229

Pressure pulsations in reciprocating pump piping

systems
Part 1: modelling

J-J Shu, C R Burrows and K A Edge

Fluid Power Centre, School of Mechanical Engineering, University of Bath

Abstract: A distributed parameter model of pipeline transmission line behaviour is presented, based on a
Galerkin method incorporating frequency-dependent friction. This is readily interfaced to an existing
model of the pumping dynamics of a plunger pump to allow time-domain simulations of pipeline pressure
pulsations in both suction and delivery lines. A new model for the pump inlet manifold is also proposed.

Keywords: reciprocating plunger pump, pipeline dynamics, pressure pulsations, distributed parameter
model

NOTATION mi ith coefficient used in approximation to fre-

quency-dependent friction
a orifice area M total number of cylinders
A area of valve face ni ith coefficient used in approximation to fre-
B bulk modulus of elasticity of fluid quency-dependent friction
Bc n effective bulk modulus of elasticity of liquid N integer (number of pipeline elements 1)=2
and air in chamber n p value of pressure at nodal points
BEFF effective bulk modulus of elasticity of fluid P pressure
c damping coefficient P9VC estimated pressure at vena contracta
cSTOP damping coefficient of valve and stop and seat Pc n pressure in chamber n
cV restrictor valve coefficient PDOWN pressure downstream of valve
p
c0 acoustic velocity BEFF =r PL pressure at entry pump inlet manifold
CD discharge coefficient POUT pressure at pipe boundary p2 N 1
CF force coefficient PUP pressure upstream of valve
CQ flow coefficient PVAP fluid vapour pressure
DP plunger diameter PVC pressure at vena contracta
f friction factor P0 atmospheric pressure
F frictional loss in pipe Q flow
FN net force acting on valve Qc total instantaneous flow drawn from inlet line
FPRE spring preload Qc n flow into chamber n
g acceleration due to gravity QIN n inlet valve flow for chamber n
k total number of terms used in approximation to QLV flow through load valve
frequency-dependent friction QNET net cylinder flow
ks valve spring stiffness QOUT delivery valve flow
k STOP stiffness of valve end stop and seat QP flow due to motion of plunger
Kc n pressure loss coefficent for chamber n QV valve flow QIN or QOUT, as appropriate
l length of connecting rod QVC vena contracta flow
lF length of fluid jet in vicinity of valve seat r crank radius
m effective valve mass r0 pipe internal radius
t time
u valve velocity; value of U at nodal points
The MS was received on 9 March 1996 and was accepted for publication U rQ=(r20 )
on 19 April 1997.

I01596 # IMechE 1997 Proc Instn Mech Engrs Vol 211 Part I
230 J-J SHU, C R BURROWS AND K A EDGE

U IN value of U at pipe boundary u0 study of pressure pulsation behaviour in delivery lines of

V volume of fluid in cylinder fluid power systems employing typically gear, vane or axial
Va n volume of air in chamber n at pressure P0 piston pumps [e.g. see references (1) and (2)] and to a lesser
Vc n volume of chamber n extent to the suction lines of these systems (3, 4). However,
Vl n volume of liquid in chamber n fluid power pumps typically employ a large number of
VTDC volume of fluid in cylinder at top dead centre pumping elements (nine cylinders are commonly used in
VV volume of vapour in cylinder axial piston machines, for example). As a consequence they
wm mth weighting function create relatively low amplitude flow pulsations and, in most
x distance along pipe instances, low-amplitude pressure pulsations with a rela-
yi value of Yi at nodal points tively high frequency content are generated. This allows a
Yi ith value of frequency-dependent friction loss linearized analysis to be adopted and predictions can be
coefficient conveniently conducted in the frequency domain.
z valve opening In contrast, plunger pumps generally have a small
number of cylinders (three or five are common) and are
valve orientation usually operated at lower speeds. This leads to very large
x pipeline node spacing flow pulsations, relative to the mean flow. It is possible that
valve seat half-angle the consequent high-amplitude pressure pulsations, particu-
0 inclination of pipe larly in resonant delivery line systems, may invalidate the
fluid viscosity use of linear theory. Hence predictions of behaviour need
r fluid density to be performed either in the time domain or by means of
crank angle relative to top dead centre an iterative scheme [e.g. see reference (5)]. A frequency
domain approach to the prediction of suction line pulsation
behaviour is likely to be invalid if cavitation is occurring as
Subscripts
the effects are highly non-linear.
e operator equation number [equations (22) and Some useful progress has already been made by a
(23)] number of workers on the mathematical modelling of the
j node number pumping dynamics of reciprocating plunger pumps.
n cylinder number Johnston (6), for example, has developed a detailed
model which accounts for both valve dynamics and
cavitation in the pump cylinders. However, inlet line
pressure is taken to be constant and the delivery line is
1 INTRODUCTION represented by a lumped parameter model. Vetter and
Schweinfurter (7) address the problems of delivery pipe-
Reciprocating plunger pumps are robust, contamination line wave propagation effects, but adopt a fairly rudimen-
tolerant and capable of efficiently pumping many types of tary pump model. Most of their predictions of pulsation
fluids at high delivery pressures. As a consequence, they behaviour are presented in terms of peak-to-peak pulsa-
are widely used in a diverse range of industrial applica- tion levels, rather than frequency spectra or time-domain
tions, including mining (for powered roof supports), chem- waveforms. Thus the accuracy of the model is difficult to
ical plant, reverse osmosis systems and food processing establish. Singh and Madavan (5) have presented a more
systems. The most common pump construction consists of detailed model of pumping dynamics which is linked to a
a small number of cylinders, usually mounted in-line, each frequency-domain model of the delivery pipeline. An
with a reciprocating piston driven by a rotating crank and iterative process is used to account for the interactions
connecting rod mechanism. During the suction stroke, flow between the pipeline and the pump. Predicted pressure
is drawn from the inlet manifold into a cylinder through a pulsation behaviour is compared with experimental data
self-acting non-return valve; various valve designs are in terms of amplitude spectra. Phase spectra are not
employed although spring-load poppet or disc valves are included in the paper, so, again, it is difficult to establish
most frequently adopted. Fluid delivery also takes place the accuracy of the model in predicting behaviour. Vetter
through a self-acting non-return valve. and Schweinfurter do not attempt to predict suction line
It is well known that the pipeline pressure pulsations pulsations and although Singh and Madavan claim that
produced by these pumps are a source of noise and their model will predict suction line behaviour, no results
vibration and may have a significant influence on the are presented.
reliability of a given installation. Consequently, it is Part 1 of this paper aims to address the inadequacy of
highly desirable to be able to predict pressure pulsations existing models by describing the development of a finite
at the design stage of an installation so that appropriate element model of pipeline dynamics (under non-cavitating
steps may be taken to minimize their levels and their and cavitating conditions) which is integrated with an
influence. existing generic model of pumping dynamics. In Part 2,
Considerable research effort has been devoted to the which follows, predictions of pressure pulsation behaviour
Proc Instn Mech Engrs Vol 211 Part I I01596 # IMechE 1997
 PRESSURE PULSATIONS IN RECIPROCATING PUMP PIPING SYSTEMS. PART 1 231

in both delivery and suction lines are compared with the

results of experimental studies on two test rigs.

2 PUMP MODEL

This study concentrates on a single-acting, in-line plunger

pump, one cylinder of which is shown schematically in
Fig. 1. A detailed model for such an arrangement has been
Fig. 2 Schematic of inlet chamber geometry used in the
developed by Johnston (6). The model accounts for flow
mathematical model
continuity into and out of each cylinder, according to
whether the inlet or delivery valve is open. Inlet pressure is
taken to be constant. In Johnston's approach the flows from
cylinders on their delivery stroke are summed together and
the flows relating to individual cylinders communicating
the resultant is used as the input to a lumped parameter
with the manifold are all assumed to occur at the same
model of the delivery line. Full account is taken of the
location. However, it has been argued (Part 2) that as a
forces acting on the inlet and delivery valves to provide
result of air release, air pockets can form in the inlet
comprehensive modelling of valve dynamics. The basic
manifold. These, combined with pressure losses, play an
equations are reproduced here in Appendix 1 for complete-
important role in the suction dynamics.
ness; full details of the model are given in reference (6). It
To model such behaviour a small chamber is assumed to
should be noted that the approach is readily adapted to suit
be present upstream of each inlet valve, as illustrated
other pump configurations.
schematically in Fig. 2. Each chamber can contain a pocket
The modifications necessary for interfacing Johnston's
of air and communicates with the inlet manifold via a
pump model to distributed parameter models of suction
square-law restrictor. This is close to many real pump
and delivery lines will now be presented.
designs where the inlet valve is located at the end of a
(usually short) passageway at right angles to the manifold.
2.1 Manifold modelling Pressure losses are introduced by the right-angled bend
and, depending on manifold geometry, air pockets could be
The flows from those cylinders communicating with the trapped near the inlet valve. Through the selection of
delivery manifold are assumed to be created at one discrete appropriate restrictor loss coefficients, some account can
location rather than being spatially distributed over the be taken of the different pressure losses likely to be
length of the manifold. This simplifies the interfacing of experienced at different locations along the manifold.
the pump model with the delivery line model. It would not The relevant equations for this model are
be difficult to develop a spatially distributed model of
flows into the delivery manifold but the agreement between p
Qc n K c n (PL Pc n ) (1)
predictions and experimental data (see Part 2) suggests that
this is an unnecessary refinement. The summed flows are
introduced at the internal end of the manifold and define and
the boundary conditions for the delivery line model
(Section 3). The manifold itself is treated as part of the dPc n Bc n
(Qc n QIN n ) (2)
delivery line. dt Vc n
Experience has shown (Part 2) that a more detailed
model may be required for the inlet manifold. Once again, Numerical integration of equation (2) gives the chamber
pressures, which on substitution in equation (1) provide the
flows drawn from the inlet manifold. The total flow drawn
from the inlet line is

X
M
Qc Qc n (3)
n1

That part of the inlet manifold not incorporated in chamber

models is included as part of the inlet line.
To account for the presence of air pockets in the
chambers, an effective bulk modulus of elasticity is
Fig. 1 Schematic of one cylinder of a reciprocating pump required. From reference (8),
I01596 # IMechE 1997 Proc Instn Mech Engrs Vol 211 Part I
232 J-J SHU, C R BURROWS AND K A EDGE

Bc n Vl n =Va n P0 =Pc n pressure P and the flowrate Q. Hence two partial dif-
(4)
B Vl n =Va n P0 B=P2c n ferential equations have to be solved:

In this instance, p0 B= p2cn  Vl n =Va n and hence Equation of continuity:

  1 @P r @Q
Vl n Pc n 2
2 0 (6)
Bc n Pc n 1 (5) c0 @ t r0 @x
Va n P0
Equation of motion:
The principal problem with this detailed model is the
difficulty in selecting appropriate values for the volumes r @Q @ P
Vl n , Va n and the pressure loss coefficient K c n . This may F(Q) r g sin 0 0 (7)
r20 @ t @x
be eased somewhat by selecting the same parameters for
each restrictor and chamber combination, albeit at the In the case of laminar flow the friction term F(Q) can be
expense of losing the ability to model cylinder-to- expressed as a quasi steady term F0 plus an unsteady term
cylinder variations. However, if this approach is accep- (`frequency-dependent friction'), for which an approxima-
table, significant gains in computational efficiency can tion has been developed by Zielke (14) and Kagawa et al.
be achieved by assuming the existence of just one (15):
chamber with which all cylinders communicate. This
chamber is linked to the manifold through one square-
1X k
law restrictor. In the case of a three-cylinder pump the F(Q)  F0 Yi (8)
error introduced has been found to be acceptably small 2 i1
since the `overlap' between two cylinders communicating
with the chamber at the same time is small, relative to where
the period of each suction stroke.
8Q
F0 (9)
r40

3 TRANSMISSION LINE MODELLING and

@Yi ni @ F0
The dynamics of distributed parameter piping systems are 2 Y i mi
described by hyperbolic partial differential equations. A @t rr0 @t
commonly used numerical scheme to solve these equations
Yi (0) 0 (10)
is the method of characteristics (9, 10) which has been
widely and successfully employed to model fluid transient
The constants ni and mi are given by Kagawa et al. (15)
behaviour such as waterhammer under non-cavitating and
and are reproduced in Table 1. The number of terms k
cavitating conditions. However, because the spatial discre-
should be selected according to the frequency range of
tization of the line is intrinsically linked to the time step
interest.
and speed of sound in the fluid, difficulties can be
encountered in obtaining compatibility with the small time
steps required to solve the differential equations describing
components connected to the line (11). For example, for
a time step of 102 ms and a speed of sound of 1000 m=s,
a line 10 m long would need to be divided into 1000
elements. Moreover, when variable time steps are required,
the calculation of intermediate values by interpolation Table 1 Values of ni and mi for use
becomes a further computational burden. in equation (10)
To avoid these problems, an alternative approach is
i ni mi
adopted in this study in which the Galerkin finite element
method (12, 13) is applied in the spatial variables only. This 1 2.63744 3 101 1.0
2 7.28033 3 101 1.16725
gives rise to an initial value problem for a system of 3 1.87424 3 102 2.20064
ordinary differential equations, allowing the time step to be 4 5.36626 3 102 3.92861
decoupled from the spatial interval. 5 1.57060 3 103 6.78788
6 4.61813 3 103 1.16761 3 101
The flow within a transmission line is to be calculated 7 1.36011 3 104 2.00612 3 101
under the assumptions of one-dimensional, unsteady com- 8 4.00825 3 104 3.44541 3 101
pressible flow. Independent variables of space and time are 9 1.18153 3 105 5.91642 3 101
10 3.48316 3 105 1.01590 3 102
denoted by x and t. The dependent variables are the
Proc Instn Mech Engrs Vol 211 Part I I01596 # IMechE 1997
 PRESSURE PULSATIONS IN RECIPROCATING PUMP PIPING SYSTEMS. PART 1 233

For turbulent flow, the quasi-steady term is replaced by U (x, t) u2 j (t)w

2 j (x) u2 j2 (t)w2 j2 (x) (15)

r f jQjQ P(x, t) p2 j1 (t)w

2 j1 (x) p2 j3 (t)w2 j3 (x) (16)
42 r50
Yi (x, t) yi,2 j (t)w
2 j (x) yi,2 j2 (t)w2 j2 (x) (17)
The unsteady friction term developed for laminar flow
[equation (10)] has been found by Vardy et al. (16) to work
well at Reynolds numbers up to about 104 and has been for j 0, 1, . . ., N 1. The unknown coefficients u, p
adopted here. For cases involving much higher Reynolds and yi are nodal values of U, P and Yi respectively.
numbers the model proposed by Vardy et al. (16) could be The weighting (or basis) functions w , w are piecewise
adopted. polynomials. Here, linear interpolation functions are adop-
ted:
( x
m2 x
3.1 Galerkin finite element method for xm < x < x m2
wm (x) x m2 xm (18)
0 otherwise
Finite element formulations based on the Galerkin method
(17) for time domain analysis have been presented by 
Rachford and Ramsey (12) and Paygude et al. (13) using a 1 wm (x) for xm < x < x m2
wm (x) (19)
conventional uniformly spaced grid system with two 0 otherwise
degrees of freedom (pressure and flowrate). In the work
that follows the method is presented for one degree of Nodal values are determined by inner products h:, :i
freedom (either pressure or flowrate), with an extension to defined as follows:
include the effects of frequency-dependent friction. The
calculation of pressure and flow at alternating nodes is x2 j3
sometimes referred to as interlacing. (L1 , w
2 j1 )  w
2 j1 (x)L1 dx 0 (20)
Equations (6) to (10) can be rearranged in terms of x2 j1

operator equations: x2 j3
(L1 , w
2 j1 )  w
2 j1 (x)L1 dx 0 (21)
1 @ P @U x2 j1
L1 (U , P, Yi )  2 0 (11)
c0 @ t @x
and, for e 2 and e 3,
@U @ P 1X k
L2 (U , P, Yi )  R9U Yi H 0 0 x2 j2
@t @x 2 i1
(Le , w
2 j)  w
2 j (x)Le dx 0 (22)
(12) x2 j

x2 j2
@Yi ni R @U
L3 (U , P, Yi )  Yi mi R 0 (13) (Le , w
2 j)  w
2 j (x)Le dx 0 (23)
@t 8 @t x2 j

where
Evaluation of these integrals results in a set of ordinary
rQ 8 differential equations which allow the calculation of the
U , R and H 0 r g sin 0 pressures at the odd-numbered nodes and flows at the even-
r20 rr20
numbered nodes. In this study it was decided to specify the
flow at one end of the line and the pressure at the other end
For laminar flow, R9 R. For turbulent flow, the quasi- (although it is equally possible to formulate solutions for
steady term may be approximated by the cases where either pressure is defined at both ends or
flow is defined at both ends). The pressure=flow boundary
f jU 1 j condition structure is appropriate for the test system studied
R9 (14)
4rr0 in Part 2 of this paper, as outlined later. In order to establish
each boundary condition it is necessary to solve the
where U1 is the value for U from the previous time step. equation relating to the end condition simultaneously
The transmission line is divided into 2N 1 equal ele- within the equation describing the behaviour of the at-
ments, each x in length. A minimum of five elements is tached component. This necessitates the use of a different
required. weighting function, spanning a single element rather than
The Galerkin method involves finding approximations to two, for the elements at each end of the line. The resultant
U, P and Yi of the form grid is illustrated in Fig. 3.
I01596 # IMechE 1997 Proc Instn Mech Engrs Vol 211 Part I
234 J-J SHU, C R BURROWS AND K A EDGE

Fig. 3 Pipeline discretization grid and weighting functions

For the interlaced nodes, the resultant ordinary differ- where

ential equations are of the following form:
0 1 u fu2 , u4 , . . ., u2 N gT
u
d@ A
p p f p1 , p3 , . . ., p2 N 1 gT
dt
y
y f y1,2 , y1,4 , . . ., y1,2 N , y2,2 , . . ., y k,2 N gT
0 1
1 1
B R9I A (
I)T C
B 16x 2 C N fn1 , n2 , . . ., nk gT , M fm1 , m2 , . . ., mk gT
B c20 C
B C
B 16x B 0 0 C f1, 1, . . ., 1gT
@ R R A
R2 M
I M
A [F 4M
(
I)T ] 0 1
16x 8 10 12 2
0 1 B 1 11 11 1 C
B C
u B 1 11 11 1 C
B C
3@ p A B
B
: : : : C
C
y AB : : : : C
B C
B : : : : C
1 0 B C
0
B
B 1 11 11 1CC
c20 1 @ 1 11 11 A
UIN @ C A POUT
16x 16x 2 12
0
0 1
0 1 0 1 12 2
D E B 11 11 1 C
B C
3@ 0 A H0@ 0 A (24) B 1
B 11 11 1 C
C
RM
D RM
E B : : : : C
B C
BB : : : : C
B C
B : : : : C
The symbol
defines the Kronecker product of a vector B C
B 1 11 11 1 C
Z fz1 , z2 , . . ., zk gT and a matrix G, i.e. B C
@ 1 11 11 1A
0 1 2 12 10
z1 G
B z2 G C
B C C f10, 1, 0, . . ., 0, 0, 0gT
Z
G  B .. C (25)
@ . A
zk G D f0, 0, 0, . . ., 0, 1, 10gT

Proc Instn Mech Engrs Vol 211 Part I I01596 # IMechE 1997
 PRESSURE PULSATIONS IN RECIPROCATING PUMP PIPING SYSTEMS. PART 1 235

E f1, 1, 1, . . ., 1, 1, 1gT 4 CONCLUSIONS

0 1
n1 I It has been argued that current models of plunger pumps
B n2 I C are inadequate in respect of the complex interactions which
B C
B : C take place between the pump and attached pipelines. These
FB
B :
C
C
B C arise because of the distributed parameter nature of the
@ : A pipelines and because of cavitation. A finite difference
nk I method for modelling pipelines, based on a Galerkin
method incorporating frequency-dependent friction, has
The nodal values at the boundary conditions are obtained been proposed. This approach circumvents the computa-
by applying the same methodology but using different tionally intensive demands associated with the use of the
weighting functions, as dictated by the grid shown in Fig. method of characteristics.
3. This results in two further ordinary differential equa- A new model for the pump inlet manifold has been
tions: developed to account for the presence of air pockets. The
pipeline models are readily interfaced to an existing model
d p0 1 d p1 3c20 of pumping dynamics, to allow time-domain simulations of
(u0 u2 ) (26) pressure pulsations. The effectiveness of the complete
dt 2 dt 4x
model is assessed in Part 2 of this paper.
(where u0  U IN ) and

du2 N 1 1 du2 N 3 ACKNOWLEDGEMENTS

( p2 N 1 p2 N 1 )
dt 2 dt 4x
R9 3 The work reported in this paper was undertaken as part of a
(u2 N 2u2 N 1 ) H 0 (27) research programme funded by the Engineering and Phy-
2 2
sical Sciences Research Council (Grant GR=G56423). The
(where p2 N 1  POUT ). For the case of the pump inlet line, authors are grateful for the Council's support. They also
the boundary condition at the reservoir is the constant gratefully acknowledge the co-operation and support of ICI
pressure source, p2 N 1 . At the pump, the total instanta- Research and Technology Centre, Dawson Downie Lamont
neous flow drawn into the cylinders defines U IN . However, Limited and CAT PUMPS (UK) Limited. Particular thanks
individual cylinder flows are dictated by the instantaneous are extended to Mr Lez Warren of CAT PUMPS (UK)
inlet valve differential pressure. Hence, equation (26) must Limited for his helpful remarks on the draft manuscript and
be used to provide p0 , thereby allowing the flows to be to Dr P. S. Keogh (University of Bath) for his valuable
calculated (see Appendices 1 and 2). comments on the formulation of the Galerkin method.
At the pump outlet, the total delivery flow is established
using the same procedure, which now provides UIN for the
delivery line. In the case of a valve terminating the delivery REFERENCES
line (as studied in Part 2 of this paper), equation (26) is
solved simultaneously with the equation describing the 1 Edge, K. A. and Johnston, D. N. The secondary source
valve behaviour. For a simple restrictor valve returning the method for the measurement of pump pressure ripple charac-
flow to atmospheric pressure, for example, the termination teristics. Part 2: experimental results. Proc. Instn Mech. Engrs,
pressure would be described by the square law relationship: Part C, 1990, 204 (C1), 4146.
2 Johnston, D. N. and Edge, K. A. Simulation of the pressure
ripple characteristics of hydraulic circuits. Proc. Instn Mech.
p2 N 1 cV Q2LV (28) Engrs, Part C, 1989, 203 (C4), 275282.
3 Edge, K. A. and de Freitas, F. J. T. Fluid borne pressure
where QLV is established from u2 N 1 at the previous time ripple in positive displacement pump suction lines. In Sixth
step. This approach may be extended to include valve International Fluid Power Symposium, Cambridge, 1981,
dynamics if required. paper D4, pp. 205217.
To accommodate the possibility of cavitation in the 4 Edge, K. A. and de Freitas, F. J. T. A study of pressure
suction line, three different pipeline cavitation models were fluctuations in the suction lines of positive displacements
pumps. Proc. Instn Mech. Engrs, Part B, 1985, 199 (B4),
considered, based on the work of Shinada and Kojima (18).
211217.
For the study reported in Part 2, a vaporous cavitation 5 Singh, P. J. and Madavan, N. K. Complete analysis and
model was adopted, following its successful implementa- simulation of reciprocating pumps including system piping. In
tion in a previous investigation employing the method of Proceedings of Fourth International Pump Symposium, 1987,
characteristics (10). In this scheme the pressure at any node pp. 5373 (American Society of Mechanical Engineers, New
is constrained not to fall below the vapour pressure. Further York).
details of the approach are given in reference (10). 6 Johnston, D. N. Numerical modelling of reciprocating pumps
I01596 # IMechE 1997 Proc Instn Mech Engrs Vol 211 Part I
236 J-J SHU, C R BURROWS AND K A EDGE

with self-acting valves. Proc. Instn Mech. Engrs, Part I, 1991, 1

V VTDC D2P
205(I2), 8796. 4
7 Vetter, G. and Schweinfurter, F. Elimination of disturbing ( " r  #)
and dangerous pressure oscillations caused by high pressure r2 2
positive displacement pumps. In Proceedings of Sixth Interna- 3 r(1 cos ) l 1 1 2 sin
l
tional Conference on Pressure Surges, 1990, pp. 309324
(BHRA). (29)
8 Hayward, A. T. J. Aeration in hydraulic systemsits assess-
ment and control. In IMechE Oil Hydraulics Conference, Thus, the rate of change of volume displaced by the
1961, paper 17, pp. 216224. plunger, QP dV =dt, is
9 Wylie, E. B. and Streeter, V. L. Fluid Transients, 1978
(McGraw-Hill, New York).
 
d 1 2 (r=l) cos
10 Shu, J.-J., Edge, K. A., Burrows, C. R. and Xiao, S. QP DP r sin 1 p
dt 4 [1 (r 2 =l2 ) sin2 ]
Transmission line modelling with vaporous cavitation. ASME
paper 93-WA=FPST-2. ASME Winter Annual Meeting, New (30)
Orleans, Louisiana, 28 November3 December 1993.
11 Edge, K. A., Brett, P. N. and Leahy, J. C. Digital computer The net cylinder flow is obtained from the continuity
simulation as an aid in improving the performance of positive equation
displacement pumps with self-acting valves. Proc. Instn Mech.
Engrs, Part B, 1984, 198 (14), 267274. QNET QP QIN QOUT (31)
12 Rachford, H. H. and Ramsey, E. L. Application of variational
methods to model transient flow in complex liquid transmis- From the net cylinder flow, the rate of change of cylinder
sion systems. Soc. Petroleum Engrs J., April 1977, 17(2),
pressure can be established:
151166.
13 Paygude, D. G., Rao, B. V. and Joshi, S. G. Fluid transients
dPC BEFF QNET
following a valve closure by FEM. In Proceedings of Inter- (32)
national Conference on Finite Elements in Computational dt V
Mechanics, Bombay, India, 26 December 1985, pp. 615
624. (This equation is integrated numerically to determine the
14 Zielke, W. Frequency-dependent friction in transient pipe instantaneous cylinder pressure.)
flow. Trans. ASME, J. Basic Engng, March 1968, 90, 109 To account for possible cavitation in the cylinder, the
115. cylinder pressure is monitored at each time step to detect
15 Kagawa, T., Lee, I.-Y., Kitagawa, A. and Takenaka, T. High whether the pressure falls below the vapour pressure of the
speed and accurate computing method of frequency-dependent fluid. Under cavitating conditions,
friction in laminar pipe flow for characteristics method. Trans.
Jap. Soc. Mech. Engrs, Ser. B, November 1993, 49(447), dPC
26382644. 0 (33)
dt
16 Vardy, A. E., Huang, K.-L. and Brown, J. M. B. A weighting
function model of transient turbulent pipe friction. J. Hydraul.
PC PVAP (34)
Res., 1993, 31 (4), 533548.
17 Galerkin, B. G. Series development for some cases of
equilibrium of plates and beams (in Russian). Wjestnik and the rate of change of the volume of vapour in the
Ingenerow Petrograd., 1915, 19, 897. Translation 63-18924, cylinder is
Clearinghouse, Federal Science Technological Information,
Springfield, Virginia. dVV
QNET (35)
18 Shinada, M. and Kojima, E. Fluid transient phenomena dt
associated with column separation in the return line of a
hydraulic machine press. Jap. Soc. Mech. Engrs Int. J., 1987, This is integrated numerically to give the instantaneous
30 (286), 15771596. vapour volume. Under non-cavitating conditions, VV 0.
The flow through the inlet and delivery valves is
determined from the orifice equation. This requires, in each
case, the instantaneous valve opening and the differential
pressure across it. The valve opening is established from
the numerical solution of the equation of motion (Appendix
2). The same equations apply for both inlet and delivery
APPENDIX 1 valves. In the case of the inlet valve, the upstream pressure
is that in the inlet manifold and the downstream pressure is
that in the cylinder. In the case of the delivery valve, the
Flow equations
upstream pressure corresponds to that in the cylinder and
All symbols are defined in the Notation. The instantaneous the downstream pressure corresponds to that in the delivery
volume of a cylinder is defined by the equation manifold.
Proc Instn Mech Engrs Vol 211 Part I I01596 # IMechE 1997
 PRESSURE PULSATIONS IN RECIPROCATING PUMP PIPING SYSTEMS. PART 1 237

To account for the effects of cavitation on flow through taken into account through appropriate adjustment of in
the valve, an estimate of the pressure at the vena contracta equation (29).
P9VC is obtained from
 2 !
r QVC 1 a2 1
P9VC PDOWN (36) APPENDIX 2
2 a C 2D A2 C 2Q
Valve motion
where PDOWN is the pressure downstream of the valve
under consideration. The flow through the vena contracta, The modelling of the valve dynamics closely follows
QVC , is either (QIN Au) or (QOUT Au), as appropriate. the approach of Johnston (6) and is applied to both inlet
These expressions account for the motion of the valve on and delivery valves. When the valve is partially open
the flow (6). (0 < z < zSTOP ),
The orifice area is obtained from
dz
" r  # FN ACF (PUP PDOWN ) mg c k s z FPRE
dt
p z
a 2z (A) sin 1 sin (2) (37) (42)
4 A
In order to simulate valve bounce, the valve seat and the
with the area taken to be a small, but finite, value for z < 0 end stop are modelled as very stiff spring=damper systems.
to account for imperfect sealing when the valve is seated. When z , 0,
It is assumed that the pressure at the vena contracta
cannot fall below the vapour pressure, i.e. FN ACF (PUP PDOWN ) mg

PVC max fP9VC , PVAP g (38) dz

cSTOP k STOP z k s z FPRE (43)
dt
The flow through the vena contracta is then given by
When z . zSTOP ,
p
CD a [2(PUP PVC )=r]
QVC p (39) dz
[1 (CD a=A)2 ] FN ACF (PUP PDOWN ) mg cSTOP
dt
Taking due account of fluid inertia in the vicinity of the k STOP (z zSTOP ) k s z FPRE (44)
valve seat, a non-linear differential equation for flow is
obtained:
This model can be further enhanced to account for the
 2 effect of slip-stick friction and coulomb friction, as
r QVC rlf dQVC described by Johnston (6), but such effects are not included
PUP PVC (40)
2 CD a a dt here because of the difficulty in establishing suitable
numerical values for the friction levels.
Finally, accounting for the motion of the valve itself on the The acceleration of the valve is
net flow,
du FN
dQV dQVC du (45)
A (41) dt m
dt dt dt
and the velocity is
Numerical integration yields the instantaneous flow.
The equations given above apply to just one cylinder. For dz
multicylinder pumps, the same equations are solved for u (46)
dt
each individual cylinder, with the effect of the relative
crankshaft angles on individual plunger positions being Numerical integration yields the valve opening, z.

I01596 # IMechE 1997 Proc Instn Mech Engrs Vol 211 Part I