Professional Documents
Culture Documents
Department of Mechanical Engineering, University of Tehran, P.O. Box 11365-4563, Tehran, Iran
Department of Mechanical Engineering, University of Tehran, Tehran, Iran
a r t i c l e
i n f o
Article history:
Received 3 August 2009
Received in revised form 24 October 2012
Accepted 14 December 2012
Available online 8 January 2013
Keywords:
kx Turbulence model
Water hammer
Energy dissipation
Dimensional analysis
Turbulent production
a b s t r a c t
Energy dissipation and turbulent kinetic energy production and its dissipation in unsteady turbulent pipe
ows due to water hammer phenomena are numerically studied. For this purpose, the two-dimensional
governing equations of water hammer are solved using the method of characteristics. A kx turbulence
model which is accurate for two-dimensional boundary layers under adverse and favorable pressure gradients is applied. The numerical results are in good agreement with the experimental data. Through an
order of magnitude analysis, two dimensionless parameters have been identied which can be used
for the evaluation of viscous and turbulent shear stress terms. The inuence of these non-dimensional
parameters on pressure oscillations, wall-shear-stress, dissipation rate as well as proles of velocity, turbulent production and dissipation are investigated. The non-dimensional parameter P, which represents
time scale ratio of turbulence diffusion in the radial direction to the pressure wave speed, is used to study
the structure and strength of turbulence. It is found that for the case of P 1, for which the values of the
non-dimensional groups are larger, the peaks of turbulence energy production and dissipation move rapidly away from the wall and turbulence structure is signicantly changed. For the case of P 1, for which
the values of non-dimensional parameters are smaller, these variations are found to be small.
2013 Elsevier Ltd. All rights reserved.
1. Introduction
In pressurized pipes and channels, perturbation in the uid ow
is produced due to several operations such as starting or failure in
pump and turbine and also fast opening or closing of the valve in
order to regulate the ow rate. These actions result in a series of
positive and negative pressure waves which move along the passage with wave speed. Regarding these positive and negative pressure waves, several problems such as pipe rupture, hydraulic
equipment damage and cavitations might occur. Water hammer
problems are usually simulated using one-dimensional water
hammer equations based on the quasi-steady friction model. The
main assumption of this method is that the head loss during water
hammer phenomenon is equal to the head loss obtained from the
steady ow. However, this assumption is not valid for most of the
water hammer problems due to existence of strong gradients and
reverse ows near the pipe wall.
The main focus of the present paper is energy dissipation in
water hammer ows. Energy dissipation in transient pipe ow
was studied by Silva-Araya and Choudhry [1]. They dened the
energy dissipation factor as a ratio between the energy dissipation
at any instant and the energy dissipation obtained by assuming
Corresponding author.
E-mail address: ariasi@ut.ac.ir (A. Riasi).
0045-7930/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compuid.2012.12.015
quasi-steady conditions. They used this factor to improve the accuracy of one-dimensional equations of water hammer. The attenuation in laminar water hammer problems was numerically studied
by Wahba [2]. He introduced a non-dimensional parameter which
controls the viscous shear stress term. Moreover, he investigated
the inuence of this parameter on pressure damping, instantaneous wall-shear-stress and Richardson annular effect.
To gain deeper insight of energy dissipation and structural variation of turbulence in turbulent water hammer problems, it is necessary to employ two-dimensional equations for water hammer
simulation using accurate turbulence modeling. For this purpose,
Vardy and Hwang [3], Silva-Araya and Choudhry [1,4], Pezzinga
[5], Zhao and Ghidaoui [6] and Wahba [7] used several algebraic
turbulence models. Recently, Zhao and Ghidaoui [8] used lowReynolds number k-e model of Fan et al. [9] for unsteady turbulent
pipe ows. More recently, Riasi et al. [10] used low-Reynolds
kx turbulence model to simulate unsteady turbulent pipe ows
emerging from water hammer problems.
The main objectives of the present work are: (i) to introduce the
non-dimensional groups for controlling viscous and turbulent
shear stress terms, (ii) to investigate the inuence of these nondimensional groups on pressure oscillations, wall-shear-stress,
dissipation rate and velocity proles and (iii) to study the radial
distribution of turbulent kinetic energy production and its dissipation in one wave cycle of water hammer.
125
Notations
a
Cl
D
e
E
f
Ftur
Fvis
Fw
g
gz
GCI
h
H
H0
k
L
M
MV
Nr
Nx
p
P
Q
r
Re
t
Td
TW
@H
@H a2 @u a2 1 @r t
u
0
@t
@x
g @x g r @r
@u
@u
@u
@H 1 @
@u
u
t
g
r qm
qu0 t0
@t
@x
@r
@x rq @r
@r
qu0 t0 qmt
@u
@r
u
u0
us
unum
U
t
t0
V
V0
Vol
W
x
y
z
Greek symbols
ax, bx, bk, rx, rk closure coefcients for k-x model
a, b
non-dimensional groups
e
dissipation per unit mass
U
dissipation function
c
specic weight of the uid
m
kinematics viscosity
mt
eddy viscosity
q
density
s
shear stress
sW
wall shear stress
sWss
steady state wall shear stress
X
specic dissipation rate
x1, x2 weighting coefcients
n
real positive parameter
C :
dH a du
a2 1 @rt a 1 @r s
dx
along
a
dt g dt
g r q @r
dt
g r @r
3
Fig. 1. Positive and negative characteristics for water hammer problem.
4a
126
C :
dH a du
a2 1 @r t a 1 @r s
dx
along
a
dt g dt
g r q @r
dt
g r @r
4b
s qm @u
qu0 t0 .
@r
ni1;j1
1 x1 C t1 jr t
1 x1 C t2 jrt
a
1 x2 C u1 juni1;j1 1 x2 C u2 j uni1;j
g
5a
n1
x1 C t1 jr tn1
C : Hn1
i
i;j1 x1 C t2 jr ti;j
a
x2 C u2 j un1
x2 C u1 jun1
x2 C u3 jun1
i;j1
i;j
i;j1
g
C u2 j C u1 j C u3 j
rj
j
X
Dr m
m1
r j r j1 rj =2
where Drm = rmrm1.
At a given time t and a location x along the pipe, Eqs. (5a) and
(5b) are valid for each rj and consequently, 2Nr linear simultaneous
n1
characteristics provided for 2Nr unknowns Hn1
; un1
i
i;1 ; ui;2 ; . . . ;
n1
n1
n1
n1
ui;Nr ; ti;1 ; ti;2 ; . . . ; ti;Nr 1 . A linear system of equations AX = B is
formed, where A2Nr2Nr = coefcient matrix, X2Nr1 = unknowns
vector and B2Nr1 = right hand side vector. Since the method of Vardy and Hwang [3] is very time consuming, Zhao and Ghidaoui [6]
converted the linear system of equations AX = B into the two sub
linear systems of equations with tridiagonal coefcient matrix.
2.1. Boundary conditions
ni1;j
1 x2 C u3 juni1;j1
a2 Dt 1
g r j Drj
aDtm mtj1 1
rj1
C u1 j
r j Drj rj rj1
g
aDtm mtj 1
rj
C u3 j
rj Dr j r j1 rj
g
C t1 j C t2 j
5b
ti;Nr 0:0
ui;Nr 1 ui;Nr
127
0:5 6 x2 6 1; 0:5 6 x1 6 1
or
0:0 6 x2 6 0:5; Dt 6
1
Dr 2
2 4x2 m mt
2
@k 1 @
@k
@u
mt
rm rk mt
bk kx
@t r @r
@r
@r
@x 1 @
@x
x @u 2
bx x2
rm rx mt
ax mt
r @r
@t
@r
k @r
k
mt
10
5
9
ax ; bx
3
9
1
1
;b
; rx ; rk
40 k 100
2
2
11
3. Non-dimensional groups
To obtain the non-dimensional form of turbulent water hammer equations, the following normalized variables are considered
u t x
r
t
q
; t ; x ; r ; t
; q
;
q0
U
L
D
nL=a
V
H
u0
t0
; u0 ; t0
H
aU=ng
us
us
u
13
nM
n
r
q
V
nMu
D Re r q @r
@t
@x
@r
@x
@r
|{z}
Viscousshearstressterm
L
1 @
0:01n M q u0 t0
14
D
r q @r
|{z}
Turbulentshearstressterm
In typical water hammer problems, as will be shown in the subsequence section, the Mach number is very small and thus convective
terms can be neglected in the k and x equations. To solve k and x
equations, the diffusion and the source terms are discretized using
the implicit central difference expression and the time derivative is
discretized using the rst order forward differencing scheme. Finally, two tridiagonal matrix systems are solved for the k and x in each
time step.
At the pipe axis, symmetric boundary conditions for k and x are
assumed: ok/or = 0, ox/or = 0, and for wall boundary condition, the
kinetic energy of turbulence, k, is set to zero. In order to facilitate
integration of the kx turbulence model through the viscous sublayer, one can show that the specic dissipation rate, x, must satisfy the following asymptotic solution close to the wall: x ? 6m/
by2.
In all computation, in order to resolve the variation of important
variables within the viscous sublayer; the y+ = usy/m value of the
rst grid node adjacent to the wall is less than 1.0.
L
O102 104 ; M O104 102 ;
D
MV O108 105 ; Re O102 106
@t
nM
|{z}
O104 102
2
@u
P mt
@r
12
e bk kx
The turbulent production represents the rate at which kinetic
energy is transferred from the mean eld to the uctuation eld.
15
u
@u
L
@u
@H
L M
1 @
@u
n M V t n
r q
@x
@r
@x
@r
D
D Re r q @r
|{z}
|{z}
O106 101
O108 1
L
1 @
0:01n M q u0 t0
D
r q @r
|{z}
16
O104 1
From Eq. (16), it can be concluded that: (1) the convective terms can
be neglected because of small Mach number, (2) the viscous and the
turbulent shear stress terms are controlled by the non-dimensional
groups nLM/DRe and 0.01nLM/D, respectively, (3) these terms are
much stronger in longer pipe of smaller diameter and also in
unsteady pipe ows with higher Mach number and (4) the viscous
and turbulent shear stresses are more signicant for larger values of
n (i.e., n > 4).
128
Note that higher Mach numbers occur in pipes with low elasticity modules like polyethylene pipes where the pressure wave
speed is less than the metallic pipes. Hereafter, these non-dimensional groups are dened as
c:v :
17
an
Lm
aD2
TW
Td
18
qmU=DpLD F v is
qaU=npD2 F W
LU
qu0 v 0 pLD
F tur
b 0:01n
Da qaU=npD2 F W
an
Lm
aD2
19
where Fvis = viscous force, Fw = water hammer force and Ftur = turbulent force. The reference surface for the viscous and the turbulent
forces is the internal surface of the pipe while the reference surface
for the water hammer force is the cross section area of the pipe.
Note that in the above equation, it is assumed that u0 0.1U and
t0 0.1U.
These non-dimensional groups are consistent with those already reported by Wahba [2] for the non-dimensional parameter
of laminar water hammer problems.
4. Energy dissipation
To study the energy dissipation in water hammer problems, the
energy balance should be considered for the standard test rig of
water hammer including reservoir, pipe and valve. For this purpose, a control volume is applied to the pipe (Fig. 3).
Accordingly, the balance of energy for this control volume is
d
dt
Z
c:v :
!
Z
V2
_
u
gz qdVol Q_ W
2
c:s:
V2
h
gz
2
q~
V d~
A
20
Z
Z
c:v :
tT
Wjt0 WjtT
Lm
L M
an 2n
D Re
aD
LU
L
b 0:01n
0:01n M
Da
D
!
V2
u
qdVol
2
"Z
c:s:
!
V2
u
qdVol
2
t0
#
!
2
V
V d~
A dt
h
q~
2
21
Z
c:v :
uqdVol
Z
c:v :
tT
Z
c:v :
!
V2
qdVol
2
uqdVol
t0
t0
c:v :
!
V2
qdVol
2
22
tT
@u
@u
qm qu0 t0
@y
@y
23
2
2
@u
@u
/ qm
qmt
@y
@y
|{z}
|{z}
Viscousdissipation
24
Turbulentdissipation
Eq. (24) represents the mean energy loss per unit volume and per
unit time. Also, in Eq. (24), the turbulence dissipation has been
approximated by the turbulence production. Integrating Eq. (24)
across the pipe area and along the pipe length gives the rate of energy dissipation
Fig. 3. Control volume applied to the pipe for a standard test rig of water hammer.
129
Mesh grid
385 47
500 62
600 80
0.00007
0.000053
0.000041
1.5
(H-H0)/(aV0 /g)
0.5
0
-0.5
-1
-1.5
at/L
Fig. 4. Effect of grid size on pressure oscillations at the valve.
/0
/ 2prdr dx
25
/00
/0 dt
The main purpose of verication (including the code and solution verication) is to check the numerical accuracy independent
of the physical modeling which is the subject of validation [15].
Regarding the solution verication of the numerical scheme, a
standard test rig data of Holmboe and Rouleau [16] has been considered. Three different meshes including a coarse grid of 385 47
points, a ne mesh with 500 62 grids nodes and a ner mesh
with 650 80 grid nodes, has been performed (Table 1). In all of
these meshes, the clustering parameter is set to 1.05 in the radial
direction.
The main characteristics of the unsteady turbulent ow including pressure, turbulent velocity prole and kinetic energy of turbulence are considered. The comparison among the results obtained
from these different meshes shown in Figs. 46. The details of
uncertainty estimation based on the grid convergence index (GCI)
are calculated and listed in Table 2 [15]. The numerical uncertainties are in the acceptable margin.
where U1, U2 and U3 are values of key variables for ner, ne
and coarse meshes, respectively; e21 = U2 U1, e32 = U3 U2, p is
observed order of method, e21 is error in non-dimension form,
GCI denotes the grid convergence index and unum is standard uncertainty in the simulation solution [15]. The following variables are
considered for the key variables: (a) the maximum deviation for
the valve pressure, (b) the maximum deviation for the steady state
turbulent velocity and the steady state kinetic energy, (c) the maximum deviation for the unsteady turbulent velocity and the unsteady kinetic energy at the middle of pipe and at t = 4.5L/a.
26
5.2. Validation
Z
Z
/00
uqdVol
uqdVol
c:v :
c:
v
:
tT
t0
!
!
Z
Z
V2
V2
qdVol
qdVol
2
2
c:v :
c:v :
t0
27
tT
1
0.8
0.6
0.6
r/R
r/R
0.8
0.4
0.4
0.2
0.2
0.2
0.4
0.6
0.8
1.2
1.4
0.2
0.4
0.6
0.8
u/V0
u/V0
(a)
(b)
Fig. 5. Effect of grid size on u-velocity prole: (a) steady state and (b) t = 4.5L/a.
1.2
1.4
130
3.5
3.5
2.5
2.5
2
k/u
k/u
2
1.5
2
1.5
0.5
0.5
-0.5
0.2
0.4
0.6
0.8
-0.5
0.2
0.4
0.6
r/R
r/R
(a)
(b)
0.8
Fig. 6. Effect of grid size on kinetic energy of turbulence prole: (a) steady state and (b) t = 4.5L/a.
Table 2
The details of uncertainty estimation [15].
U1
U2
U3
e21
e32
p
e21
GCI
unum
Table 3
The detail of test cases examined.
Valve
pressure
Steady state
turbulent
velocity
Unsteady
turbulent
velocity
Steady state
kinetic
energy
Unsteady
kinetic
energy
1.015
1.01226
1.00833
0.00274
0.00393
1.3747
0.27%
0.7769%
0.3884%
1.1795
1.1848
1.1908
0.0053
0.0059
0.4194
0.45%
4.8558%
2.4279%
0.7327
0.7443
0.7597
0.0116
0.0154
1.0636
1.58%
6.1744%
3.0872%
3.6466
3.5577
3.4338
0.0888
0.1239
1.2699
2.43%
7.6999%
3.8499%
2.0209
2.0627
2.1197
0.0418
0.0569
1.1789
2.06%
7.1355%
3.5677%
6166
6166
6166
Computed
Input Data for Laminar water Hammer:
L=36.09 m
D=0.025 m 3
=878.4 Kg/m 2
=0.03484 N-s/m
a=1324 m/s
Re=82
1.5
0.5
0
-0.5
-1
-1.5
-2
(H-H0)/(aV0 /g)
(H-H0)/(aV0 /g)
135
1350
6750
1.5
1
0.5
0
-0.5
-1
-1.5
10
12
b 0:01n DL M
are 0.0535 and 0.0742 for Fig. 7a and b, respectively. Since, the
standard uncertainty in the experimental data is unknown; it is
not possible to estimate the validation standard uncertainty.
Regarding Fig. 7, the events occurs in a working cycle of water
hammer (T = 4L/a) are fully described in the Section 3.
M
a n DL Re
14
-2
at/L
at/L
(a)
(b)
10
12
14
Fig. 7. Pressuretime history at the pipe midpoint: (a) laminar water hammer, Re = 82 and (b) turbulent water hammer, Re = 6166.
131
=8.4710-5, =0.00538
-5
=8.4710 , =0.00538
-4
=4.3710 , =0.0269
30
-4
=4.3710 , =0.0269
=4.3710-3, =0.269
-3
=4.3710 , =0.269
25
20
' /' ss
(H-H0)/(aV0 /g)
0.5
15
10
-0.5
5
0
-1
0
10
12
at/L
Fig. 8. Pressuretime history at the pipe midpoint for various values of a and b.
-4
-3
=4.3710 , =0.269
30
First step
Test case
number
t (s)
U00v is (J)
U00tur (J)
1
2
3
3.2
0.32
0.064
0.285
0.1185
6.61 102
1.44 102
5.335 103
1.79 103
Second step
10
W /Wss
12
Table 4
Energy dissipation in a three cycle of water hammer.
=4.3710 , =0.0269
ratio of the diffusion time scale to the wave time scale in water
hammer problems and dened as [17]
-10
-20
-30
-40
10
Fig. 11. Energy dissipation time history at the pipe midpoint for various values of a
and b.
=8.4710-5, =0.00538
20
at/L
10
12
at/L
Fig. 9. Wall-shear-stress time history at the pipe midpoint for various values of a
and b.
p
2D=2us
L=a
28
Fig. 8 shows the head pressure at the middle of the pipe for the
cases (13). It is evident that the pressure damping is much stronger for test case (1) that the values of a and b are larger than others.
Two asymptotic limits of the momentum equation can be distinguished as:
(a) When a, b ? 0 and as a result the viscous and turbulent
terms are neglected and the pressure damping tends to be
zero. On the other hand, the condition of inviscid water
hammer is achieved.
(b) When a, b ? 1 and consequently the unsteady and pressure
terms are neglected and the period of the transient ow
tends to innity and as a result the transient nature of problem is vanished in the recent condition.
0.95
0.95
0.95
0.9
0.9
0.9
0.85
r/R
r/R
r/R
0.85
0.85
0.8
0.8
0.8
0.75
0.75
0.75
0.7
0.7
-1.5
-0.5
0.5
1.5
-1.5
-0.5
0.5
1.5
0.7
-1.5
-0.5
u/V0
u/V0
u/V0
(a)
(b)
(c)
0.5
1.5
Fig. 10. Unsteady velocity proles for various values of a and b: (a) a = 4.37 103, b = 0.269, (b) a = 4.37 104, b = 0.0269 and (c) a = 8.47 105, b = 0.00538.
132
0.15
0.1
0.6
0.8
-0.1
-0.05
0.2
0.4
0.6
0.8
-0.1
-0.15
-0.15
0.1
(a)
0
4
0.4
0.15
0.05
r/R
/u
0.2
r/R
/u
/u
-0.05
0.1
0.05
0.05
0
0.2
4
0.15
P/u
0.2
4
0.2
P/u
P/u
0.2
0.4
0.6
0.8
r/R
-0.05
-0.1
-0.15
(b)
(c)
Fig. 12. Proles of turbulence production and turbulence dissipation for various values of a and b: (a) a = 4.37 103, b = 0.269, (b) a = 4.37 104, b = 0.0269 and (c)
a = 8.47 105, b = 0.00538.
are considerably changed from their steady state values and the
peaks move away from the wall. This variations are small for case
of P 1, for where the values of a and b are smaller.
6. Conclusions
In this paper, the energy dissipation and turbulent production
and dissipation in unsteady turbulent pipe ows due to water
hammer problems, has been numerically investigated. For the both
steady and unsteady turbulent pipe ow, a good agreement between the numerical and the experimental results is observed.
The results show the kx turbulence model has reasonable capability in prediction of unsteady pipe ows due to water hammer
problems. Through dimensional analysis and order of magnitude,
two non-dimensional groups are identied that can be used for
modeling the viscous and turbulent shear stresses. The result
shows that when a and b decrease: (a) the pressure damping decreases, (b) the sudden rise of wall shear stress and dissipation
function increase and (c) the effect of Richardson annular ow is
more pronounced. Moreover, the non-dimensional parameter, P,
decreases as the values of a and b increase. In case of P 1 and
for a cycle of water hammer, the turbulence structure is signicantly changed and the peaks of turbulent production and dissipation move rapidly from the wall to the pipe axis. This variations are
small for case of P 1, for which the values of a and b are smaller.
Acknowledgment
The authors wish to extend their utmost gratitude to the staff of
Hydraulic Machinery Research Institute (HMRI) for their cooperation and support.
References
[1] Silva-Araya WF, Chaudhry MH. Computation of energy dissipation in transient
ow. J Hydraul Eng 1997;123(2):10815.
[2] Wahba EM. Modelling the attenuation of laminar uid transients in piping
systems. Appl Math Model 2008;32:286371.
[3] Vardy AE, Hwang KL. A characteristic model of transient friction in pipes. J.
Hydraul Res 1991;29(5):66985.
[4] Silva-Araya WF, Chaudhry MH. Unsteady friction in rough pipes. J Hydraul Eng
2001;127(7):60718.
[5] Pezzinga G. Evaluation of unsteady ow resistances by quasi-2D or 1D models.
J Hydraul Eng 2000;126(10):77885.
[6] Zhao M, Ghidaoui MS. Efcient quasi two dimensional model for water
hammer problems. J Hydraul Eng 2001;129(12):100713.
[7] Wahba EM. RungeKutta time-stepping schemes with TVD central
differencing for the water hammer equations. Int J Numer Methods Fluids
2006;52(5):57190.
[8] Zhao M, Ghidaoui MS. Investigation of turbulent behavior in pipe transient
using a ke model. J Hydraul Res 2006;44(5):68292.
133
[15] Standard for verication and validation in computational uid dynamics and
heat transfer. The American Society of Mechanical Engineering; 2009.
[16] Holmboe EL, Rouleau WT. The effect of viscous shear on transients in liquid
lines. Trans ASME 1967;89(1):17480.
[17] Ghidaoui MS, Mansour SGS, Zhao M. Applicability of quasisteady and
axisymmetric turbulence models in water hammer. J Hydraul Eng
2002;128(10):91724.
[18] Kucienska B. Friction relaxation model for fast transient ows. PhD thesis.
Belgume: University of Catholique de Louvain; 2004.