 :
176: SAK (NPIPES) *VMINUS*ABS (V.MIN'.’S)
177: IF(I.EQ.NPIPES) GO TO 20 261 :
IF (T.GE.TC2) GO TO 5*1
178: H(I+1,1)«H(I,K) 262 :
IF(T.GT.TCI) GO TO 49
179: V (1+1,1)’VZERO(1 + 1) 263 :
OPEM’l00,-(100.- PC 1)*T/TC1
180 : 20 CONTINUE 264 :
265 : GO TO 48
181: DO 23 1=1,NPIPES 49 OPEN = PC 1 * (TC2.-T) / (TC2-TC1 )
182: 266 :
K«NPAR(I)+1 48 DO 45 1=1,11
267:
183: DO 23 J = 1,K ITEST=(OPEN-PCT(I))*.1P
268 :
184 ; HLOW(I,J)«H(I,J) IF (ITEST.EQ.fl) GO TO 46
269:
185 : 23 HHIGH(I,J|»H(I,J) 45 CONTINUE
270:
186: PHMAX —103. 271 : 46 FACT=(OPEN-PCT(I))*.10
187; PHMIN’100300. KLI=KI(I)+FACT*(KI(I+1)-KI(I))
C 272 :
188: CCC=64.4*iHDOWN-CC/C(NPIPES))
C ** WRITE OUT STEADY STATE CONDITIONS “ 273 :
189; CB = 64 . 4/C(NPIPES)
C 274:
190 : IF(CCC.GT.0,) GO TO 43
WRITE(n,2B4) 275 :
191:
276: VNEW(MP IPES,NV)=0.5*KLI*CB*(-1.0+SQRT(1.0-4.*CCC/(KLI*CB*CH)) )
192: WRITE(6,2fl5) T
DO 22 I’l.-NPIPES 277: GO TO 51
193:
194 : K’NPAR(I)-1 278 : 4 3 VNEW(NPIPES,NV)=0.5*KLI*CB*{ 1.0-SQRT(1.0+4.*CCC/(KLI*CB*CB) ) )
195 : WRITE(6,2«6) I,(X(I,J),HEAD(I,J),H(I,J), V(I,J),J=1,K) 279: GO TO 51
196: 22 CONTINUE 280: 50 VNEW(NPIPES,MV)=0.
197: C 281 : 51 HNEW(NPIPES,NV) =(CC-VNEW(NPIPES,NV))/C(NPIPES)
198 : 282: C
199: C BEGIN TRANSIENT ANALYSIS 283 : C ** LOCATE MAXIMUM AND MINIMUM HEADS AND H-VALUES **
200 : C 28 4 : C
201 : c---------------------------------------------------------------------------------------------------------------------------------------------------------- 285 : DO 61 1=1,NPIPES
202: DO 99 11=1,INDEX 286 : K=NPAR(I)+1
203: T=T+DELT 287 : DO 61 J=1,K
204 : C 288 : I F(HNEW(I,J) .LT.H LOW(I , J) ) H LOW(I,J)= HNEW(I,J)
205 : C ** COMPUTE H AND V AT INTERIOR MODES ** 289 : IF(HNEW(I,J) .GT. UH IC,II (I,J) ) IHIIGU (I, J) =HNEH ( I , .7 )
206 : C 290: HEAD(I,J)=HNEW(I,J)-(PIPEZ(I)+X(I,J) *L(I) *S INE(I) )
207; DO 30 1=1,NPIPES 291: IF<BEAD(I,J).LT.PHMAX) GO TO 66
208 : K=NPAR(I) 292 : IPMAX=I
209 : DO 30 J=2,K 293 : XMAX=X(I, J)
210 : RA’RATIO(I)*A(I) 29 4 PHMAX’HEAD(I,J)
211: VMINUS’V(I,3)-PA*(V(I,J)-V(I,J-l)) :
TTHAX=T
212: VPLUS’V(I,J)-RA*(V(I,J)-V(I,J+1)) 295:
GO TO 67
296 :
213 : IIMINUS’H (I, J) - RA* (H (I ,J) - H (I, J-l) ) 66 IF(HEAD(I,J).GT.PHMIN) GO TO 67
214 : HPLUS’H (I, J) -RA* (II (I , J)-H(I, J+l) ) 297 :
IPMIN=I
215 : VNEK(I,J)=0.5*(VMINUS+VPLUS+C(I)*(HMINUq-HPLUS)+C(I)*DELT*SI NR(I) 298 :
XMIN=X(I,J)
216: 5* (VMINUS-VPLUS)-AK (I) * (VM INU.S * AF?S (VM INUS) +VPLUS* ABS (VP LUS ) ) ) 299:
PHM I N = HEAD(I,J)
217: HNEW (I, J) =2.5* (HM INUS + ll PLUS+( VM INUS-VPLU 5)/C ( I) +DELT*S I NE (I) * 300:
301: TTMIN=T
218: $ (V.MINUS + VPLUS) - ( AK ( I)/C (I) ) * (VMINUS * ABS (VM INUS) - VPLUS * A!*S (VP LUS) ) ) 67 CONTINUE
30 CONTINUE 302 :
219: I F(HEAD(I,J) •GT,-HATH) GO TO 62
C 30 3 :
220 : WRITE(6 ,214) T,I,X(I,J)
C ** COMPUTE H AND V AT INTERIOR JUNCTIONS ** 30 4 :
221: FAIL’.TRUE.
C 305 :
222: 62 CONTINUE
KK=NPIPES-1 306 :
22 3 : 61 CONTINUE
DO 40 1=1,<K 307 :
2 24 : C
K=NPAR(I)+1 308 :
2 25 : C ** WRITE OUT H,V,AND HEAD VALUES AND TEST FOR TMAX **
309:
226: VMINUS’V(I, X)-RATIO(I)* A(I)*(V(I,K)-V(I,X-1))
310 : C
227: ll-MINUS = H (I , K ) -RATIO (I) * A (I) * ( H (I, K) -H ( I, K-1) )
311 : 70 IF(FAIL) GO TO 71
229 : CCC’VMINUS+C(I)*HMINUS+C(I)*VMINUS* DE LT*SINE(I)-AK(I)*VMINUS*
312 : I F (MOD(11,IOUT) ,NE . 0) CO TO 72
2 29: SA3S(VMINUS
313 : 71 WRITE(6,205) T
230: VPLUS’V(I+_,1)-PATIO(I+1)*A(I+1)*(V(I+1,1)-V(1+1,2))
HPLUS’H(I+_,1)-RATIO(I+1)*A(I+1)*(H(I+1,1)-H(1+1,2)) 314 : DO 73 1=1,NPIPES
231:
CC = VPLUS-C : I + 1)*HPLUS-C(1 + 1)*DELT*VPLUS*SIME(I + 1)-AK(1+1>*VPLUS* 315 : K=NPAR(I)+1
2 32 :
233 : SABS(VPLUS) 316 : WRITE (6,20 6) I, (X(I,J) , HEAD (I, J) ,HNEW(J,K) ,VNEW(J,K),J’l,K)
2 34 : HNEN(I,K) = ’-CC*AREA (1+1)+CCC*A REA (I) )/(C <I+L)* ARFA(1+1)+C(I)* 317: 73 CONTINUE
235: SAREA(I)) 318: IF(FAIL) GO TO 400
2 36: HNEW(1 + 1,1J =HNEW(I,K) 319: 72 CONTINUE
2 37: VNEW(I,K) =CCC-C(I) *IINEW(I,K) 3 20 : C
VNEW(1+1,1)=CC+C(1+1) *'!NEW( 1+1,1) 321: C ** PREPARE FOR NEXT TIME STEP COMPUTATION **
2 3 8 :
40 CONTINUE 322: C
239:
C 323 : DO 80 1=1,NPIPES
240 :
C ** COMPUTE H AND V AT UPSTREAM END ** 324 : K=NPAR(I)+1
241: C —THIS BOUNDARY CONDITION IS FOR A CONSTANT HEAD RFSFRVO'R _______________ 325 : DO 80 J=1 ,K
2 4 2 : C
2 4 3 : 326 : V ( I, J)’VNEW(I,J)
KNEW (1 , =HZERO 327 :
24 4 : VPLUS’V'1,J)-RATIO(1)*A(1)*(V(1,1)-V(l,2) > 80 H(I,J)=HNEW(I,J)
328: IP(T.GT.TMAX) GO TO 400
2 4 5 : HPLUS’H [1 , ] )-RATIO (1) *A (1) * (11 (1 , 1 )-H(l , 2) ) 329 : 99 CONTINUE
2 4 6 : VNEW(1,1)=V’LUS+C(1)*(HNEW(1,1)-HPLUS)-C(l)*DELT*VPLUS*S~NE(ll 330: C
2 4 7 : S-AK(l)*VPLLS*APS(VPLUS) 331:
2 4 8 : C C -----------------------------------------------------------------------------------------------------------------------------------------------
332:
2 49 : C ** COMPUTE H AND V AT DOWNSTREAM END ** C END OF TRANSIENT ANALYSIS
250 : 3 3 3 :
C ----------- ---- --------------------------------------------------------------------------------------------------------------------- --------
251: C —THIS BOUNDARY CONDITION IS FOR A GATE VALVE CLOSING AT TWO 33 4 :
c
252 : C SEPARATE FATES AND FLOWING INTO A RESERVOIR— 335 :
c
336:
C ** WRITE OUT MAXIMUM AND MINIMUM HEADS AND H-VALUES **
C
240 UNSTEADY FLOW IN PIPELINES
241
242 UNSTEADY FLOW IN PIPELINES
J 13.3 24 6J ’
I1/. 30 20 3.6
4
6
_
484
200
541
237
156
120
140
83
4.9
5.8
4.1
4.2 APPENDIX II 243
FLOW CHARTS
The data presented in the charts or these pages have been The flow rate ranges shown in the charts are approximate since-the
carefully authenticated. It should be clearly remembered that the proper range of flows for any particular type of valve may vary
figures given are for valves.in the fully open position, in some considerably. Interpolation to find pressure loss at other flow
types of valves such as check val-es, pressute regulators, and rates is permissible except at the low end of the curve. Values given
certain throttling valves the fully open condition is rarely reached. are based on fresh water at 6Q F. Convenient factors for head loss
PRESSURE W0P-
For these valves the figures m the charts should be taken comparison, and for calculating and plotting total system head
only as a guide to proper size selection. curves are given in the following table,
P.S.I.
EQUIV. LGTH.
Cv FACTOR OF PIPE K FACTOR
GLOBE ANGLE GLOBE ANGLE GLOBE ANGLE
VALVE
1‘4
8 Vi ?44 326 ~99029 260
387 1*61
39~?5 ~ 6 37
76 3.9
58 7.1
SIZE
10
2 1126
488-5 _ 1575
60~ ’~n375
~
65 ~197
<0 ~~ ' 6.77.1
5.7 4?S
. I? 1721_____
2500*
80 ____42592______
194-
59 ____6.345 _____25,y2 3.9
J* _____ 2112
3 ______ 3060*
i?5_______475_ 147 220’75____
____6,251____ 2.8- 5.1
_J_6 3.7 2837 ___42Q0* __525 22f 5.7 2.6*
24 6884 - 800 '49
ESTIMATEO ' ------------------
FLO* RATE-GALLONS PER MINUTE
(WATER)
ANGIE PATTERN
(1) Cv factor is defined as the number of gallons per minute of water at 60°F. which
will flow a: a one pound per square meh differential.
(2) Equivalent lengths of pipe are based on data contained in Hydraulics Institute Pipe
Friction Manual. 1954 Edition, Tables 4 through 20. In general, this data is for new,
clean wrought iron or steel, schedule 40 pipe with no allowance made for age, differ-
ences m diameter, or any abnormal condition of the interior surface. For further details,
refer to the above manual.
KVJ
(3) The value for K is calculated from the formula h = —
. 2g
where h - -rictional resistance in feet of fluid
V-jverage velocity in feet/second in a pipe of corresponding diameter
g = 32.17 (eet/second/second
K= Resistance coefficient for valve.
(4) FormJas for computing Cv Factor, Flow (Q) and Pressure Drop (&P):
.......................................................
CV 8" = 740
CV
CV 4" = 205
CV
GLpBE
OPEN
3”=
100
-
40
60
80
20
VALVE
o
128
423
40
60
GPM = CV /AP
001 08
VEE
I 2 3 d 5 6 7 69 10 2 3 456789
100
8
7
€
T7
8
7
6 "I
LOSS OF HEAD IN FEET
... i..
7
6
3
A
A
APPENDIX II 245
~
\ 100 3 4 5 6 8 1000 2 3 4 5 6 8 10000 2 34568
>1
Index
2 , 3 .45,t 7 «>
to to
to to
to to to
appurtenances
OO
r- o
history Characteristic lines
sizing 221 grid 52
oslOOO i 3
branching pipes 97
to
interior major losses 101
Booster pumps 167 interior valves 103
Boundary conditions 51 interpolation 85
Courtesy of Golden-Anderson Valve Specialty Co.
•—‘■*□00
four-pipe junctions 100
OMw
Hardy Cross analysis 1
three-pipe junctions 97,140
linear theory analysis 123
nxrr
248 UNSTEADY FLOW IN PIPELINES INDEX 24!
Ln O'.
series pipe with steady flow friction factor 5 computerizing solution
calculation built in 236-239 formula 5 stability
source pump power failure 174 interior major losses 101 Length increment As
minor losses 25
o
surge relief valve program 209 computing As 52,70
Computer output wall shear 43 effect on solution accuracy 75
(see computer programs) interpolation 68 One-way surge tanks 21
Concrete pipe Linear theory boundary conditions 21
air
(seechambers
pipe) 216 computer program 123
L,
C
Control
fillingdevices and techniques 207
empty lines example 122
one-way surge tanks 214 Gate valves theory 120
removing air from lines 212 in pipe networks 140 Loss coefficients Parallel pipes 2
surge anticipation valves 231 loss coefficients 107 interior of lines 102 Pipe elasticity 3
surge relief valves 206 Grid of characteristics valves 103,106 Pipe junctions 97,100,140,15
valve movement control 205 (see characteristic lines) Pipe networks
M Hardy Cross analysis
1
Mass, conservation of 30,44,46 1
Elastic theory Hardy Cross analysis 116 Method of characteristics linear theory analysis 12
definition 1 computer program 119 approximate Newton-Raphson analysis 12
Ox -Q
complete Pipe restraint 1132-
OX
moduli example 117 steady state analysis
(see moduli of elasticity) procedure 118 Minor losses Pipe stress
water hammer in networks 13 35,33
Elasticity Head increase in unsteady flow 25,102 relation to pressure
effects 30 (see pressure head increase) Modulus of elasticity Pipes
liquid 30,45 in pipe deformations 31,45 concrete pipes 4
Pipe 31 I pipe materials 35 elastic stretching J
Equivalent pipes water 30,34 moduli of elasticity
minor losses 25 Im p u Ise-m ome n tu m 28 Momentum 28,43,46 Poisson’s ratios 2
parallel pipes 21 Initial conditions 51 restraint in thick-walled
z
series pipes 19 Interpolation pipes
Euler equation 10,43,46 complete method of Network analysis restraint under action of
Experimental verification 77 characteristics 68 Hardy Cross water hammer 32-1
O OX
computer code for linear linear theory stress: in wall 31,1
to
interpolation 11 Newton-Raphson 12 volume changes 31,36/
L
computer code for parabolic 0 Network water hammer 2 wall shear stress
Finite difference equations interpolation examples 13 Poisson’s ratio
approximate characteristics for valve loss coefficients 11 Newton-Raphson method 3 in pipe elasticity 31,1
method 51 linear 2 example 14 values for pipe materials
complete characteristics parabolic in one dimension 2 Power failure, pump
method 67 in series pipes 10 in pipe networks (see pump rundown)
00 OO
CM
o
IZh OO
controlled movement 3
relation to pipe elasticity 32 air chamber 216 effect of L/a
Pressure reducing valves (PRV’s) one-way 214 effect of closure rate
quick response 112 open end 212 on pressure 111
slower acting 115 Rectangular grid 70 Surge theory effect of movement 200
Pressure relief valves Reflections (see wave transmission (see rigid water column theory) gate valves 107,140
(see surge valves) and reflection) Surge valves in line interior 103
Problems, causes of Reinforced concrete pipe effect on pressure head loss coefficients 103,106,107
check valves 200 (see pipe)
CM CM
OO
O kO CM
at valve loss coefficient as function
filling empty lines 201 Released air of opening 109
example
pump power failure 203 causes 152 pressure generated 16,17
effect on analysis 152 history
pump startup 202 surge anticipation 231 pressure reducing 112
removing air from 202 Reservoirs sudden closure 5,50
boundary conditions, surge relief 206
lines 200 surge anticipation 231
valve movement 164 simple pipe 54 surge relief 206
Pump rundown 167 Reverse flow Volume changes, 31,36,45
boundary conditions 217 in pipes 105 Thick-walled pipes pipes
37
with air
with backflo
chamber 192 in pumps 192 W
Time increment At
with one-way surge tanks 214 Rigid water column theory computing At 52,70,85
Pump startup 197 definition 1 Water hammer
effect on solution accuracy 74
Pump tripout 190 examples 15,17,23 effect on stability description
76
bO tn
Pumps flow establishment 13 Time of travel history
kD Ok
basic theory parallel pipes 21 in pipe networks
Tunnels, wave speed in 133
IT) kO
co
booster pressures from valve closure 17 in pumped pipelines 164,21
cavitation 163 series pipes 19 in series pipes 2
valve closure 16
kO kO
c-
Unsteady flow
characteristics at different (see pump rundown) basic equations Wave propagation
Ch
Vi
speeds compared to steady flow time of travel
166,19 definition Wave propagation speed
bJ ch Ch to o\
equation for
00
2 elastic effects 2
co
computerized constant speed equation of continuity 7 history
boundary conditions 57 Series pipes flow establishment in concrete pipes
43,4 40
computer program
OO 00
r-
co co
r--
co co co TJ-
Ch kn ch c\)
6
efficiency
conditions 55 in rigid water column theory one-dimensional equation in thin-walled pipe
55 1
internal boundary conditions
ch
3
k© k©