4

NATIONAL AERONAUTICS AND SPACE ADMINISTRATION

WAVE-PLAN ANALYSIS OF UNSTEADY FLOW

by Don J. Wood, Robert G. Dorsch,
and Gharlene Lightner
Lewis Research Center
Cleveland, Ohio
GPO PRICE

,
S

CFSTl PRICE(S) $
-

Prepared for
Journal of the Hydraulics Division
Proceedings of the American Society of Civil Engineers

, .

July 12, 965

*-

65-48

WAVE-PLAN ANALYSIS OF UNSTEADY FLOW

by Don J. Wood, Robert G, Dorsch,

m d Charlme Lightnor
Lewir Research Center
Cleveland, Ohio
1

ABSTRACT: A digital distributed parameter model for computing unsteady

flow in liquid-filled fluid systems is presented. The wave-plan method employed in the model involves essentially the synthesis of incremental p r e s s u r e
pulses. The analysis is presented i n a form general enough to be applied to a
variety of hydraulic systems. T o illustrate the application of the method to a
specific system, the response of a long, straight liquid-filled line to sinusoidal inlet flow and p r e s s u r e perturbations is computed. Both a constantcross-section line and a tapered line are analyzed in the example.
I

KEY WORDS: hydraulics, unsteady flow, distributed parameter, nonlinear,

I

digital.

WAVE-PLAN ANALYSIS O F UNSTEADY FLOW

M,, A. S.C . E. a Robert G. Dorsch2,

and Charleno Lightnsr3

by Don J. Wood, A.

SYNOPSIS
An analytical method for computing unsteady flow conditions in liquidfilled flow systems is developed. The method which is called the wave plan incorporates distributed parameter and nonlinear effects including the effects of
viscous resistance. The wave plan is essentially a solution synthesized from
the effects of incremental step p r e s s u r e pulses. The p r e s s u r e pulses a r e generated because of incremental flow-rate changes that originate in a hydraulic
system from a variety of sources, including the mechanical motion of the system structure. The p r e s s u r e pulses propagate throughout the system at eonic
velocity and a r e partially transmitted and reflected at each discontinuity. The
velocity change caused by each pressure pulse i s obtained from the Joukowski
relation. P r e s s u r e and velocity time histories at any point in the system a r e
obtained by a timewise summation of the contributions of the incremental pressure pulses passing that point.
The analysis is presented in a f o r m general enough to be applied to a
variety of hydraulic systems. To illustrate the application of the method to a
specific system, the response of a straight hydraulic line to a sinusoidal
orifice-area variation of an upstream valve is computed. Both a constantcross-section line and a tapered line a r e analyzed in the examples, and various
nonlinear effects evaluated. Comparisons a r e carried out with experimental
data obtained for the constant-diameter line and good agreement is shown to
exist.

INTRODUCTION
Hydraulic systems employed in present day applications often require
high performance with a small permissible range of variation from design
flows and p r e s s u r e s . In addition, many hydraulic systems a r e structurally
complicated. It is essential to be able t o predict the dynamic response of these
systems under transient and forced periodic conditions.
Asst.

Prof. of Civil Engrg., D u k e University, Durham, North

Carolina.

. Ohio.

Head,

Dynamics Section, NASA, Lewis Research Center, Cleveland,

3h4athematician, Instrument and Computing Division, NASA, Lewis Res e a r c h Center, Cleveland, Ohio.

The basic partial differential equations for unsteady flow a r e derived

f r o m continuity and momentum relatione. Closed-form distributed parameter
solutions for small sinusoidal flow perturbations in long lines having negligible
fluid damping are available and a r e in good agreement with experimental
t4ata4-6,
#anoia, howavap, a0 nonlinear partial dittsrentirl squrtionr ernnot be solved without resorting to tediour numerical o r graphic techniques.
An analytical study was therefore undertaken at the NASA Lewis Research Center to develop a numerical distributed-parameter solution for unsteady flow in hydraulic systems that would be in a form readily lending itself
to digital-computer computation. The analysis technique s which were developed will be referred to herein as the wave-plan, The wave-plan method has
some similarities to a method of characteristics solution due to the technique
of tracing eonic disturbances throughout the systerh; however, the wave-plan
. method i e more easily applied to complex unsteady flow systems in addition to
lending itself better to physical interpretation.
This paper presents the details of the wave-plan analyeis along with the
basic elements of the corresponding digital-computer program required t o .calculate unsteady flow in liquid systems. The analysis is presented in a f o r m
genera3 enough to be applied to a variety of fluid systems. Examples are given
to illustrate the application of the method to specific hydraulic systems.
I

._

ANALYSIS

General

Basic equations. The pressure and velocity of the liquid within a line as a function
of position and time can be obtained from the equations of momentum and continuity.
The momentum equation for a one-dimensional elastic fluid of constant mean density
is7

where f ( v )is the resistance due to fluid viscosity, which is some function of velocity.
(Symbols are defined in apendix D.)
- ---._I___
The continuity equation for an elastic liquid in an elastic line ie7
+

Equations (1) and (2) a r e nonlinear partial differential equations which,

when solved along with the boundary condition8 in a liquid-filled line, give relations for velocity and premsure in the line,
Method of rolutioa. The direct rolution Q the continuity urd mornatrtum
rqwllonr 60 obuln urloolthr and prarrurrr ir iumault w e n when the boundary
condition8 are relatively simple (open or clored endr). In addition, exact eoluc
tions have been found only for a system where the resistance t e r m ie linearilred
or neglected entirely4=6

Acceptable solutione of nonidealilred unr teady flow pr oblems including

nonlinear effects require the use of numerical methods and a digital computer,
The continuity and momentum equations have been eolved by a numerical tech-'
nique based on the method of characteristics and adapted for digital computer
use'.
Identical eolutions may be obtained more readily by a n alternate process. This method, referred to a s the wave-plan solution, is developed in this
A wave-plan solution i s obtained as follows: At the point o r points in a
liquid system where a disturbance is introduced (such as an oscillating valve
or a moving impedance change), an incremental change in liquid flow rate due
to the disturbance over a very short interval of time is computed. The incremental p r e s s u r e pulse accompanying the flow-rate change is then computed.
This p r e s s u r e pulse is propagated throughqut the system a t sonic speed. The
p r e s s u r e pulse is partially reflected and partially transmitted at all geometrical and physical discontinuities in the fluid network. P r e s s u r e and velocity
time histories a r e computed for any point i n the system by summing w i t h time
the contributions of incremental waves.
Characteristic impedance.
The relation (characteristic impedance) between p r e s s u r e and velocity changes caused by a pulse traveling in the liquidfilled line is computed from momentum considerations. Figure 1 shows pressure and flow conditions, as a pressure wave propagates in a liquid-filled line,
that exist at time t and a t time t + At. The wave takes the time At to travel
the distance Ax from A to B. During this time there is a p r e s s u r e P + A P
on the left end and a p r e s s u r e P on the right end of the liquid contained between points A and B. This unbalanced pressure causes the fluid to accelerate.
Newton's second law gives

V+AV
P+AP

I!

V+AV

P*AP

I :

(Pt A P

l+bl

- P ) A = P A AX-AV
At

Canceling and rearranging yield

A P = p A V -Ax
At

Tlw t o r i &/At ir tho propyruOn &#ad d tho prorwro w m e 'Elw wlw #modL
qual to the rani0 velocity C in the ayotem if the mean velocity ofthe liquid in the line
is neglected. Gince the mean velocity of the liquid la usually much o d e r than the
acouBtlc velocity, thie is usually permieeible. Thus,

or in terms of head of liquid

C
AHz-AV
g

The sonic speed C for a liquid flowing within a line is influenced by the elasticity of the
line wall, and for a system that is axially unrestrained7'9 can be'calculated
from

Equation (3), which is the well known Joukowski equation, states that the pressure
change at any point in a line is equal t o the product of the velocity change at the point
times the characteristic acoustic impedance pC of the liquid in the line. This relation
may also be derived by employing continuity and energy principles8 8 9 .

Generation of Pressure Waves

Pressure waves may originate in liquid flow systems in a variety of ways. Two
c o m m a sources are an orifice o r valve with a varying open area and a moving point in a
line where there is an impedance change. In order to apply the wave-plan method to
varying area change or moving impedance change, the corresponding disturbance function is approximated by a series of discrete changes over small equal time intervals.
Figure 2 shows this approximation.
Each of these changes is aeeumed to occur in&antaneouely at Borne time during tho
time interval. The pressure perturbations associated with each of these changes
4

are then computed as shown in the following examples.

varying area orifice. A pressure wave is generated at an orifice that experiences a change in opening area. The magnitude of the generated wive is a
function ofconditiwna before the area c k q p and tba
mugnwh ol tha wan ahurgr,
T b flow through an oriflee L meumd to obey the
eqwre-law relation:

Figure 2 Step wroxitWon d disturbing

fundkn.

V = B G

The orifice coefficient B i s a function of the open area, the line area, and the diecharge coefficient and is easily determined for a particular orifice. For the case of an
open area thst i a a small percentage of the line area (A,, << A), the orifice coefficient ie
given closely by

where the discharge coefficient CD is a function of Reynolds Number (and hence, velocity). For small velocity perturbations, however, the discharge coefficient varies only
slightly, and hence the orifice coefficient m a y be considered to vary linearly with area
changes. The examples in this report deal with this case.
Figure 3 shows conditions at an orifice before and after a small orifice-area change.
The flow is from left to rig&. The external pressure head m a y be varied in a prescribed manner.
The momentum equation across the wave front gives

The flow out the orifice after the oriice area change is given by
v 1 1 = 8 2 ,"1+

AH11

- H22

Solving the momentum and orifice equations simultaneously gives the following quadratic in Vll

Figure 3. E M d dep change In orftke oglflclent at tarmlnd aifla.

where
5

B
:
C
bn8

The positive root of this equation gives the desired value for outflow velocity. Substituting this value for Vll back into the momentum equation gives the magnitude of the
pressure wave generated by a sudden change in orifice coefficient from B1 to BZ.
Moving terminal orifice. The analysis of the magnitude of a pressure wave generated by the motion of a component in a fluid line is exemplified by considering the
motion of a terminal orifice. The lateral velocity of the orifice is so represented by a
series of step changes that the velocity of the orifice remaine canstant Over short time
intervals. Ffgure 4 shows conditions at a terminal orifice just before and a ebrt time
after a step change in velocity. The orifice coefficient need not be constant.
The momentum equation across the generated wave is

.
The displacement shown in figure 4 takes place in the time increment At. The continuity equation states that the net inflow across boundary M equals the net flow out the
orifice plus the storage that takes place during the time interval. Therefore
VllA At = YE$

At

+ VOA At

or

@l
v1* H1

H2

flme

fice end during the time interval, and VO is

for the resulting line velocity gives the following quadratic in VO.
V02 + bVO+ c = 0

where

and

:[

C
+ -(VI

c = -B H
1

- VE2) -

Effect of Viscous Resistance

After the generation of a pressure wave at a perturbing point in a liquid system, the
wave propagates with sonic velocity throughout the system. The viscous resistance of
the liquid medium influences the propagation of this wave.
The effect of viscous resistance on pressure pulse propagation can be neglected
without appreciable effect for short, large-diameter lines where such losses are small;
however, for longer lines or small-diameter lines carrying viscous liquid, the resistive
losses are not negligible and should be included in an unsteady flow analysis.
The exact solution of the partial differential equations (1) and (2) is limited by the
necessity of linearizing or neglecting the friction term entirely. For graphical analyses
it has been necessary to consider the friction losses as lumped at one or more points in
the pipeline if a manageable solution is to be obtained.
To include the effects of viscous resistance in a wave-plan solution, the extent that
fluid viscosity influences the propagation of a pressure pulse must be determined.
There is presently no experimental or analytical information available which is in a
form that can be used for the prediction of the effect of viscous resistance in unsteady
line flow. Because of this limitation it is necessary to make a quasi-steady approximation and to assume that viscous losses in a small line length dx are given at any instant
by the Darcy equation as
Aht =

f Axv2

The form of equation (7)is identical to that of an internal square-law orifice if the
orifice coefficient is given by
7

This equation implies that viscous losses over a small length of line can be lumped
ut en internal aquare-law orifice with a properly choeen orifice coefflctent, TMg regreUlb & PaPhfCd W Y d I hU
@@8Wlotl
k F@&Wcrdt@ U tb * * O P m I Wkb~,**
Wld
been suggested by Bergeronl'.
Line losses can be distributed at many discrete points along a line by inserting a
large number of correctly chosen square-law (friction) orifices. Impinging waves are
then reflected and transmitted at these friction orifices in a manner similar to that of
reflection and transmission through a small region of flowing viscous liquid. The equations governing wave reflection and transmission at a friction orifice are developed in
the next section.

Reflection and Transmission of Pressure Waves at System Discontinuities

It is necessary at each discontinuity to compute the magnitude of transmiesion and
reflection of each pressure pulse in terms of initial conditions at a diecontinulty, the
nature of the discontinuity, and the magnitude of the impinging waves. These derivations are all made by employing the three fundamental fluid flow relations (continuity,
momentum, and energy).
Figure 5 shows two waves impinging and reflecting simultaneously at a disconWty
in a liquid system. The notation used in this figure is employed in subsequent specific
examples. Liquid moving from left to right has a positive magnitude of veldcity.
The momentum equations across the wave front8 yield

AH1 = g
Ll
(Vr Vl)

AH2 =
G

C,

(V2 V")

Figwe 5. Nomendahre for conditions before and after wave

action it discontinuity.

(See fig. 5 for definition of Vr and V".) Equations (8)and (9) combine t o give
C.

and equatione (10) and (11) give

C,

Terminal orifice. A terminal orifice is considered to be bounded by a pressure

reservoir. The pressure in t b reservoir m a y be changing in a prescribed manner. In
addition, the orifice coefficient m a y also be changing in a prescribed manner. Figure 6
shows conditione before and after reflection from an orifice bounded at the inlet by a
pressure reeemir.
If the discharge after wave action i a in the podtive direction (from the reservoir to the line), the

Figure 6. Pressunpuise rdsdion 8l WminJ

.Solving this equation simultaneouoly with the

momentum equations (13) for the velocity gives the
following quadratic in V22:
V i 2 + bV22 t c = 0

where

and

Because the resulting direction of flow was taken as positive, the positive root of

equation (16) is the desired result.

The magnitude of the reflected pressure wave is obtained from equation (13). The
new pressure at the orifice is given by equation (15).
If the direction of flow is into the reservoir after wave action (negative direction),
the velocity head relation ie given bp

Simultaneous solution with the momentum equations give8 a quadratic in V2$

2
V22 + bV22 + c

where

The negative root of equation (17) is the desired result.

It is necessary to determine the resulting direction of flow so that the correct equation (eq. (16) or (17)) can be applied. Inspection of equation (16) discloses that the term
Hll
H2 2 AH2 + C2V2/g must be positive to yield the necessary positive root. If
this term is negative, the flow must be into the reservoir, equation (17) must be used,
and the negative root must be computed.
Friction orifice. Figure 7 shows conditions before and after pulse-wave action at
a friction orifice. Because of identical conduits on both sides of the orifice, the headvelocity relation for the orifice after wave action for flow from left to right is

V i 1 = V22 = BF fH1+ AH1 + AH11

- H2 - AH2 - AH22

Noting that C1 = C2 = C and solving this equation

simultaneously with the momentum equations (12)
and (13) give
- w 1

V'

"11. H11

'

HZZ. v22

Y'
-c

t+

where
10

The positive root of this quadratic equation gives the resulting velocity through the
friction orifice. The term I+1
2 AH1 HZ 2 AH2 + 2CVl/g muet be positive to yield
a positive root. If this term is negative, the resulting flow i s from right to left, a d the
head-velocity relation for the orifice is

- -

This equation combines with the momentum equation8 to give the following:

where the coefficients b and c are of the same magnitude and of opposite sign from
the corresponding coefficients in equation (18).
The negative root of this equation gives the resulting velocity in the negative direction. The magnitudes of the resulting pressure waves and the pressures after wave
action are given by equations (12) to (15).
Diameter discontinuity. The relation between net head and discharge for a diameter change in a conduit is derived by utilizing energy relations. Figure 8 shows the
notation for this derivation.
If A1 > AZ, the diameter discontinuity is an abrupt constriction, and the energy
equation for flow through the constriction from left to right after wave action is

I
"

The area ratio is denoted as

Then the continuity equation is

11

(W

v22 = RVll

If theee relatione are substituted into the energy equatlon ond the torme are comb&d,
the following net head-velocity relation fe obtained.

If Ai C Aa, the diameter diecontinuity is an abrupt expansion, The energy equntfon

for flow through the expansion from left to right after wave action is

This equation gives the following velocity-head relation for an -ion:

Hll

- H22 = Vil("?;

R)

The case of a lossless diameter discontindty is also important because step diameter changes can be employed to simulate a &?red line. In the continuously tapered
line the losses are small and sometimes can be neglected. The energy equation without
the loss term ia

The resulting velocity-head relation i6

;;

H22=%1(
2 R

1)

The final net head after wave action can be written as

or 12

.t'

The relation is combined with the continuiw equation and the velocity-head rehtion
to obtain second-order polpmialo for .the Una velsctty after warn &ctioa, Tho rrqulattorr
48

-aVll

.) where

+ bVll + c = 0

(W

The coefficient a depends on the type of discontinuity and whether emrgy loeses
are included. For a contraction with losses included the coefficfent is
a = 3R2

- 2R - 1

(27)

4e

For an expansion with losses included the coefficient ie

a=- R2 R
!

For either Qpe of discontinuity with no losses included the coefficient is

a=- R2- 1

For each case the resulting velocity is the positive root of equation (26). Therefore, the term

13

must be positive. Sf this term is negative, the resulting flow is from right to left, and
the equations are reformulated by putting the loss term on the correct side of the energy
a

eq-m

Junctions of three or more legs. The previously presented*mathematical analyses

pertaining to the reflection of a pressure pulee at a diameter discontinuity have taken

into robowrt energy lmrop iatrodwrd bp tha dkoonthuity, With j ~ o t i o n d

r UImo or
more legu, the incluaian of enegg loor tarmr beearner Incrersingly difficult b e c w e d
the irreversibility of frictional effects, T b , the treatment employed herein w i l l be
similar to the one preeented in etandard references on fluid dynamics and w i l l xmt be
derivedl 1.
The relatiom for computing the percentages of magnitude of a preerure wave trammftted and reflected at a junction are baaed on the following:
(1) The Joulcowski equation applied acme each wawe
(2) Continuity of flow at the junction
!
(3) Continuity of pressure at the junction
The magnitude of the wave that is transmitted to all the other legs of a junction of
n legs due to a wave of magnitude a impinging in line i is given by T(i)m,*re
the transmission coefficient T(i) is given by

The magnitude of the wave reflected in leg i is given by F?(i)AH when the reflection coefficient R(i) is given by
R(i) = T(i)

-1

Analysis of the Dynamics of a Liquid System by Wave Plan

The wave-plan analysis supposes a system composed of a discrete number of discontinuities connected by lossless line segments, which serve only to transmit pressure
pulses. The discontinuities include geometric ones, such aa terminal orifices and diameter discontinuities, and artificial ones, such as friction orifices. A typical simulated
fluid system is shown in figure'9,
14

' I
I

'Phrrrcl Ir R vurlrrhlo~rrorsrWc,at A (vrlvo) ud a fimd srlflee at b, At 8 there ir

an abrupt change in line diameter. Friction orifices are ineerted between A and C and
between C and B to eimulate viecaue reeietance.
The equations developed thus far have applied only to c a n d i t f o ~at a particular point
in a eyetern at a particular instant of time. Through the use of these equations it ia
possible to compute velocity, pressures, and magnitudes of pressure pulses leaving a
p c u l a r discontinuity in terms of the magnitudes of impinging waves and conditio^
Ojrior to wave action.
To analyze a system, however, the equations must be both time and position dependent. This dependency is indicated by the form of the basic partial differential equations (eqs. (1) and (2)). It ie apparent that the pressure wave impinging at one diecontinuity in a liquid system is exactly that wave which left an adjacent discontinuity at some
time in the past (because of the lossless line connector simulation). Thus, the waves
are related by position and time. In general, variables have the form f ( x , t). Specifically,
t), V(x,t), and AH(x, t) denote the pressure head, velocity, and pressure
pulse as a function of position and time, with x and t being the position and time subscripts, respectively.
The wave-plan analysis provides for information at specified points in the system
(discontinuities) and at discrete time intervals. New information is available only at
discrete time intervals because all of the disturbing functions are approximated by a
series of small step changes occurring at specified times. The system response to
these disturbing functions will also be in the form of step changes. Because of the twoparameter dependence of the dependent variables, it is advantageous to denote position
and time subscripts.
Position subscripts. In order to apply the analysis equations that have been developed for a discrete point to a liquid system, it is necessary to introduce a method of
denoting position. This is done ty numbering adjacent discontinuities in a consecutive
manner as indicated in figure 9, (Thisarrangement is modified at junctions of three or
more legs). The position subscript is denoted as L.
In addition, because pressure and velocity states may differ across discontinuities,
it is necessary to denote a left and a right side of a ciiscontinuity. This is done by adding
an L for left and an R for right to the nonaubscripted part of the variable.
Thus HL(L, t) denotes the pressure head to the left of diecontinuity L at time t.

ax,

15

I*
fl

Time subscripts.

- The time subscript J is defined aa followe:

t=JAt

where t ie the time and At is the working time increment. Thus HR(L, J) denote8 o#r
pFOWW@bd tO Un Pi# Oi dtrCaMdQ L & #Iru t J 01,
gelectlon of worldng time increment. The working time increment At i o the time
interval between successive computatiow.
Its selection is determined by two factore. First, the time increment must be
small enough to approximate accurately all disturbing functions by a seriee of step
changes. Second, it is necessary that a l l reflections of pressure pulses in the system
take place at an integer number of time intervals. The wave travel times between all
adjacent discontinuities w i l l then be an integer number of working time increments. If
this were not the caee, waves would be impinging on a diecontinuity in a completely
arbitrary fashion and would make the solution unmanageable.
The selection of the time increment is simplified because the majority of diecontinuities in a system will be friction orifices, and these can be placed where desired.
For example, the selection of the working time interval for the system shown in figure 9
is made as follows:
(1) The wave travel times between adjacent geometric discontinuities are determined:
(a) tac wave travel time between A and C
(b) tcb wave travel time between C and B
Within the desired limits, the largest time interval of which these travel times are
integer multiples is determined. This time interval represents the largest working time
increment possible.
(2) It is then determined if the integer number of working time increments necessary
for travel between A and C and between C and B is large enough to insert the desired
number of friction orifices that must, of course, be separated in time by at least one
working time increment. If not, the increment can be divided by any integer to obtain a
smaller working time increment.
(3) It must be determined if the working time interval necessary to assure an integer
number of time intervals between all discontinuities is small enough to approximate
accurately the disturbing function (variable-area orifice at A) by a series of step changes.
If not, the working time increment is further divided by an integer necessary to get a
suitable approximation. The integer number of time increments between discontinuities
is increased by this factor.
Subscripted notation for analysis equations. The analysis equations are easily extended for application to a liquid system by the we of the subscripted notation. Figure 10 shows conditions at a discontinuity before and after wave action when the subscripted notation is employed.

16

The time that the waves reach discontinuity L is t 5: J At. C O I I ~ ~ ~before

~ O Mand
'after that time are constant, and step changes occur at t = J At.
A comparison of the nornubscripted natation in figure 5 to the subecripted notation
in figure 10 gives the identities necessary to make the analysis equrrtione general e~qmee8 i O M for a Uquid system. These identities are au follows:
VL(L, J 1) V1 velocity to left of discontinuity L
HL(L, J 1) = HI = pressure head to left of discontinuity L
VR(L, J 1) = V2 = velocity to right of discontinuity L
HR(L, J 1) = H2 = pressure head to right of discontinuity L
AHE(L 1, J KX) = AHt = pressure pulse coming from adjacent discontinuity to
left. KX is the number of working time intervals it takes a sonic disturbance
to travel between the two disccmtinuities.
AHL(L + 1, J KY) = AH2 = pressure pulse coming from adjacent discontinuity to
right. Ky is the number of working time intervals it takes a sonic disturbance
to travel between these two discontinuities.
The conditions after wave action are as follows:
VL(L, J) = Vll = velocity to left of discontinuity L
HL(L, J) = Hll = pressure head to left of discontinuity L
VR(L,J) = V22 = velocity to right of discontinuity L
HR(L, J) = H22 = pressure head to right of discontinuity L
AHL(L, J) = AHl1 = pressure pulse leaving discontinuity L and moving to left
A W L , J) = AHz2 = pressure pulse leaving discontinuity L and moving to right

- -

- -

Substitution of these identities into the analysis equations (eqs. (8) to (29)) yields a
perfectly general set of algebraic equations applicahle to any liquid system.
Method of computation. The solution is carried out by using the subscripted equa-

t i O M a8 fOllOWS:

(1) A working time increment is selected.

(2) Mtiol conditions a r o computed from glrton rtody-rtrto valuor that rrdrt in the
17

system prior to the initiation of a disturbance. These conditions include velocity and
pressure to the left and right of each discontinuity and are denoted at t = 0 by VL(L,O),
VR(L,O), HL(L,O), and HR(L,O).
(3) Computations are then carried out to determine conditions at each discontinuity
at the end of the first working time interval by using the conditions from step (2) as Mtu conditiono, It should be noted that the anatgrir qurtionr will pradiot no crhrngrr at UI
undisturbed discontinuity that is not subjected to impinging waves. Thus, conditions will
only change at the end of the first time interval at a disturbance source (point A, fig. 9):
At the disturbance source conditions change, and waves are generated that w i l l subsequently alter conditions at adjacent discontinuities.
(4) Each discontinuity is analyzed at the end of each interval. The computations are
carried out as long as desired to give pressures and velocities at each discontinuity at
the end of each working time interval.

Digital Computer Programing

The wave-plan analysis consists of the sequential solution of many equations. The
dynamic analysis of even a rather simple liquid system for a reasonable number of time
intervals involves a large number of computations. A digital computer must be employed
if the solution is to be obtained in a reasonable length of time,
i.
To make the digital analysis as general aa possible, computer segment programs
(subroutines) have been written for the most common types of discontinuities. These
discontinuity solutions can then be combined (incorporating junction equations, if necessary) to obtain digital computer models for various liquid flow systems,
Each discontinuity subroutine computes conditions at the discontinuity for some
point in time as a function of the local velocity and pressure-head conditions a time interval earlier, the magnitude of impinging waves (from adjacent discontinuities), and the
step changes in disturbing functions during the time interval.
The following subroutines have been written to be used in the digital model representation:
(1) Terminal orifice subroutine
(2) Internal (friction) orifice subroutine
(3) Diameter discontinuity subroutine
Each subroutine uses the general relations given by equations (12) to (15). Also
each subroutine initially assigns nonsubscripted identities to their subscripted counterparts and finally assigns the nonsubscripted results to the subscripted terms (as indicated in the section entitled Subscripted notation for analysis equations). A description
and computer printout8 of the digital computer subroutine8 are included in appendix A.
A l l digital computer programming has been done in Fortran IV.
I

18

The subroutines are combined to form digital computer models for the dynamic
analysis of liquid flow systems. Two examples are included in the next section.

RESULTS AND DISCUSSION

The equation6 and program8 developed in the ANAL=

section form a baais for the

analysis of unsteady flow conditions in a liquid flow network. The equations and program8 have been kept general 80 that they may be applied to the analysis of a variety of
liquid systems. In general, the wave-plan digital-computer analysis has distinct advantages over other methods such as analog-computer models that substantially linearize or
@mp parameters and small perturbation techniques that linearize around mean-line conPtions. The following features of the wave-plan analysis should be noted:
'i
(1) Viscous friction effects are easily included.
(2) Perturbing functions of any form may be used.
(3) Nonlinear relations a r e easily included. (Linearization of parameters is of little
or no advantage.)
(4) Both transient and steady-state responses to disturbing functions are given.
(5) Complex networks can be analyzed.
Two examples will be used to illustrate the application of the wave-plan method to
the construction of digital models of particular liquid flow systems. The first example
will illustrate the methad of computing the transient and steady-state respowe of a long,
straight hydraulic line to a periodic sinusoidal flow dbturbance. In the second example
the response of a long tapered line wfll be computed.

Example 1: Long Hydraulic Line With Oscillating Inlet Orifice

This example w a s chosen because the computer results can be compared
directly with experimental data from the Lewis Research Center line dynamics
rig406. Schematics of the liquid flow system a r e given in figure 11. Figu r e s l l ( a ) and (b) show the analytical model and the experimental rig, respectively. The system consists of a 68-foot-long, 7/8-inch-inside-diameter line
with p r e s s u r e r e s e r v o i r s at both ends. The test fluid was JP-4 fuel o r another
hydrocarbon, depending on the liquid viscosity desired. Line p r e s s u r e s
ranged from 200 to 400 pounds per square inch. A variable-area throttle valve
that perturbs the system i s located at the upstream end of the line (point A),
and a fixed orifice is located a t the downstream end (point B). The details of
constructing the digital model a r e given i n appendix B along with the program
constants used in the calculations.

E-2587
.

_."_
.

19

I
I

_-I_-

.+

Boostpump
Boostpump

Dum0

FluMsupplytmk
FluMsupplytmh
(b) Schrmrtlc of eqmrlmentrl

Figure 11.

- t$dfaullc

line wlth osclllllng inlet alflrn.

- _

rig.

Effect of amplitude of input perturbations. Most line-dynamics analyses are based

on small-perturbation techniques that permit the use of linearized terminal impedances
as an approximation for nonlinear pressure-flow relations at the end of the line. Therefore, in taking experimental line-dynamics data406 the practice is to limit the
amplitude of the oscillating throttle disturbance generator. The minimum usable amplitude, however, for a given line condition and terminal impedance is limited in practice
by the ratio of the sine wave signal to the nonharmonic noise which decreases with decreasing amplitude. The permissible maximum amplitude is limited in practice by the
appearance of harmonics in the sinusoidal pressure signal as the amplitude is increased.
The system must be operated between these two limits.
The relation between the disturbance amplitude and the degree of nonlinearity of the
pressure and velocity perturbations was determined analytically by varying the amplitude
of the sinusoidal input perturbations for the analytical line model. The mean orifice
coefficienta were 0.8 at the inlet and 1.0 at the outlet. The steady line velocity wa8
14 feet per second. For the digital computer program (appendix B) the amplitude of the
input orifice coefficient perturbation BA waa varied at 40 cps 88 shown in table 1.
20

- AMPI.JTUD& OF ORIFICE

The calculated steady-state responses for the

pressure and fluid velocity at the inlet and outlet of
COEFFICIENT PERTURBATIONS
the line (points A and B, fig. ll(a)) are shown in
figure 12 (p. 22). The velocities and pressures are
coemcient
rtcrtt orifice
dimensionlese, having been divided by steady line
valuer, ~ i ~ ri ac a,how thrt, w tha rwutudr oi
the disturbing function i a decreased, the system
responses are more nearly described by a
sinusoidal-type periodic function; however, for the
particular line condition and terminal impedance
used in the example, nonlinearities are evident
5
even for relatively small perturbations. For example, the inlet velocity perturbation is very clearly nonsinwoidal for an 18,75-percent
disturbing amplitude even though the perturbation amplitude'ie only 2 percent in the positive direction and 4 percent in the negative direction.
Transient response, Dynamic systems analyses are usually based on eteady-state
responses to steady sinusoidal inputs. This analysis technique does not, however, predict the transient response to the initiation of a disturbance or a change in the dbturbance characteristics in a liquid flow system. The wave-plan solution gives both traneient and steady-state solutions. This point is clarified by examining both the transient
and final steady-state responses to the initiation of a sinusoidal disturbing function in a
liquid system under steady mean-flow conditions. A value for BA of 0.15 (case 4, table I)
was chosen for the amplitude of the disturbance function. The line response starting with
the initiation d the disturbing function is given in figure 13 (pp. 23 and 24). For the case
studied several cycles were required for the perturbation velocities and pressures to approach the steady-state values. In some cases the amplitudes of the transient responses
are considerably different from the steady responses. For example, during the early
part of the transient period the magnitude of the inlet velocity perturbation was nearly
double the steady-state value.
Comparison of analytical and experimental results. The experimental hydraulic
line (fig. ll(b)) was operated with the upstream throttle valve sinusoidally perturbed at
70 cps with area perturbation amplitudes of 12.3, 33.8, 55.3, and 76.8 percent of the
open area. The mean value of the coefficient for the upstream orifice was 0.65, and the
corresponding value for the downstream orifice waa 0.42. The downstream orifice was
slightly open with respect to characteristic. Figure 14 (p. 25) shows the resulting experimental pressure-time oscilloscope traces for the quarter-length point of the line
(measured from the upstream throttle valve).
The experimental Conditions were duplicated by the digital model, and figure 15
(p. 25) shows the theoretical results for these four different input amplitudes. The
TABLE I.

'

21

4I

1.1

LO
.9

.7
.6

.5
.I

rn

12s

135

I40

145

150

i55

Time increment
Figure P -Effect of anplitudc of inld orifkc area prturMion on stealy-state response of long line.

agreement between the wave shapes determined analytically and e x p e r i m b l y is very

good for all cases. In addition the analytical and experimental magnitud$h agree to
within the accuracy of the measurements (approx. 3 percent).
This example illustrates that the wave-plan analysis for unsteady liquid flow is
capable of accurately including nonlinear effecta and giving results in good agreement
with experimentally observed values.

22

20

io

30

60

w)

Time increment

no

w i w i i o

Figure U. Translent responseof long line to initiation of sinusoidal inlet valve area perturbaim.

Example 2: Response of Long Tapered Line to Sinusoidal Input Disturbance

The analysis of the response of a liquid system composed of lines of nonuniform
diameter (tapered) to a periodic disturbing function is difficult if classical methods are
employed; however, tapered lines can easily be included in a wave-plan analysis by
approximating the tapered line by a line composed of a discrete number of diameter
changes (fig. 16, p. 26).
In order to evaluate the quantitative effect of tapered lines, the system shown in
figure 17 (p. 26) was studied in detail. This system consists of two pressure reservoirs
connected by a tapered line. At the end of the tapered lines, relatively large resistive
lossee were introduced in the form of square-low ortfises. The throttle valve at point A
har a rtrnrroldally varying area, and t b orifice, at point B ir fixad.
23

24

1 raho In.-\

(a) Orifla area pertutMionamplituda, 123 percent of open

(c) Orifice area perturbation amplitude, 55.3 percent of opn

area

am.

25

This system was studied for variaue

degrees of taper. The configurations

were chosen to give the diameter at the
inlet ae'

and the diameter at the outlet as

The diameter of the tapered conduit

is assumed to vary linearly between the
ends, and the average diameter is alwaye
D feet.
If the resistances of the terminal orifices are large compared to the line loss, the
steady discharge of the system is practically independent of 8 and would depend mainly
on the reservoir pressures and the mean line diameter. Even for lines with much
smaller end-resistive losses the system discharge would be only slightly dependent on
the degree of taper, because line losses consisting of friction and effdansion or contraction losses would be mainly dependent on line length and average diameter. Mathemtically this can be written a8

Q2 =

HEN
Ken

- HEX
%

Kex

where Ken, K1, and Kex are the loss coefficients of the entrance orifice, the line, and
the exit orifice, respectively.
When this system is used with large values of Ken and ICex a8 compared to Kl,
it is possible to compare directly the dynamic response of systems that are essentially
~taticallyequivalent. Pressure and flow perturbations can be quantitively compared as
a function of degree of taper. The straight conduit (6 = 0) is a special case that is included in the analyses.
The tapered conduit was chosen so that the wave velocity would be constant; thus the
ratio of diameter to wall thickness remained constant along the length of the line.
For the digital model the tapered line was approximated by N - 1 short straight
sections of conduit, as shown in figure 17. In this manner the diameter of each approximating section I is computed by the equatson
26

/
276 R

0.mn

t
-

FIgwe lR Pressure nrd velocity pertuMkn responses in tqered line.

lnld dianctcr. ab foat: exlt diameter, L4 feet; frequency, 17.5 cps.

D O = D A + (D B - D A

)( )
21-1

N-1

(33)

The details of the digital model along with-the computer flow diagram
and program constants a r e given i n appendix C. One basic system w a s studied
in detail, a system having a high-resistance orifice (three-fourths to seveneights closed geometrically but slightly open with respect to characteristic impedance) at the inlet with a 10-percent sinusoidal variation in the a r e a of the
orifice opening, The outlet orifice w a s chosen of even higher resistance than
the inlet and represents an end that is closed with respect to characteristic
impedance. Nine degrees of taper were studied, the average diameter D
being 1 foot for each case. The values of 6 that were used were -0.3, -0.2,
-0.1, 0 , 0.1, 0 . 2 , 0.3, 0.4, and0.5foot.

The quarter-wave resonance frequency for the nontapered line is 15 cps. Frequencies of 5, 7.5, 10, 12.5, 15, 17.5, 20, and 25 cps were studied for each case. In all
cases the pressure and flow perturbations were slightly nonlinear even though they were
small. In all cases they were within 2 percent of mean-line values. Figure 18 shows
typical pressure and flow perturbations for one cycle,
27

0,

5.0

Figure 19.

7.5

10.0

12.5

15.0

17.5

20.0

2 2 5 25.0

- Frequency dependence ol trgered-line dynmk response.

In spite of the nonlinearity of the perturbations, they were steady and repeatable
after several cycles (during which the transient died out) and resembled sinusoids. It
was considered appropriate to make comparisons based on positive amplitudes. This
procedure is followed in the ensuing discussion.
The analytical results are displayed graphically in figure 19. Figure 19(a) shows
the variation of the amplitude of the outlet pressure perturbation with frequency for the
expanding taper (line that expands from inlet to outlet). Figure 19(b) shows the pressure
perturbatim gain (ratio of outlet pressure perturbation amplitude to inlet pressure perturbation amplitude) for the expanding taper, Figures 19(c) and (d) show the same results for contracting tapers. The most striking result apparent in figure 19 is the shift
in the resonant frequency for tapered lines. This shift is shown in figure 20.
Another result is that the maximum pressure perturbations at the outlet have a
strong dependence on 6 and vary considerably over the entire frequency range. Thus a
certain amount of control can be exerted over the response of this system by the judicious choice of tapers. For example, in the case of an expanding taper of 6 = 0.3 the
amplitude of the outlet pressure perturbation is significantly reduced when compared to
a straight-line response for the frequencies shown except in the range from'6 to 1 2 cps.
A line with a contracting taper of 6 = -0.3 differs little from a nontagered line (6 = 0)
up to 15 cps. For frequencies greater than 15 cpe the outlet preaaure amplibde pertur-

bation is considerably greater.

This example indicates that the
dynamic response of a liquid system
can be altered significantly with a certain amount of control without signiffoantly rlteriag the rtrtic rarponre OX
the system; however, the primary
intent of the example is to illustrate
the use of the wave-plan analysis.
Although the example was applied to
a system with a linear taper and a
sinusoidal disturbing function, a system of any arbitrary taper with any
disturbing function could have been
analyzed with no additional difficulty.

CONCLUSIONS
The analytical method presented in this paper provides a means of obtaining distributed parameter solutions to a variety of unsteady flow problems
for liquids flowing in conduits. An advantage of the method is that the complete solution is obtained. F o r example, both the transient and the steadystate response to a suddenly imposed periodic flow disturbance is obtained.
Furthermore, the disturbing function c a n be of a r b i t r a r y f o r m and need not be
periodic. Nonlinear effects a r e easily included. In addition, the wave-plan
method is advantageous in making certain types of dynamic response calculations. For example, the response of liquid-filled lines having types of axial
cross-sectional area distributions f o r which there would be little hope of obtaining closed-form analytical solutions can be easily handled.
The wave-plan analysis can readily solve problems in which there is
interaction between the structural motion of the conduits and the perturbations
in the fluid flowing within the conduits. Interactions of this type often occur
in hydraulic systems; however, the motion is usually neglected in the analyses.

29

APPENDIX A
COMPUTER SUBROUTINES

Piglkl-.oomputss rwtinae have been written te, rolve fnr aonditfanr%at vwfour h e
discontinuities after wave action in terms of the conditions at the discontinuity prior to

wave action, the magnitudes of impinging waves , and the physical characteristics of the
discontinuity. These routines have been formulated in such a manner that they may be
easily incorporated into a computer program to synthesize different liquid flow systems.
In all cases the computer routines have been written in nonsubscripted notaion,
which corresponds exactly to the notation presented in the text of this paper. In the
calling vector that calls the subroutine into the main program, the subscripted counter: parts of the nonsubscripted subroutine variables are identified.
This mechanism of identifying subscripted variables in a calling vector
makes i t possible to use the analyses presented in the text for specific orientations and flow conditions to solve for all orientations and flow conditions. In
this manner the analyses presented in the text of this paper have been generalized. Proper identification of the variables in the calling vector is tantamount
to orienting the discontinuity to correspond to the case analyzed in the text. A
table incorporated into each subroutine gives the proper subscripted-tononsubscripted dependence for each case.
Each of the analyses presented in the text resulted in a polynomial that
is quadratic in a velocity term. For several reasons these equations were
solved by an iterative manner by employing Newtonian extrapolation. The roots
of the polynomials were determined by looping the following equations:

error = aV112 + bVll + c

2aVll+ b

V11 = V11

- error

The looping is continued until the error is of sufficiently small magnitude. An acceptable solution can be obtained within a very few cycles because the wave-plan solution
deals with small changes, and the value of the velocity before wave action is a good
approximation of the final value. In addition, the coefficients b and c are generally
large and always much larger than a. These conditions lead to a rapidly diminishing
error term.
A direct crofution of the gu&rlrticprquakianr by tho quadratio formula rerultr irt a
30

solution that is a small difference between two large numbers. In soqe cases sufficient
accuracy cannot be obtained even when double-precision computing teqhniquee are employed. The computer programs which use notation identical to that appearing
in the text of this paper didactically illustrate the logic of the subroutines and .
a r e presented in lieu of flow diagrams, + A brief description of the program and
the Fortran I V program f o r each computer routine followo

Te rm inaI 0 rifice S u brout ine

Equations (16) and (17) a r e the basic equations employed in this analysis. They axe
derived in the text for an orifice on the left end of a line. The table at the b e g h h g of
the subroutine gives the identities that must be substituted into the calling vector depending on whether the orifice is on the right or left end of the line. Before the eubroutine
is entered, the following terms must be identified with a numerical value:
B
orifice coefficient
HEN reservoir head for left end
HEX reservoir head for right end
J
time counter
number of time increments to nearest discontinuity
K
L
position counter
The comptter program for a terminal orifice is a8 follows:
SUBROUTINE T O R ( H 2 , D H 2 r V 2 , H l l , H 2 2 , D H 2 2 , V 2 2 )
S U B R O U T I N E TERMINAL O R F I C E
COMMON/BOX/A,B ,C, GvBF ,ERROR
LEFT END
R I G H T END
H L ( L 9 J-1 1
H2
HR ( L p J - 1 1
-V(L,J-l)
v2
V(L9J-11
DHR(L-19J-K)
OH2
DHL 4 L + l 9J-K 1
-V(L,J)
v22
V(L,J)
HL(L,J)
H2 2
HR(L9J)
DHL L 9 J 1
DH22
DHR ( L 9 J 1
HEX
H11
HEN
BB = C*B*B/G
CC=-B*B+(H11-H2-2.*DH2+C~V2/G~
I F ( C C ) 10,1094
4 66 = -BB

cc = -cc

10 v 2 2 = v 2
11 ER = ( V 2 2 * V 2 2 + B B * V 2 2 + C C ) / ( 2 . * V Z 2 + 8 8 )
V 2 2 = V22-ER
IF ( A B S ( E R ) - E R R O R ) 12,12911
1 2 DH22=DH2+C/G*(V22-V2)
H22=H2+DH2+DH22
RETURN
END

31

Friction Orifice Subroutine

The basic equqions for this routine a r e equations (18) and (19).
is entered, the foll6wing terms must have a numerical value:
1

BF
Y

Before thie routine

friction ofifice coefficient

ttm oountrr
K
number of time increments to neareat dieccmtfrmity
L
position counter
The computer routine is the following:
C
C
C
C
C
C
C
C
C
C
C

3
4
5

SUBROUTINE F O R f H l ~ D H l ~ D H 2 ~ H ~ ~ V l ~ H l l r D H 2 2 ~ D H l l r H 2 2 ~ ~ ~ ~
SUBROUTINE FRICTION ORFICE
H1
HL ( L 9 J-1 1
DH 1
DHR(L-19J-K)
DHL(L+lrJ-K)
DH 2
H2
HR ( L 9 J-1 )
v1
V(L9J-1)
Hll
HL IL 9 J )
DHll
DHL L 9 J 1
DH22
DHR ( C 9 J 1
H22
HRILgJ)
v11
VILpJ)
COMMON/BOX3/ A i 8, C,GiBFtERROR
BB = 2e*BF+BF*C/G
CC=-(H1+2e+DH1+2.+C/G*Vl~(H2+2,+OHZ11+8F*BF
IFfCC) 3 9 3 9 2
BB = -86
cc = -cc
v11 = v1
ER = (V11+V11+BB+Vll+CC)/~8B+2.*Vll)
Vll = V11-ER
IF (ABS(ERI-ERROR 1 5 , 5 9 4
DHll = D~l+C/G+fVl-Vll)
H11 = Hl+DHl+DHll
DH22 = DHZ+C/G*(Vll-Vl)
H22 = H2+DH2+DH22
RETURN
END

Diameter Discontinuity Subroutine

Equations (20) to (29) are employed in the diameter discontinuity subroutine. These
equations are derived for flow in the positive direction (to the right) for both an expansion and a contraction. Before the subroutine is entered, the direction of flow after
wave action is determined, and this direction in turn determines which subscripted variables are inserted into the calling vector.
The subroutine also includes the case where no viscous loss is considered across
the discontinuity. To utilize this case, it is necessary to assign the term LOSS a value
of zero. Additional terms that must be defined are the following:
32

J timecounter
K number of time increment6 to nearest discontinuity
L position counter
The computer routine is as follows:

lU8ROUTINI D l ) C l H l , V I q ~ H l ~ b 4 8 t V t t b H l ~~ CC ll q A ,lA I , H l l t b W l l e n 8 8 , b n 8 ~ 0

C
C
C

C
C
C
C
C
C
C
C
C
C
C

l V 1 1tV22
OISCONTINUITY
SUBROUTINE
COMMON/BOX/A,BICI
GtBFtERROR
COMMON/BOX2/ L O S S
FLOW TO R I G H T
FLOW TO L E F T
CCC GREATER T H A N 0
CCC L E S S T H A N 0
H1
HL(A9J-1)
H R f L tJ-1)
v1
V L ( L 9 J-1 1
-VR ( L 9 J - 1 )
D H1
DHR ( L-19 J - K 1
DHL(L+l,J-K)
HZ
HL 4 L t J-1
H R ( L 1 J-1 1
v2
VR(L9J-11
VL(L9J-1)
OH2
DHL(L+l*J-K)
DHRtL-1tJ-K)
HR(L9J)
H11
HLILwJ)
v11
V L (L 9J)
-VR ( Le J 1
O H 11
DHL ( L 9 J1
DHR ( L 9 J)
HLlLtJ)
H2 2
HR ( L 9 J )
- V L ( L 9 J1
v22
VR(L9J)
DHL ( L 9 J 1
OH22
DHR(L9J)
CR(L)
c1
CL(L)
CL(L)
c2
CRIL)
AR(L)
A1
AL1L)
AL(L)
A2
AR(L)
R=Al/A2
CCS - ( H 1 + 2 . + D H l + C l + V l / G - H 2 - 2 . + O H 2 + C 2 * V 2 / G )
BB = C l / C + C 2 + R / C
I F (LOSS) 4 9 3 9 4
3 AA
(R*R-l.)/(2o*G)
GO TO 10
4 IF(R-1.1
1 9 1 9 2
1 AA = ( R * R - R ) / G
GO TO 10
2 AA = ~ 3 o * R * R - 2 0 + R - l 0 ) / ( 4 o * G )
10 v11= v 1
S ER = ( A A * V 1 1 * V 1 1 + B 6 * V l l + C C ~ / ( 2 . + A A * V 1 1 + 6 B ~
VllSVll-ER
IF ( A B S ( E R ) - E R R O R ) 6 , 6 9 5
6 D H l l = DHl+Cl/G+(Vl-Vll)
V22=R*Vll
DH22=DH2+C2/G*(V22-V2)
~ 1 1 = ~ 1 + ~ ~ 1 + ~ ~ 1 1
H22=H2+DH2+DH22
RETURN
END
a

33

APPENDIX B

DIGITAL MODEL OF HY D RAULlC LINE WITH OSCILLATING INPUT ORIFICE

A rerchemtic drawing of the styrrtorm chooen for this expmple i a sham in figure 11

(P. 20).
The system consists of a variable-area orifice at A that perturbs the system and a
fixed orifice at B. The conduit is bounded by pressure reservoirs. (The reeervoir

pressures can vary. )

To keep the solution general, N 2 friction orifices are inserted, which result in
N discontinuities. A time interval is 80 chosen that each discontinuity ie one time int e d from adjacent discontinuities. Thus

This time increment is a multiple of the working time increment At, which fe
chosen small enough so that all disturbing functions can be accurately described by step
*

functions.
Initial values for velocity through each discontinuity and the pressure to the right
and left of each discontinuity must be inserted into the computer a8 data or the computation of these quantities from the initial steady condition must be integrated into the computer program.
The disturbing function can either be read into the computer a8 a function of time or
it can be computed. For this case a sinusoidal orifice coefficient for the Met (at A) is
taken to be

B = BO + BA SIN(2aFJAt)
The orifice coefficient for the outlet is considered to be constant and is denoted

by BN.
The following data are needed aa input to the computer:
C

L
BO

BO1
BA
F

wave velocity in conduit, ft/sec

length of line, ft
mean orifice coefficient when subjected to sinusoidal perturbations
steady-state inlet orifice coefficient
amplitude of inlet orifice coefficient perturbation
frequency of pressure perturbotion
34

BN

vo
HEN

HEX
g

outlet orifice coefficient

steady 1i+e velocity, rt/sec
pressure head of inlet reservoir, ft
pressure head of outlet reservoir, ft
acceleration due to gravity, ft/sec 2

number of discontinuitieo

number of working time increments for wave to travel length of conduit

K
number of working time increments between discontinuities
M
total number of time increments for which computations will be made
N2
number of friction orlfices for which pressure and velocity data w i l l be
printed
In setting up the computer program the notation j = x, y, e is used. The notation
means that computations are carried out for j starting at j = x. The computatiom are
repeated for values of j that are increased by y until j > e,
The following output parameters are printed out by the computer:
w , 4 , "1,J)
NlQ

w , 4 , WN,4
V W , J ) , H"2,J),

W N 2 , J)

For a specific example, the fluid system constants appearing in the computer program were chosen nominally to match experimental conditions in the Lewis linedynamics facility shown schematically in figure ll(b) (p. 20). Two sets of line conditions were used. The first (case I)
TABLE II. CONSTANTS FOR EXAMPLE
wat3 chosen in conjunction with an
analytical study of the effect of the
size of the input amplitude on the
C , ft/sec
linearity of the perturbation imposed
L, ft
on the mean flow and pressure in the
0.8
0.65
BOl=BO
line. The second (case II) was chosen
BA
0.42
1.0
BN
to match conditions frequently obtained
in the line.
0.5
14
VO, ft/eec
360
520
HEN, ft
The constants listed in table 11
0
0
HEX
were employed. A simplified

g,

ft/sec2

K
M

"
F,

CPg

N2

ERROR

32.2
17

1
170
10
40
5
0. ooool,

32.2
17
2
400

32
70
5
0.00001

35

computer flow diagram for this

example i s presented in figure 21.
A complete listing of the Fortran
IV program for this example is
available 12

APPENDIX C

DIGITAL MODEL FOR TAPERED HYDRAULIC LINE

T b oyottrm chorrn for tho to;inrod iay&rUrio Una k ohown in fi(ur0 17 (pa IS). Thk
system conoisb of a variable-area orifice at A and a fixed orifice at B connected by a
tapered line with a linear diameter variation between A and B. The line is bounded by
pressure reservoirs. The tapered line was approximated by a line composed'of a dieCrete number of constant-diameter sections of decreasing (or increasing) diameter.
To keep the solution general, N 2 diameter discontirmities are inserted, which
.result in N discontinuities and N 1 sections of straight conduit.
; The conduit is so divided that the wave travel time8 between adjacent discontinuities
are multiples of the working time interval, and the diameter of the approximating
straight section of conduit is equal to the average diameter of the section that it is
approximating.
It i s assumed that the wave velocity remains constant over the entire conduit, which
is equivalent to assuming that the ratio of the conduit diameter to wall thickness remains
constant (a re5L8onable assumption from a structural viewpoint). The conduit ie thus
apprordmzrted by N 1 sections of equal length. The diameter of the approximating
straight section is given by

Initial values for velocity and pressure to the right and left of each discontinuity
must be inserted into the computer, or the equations for the computation of these quantities from initial steady conditions must be integrated into the computer program.
The disturbing function can either be read into the computer aa a function of time
or it can be computed. For this mdel a sinusoidal orifice coefficient for the inlet (at A)
is taken to be

B = BO + BA SIN(2nFJAt)
The orifice coefficient for the outlet is constant and denoted by BN.
The following data are needed a8 input to the computer:
DA diameter at A, ft
DB diameter at B, f t
CA wave velocity in conduit, Wrec
-.
36

DA

DB

BO

BA

BN

1
1.1
- 8 1.2
. 7 1.3
.6
1.4

1
1.235
1.562
2.041
2.78

0.1
.1235
.1562
.278

0.2
.I65
-139
.I185
102

1.5

4.0
.825
.695
,583

.4
,0825
.0695
,0593

.089
,247
.a13
,408

-9

e 5

1.1
1.2
1.3

-9

e8
e7

.mi

N = 11
NN- 10
M = 300
Table IlI gives the end diameters and conditions
for the various tapers studied.
These cases were studied for frequencies of
5, 7.5, 10, 12.5, 15, 17.5, 20, 22.5, and 25 cps.
Figure 22 gives a simplified computer
flow diagram for this example. A listing
of the Fortran I V program for this example

APPENDIX D

SYMBOLS
uR(I WI1, q ft
line area of 1, q ft

90

'

o h t i o modulur ot oonvoyar,
Ib/#q ft

elastic moduhw of fluid, Ivaq ft

line area of j, sq ft
orifice area, sq f t

Darcy friction factor

line area to left of diameter diecontinuity, sqft

line area to right of diameter
discontinuity, eq ft
--_

function

acceleration due to gravity,

' ft/sec2

polynomial coefficients

HEN

pressure head of inlet reservoir,

ft

friction orifice coefficient,

(ft/secq 'I2

HEX

pressure head of outlet reservoir, ft

orifice coefficient before wave

action, (ft/sec 2) 1/2

H1

pressure head to left of discontinuity before wave action, ft

orifice coefficient after wave

action, (ft/sec?'/'

HZ

pressure head to right of discontinuity before wave action, ft

sonic velocity in line, ft/sec

H1l

pressure head to left of discontinuity after wave action, f t

H22

pressure head to right of discontinuity after wave action, f t

orifice coefficient, (ft/sec?l/'

orifice discharge coefficient

wave velocity in line i, ft/sec
wave velocity in line j, ft/sec

AH

wave velocity to left of discontinuity, ft/sec

*H1

wave velocity to right of discontinuity, ft/sec

line diameter, ft

AH2

mean diameter of tapered line,

ft

38

pressure head, ft

pressure head change across

wave or orifice, ft
wave impinging from left of discontinuity before wave action,
ft
wave impinging from right of
discontinuity before wave
action, ft

.
0

*%l
LIH22

velocity in h e , ft/~ec

wave leaving discontinuity on

left side after wave action, ft

. YE1

wave leaving discontinuity on

right side after wave action,
ft

V*a

prorruro Mad lor#wo?amall

VO

line length, it

average v e k i t y of moving end

before wave action, ft/eec
average velocity of moving end

Mor WIva wtim, rt/rro

velocity in line adjacent to moving oritice, relative to i f i c e ,
fi/rec

time aubrcr\pt
1

number of working time increments between diecontinuitieo

V1

velocity in line to left of diecontinuity before wave action,

ft/sec

v2

velocity in line to right of discontinuity before wave action,

ft/eec

line length, ft
pressure in line, lb/sq ft
pressure wave, lb/eq t
volume flow rate, cu ft/sec

vll

area ratio at diameter discontinuity

reflection coefficient in line i

v22

transmission coefficient in line i

time, sec
incremental time interval, sec

velocity in line to left of discontinuity after wave action,

ft/sec
velocity in line to right of discontinuity after wave action,
ft/sec

position, ft

short time before wave action,

sec

line taper factor (eqs. (31)),ft

mass density of fluid, slugs/ft3

short time after wave action, see

wall thickness of conduit, ft

--.
>

39

FOOTNOTES AND REFERqNCES

!

1. Asst,

Prof. of Gtvil Engrg. auks University, Durham, North

GaPOlim

2. Head, Dynamic8 Pecgon, NASA, Lewis Reeearch Center , Cleveland, Ohio. .

30 Mathematician, Instrument and Computing Division, NASA, Lewis Research
Center , Cleveland, Ohio.

I.

4. RegetZ, John

D. 8 Jr.

: An Experimental Determination of the Dynamic Re-

NASA TN D-576, 1960.

sponse of a Long Hydraulic Line.

5.--Blade, Robert J. i Lewis, William; and Goodykoontz, Jack H. : Study of a

Sinusoidally Perturbed Flow in a Line Including a 90 Elbow with Flexible
Supports. NASA TN D-1216, 1962.

6. Lewis, William; Blade, Robert J. ; and Dorsch, Robert G. : Study of the

Effect of a Closed-End Side Branch on Sinusoidally Perturbed Flow of
Liquid in a Line. NASA TN D-1876, 1963.

7. Streeter, Victor L.

i and Lai, Chinter: Water-Hammer Analysis Including

Fluid Friction. Proc. Am. Soc. Civil Eng. (Jour. Hydraulics Div.),
vol. 88, no. HY3, May 1962, pp. 79-112.

8. Wood, D. J.; and Stelson, T. E. : Energy Analysis of P r e s s u r e Surges in

Closed Conduits. Developments in Theoretical and Appl. Mech. , Vol. 1,
Plenum Press, 1963.

9. Joukowski, N. (0. Simin, trans.): W a t e r Hammer.

Works ASSOC., ~ 0 1 .24, 1904, pp. 341-424.

Proc, Am. W a t e r

10. Bergeron, L. : Water Hammer in Hydraulics and Wave Surges in Electricity.

John Wiley & Sons, Inc., 1961.
11. Rouse, H. : Engineering Hydraulics.
pp. 479-4820

John Wiley & Sons, Inc. , 1950,

12. Wood, Don J. ; h r S C h , Robert G. i and Lightner, Charlene: Digital Distributed P a r a m e t e r Model for Analysis of Unsteady Flow in LiquidFilled Lines. NASA TN D-2812, i965.
.

.'

40

- -

-Compute or read
initial conditions

Read

program
data

Thrwgh
A
J = 1,l. M

conditions at
lefl end
L-l

Thrwgh
B
L-2,1, N - 1

Figure 21.

- Read

prqlram

data

end

- Computer flw diagram for hydraulic line with oscillating input orifice.

read^

Compute or
initial conditions

Through
A
J =l,l,
M

Conditions at
left end
L- 1
B = BO + BA sin(216J At)

end

Figure 22.

- Computer flaw diagram for tapered hydraulic line with oscillating input orifice.