ournal of pplied luid echanics, Vol. 1, No. 1, pp. 55-61, 2008.

Available online at www.jaImonline.net, ISSN 1735-3645.

Effects of Air Entrainment on Fluid Transients
in Pumping Systems
T.S. Lee, H.T. Low and D.T. Nguyen
Department of Mechanical Engineering, National Universitv of Singapore, Singapore, 119260
Email: mpeleetsnus.edu.sg
(Received February 27, 2007; accepted May 27, 2007)
ABSTRACT
In pumping installations, Iluid transient computations are necessary to achieve saIety, eIIiciency and
economy in design and operation. In some systems, where air content and air entrainment exist, such
computations become highly inaccurate when constant wave speed is assumed. In this paper, a numerical
model and a computational procedure have been developed to investigate the eIIects oI air entrainment on
the pressure transient in pumping systems. Free gas in the Iluid and cavitation at the Iluid vapour pressure
were modeled in the Iorm oI variable wave speed model, which was numerically solved by the method oI
characteristics. This model was tested Ior the case oI pump trips due to power Iailures. The pressure
transient results obtained by this variable wave speed model were analyzed and compared with those results
obtained by constant wave speed model and with the experimental results oI other investigators.
Keywords: Pressure transient, Air entrainment, Variable wave speed
NOMENCLATURE
a wave speed
A
1
, A
2
, A
3
constants Ior pump H-Q curve
B
1
, B
2
, B
3
constants Ior pump T-Q curve
C
1
, C
2
, C
3
constants Ior pump q-Q curve
c
l
parameter describing pipe constraint
D mean diameter oI pipe
E modulus oI elasticity
e local pipe wall thickness
] Iriction Iactor
g gravitational acceleration
H gauge piezometric pressure head
I pump set moment oI inertia
k time level
K bulk modulus oI elasticity
N
i
k
pump speed in rpm
N total number oI node points
n
p
number oI pumps in a pumping station
P pressure inside the pipe
Q Iluid Ilow rate
R C

line intercept on x-axis

S C
-
line intercept on x-axis
T pump torque
t time
V Ilow velocity
x distance along pipeline
o pipeline inclination
q pump eIIiciency
At
k
time step at k
th
time level
Ax node point distance along pipeline
c Iraction oI gas in liquid
c
0
initial air void Iraction
c
g
Iraction oI dissolved gas in liquid
density oI Iluid
Subscripts
d dissolved
e equivalent
r released
0 initial
1. INTRODUCTION
In pumping installation, Iluid transient computation
is necessary to predict excessive transient pressures
which may cause collapse oI pipelines, and damage
oI hydraulic components. In these systems, air
content and air entrainment always exist and aIIect
the pressure transient. Air in pumping systems
comes Irom three primary sources. The Iirst source
Archive of SID
www.SID.ir
T.S. Lee et al. / , Vol. 1, No. 1, pp. 55-61, 2008.
56
oI air is trapped air pockets at the top oI the pipe
cross-section at high points along pipe proIile due
to the incomplete removal oI air or the progressive
upward migration oI air pockets. The second source
oI air is Iree gas, dissolved gas in the Ilow which
contains approximately 2 volume oI water. The
third source oI air comes Irom mechanical
equipment in systems. This air may be Iorced into
the systems as a result oI Ialling jets oI sewage into
the pump sump, attached vortex Iormation, and
adverse Ilow path towards operating pump. This air
may also be admitted though packing, air valves,
air vessel, etc. under vacuum conditions.
The eIIects oI air entrainment on the pressure
transient in pumping systems were Iirstly studied
by Whiteman and Pearsall (1959, 1962) in their
pump shut-down tests. In practice, analysis oI
pressure transient with air entrainment Iace many
diIIiculties owing to lack oI inIormation such as:
the location and size oI air pockets in the pipelines;
the amount oI Iree air bubbles distributed in the
liquid, and the physical process oI release and
absorption oI air in the liquid. Numerical modelling
oI Iluid transients with air entrainment has been the
subject oI much research. The concentrate
vaporous cavity model (Brown 1968, Provoost
1976) and the air release model (Fox 1972, Wylie
1980) have been proposed in literature showing
reasonable prediction oI pressure transient
behaviours in pipeline systems. The variable wave
speed model used by Lee (1991), Borga et al
(2004) can improve the simulation oI experimental
observations in terms oI shape oI the pressure
peaks, the Irequency oI the oscillations and the rate
oI decay. When air is entrained such that the gas
void Iraction is signiIicant and two phase motion
occurs, it become necessary to introduce multi-
phase modeling (Huygens et al, 1998, Fujii and
Akagawa, 2000 and Lee et al, 2004).
This paper aims to investigate the eIIects oI air
entrainment on pressure transient in a typical
pumping system. The system consists oI a lower
reservoir, a group oI three pumps in parallel which
has a check valve in each branch, and a pipeline
system discharging into an upper reservoir. The
most dangerous case oI pressure transient is the
stop oI all three pumps in the station due to a power
Iailure. In this case, the Iollowing events take
place: (i) the Ilow rapidly diminishes to zero and
then reverses, (ii) when the Ilow reverses, positive
pressure waves propagate downstream oI the pump
towards the reservoir, and negative pressure waves
propagate upstream oI the pipe towards the suction
oI the pump, (iii) the pump rapidly losses its
Iorward rotation, and reverses, (iv) to prevent
reverse Ilow through the pump, when the Ilow
reverses the check valve is activated and closed. A
large pressure transient occurs in the pipeline. A
variable wave speed model is proposed to include
the eIIects oI Iree gas in the Iluid and cavitation at
the Iluid vapour pressure on the pressure transient
in the pipeline. This model is solved numerically
by using the method oI characteristics.
2. VARIABLE WAVE SPEED MODEL
In this paper, the variable wave speed model
assumes the existence oI Iree entrained air content
c
0
and dissolved gas content c
g
in the liquid at
atmospheric pressure. Some assumptions were
applied in this model: (i) the gas-liquid mixture is
homogeneous, (ii) Iree gas bubbles in the liquid
Iollow a polytrophic compression law with n 1.2-
1.3, and (iii) the pressure within air bubbles is in
equilibrium with the local Iluid pressure. When the
local pressure Ialls below the liquid vapour
pressure, an amount oI dissolved gas content o
r
c
g
is
assumed to release instantaneously. The local
pressure is kept constant and is equal to the vapour
pressure. When the local pressure recovers to a
value higher than the vapour pressure, an amount
oI gas content o
d
c
g
is assumed redissolved into the
liquid.
The eIIective bulk modulus K
T
oI the gas-liquid
mixture, including the pipe distensibility eIIect and
pipe constraint condition c
l
is given by (Fox 1984,
Lee 1991):
eE
D c
np K K
l
T
+ + =
c 1 1
(1)
Along the pipeline, the Iraction oI air content
depends on the local pressure and local air volume.
ThereIore, in variable wave speed model, wave
speed is not constant and is calculated Ior each
point i which has local pressure p
i
and air Iraction
content c
i
:
( )
2
1
1
1

(

|
|
.
|

\
|
+ + =
eE
D c
np K
a
l
k
i
k
i k
i w
k
i
c
c
(2)
The initial Iree air Iraction c
0
and dissolved gas
Iraction c
g
must be speciIied. Then, the initial
variable wave speed along the pipeline is computed
through the absolute pressure distribution. This
value is then used to calculate the pressure
distribution Ior the next time level. The air Iraction
Ior the next time level is given by (Lee 1991):
k
i
n
k
i
k
i k
T
p
p
c c
/ 1
1
1
|
|
.
|

\
|
=
+
+
and
0
/ 1
1
0 1
0
c c
n
k
i
k
p
p
|
|
.
|

\
|
=
+
+
(3a)
For
g
k
i
p p >
+1
and
g r
k k
T
c o c c + s
+ + 1
0
1
:
Archive of SID
www.SID.ir
T.S. Lee et al. / , Vol. 1, No. 1, pp. 55-61, 2008.
57
1 1 + +
=
k
T
k
i
c c (3b)
For
g
k
i
p p >
+1
and
g r
k k
T
c o c c + >
+ + 1
0
1
:
( )
g r
k
i
n
k
i
k
i k
i
p
p
c o c c
|
|
.
|

\
|
=
+
+
/ 1
1
1
(3c)
For
g
k
i
p p <
+1
:
( )
g r
k
i
n
g
k
i k
i
p
p
c o c c +
|
|
.
|

\
|
=
+
/ 1
1
(3d)
The new air Iraction content is then substituted into
Eq. (2) to compute the variable wave speed along
the pipeline Ior the next time level.
3. THE METHOD OF CHARACTERISTICS
AND BOUNDARY CONDITIONS
By using the method oI characteristics, the
governing equations Ior transient Ilow can be
described by the respective C

, and C
-
characteristic equations:
0
2
sin = + +
D
J J f
J
a
g
dt
dJ
dt
dH
a
g
l
o
(4, 5)
a J
dt
dx
=
(6, 7)
Then, the characteristic equations can be
approximated by Iinite diIIerence expressions:
k
R
k
i
k
R
k
i
R
t
J J
t
H H
a
g
A

+
A

+ + 1 1
0
2
, ,
sin = + +
D
J J f
J
a
g R R l
i R
R
R
o
(8)
R R k
R i
a J
t
x x
+ =
A

(9)
k
S
k
i
k
S
k
i
S
t
J J
t
H H
a
g
A

+
A

+ + 1 1
0
2
, ,
sin = +
D
J J f
J
a
g S S l
i S
S
S
o (10)
S S k
S i
a J
t
x x
=
A

(11)
where i denotes the regular x-mesh point value at
location x (iAx) and k denotes the irregular time
level corresponding to the time at t
k
E(At
k
). Time
step At
k
at each time level is determined by the
Courant-Friedrichs-Levy (CFL) criterion:
( ) | |
i i i
k
a J x k t + A = A / min Ior i 0, 1, ., N (12)
where k
i
is a constant less than 1.0.
The value oI velocity and pressure head at the point
oI interception R and S can be evaluated Irom
velocity and pressure head conditions known at A,
B, and C by a linear interpolation procedure. The
conditions at R and S are then substituted into Eqs.
(8)-(11) to compute the solution at the (k1)
th
time
level at point i.
The pump characteristics in pumping station during
pump stoppage and pump run-down can be
described by the homologous relationship Ior n
p
pumps as:
( ) ( )( )
1
0
1
2
2
1
1
1
/
+ + + +
+ =
k k
p
k k
e
Q N n A N A H
( )( )
2
1
0
2
3
/
+
+
k
p
Q n A (13)
( )( ) ( )
1
0
1
2
2
1
1
1 + + + +
+ =
k k k
p
k
e
Q N B N n B T
( )( )
2
1
0 3
/
+
+
k
p
Q n B (14)
( )( )
1 1
0 2 1
1
/ /
+ + +
+ =
k k
p
k
e
N Q n C C q
( )( )
2
1 1
0
2
3
/ /
+ +
+
k k
p
N Q n C (15)
dt d I T
e e
/ e = (16)
where H
e
k1
H
0
k1
, I
e
n
p
I, e 2tN, Q is the
Ilow rate, n
p
is the number oI pumps, A
1
, A
2
, A
3
, B
1
,
B
2
, B
3
, and C
1
, C
2
, C
3
are single pump constants.
H
e
, T
e
, and q
e
are the equivalent pump variables.
At time t
k
k ^t, N
k
, Q
k
, T
k
, H
k
, and p
k
are assumed
known. For time t
k1
(k1)^t the computation
Ior pump run down characteristic is given by the
Iollowing procedure:
1) Estimated a new pump speed:
I
t
T N N
k k k
t 2
60
1
0
A
=
+
(17)
2) Estimated H
k1
, Q
k1
by solving Eq. (13)
together with the C
-
characteristic line Eqs. (10)-
(11).
3) Obtained T
k1
and p
k1
by solving Eqs.
(14)-(15)
4) Improved estimation oI pump run-down
speed:
Archive of SID
www.SID.ir
T.S. Lee et al. / , Vol. 1, No. 1, pp. 55-61, 2008.
58
I
t T T
N N
k k
k k
t 2
. 60
2
1
0
1
1
A
|
|
.
|

\
| +
=
+
+
(18)
5) II ,N
1
k1
- N
0
k1
, larger epsilon (0.1RPM), set
1
1
1
0
+ +
=
k k
N N and repeat steps 2) to 5)
6) Else N
k1
N
1
k1
and obtain Q
k1
, H
k1
,
T
k1
and q
k1
which are deIined at t
k1
(k1)At.
When reverse Ilow is encountered in the pumps,
check valve is assumed closed. From this instant,
J
0
k1
is assumed to be zero Ior the C
-
characteristic
line at i 0 Ior all subsequent time levels. The
downstream boundary condition is assumed a
constant head reservoir, i.e. H
N
k1
constant Ior all
time levels and this is solved with the C

characteristic line Eqs. (8)-(9) Ior J

N
k1
Ior each
time level.
4. RESULTS AND DISCUSSION
In this paper, the eIIects oI air entrainment on
pressure transients generated by simultaneous
pump trip at pumping station were studied using a
pipeline contour in Fig. 1. The pumping station
uses three parallel centriIugal pumps to supply
water to an upper reservoir through a 0.985m
diameter pipe. The value oI variables downstream
oI pump (check valve position) is specially noticed
in this paper.
Figure 2 shows the general eIIects oI air
entrainment on pressure transient in pumping
system aIter pumps trip. For an unprotected
pumping station, a study oI many corresponding
numerical experiments using variable wave speed
model shows distinct characteristic diIIerences oI
pressure surges. (i) The Iirst pressure peak is higher
than that predicted by the constant wave speed
model, and the time to get the Iirst peak as well as
the time to complete a pressure surge period is
longer. (ii) The damping oI the pressure surges is
noticeably Iaster in comparison with the damping
in constant wave speed model. (iii) The pressure
surges are asymmetric with respect to the static
head, while the pressure surges Ior the constant
wave speed model are symmetric with respect to
the static head. (iv) With the variable wave speed
model, the pressure transient demonstrates longer
periods oI down-surge and shorter periods oI
upsurge in comparison with the pressure transient
oI no air content case.
Figures 3, 4, 5 and 6 show that the pressure
transient behaviour and wave speed vary
signiIicantly with the initial amount oI air within
the pumping system. When the initial air void
Iraction is increased, the pressure head oI the Iirst
pressure peak grossly increases to a maximum
value then slightly decreases; the pressure surge
periods increase. The presence oI air content in
Iluid greatly reduces the wave speed. The increase
oI initial air void Iraction leads to the decrease oI
the wave speed. The dependence oI wave speed on
the initial air void Iraction implies that the eIIects
oI air entrainment on pressure transient are more
signiIicant under low-pressure conditions, where its
volume is greater than under high-pressure
conditions. This also explains the asymmetric and
diIIerent periods oI down-surge and upsurge in
variable wave speed model. These characteristics
were also observed experimentally by previous
investigators such as Whiteman and Pearsall (1959,
1962), Dawson and Fox (1983), Jonsson (1985),
etc.
Previous studies have also given several
explanations Ior above eIIects oI air entrainment on
pressure transient. For the increase in peak
pressure, Jonsson attributed the increase in peak
pressure to the compression oI an isolated air
pocket` in the Ilow. Dawson and Fox proposed the
cumulative eIIect oI minor Ilow changes during
the transient`. Lee et al. (1991, 2004) explain by the
lapping oI the eIIects oI two Iactors: (i) the delay
wave reIlection at reservoir, and (ii) the change oI
wave speed. Free air in the Iluid increases the
eIIective bulk modulus and thus lowers the average
wave speed. As a result, the wave reIlection at
reservoir is delayed; thus a more complex variation
in pressure interaction occurs in the system,
culminating in a peak at a speciIic transient
interval. Meanwhile, the reduction oI the wave
speed oI the mixture directly causes changes in the
strength oI pressure oscillations. When the positive
eIIect Irom the delayed wave reIlection couldn`t
compensate Ior or exceed the negative eIIect Irom
the reduction oI wave speed, a suppressed pressure
peak happens. This explains why the Iirst peak
pressure increases initially and then decreases with
the increase oI the initial air void Iraction. Beside,
in the variable wave speed model, air content is an
important source oI damping beside Iriction and
minor losses. Damping eIIect oI air content is
suggested in the Iorms oI (i) direct damping owing
to the increased eIIective bulk viscosity oI the
Iluid-gas mixture, (ii) losses due to slip between air
bubbles and water, (iii) thermodynamic losses, and
(iv) indirect damping due to partial wave reIlection.
This explains the Iast damping oI the pressure
surge in the variable wave speed model in
comparison with the constant wave speed model.
5. CONCLUSIONS
A variable wave speed model has been introduced
to study the eIIects oI air entrainment on pressure
transient in pumping system Iollowing power
Iailure and check valve closure when the Ilow
reverses. Free gas in the liquid and cavitation at the
Archive of SID
www.SID.ir
T.S. Lee et al. / , Vol. 1, No. 1, pp. 55-61, 2008.
59
vapour pressure are modeled. Numerical
experiments show that where gas content and air
entrainment exist, the Iirst peak pressure is
ampliIied; the damping eIIect is Iast; and pressure
surges are asymmetric with respect to static head.
The pressure transient shows a longer period Ior
down-surge and a shorter period Ior up-surge. The
paper also introduced some explanations Ior above
eIIects oI air entrainment on pressure transient
behaviours oI pumping systems. However, the
results still remained some doubts such as the
precise physical cause oI large surge damping, the
physical process oI gaseous diIIusion in a closed
conduit subjected to unsteady Ilow, etc. This
limitation is partly due to the random nature oI
bubble nucleation, coalescence and growth in Ilow
Iields. ThereIore, much more research in this Iield
is needed.
REFERENCES
Borga, A., Ramos, H., Covas, D., Dudlick, A. and
Neuhaus, T. (2004), Dynamic eIIects oI transient
Ilows with cavitation in pipe systems, Proceedings
of the 9
th
International Conference on Pressure
Surges The practical application of surge
analvsis for design and operation, British
Hydromechanics Research Group (BHRG),
Chester, UK, pp. 605-617.
Thorley, A. R. D. (2004), Fluid transients in
pipeline svstems. a guide to the control and
suppression of fluid transients in liquids in closed
conduits, ProIessional Engineering, London,
second edition.
T.S Lee, H. T. Low and W. D. Huang (2004),
Numerical study oI Iluid transient in pipes with air
entrainment`, International Journal of
Computational Fluid Dvnamics, Vol. 18 (5), pp.
381-391.
Fujii, T. and Akagawa, K. (2000), A study oI
water hammer phenomena in a one-component
two-phase bubbly Ilow`, JSME Int Journal Series
B-Fluids and Thermal Engineering, Vol 43, Pt 3,
pp. 386-392.
Huygens, M., Verhoeven, R. and Van Pocke, L.
(1998), Air entrainment in water hammer
phenomena`, Advances in Fluid Mechanics II, Vol
21, pp. 273-282
E. B. Wylie and V. L. Streeter (1993), Fluids
Transients in Svstems, Prentice Hall, Englewood
CliIIs, NJ 07632.
T. S. Lee (1991), Numerical computation oI Iluids
pressure transient in pumping installations with air
entrainment`, International fournal for numerical
methods in fluids, Vol. 12, pp. 747-763.
L. Jonsson (1985), Maximum transient pressures
in a conduit with check valve and air entrainment`,
Proc. Int. Conf. on the Hvdraulics of pumping
stations, Manchester, BHRA, CranIield, pp. 55-76.
J. A. Fox (1984), Hvdraulic analvsis of unsteadv
flow in pipe network, Macmillan, London
P. A. Dawson and J. A. Fox (1983), Surge
pressures at Riding Mill pumping station: actual
values and theoretical predictions`, Proc. 4
th
Int.
Conf. on Pressure Surges, Bath, BHRA, CranIield,
pp. 427-445.
E. B. Wylie (1980), Free air in liquid transient
Ilow`, Proc. 3
rd
Int. Conf. on Pressure surges,
Canterbury, BHRA, CranIield, pp. 12-23.
E. B. Wylie and V. L. Streeter (1978), Fluids
Transients, McGraw-Hill, New York
G. A. Provoost (1976), Investigation into
cavitation in a prototype pipeline caused by
waterhammer`, Proc. 2
nd
Int. Conf. on Pressures,
London, HBRA, CranIield, pp. 35-43.
J. A. Fox (1972), Pressure transients in pipe
networks - a computer solution`, Proc. 1
st
Int. Conf.
on Pressure Surges, Canterbury, BHRA, CranIield,
pp. 68-75.
R. J. Brown (1968), Water-column separation at
two pumping plants`, J. Basic Eng., ASME, 521-
531.
I. S. Pearsall (1965/1966), The velocity oI
waterhammer waves`, Proc. Inst. Mech. Eng., 180,
pt. 3E, 12-20.
K. J. Whiteman and I.S. Pearsall (1962), ReIlux
valve and surge tests at a station`, Fluid Handling,
XIII, 248-250, 282-286.
K. J. Whiteman and I. S. Pearsall (1959), ReIlux
valve and surge tests at Kingston pumping station`,
Brit. Hvdromech. Res. Assoc./National Engineering
Laboratorv Joint report 1.
Archive of SID
www.SID.ir
T.S. Lee et al. / , Vol. 1, No. 1, pp. 55-61, 2008.
60
90
95
100
105
110
115
0 1000 2000 3000 4000 5000
Chainage (m)
E
I
e
v
a
t
i
o
n

(
m
)
Sump level = 92.8m
Peak level = 107.5m
nlet level = 106.0m
Resevoir level = 112.5m
Check valve
Fig. 1- Pumping station pipeline proIile.
0
5
10
15
20
25
30
35
40
0 50 100 150 200 250
Time (seconds)
P
r
e
s
s
u
r
e

h
e
a
d

(
m
e
t
e
r
s
) Constant wave speed
Variable wave speed
Fig. 2- Pressure head downstream oI pump
0
5
10
15
20
25
30
35
40
0 50 100 150 200 250
Time (seconds)
P
r
e
s
s
u
r
e

h
e
a
d

(
m
e
t
e
r
s
)
eps = 0.0001
eps = 0.001
eps = 0.01
Fig. 3- Pressure head downstream oI pump with diIIerent initial air void Iraction
25
27
29
31
33
35
37
39
0 0.005 0.01 0.015 0.02
InitiaI air void fraction
P
r
e
s
s
u
r
e

h
e
a
d

o
f

f
i
r
s
t

p
e
a
k

(
m
)
Fig. 4- Pressure head oI Iirst pressure peak with initial air void Iraction
Archive of SID
www.SID.ir
T.S. Lee et al. / , Vol. 1, No. 1, pp. 55-61, 2008.
61
0
200
400
600
800
1000
1200
1400
0 50 100 150 200 250
Time (seconds)
W
a
v
e

s
p
e
e
d

(
m
/
s
)
eps = 0.0001
eps = 0.001
eps = 0.01
Fig. 5- Wave speed downstream oI pump with diIIerent initial air void Iraction
25
225
425
625
825
1025
1225
1425
0 0.005 0.01 0.015 0.02
InitiaI air void fraction
W
a
v
e

s
p
e
e
d

(
m
/
s
)
Maximum wave speed
Minimum wave speed
Fig. 6- Wave speed downstream oI pump varies with initial air void Iraction
Archive of SID
www.SID.ir