Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Holistic Unsteady-Friction Model for Laminar Transient

Flow in Pipeline Systems


SangHyun Kim, M.ASCE1

Abstract: This paper proposes a holistic unsteady-friction model for the transient simulation of the condition of laminar initial flow. The
effect of wall shear stress is considered through the introduction of radial-velocity distribution with kinematic viscosity from a simplified
Downloaded from ascelibrary.org by Louisville, University Of on 09/24/13. Copyright ASCE. For personal use only; all rights reserved.

two-dimensional Navier-Stokes equation. The model simultaneously addressed the frequency-dependent friction and the unsteady friction
associated with local and convective acceleration on the platform of the impulse-response method. A widely used hypothetical example
demonstrated that the proposed model appropriately and simultaneously expressed the effect of pressure damping from both the radial-
velocity variation with viscosity and acceleration terms. The nonlinear mutual effects were examined between two distinct energy-dissipation
mechanisms by comparing the behavior of the pressure head between the proposed and existing models. Comparisons of the experimental
results illustrated that the combined model exhibited predictive and fitting capabilities for the pressure-head oscillation and shape for both the
end and middle positions of a reservoir-pipeline-valve system. The proposed model showed significantly improved predictability over
the frequency-dependent friction model for the early pressure associated with a water hammer. The comparison of the performance between
the developed model and the traditional method of characteristics approach demonstrated the robustness of the proposed method in both its
computational efficiency and its representation of a real-life system. DOI: 10.1061/(ASCE)HY.1943-7900.0000471. © 2011 American
Society of Civil Engineers.
CE Database subject headings: Transients; Pipe flow; Friction; Hydraulic models; Pipelines.
Author keywords: Transients; Pipe flow; Impulse-response method; Unsteady-friction model.

Introduction two approaches have improved the computational cost and the
shape of transient flow (Vardy and Brown 2004; Pezzinga
The modeling of fluid transients in pipeline systems has been 2009).
simulated to improve their accuracy and efficiency and has been Although the unsteady shear stress can be approximated from
an ongoing and challenging issue for the previous few decades the instantaneous acceleration as a function of the Reynold number
(Chaudry 1987; Wylie and Streeter 1993). Unsteady-friction models (Vardy and Brown 2003; Brunone et al. 2004), the restriction in
of water hammers in pipeline systems have been developed using the timescales of wave propagation and valve closure, as well as
two-dimensional (2D) models to accurately realize the energy- the unachievable assumption of uniform acceleration for the ball
dissipation mechanism of the friction effect. However, the computa- valve maneuver in most real-life systems, leads to the notable var-
tional burden of 2D models causes difficulty in the application of iations in predictions of the water hammer between the two differ-
such models in field pipeline systems (e.g., Brunone et al. 1995; ent approaches (Bergant et al. 2001; Vítkovský et al. 2006a). Most
Silva-Araya and Chaudry 1997; Pezzinga 1999, 2000, 2002; Zhao unsteady-friction models possess differing behaviors, and these
and Ghidaoui 2006). Approximations of one-dimensional (1D) differences are more pronounced between the acceleration-based
unsteady friction have been explored in two different ways: through models and convolution-based approaches (Adamkowski and
the introduction of weighting functions for the past flow-velocity Lewandowski 2006). Even in several recent investigations, the de-
history (Zielke 1968; Trikha 1975; Schohl 1993; Vítkovský et al. velopment of unsteady-friction models has evolved separately in
2006b), and based on the instantaneous acceleration of flow two different directions (Abreu and Almeida 2009; Pezzinga
(Daily et al. 1956; Brunone et al. 1991; Ramos et al. 2004). Exper- 2009; Vardy and Brown 2007, 2010). The difficulty in merging
imental studies have evaluated several unsteady-friction models two distinct philosophies into an integrated modeling platform
related primarily to these two distinct methods (Bergant et al. 2001; could likely be responsible for contemporary research initiatives.
Adamkowski and Lewandowski 2006). The theory of extended irre- Considering the mutually exclusive strengths and weaknesses that
versible thermodynamics proposed by Axworthy et al. (2000) pro- exist among the acceleration-based approach, weighting function
vided theoretical justification for instantaneous acceleration-based models, and even the 2D models, the simultaneous consideration
modeling. Attempts to relax the limitations imposed by these of distinct mechanisms is required to comprehensively address
distinct energy-loss processes. In actuality, a combined approach
1
Professor, School of Civil and Environmental Engineering, Pusan merging the convolution-based model (Zielke 1968) and an accel-
National Univ., Pusan 609-735, South Korea (corresponding author). eration term with a momentum correction factor was explored on
E-mail: kimsangh@pusan.ac.kr the basis of the method of characteristics (MOC) (Bergant et al.
Note. This manuscript was submitted on October 6, 2010; approved on
June 6, 2011; published online on June 8, 2011. Discussion period open
2003). However, a number of engineering approaches based on
until May 1, 2012; separate discussions must be submitted for individual numerical schemes for solving transients such as water-hammer
papers. This paper is part of the Journal of Hydraulic Engineering, problems [such as MOC and Godunov-type schemes (GTS)] are
Vol. 137, No. 12, December 1, 2011. ©ASCE, ISSN 0733-9429/2011/ severely restricted by the issue of pipeline discretization (e.g.,
12-1649–1658/$25.00. Ghidaoui et al. 1998; Wood et al. 2005; Leon et al. 2008) and also

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / DECEMBER 2011 / 1649

J. Hydraul. Eng. 2011.137:1649-1658.


cause the numerical dissipation in the instantaneous acceleration- The k in Eq. (4) includes a momentum correction factor, η, is not
based model (Vítkovský et al. 2006b). As an alternative method equal to zero, and also varies with time for the 1D model (Brunone
for pipeline transient modeling, based on discretization schemes, et al. 1991).
the impulse-response method (IRM) can address the effect of The acceleration components and mean piezometric head in
frequency-dependent friction in the frequency domain (Holmboe Eq. (4) can be represented by ∂v=∂t, ∂v=∂x and ∂h=∂x, as the radial
and Rouleau 1967; Suo and Wylie 1989). The incorporation of distribution of velocity and head are integrated over cross-sectional
acceleration-based unsteady friction by Kim (2005) showed greater area and then divided by the corresponding area
potential than the traditional discretization methods in computa- R
tional cost and system representation. ∂v ∂h 32ν R v2πrdr ∂v
ð1 þ k1 Þ þ g þ 2 0  ak 2 ¼0 ð5Þ
In this study, a combined unsteady-friction model was devel- ∂t ∂x D πR2 ∂x
oped on the platform of IRM, comprising elements from both
where k 1 , k 2 = momentum-corrected unsteady-friction decay coef-
the frequency-dependent friction model and the instantaneous
ficients (see Appendix).
acceleration-based approach. The proposed model was tested for
However, if the radial-velocity distribution with a kinematic vis-
Downloaded from ascelibrary.org by Louisville, University Of on 09/24/13. Copyright ASCE. For personal use only; all rights reserved.

a hypothetical pipeline system and an experimental reservoir-


cosity is considered, then the k ¼ η þ φ (see Appendix) could be
pipeline-valve (RPV) system. Comparisons of MOC simulations
replaced as a parameter, C k , which represents the effect of the unre-
and experimental results are presented to discuss the behavior of
alistic assumption of axial symmetric pipe flow, the effect of a small
the combined model versus other existing approaches. Example
inertial length near the orifice discharging into the reservoir at the
simulations with randomized parameters have demonstrated the
end of the pipeline system (Funk et al. 1972), and the contributions
robustness of the proposed model.
from the second and fourth terms in Eq. (1).
A new momentum equation is proposed by combining the con-
tributions of the acceleration-based terms and the radial distribution
Development of Combined Unsteady-Friction Model
of velocity with a kinematic viscosity as two distinct friction
The following 2D Navier-Stokes equation for axisymmetric tran- considerations
sients in a horizontal circular pipe with a constant cross section  
∂v ∂v ∂h 1∂ ∂v
expresses the equations of momentum and continuity: ð1 þ C k Þ  C k a þ g  ν r ¼0 ð6Þ
∂t ∂x ∂x r ∂r ∂r
 2  
∂v ∂v ∂h 4∂ v 1 ∂ ∂v
þv þg ν· þ r ¼0 ð1Þ Considering the difference in contributions to wave oscillation
∂t ∂x ∂x 3 ∂x2 r ∂r ∂r
damping and phase for the local acceleration and convective accel-
eration terms (Ramos et al. 2004), the generic form of Eq. (6) can
∂h a2 ∂v
þ ¼0 ð2Þ be expressed as follows:
∂t g ∂x R
∂v ∂v ∂h 32ν 0R v2πrdr
where v = velocity and h = piezometric head, which are functions of ðCk1 þ 1Þ  Ck2 a þ g þ ð1  k 3 Þ 2
∂t ∂x ∂x D πR2
time (t), axial distance (x), and radial distance (r); ν = kinematic  
1∂ ∂v
viscosity; g = gravitational acceleration; and a = wave speed.  k3 ν r ¼0 ð7Þ
One of the most widely used approximations for Eq. (1) is the r ∂r ∂r
following simplified momentum equation proposed by Zielke
(1968): where k3 ¼ 1 or 0, depending on the model in Table 1. Table 1
illustrates that Eq. (7) can address most unsteady-friction models.
 
∂v ∂h 1∂ ∂v To develop the most general development from Eq. (7), the
þg ν r ¼0 ð3Þ parameter conditions for Model 6, such as C k1 ≠ C k2 ≠ 0 and
∂t ∂x r ∂r ∂r
k3 ¼ 1, were used in Table 1. The momentum equation can be
The other 1D modeling approach was based on acceleration expressed as follows:
terms (Daily et al. 1956; Brunone et al. 1991)  
  ∂v ∂v ∂h 1∂ ∂v
∂V ∂H 32ν ∂V ∂V ðC k1 þ 1Þ  C k2 a þ g  ν r ¼ 0 ð8Þ
þg þ 2 V þk a ¼0 ð4Þ ∂t ∂x ∂x r ∂r ∂r
∂t ∂x D ∂t ∂x
The velocity, v, and the piezometric head, h, in Eqs. (2) and (8)
where V = average velocity; H = mean piezometric head; and k = can be divided by two terms: the average and the oscillatory. Intro-
friction decay coefficient, which can be determined either analyti- ducing the steady oscillatory flow assumption can eliminate the
cally by using Vardy’s shear decay coefficient (Vardy and Brown mean components for velocity and head, such as ∂v=∂x ¼ ∂v=∂t ¼
2003) or experimentally by using the trial-and-error method (Wylie ∂ h=∂t ¼ 0, where v and h are the mean velocity and head, respec-
1996; Bergant et al. 2001; Adamkowski and Lewandowski 2006). tively. A linearization procedure for a steady laminar flow provides

Table 1. Unsteady Friction Models Addressed by Generic Momentum Equation


Model number Condition Time domain model (discretized model) Frequency domain model
1 C k1 ≠ C k2 ¼ 0, k3 ¼ 0 Daily (1956) Eqs. (7) or simplified model from Kim (2005)
2 Ck1 ¼ C k2 ¼ 0, k3 ¼ 1 Zielke (1968) Eq. (7), Suo and Wylie (1989)
3 C k1 ¼ Ck2 ≠ 0, k3 ¼ 0 Brunone (1991) Eq. (7), Kim (2005)
4 C k1 ≠ C k2 ≠ 0, k 3 ¼ 0 Ramos et al. (2004) Eq. (7) or modified model from Kim (2005)
5 C k1 ¼ Ck2 ≠ 0, k3 ¼ 1 Not applicable Eq. (7)
6 C k1 ≠ C k2 ≠ 0, k 3 ¼ 1 Not applicable Eq. (7)

1650 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / DECEMBER 2011

J. Hydraul. Eng. 2011.137:1649-1658.


∂ h=∂x ¼ ν=ðgÞ · 1=r · ∂=∂rðr · ∂v=∂rÞ, and Eqs. (2) and (8) can be The particular solutions for the oscillatory components of aver-
expressed in terms of the oscillatory discharge and head, v0 and h0 as age head and discharge, respectively, can be expressed as follows:
follows:
H 0 ¼ C A · eγA x · est þ C B · eγB · est ð12Þ
 
∂h0 ðC k1 þ 1Þ ∂v0 ∂v0 ν 1 ∂ ∂v0
þ  CM A  r ¼0 ð9Þ
∂x g ∂t ∂x g r ∂r ∂r CC A s γA st CC B s γB st
Q0 ¼  e ·e þ e ·e ð13Þ
γA γB
∂v0 C ∂h0
þ ¼0 ð10Þ
∂x A ∂t
pffiffiffiffiffiffiffi as a complex variable s ¼ σ þ iω;
where s = Laplace variable
σ = decay factor; i ¼ 1; and ω = frequency.
where C M ¼ C k2 a=gA; and C ¼ gA=a2 is the capacitance. The integration constants C A and C B , which are functions of
The application of further linearization procedures to Eqs. (9) H U = upstream complex head, and QU = upstream complex dis-
Downloaded from ascelibrary.org by Louisville, University Of on 09/24/13. Copyright ASCE. For personal use only; all rights reserved.

and (10), as well as the wave-transmission relationship obtained charge, can be written as follows:
from the Fourier transformations of the motion equations for un-
steady flow, can be implemented in the frequency domain. Z CA
The series impedance, ZðsÞ, is defined as the effect of volume CA ¼ ðH  Z CB · QU Þ ð14Þ
Z CA þ Z CB U
flow on the pressure gradient (Brown 1962)

∂HðsÞ Z CB
ZðsÞ · QðsÞ ¼  ð11Þ CB ¼ ðH þ Z CA · QU Þ ð15Þ
∂x Z CA þ Z CB U
where QðsÞR = Fourier transformation of the volumetric flow;
QðsÞ ¼ 2π 0R r · vs dr, where vs = Fourier transform of v; and where Z CA ¼ γA a2 =ðgAsÞ; and Z CB ¼ γB a2 =ðgAsÞ.
HðsÞ is the corresponding transformation of h. The propagation constants γA and γB , respectively, are expressed

8 2sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 39
 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1ffi =
1< s 2 2J ðiR sð1 þ C Þ=ν Þ
γA ¼ CC M s þ 4 ðCC M sÞ2 þ 4 2 1  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
k1 5 ð16Þ
2: a iR sð1 þ C k1 Þ=ν · J o ðiR sð1 þ C k1 Þ=vÞ ;

8 2sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 39
 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1ffi =
1< s 2 2J ðiR sð1 þ C Þ=ν Þ
γB ¼ CC s þ 4 ðCC M sÞ2 þ 4 2 1  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1
p k1
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 5 ð17Þ
2: M a iR sð1 þ C k1 Þ=ν · J o ðiR sð1 þ C k1 Þ=vÞ ;

where J 0; J 1 = Bessel functions of the first kind for zeroth and first Although the hydraulic impedance defined in Eq. (20) for an
orders, respectively. RPV system can be expressed as a function of the upstream char-
The relationship between the upstream complex head and the acteristics, the water hammer generated from the downstream valve
discharge to a downstream complex head for a homogeneous pipe- should include downstream information between the designated
line segment can be expressed in the following equation: point, x, and the transient initiation location. The response func-
tions for the transient computation need to be defined by addressing
HðxÞ ¼ C A eγA x þ C B eγB x ð18Þ the comprehensive characteristics of the pipeline system. Assuming
a constant upstream head reservoir in an RPV system, the pressure-
where x = distance between two points. head response at the downstream valve, or the head and discharge
The complex discharge can be expressed as follows: response functions at distance x from the downstream valve in Fig. 1
can be expressed as follows:
QðxÞ ¼ 
C A γA x C B γB x
e þ ð19Þ Z 
e 1 ∞ Z Z er A L þ Z Z eγB L
Z CA Z CB r DH ðtÞ ¼ R CA CB CA CB
e iωt
dω ð21Þ
π 0 Z CA eγB L þ Z ZB eγA L
On the basis of the definition of hydraulic impedance, i.e.,
ZðxÞ ¼ HðxÞ=QðxÞ, the impedance for a RPV system (Fig. 1) Z 
1 ∞ Z eγB x þ Z eγA x
can be defined as follows: r xH ðtÞ ¼ R CA CB
γA x γB x
π 0 ðZ CA þ Z CB Þe e
Z CA Z CB erA l þ Z CA Z CB eγB l Z CA Z CB erA L þ Z CA Z CB eγB L
ZðxÞ ¼ ð20Þ
Z CA eγB l þ Z ZB eγA l ·
Z CA eγB L þ Z ZB eγA L
 
Z Z ðerA x  eγB x Þ iωt
where l = distance from upstream reservoir to any designated point þ CA CB e dω ð22Þ
along the pipeline. ðZ CA þ Z ZB ÞeγA x eγB x

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / DECEMBER 2011 / 1651

J. Hydraul. Eng. 2011.137:1649-1658.


Z 
1 ∞ Z CA eγA x þ Z CB eγB x erBA x  eγA x
rxQ ðtÞ ¼ R γA x γB x 
π 0 ðZ CA þ Z CB Þe e ðZ CA þ Z ZB ÞeγA x eγB x
 
Z CA Z CB erA L þ Z CA Z CB eγB L iωt
· e dω ð23Þ
Z CA eγB L þ Z ZB eγA L

where L = distance of the pipeline in RPV system.


The pressure head at the downstream valve and the head and
discharge variations at point x can be obtained by convolving
the response functions with a discharge pulse at the downstream
valve, written
Z t
Downloaded from ascelibrary.org by Louisville, University Of on 09/24/13. Copyright ASCE. For personal use only; all rights reserved.

H D ðtÞ ¼ H 0 þ r Dh ðt  τ ÞΔqD ðτ Þdτ ð24Þ


0

Z t
H x ðtÞ ¼ H 0 þ r xh ðt  τ ÞΔqD ðτ Þdτ ð25Þ
0

Z t
Qx ðtÞ ¼ Q0 þ r xq ðt  τ ÞΔqD ðτ Þdτ ð26Þ
0

where H 0 = the steady initial pressure head; Q0 = initial mean flow


rate; and ΔqD ðτ Þ = the discharge impulse obtained by the appli-
cation of discrete convolution to the valve boundary condition Fig. 2. Normalized pressure heads using steady friction, frequency-
(Suo and Wylie 1989; Wylie and Streeter 1993). dependent (FD) friction, instantaneous acceleration-based (IAB) un-
steady friction, and combined unsteady friction: (a) at the end point
of the pipeline system; (b) at the middle point of the pipeline system
Application Examples with Discussions

Hypothetical Pipeline for Hagen-Poiseuille Flow


capacitance (C) was gA=a2 , the inertance (L) was 1=gA, and the
A widely accepted hypothetical pipeline example (Fig. 1) was used resistance (Ω) was 32ν=ðgAD2 Þ for laminar flow (Wylie and
to test the application of the proposed unsteady-friction model Streeter 1993).
(Wylie and Streeter 1993; Prado and Larreteguy 2002). The pipe Considering frequency-dependent
pffiffiffiffiffiffiffi (FD)p friction,
ffiffiffiffiffiffiffi theppropagation
ffiffiffiffiffiffiffi
length was 36 m with a diameter of 2.54 cm; the wave speed constant, γ ¼ s=af1  ½2J 1 ðiR s=ν Þ=½iR s=ν J 0 ðiR s=ν Þg0:5 ,
was 1;324 m=s; kinematic viscosity was 3:97ð105 Þ m2 =s; and pffiffiffiffiffiffiffi
and the characteristic impedance, Z c ¼a=ðgAÞf1½2J 1 ðiR s=ν Þ=
the initial steady velocity was 0:128 m=s. An upstream reservoir pffiffiffiffiffiffiffi pffiffiffiffiffiffiffi
½iR s=ν J 0 ðiR s=ν Þg0:5 , were used to represent the effect of
had a constant head of 17.28 m, and the water hammer was gen-
erated through the rapid closure of a control valve in 0.0034 s kinematic viscosity on the variation in the velocity in the radial
direction (Suo and Wylie 1989). Two propagation constants,
at the downstream end. The impulse-response functions were de- pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
fined in the frequency domain, with the maximum frequency range γ1;2 ¼ 0:5 · f∓MCs þ ½ðMCsÞ2 þ 4CsðLs þ RÞg, where M ¼
truncated to 1;850 rad=s. The number of samples necessary for a ak=ð2gAÞ, and two characteristic impedances, Z C1;C2 ¼ γ1;2 =Cs,
fast Fourier transform is 2,048. were defined for acceleration-based (AB) unsteady friction (Kim
Figs. 2(a) and 2(b) show the normalized head variations com- 2005). Considering the revised Vardy’s shear decay coefficient,
puted using several friction models for water-hammer events at C  ¼ 12:86=Rκ , where κ ¼ log10 ð15:29=R0:0567 Þ (Vardy and
the end and middle of an RPV system, respectively. The char- Brown 2003) for laminar flow (R ¼ 2;000), k ¼ 0:041 was used
acteristic impedance, Z c ¼ γ=ðCsÞ, and the propagation constant,
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi for both of the AB unsteady-friction models. The combined friction
γ ¼ CsðLs þ ΩÞ, were used for the steady friction where the coefficients from Eq. (8), Ck1 and C k2 , were determined through
trial and error. This was done because the combined model simul-
taneously considered the effect of kinematic viscosity on variation
in the velocity in the radial direction and the acceleration-based
approximation. Considering the heuristic practices for the experi-
mental test in this study, C k1; C k2 ¼ 0:02 (Model 5 in Table 1) was
used for the combined unsteady-friction modeling. Both Figs. 2(a)
and 2(b) indicate that the combined friction model showed slightly
more damping than the FD friction model. This may compensate
for the minor discrepancy between the frequency-dependent fric-
tion and the experimental head fluctuations in Zielke (1968) and
Prado and Larreteguy (2002), which can be attributed to the addi-
tional effect of AB friction. The association among the local and
convective acceleration terms and the radial-velocity distribution of
kinematic viscosity was investigated to provide an understanding of
Fig. 1. Schematic of the reservoir-pipeline-valve system
the nonlinear interactions of the two distinct energy-dissipation

1652 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / DECEMBER 2011

J. Hydraul. Eng. 2011.137:1649-1658.


mechanisms in Eq. (8). The associative effect of the AB friction dHðtÞν ¼ ðHðtÞc;r  HðtÞc;25%k Þ  ðHðtÞAB;rk  HðtÞAB;25%k Þ
term on the FD friction contribution in Eq. (8) can be evaluated ð28Þ
by comparing the FD friction (Suo and Wylie 1989) and the com-
bined model. where HðtÞc;25%k = computed head variations using a combined
Differences in the head variations among the referenced and var- model with increased or decreased decay coefficients, such as
ied coefficients were computed for the FD friction and combined C k1; C k2 ¼ 0:025 or 0.015; and HðtÞAB;rk and HðtÞAB;25%k =
models as follows: the corresponding head variations with AB friction models.
Fig. 3(b) presents the normalized head differences at the down-
dHðtÞν ¼ ðHðtÞc;r  HðtÞc;25%ν Þ  ðHðtÞFD;rν  HðtÞFD;25%ν Þ stream valve between the combined and AB friction models (Kim
2005) for head differences between varied decay coefficient k by
ð27Þ
25% and the reference decay coefficient, respectively. The refer-
ence kinematic viscosity, ν ¼ 3:97ð105 Þ m2 =s, was used for both
where HðtÞc;r = head variations computed using the combined models. The effect of FD friction on AB friction in Eq. (8) was
model with referenced parameters, such as ν ¼ 3:97ð105 Þ m2 =s
Downloaded from ascelibrary.org by Louisville, University Of on 09/24/13. Copyright ASCE. For personal use only; all rights reserved.

significantly different from the effect of AB friction on FD friction.


and C k1; C k2 ¼ 0:02; HðtÞc;25%v = computed head variations using Substantial differences were found in close proximity to each wave
the combined model with increased or decreased kinematic viscos- reflection time, which were multiples of 2L=a, such as 0.0544 and
ities, such as ν ¼ 4:96ð105 Þ or 2:98ð105 Þ m2 =s; and HðtÞFD;rv 0.1088 s, in Fig. 3(b). This was because the local acceleration term
and HðtÞFD;25%v ¼ corresponding head variations with the FD caused by AB friction can cause phase movement in the wave trans-
friction models. mission. In fact, the difference in the appearance of the first wave
Fig. 3(a) illustrates the normalized head differences at the was observed between C k1; C k2 ¼ 0:015 and C k1; C k2 ¼ 0:025 in
downstream valve between the combined and FD friction models Fig. 3(b). Both Figs. 3(a) and 3(b) indicate not only significant
for the head differences between the varied viscosities and refer- nonlinear mutual interactions between FD friction and AB friction
ence viscosity, respectively. The contributions of viscosity with but also a substantial variation in the energy-dissipation mechanism
a radial-velocity distribution to the head dissipation process are between the two distinct friction-modeling approaches. The approx-
similar in the first shock wave of the water-hammer event; however, imation using the acceleration-based approach to the frequency-
the difference generated from the AB decay mechanism appears to dependent approach (Vardy and Brown 2003) appears to be unfeasible,
be amplified as the elastic wave is fed back to the downstream considering the difference between the two distinct mechanisms in the
location. The head variations of FD friction caused by AB friction amplitude and phase, as shown in Figs. 3(a) and 3(b).
are similar within multiples of 2L=a periods, such as 0.0544 and
0.1088 s, in Fig. 3(a). Experimental Test for the Combined Friction Model
The effect of FD friction on the AB friction term can be evalu- A pilot-scale pipeline system equipped with multiple water tanks,
ated as follows: valves, and a pressurized chamber along a pipeline was designed
and constructed for use in this experiment [Fig. 4(a)]. The length of
the total pipeline system was 151.29 m. The internal diameter of the
stainless steel pipeline was 0.02 m. The distance between the pres-
surized water tank and the internal valve was 87.22 m, which is
identical to the length in the pipeline water-hammer test. To accu-
rately secure the designated steady flow, water was constantly sup-
plied to water tank A, with free surface elevations maintained in the
water tanks, thereby allowing overflow through multiple valves
(16 for each water tank) in corresponding water tanks [see Fig. 4(b)].
A steady flow was controlled by the difference in the opening valve
elevations between two water tanks. Steady flow rates were calcu-
lated using multiple volumetric measurements. No difference was
found in the steady flow rates during the three tests. The water
hammer was introduced through rapid closure of the ball valve
(in approximately 0.1 s). Pressure transducers were used to monitor
the head variations at the point of the ball valve and at 65.43 m up-
stream of the ball valve [see Fig. 4(a)]. Pressure data were recorded
1,000 times per second using a data acquisition card 6024E from
National Instruments.
Fig. 5(a) compares the time series of the measured pressure head
and the predictions of the AB, FD, and combined friction models at
a downstream valve on the platform of the IRM. A discharge for the
initial steady flow of 3:11ð105 Þ m3 =s was used, with a kinematic
viscosity of ν ¼ 1:12ð106 Þ m2 =s. The Reynolds number for
initial steady flow (Ro ) was 1,774. The wave speed was esti-
mated to be approximately 1;410 m=s using the measurement of
Fig. 3. Differences in normalized head differences: (a) between
the wave propagation time for two designated pressure trans-
combined and FD friction models for increasing and decreasing the
ducers. Considering the extended simulation time and prediction
viscosity by 25% from that of the reference viscosity; (b) between
of temporal resolution, the maximum frequency range was trun-
the combined and IAB friction models for increasing and de-
cated to 628:31 rad=s, with a number of the fast Fourier trans-
creasing the decay coefficient by 25% from that of the reference decay
form of 8,192. The decay friction coefficient k ¼ 0:041 in
coefficient
laminar flow was used for the AB friction model prediction

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / DECEMBER 2011 / 1653

J. Hydraul. Eng. 2011.137:1649-1658.


Downloaded from ascelibrary.org by Louisville, University Of on 09/24/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 4. (a) Schematic of pilot-scaled pipeline system; (b) pictures of the water tanks with multiple valves

(Vardy and Brown 2003). Considering the contributions from two unsteady models. Both the FD friction and the combined models
distinct mechanisms, several heuristic estimations were performed accurately predicted pressure variations; however, the combined
for the AB decay coefficient in the combined model using model only permitted calibration for closer fitting to measurements.
C k1; C k2 ¼ 0:02. Figs. 6(a) and 6(b) compare the measured and predicted pressure-
Fig. 5(a) indicates that all the friction models demonstrated pos- head variations at a distance of 65.42 m upstream of the location
itive agreement for measurements in the highest and lowest pres- of the valve. The combined model demonstrated flexibility to better
sure heads for the first reaction of the pressure wave; however, fit the measurements than the FD or AB models, as shown in
discrepancies among the measured and predicted traces tended Fig. 6(b).
to be amplified in the later time steps. An enlargement of Fig. 5(a), To compare the performance of predicting pressure between the
between 1.0 and 1.4 s, is presented in Fig. 5(b). Predictions of the FD friction and the combined models, the sum of the square ratio
pressure wave using the AB friction model provided less useful (SSR) using measured pressure head and calibrated pressure heads
results than the other models, which is similar to the results is defined as follows:
reported by Adamkowski and Lewandowski (2006). Brunone
et al. (2004) demonstrated that the constant k cannot be used in PED P8l=a
½hm ðtÞ  hCM ðtÞ2
Brunone’s model. This was mainly because the radial-velocity SSR ¼ PMD Pt¼1
8l=a
ð29Þ
t¼1 ½hm ðtÞ  hFD ðtÞ
ED 2
distribution was not properly addressed in the structure of 1D MD

1654 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / DECEMBER 2011

J. Hydraul. Eng. 2011.137:1649-1658.


where hm ðtÞ = measured pressure head; hCM ðtÞ = calibrated pres-
sure head with the combined model; hFD ðtÞ ¼ computed pressure
head with the FD model; subscripts MD and ED indicate the meas-
urement points at the middle and end points of the pipeline system
in Fig. 4(a), respectively; and 8l=a represents the duration of cal-
ibration, which was deliberately assigned to be twice the theoretical
period of the pipeline system (Wylie and Streeter 1993) to facilitate
the feasible calibration of the pipeline parameters, such as friction
and leakage.
Model 5 and Model 6 in Table 1 were both used to estimate the
SSR, with the experimental results obtained from the pipeline sys-
tem in Fig. 4(a). The SSR for Models 5 and 6 were 0.787 and 0.783,
respectively. Models 5 and 6 in Table 1 both provided better agree-
Downloaded from ascelibrary.org by Louisville, University Of on 09/24/13. Copyright ASCE. For personal use only; all rights reserved.

ment with the measurements than the FD model. However, the dif-
ference in the performance of the combined model attributable to
difference in parameters between C k1 and C k2 was minor. The con-
tribution of the radial-velocity distribution with kinematic viscosity
to the combined model may be responsible for the lesser effect of
the two different shear decay factors in predicting the pressure
variation (Ramos et al. 2004).

Comparison of the Combined Model and MOC


The convolution of the velocity difference with the weighting func-
tion has been the most frequently used approach, based on MOC
(Zielke 1968). To apply MOC, the experimental pipeline system
was divided into 40 segments, with a computational time step de-
termined to be 0.00155 s to provide a Courant number equal to 1.
Fig. 5. (a) Comparison of predicted and measured pressure traces at the Fig. 7(a) compares the normalized pressure-head responses esti-
valve for Ro ¼ 1;774; (b) enlargement of (a) between 1.0 and 1.4 s mated by MOC, the FD friction, and the combined models. Appar-
ent differences were observed among the combined models and
the two FD models using MOC and IRM, with minor differences
observed between MOC and IRM that may be associated with the
accumulated errors generated from the tremendous number of
convolution operations with an adequate timescale.
Fig. 7(b) shows the computational costs for the three models,
which were evaluated for a CPU using a 1.73 GHz Intel processor.
The combined model requires a computational cost of approxi-
mately twice that of the IRM-based FD friction model. This is
the computational demand for the evaluation of complicated propa-
gation constants and the hydraulic impedance of the combined
model. The combined model requires less than 0.4% of the com-
putational cost of the MOC-based FD friction model (Zielke 1968)
for a 10 s transient simulation. There are several approximate meth-
ods to account for weighting functions that are more efficient than
the computationally expensive convolution technique (Schohl
1993; Trikha 1975; Vardy and Brown 2004). The strength of the
combined model, in terms of the computational cost, over that
of the traditional MOC-based method was more pronounced in
the later computational time steps because of the greater burden
of evaluating the weighting function convolutions of MOC.
Another primary strength of this approach is its robustness in
representing the system. The characteristics of the system for
geometric parameters, such as length and diameter, or material
properties (i.e., wave speed) can be accurately expressed on the
platform of IRM. Fig. 8 presents 100 randomized combinatory
pairs of pipeline lengths and wave speeds to test the performances
of the models. To evaluate the computational capability under
extreme conditions, the maximum specified frequency range was
3;141:6 rad=s, with 8,192 as the fast Fourier transform number.
Transients were computed for a simulation time of 8 s, with
8,000 time steps for each computation. The maximum and mini-
Fig. 6. (a) Comparison of the predicted and measured pressure traces at mum computational time costs for 100 water-hammer simu-
65.42 m from the valve for Ro ¼ 1;774; (b) enlargement of (a) between lations were 14.68 and 13.66 s, respectively. The average time cost
1.0 and 1.4 s and standard deviation were 13.77 and 0.15 s, respectively. The

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / DECEMBER 2011 / 1655

J. Hydraul. Eng. 2011.137:1649-1658.


1.5
convective acceleration terms unless the grid size is infinitely small.
The combined model fundamentally resolved the spatial scale
1
issue associated with a numerical scheme (Ramos et al. 2004)
and its numerical diffusion (Vítkovský et al. 2006b), which may
0.5
also be related to the restriction of the Courant number (Kim
Hg/aV0

2005, 2007).
0
0.0 0.5 1.0 1.5 2.0

-0.5 Conclusion
-1 The modeling of rapid transients is an important academic and
practical issue in the field of pipeline engineering. Many studies
-1.5 have been performed to improve the predictability of the water-
Time (s)
hammer phenomenon and to resolve existing problems associated
Downloaded from ascelibrary.org by Louisville, University Of on 09/24/13. Copyright ASCE. For personal use only; all rights reserved.

(a) MOC (FD friction) IRM (FD friction) Combined model


with the computational cost. Although the governing equations for
1000 unsteady pipe flow originated from the Navier-Stokes equation, de-
velopments were determined by a few earlier attempts that were
restricted to corresponding modeling philosophies and assump-
tions. In this paper, a new holistic model for unsteady friction in
100
transients is developed by merging acceleration-based approaches
Time Cost (s)

and the convolution-based model. The structure of the combined


model provides a generic platform to address various modeling ap-
10
proaches while circumventing the issue of numerical discretization.
The combined model yielded better prediction than other existing
unsteady-friction models.
1 The controversial issues associated with cost and accuracy have
been points of contention in the modeling of pipeline hydraulics.
The progress in the modeling of pipeline transients achieved during
0.1 the past half-century still does not appear to fully address the im-
3 4 5 6 7 8 9 10
Simulation Time (s)
portant issues in real-life systems, such as diagnosing pipeline
(b) MOC (FD friction) IRM (FD friction) Combined model problems and accurately predicting water quality. Advancements
in computational power developed by the computer industry may
Fig. 7. (a) Normalized pressure heads computed using MOC with FD not be sufficient to overcome the restrictions associated with the
friction, IRM with FD friction, and the combined model; (b) the com- existing water-hammer modeling paradigm. This study explores
putational costs for the three methods the fundamentals in model structures and attempts to comprehen-
sively address these issues on the basis of the impulse-response
method.
Recent advances in pipeline transient modeling based on the
1450
impedance-based approach, including this study, have illustrated
the potential of the frequency-domain approach as an alternative
method to other existing grid-based methods. The time-domain ap-
proach has several obvious advantages over its counterparts, such
Wave Speed (m/s)

1350
as better comprehensive representation from a spatial perspective
and a more feasible expression of the hydraulic structure because
of less restriction in its linearization. Therefore, further work will
1250 be required to relax the fundamental assumptions of the frequency-
domain approach. In addition, model developments and experimen-
tal verification will need to be enacted under more complicated
1150 circumstances, such as for complex and heterogeneous pipe net-
20 300 580 860 1140 works, hydraulic structures, turbulence, and the uncertainty of field
Pipe Length (m)
pipeline systems.
Fig. 8. Combinatory pairs of randomized pipeline lengths between
20 and 1,100 m and randomized wave speeds between 1,150 and
1;450 m=s
Appendix. Relationship between 1D and 2D
Momentum Equations

The relationship between Eq. (4) and Eq. (5) can be derived as fol-
lows. If terms in Eq. (5) are multiplied by 2πrvdr
additional costs for modifying system parameters in the RPV
system were negligible. ∂v ∂v
2πrvdrð1 þ k 1 Þ þ g2πrvdrV  k2 2πrvdra ¼ 0 ð30Þ
The additional subsidiary strength of the combined model was ∂t ∂x
also associated with the absence of numerical discretization.
where
Many numerical unsteady-friction models (MOC-based) require RR
the computation of VdV=dx, dV=dx or signðVdV=dxÞ. The spatial 2πrvdr
V¼ 0
scale of discretization may influence the evaluation of the πR2

1656 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / DECEMBER 2011

J. Hydraul. Eng. 2011.137:1649-1658.


Eq. (30) can be integrated over the flow, and the operations of x = axial distance;
integration and differentiation are interchanged as the relationships ZD = hydraulic impedance at the downstream valve;
among r, x and t are independent ZðxÞ = hydraulic impedance;
γA , γB = propagation constants;
Z Z Z R ρ = density of fluid; and
1 þ k1 ∂ R ∂ R 32v ν
2πrv dr þ g
2
2πrvhdr þ 2 V 2πrvdr = kinematic viscosity.
2 ∂t 0 ∂x 0 D 0
Z R
k ∂
 2α 2πrv2 dr ¼ 0 ð31Þ References
2 ∂x 0
Based on the procedure of Brunone Abreu, J., and de Almedia, A. B. (2009). “Timescale behavior of the wall
R et al. (1995), assuming the
momentum correction factor, β ¼ A v2 dA=V 2 A, and dividing shear stress in unsteady laminar pipe flows.” J. Hydraul. Eng., 135(5),
Eq. (31) by a discharge, Q, the momentum equation for entire flow 415–424.
Adamkowski, A., and Lewandowski, M. (2006). “Experimental examina-
Downloaded from ascelibrary.org by Louisville, University Of on 09/24/13. Copyright ASCE. For personal use only; all rights reserved.

can be obtained as follows:


tion of unsteady friction models for transient pipe flow simulation.”
∂ ∂H 32v ∂ J. Fluids Eng., 128(6), 1351–1363.
ð1 þ k1 Þ ðβVÞ þ g þ 2 V  k 2 a ðβVÞ ¼ 0 ð32Þ Axworthy, D. H., Ghidaoui, M. S., and McInnis, D. A. (2000). “Extended
∂t ∂x D ∂x thermodynamics derivation of energy dissipation in unsteady pipe
where
R k1 ¼ ð1 þ k  βÞ=β: k2 ¼ k=β; k ¼ η þ φ; η ¼ flow.” J. Hydraul. Eng., 126(4), 276–287.
ð A v2 dA=V 2 AÞ  1; and φ = parameter used to consider the effect Bergant, A., Simpson, A. R., and Vítkovský, J. (2001). “Developments
in unsteady pipe flow friction modeling.” J. Hydraul. Res., 39(3),
of difference between steady and unsteady frictions (Brunone
249–257.
et al. 1991). Bergant, A., Tijsseling, A., Vítkovský, J., Covas, D., Simpson, A., and
Lambert, M. (2003). “Further investigation of parameters affecting
waterhammer wave attenuation, shape and timing. Part 1: Mathematical
Acknowledgments tools & Part 2: Case studies.” 11th Int. Meeting of the Work Group
on Behaviour of Hydraulic Machinery Under Steady Oscillatory Con-
This research was supported by Basic Science Research Program ditions, Int. Association for Hydraulic Research (IAHR), Madrid,
through the National Research Foundation of Korea (NRF) and Spain.
Brown, F. T. (1962). “The transient response of fluid lines.” J. Basic Eng.,
funded by the Ministry of Education, Science and Technology
84(4), 547–553.
(2010-0021511). Brunone, B., Ferrante, M., and Cacciamani, M. (2004). “Decay of pressure
and energy dissipation in laminar transient flow.” J. Fluids Eng., 126(6),
928–934.
Notation Brunone, B., Golia, U. M., and Greco, M. (1991). “Some remarks on the
momentum equation for fast transients.” Proc., Int. Meeting on Hy-
The following symbols are used in this paper: draulic Transients and Water Column Separation, International Asso-
A = cross-sectional area of pipe; ciation for Hydraulic Research (IAHR), Madrid, Spain, 201–209.
a = wave speed; Brunone, B., Golia, U. M., and Greco, M. (1995). “Effects of two-
C A , C B = integration constants; dimensionality on pipe transients modeling.” J. Hydraul. Eng., 121(12),
C k , C k1 , C k2 = decay coefficients for local and convective 906–912.
Chaudry, M. H. (1987). Applied hydraulic transients, 2nd Ed., Van
acceleration;
Nostrand Reinhold, New York.
D = pipe diameter; Daily, J. W., Hankey, W. L., Olive, R. W., and Jordan, J. M. (1956). “Re-
g = gravitational acceleration; sistance coefficient for accelerated and decelerated flows through
H = piezometric head; smooth tubes and orifices.” Trans. ASME, 78(7), 1071–1077.
HðtÞ, hðtÞ = pressure-head time series; Funk, J. E., Wood, D. J., and Chao, S. P. (1972). “The transient response of
HðxÞ = complex head; orifices and very short lines.” J. Basic Eng., 94(2), 483–491.
H 0 = oscillatory head; Ghidaoui, M. S., Karney, B. W., and McInnis, D. A. (1998). “Energy es-
J 0 , J 1 = Bessel functions of the first kind with zeroth and timates for discretization errors in water hammer problems.” J. Hydr.
first orders; Engrg., 124(4), 384–393.
k = Brunone’s friction coefficient; Holmboe, E. L., and Rouleau, W. T. (1967). “The effect of viscous shear on
transients in liquid lines.” J Basic Eng., Trans. ASME, 89(1), 174–180.
k1 , k 2 = momentum-corrected unsteady-friction decay
Kim, S. (2005). “Extensive development of leak detection algorithm by
coefficients; impulse response method.” J Hydraul. Eng., 131(3), 201–208.
k 3 = model coefficient for the general unsteady-friction Kim, S. (2007). “Impedance matrix method for transient analysis of com-
model; plicated pipe networks.” J Hydraul. Res., 45(6), 818–828.
p = pressure; Leon, A. S., Ghidaoui, M. S., Schmidt, A. R., and Garcia, M. H. (2008).
QðxÞ = complex discharge; “Efficient second-order accurate shock-capturing scheme for modeling
Qx ðtÞ = discharge responses; one and two phase water hammer flows.” J. Hydraul. Eng., 134(7),
Q0 = oscillatory discharge; 970–983.
R = Reynolds number for initial flow; Pezzinga, G. (1999). “Quasi-2D model for unsteady flow in pipe net-
r = radial distance; works.” J. Hydraul. Eng., 125(7), 676–685.
Pezzinga, G. (2000). “Evaluation of unsteady flow resistances by quasi-2D
rDh ðtÞ; rxh ðtÞ = pressure-head response functions at the
or 1D models.” J. Hydraul. Eng., 126(10), 778–785.
downstream valve and point x; Pezzinga, G. (2002). “Closure to ‘Evaluation of unsteady flow resistances
rxq ðtÞ = discharge response functions at point x; by Quasi-2D or 1D models’ by Giuseppe Pezzinga.” J. Hydraul. Eng.,
t = time; 128(6), 647–648.
V = average velocity; Pezzinga, G. (2009). “Local balance unsteady friction model.” J. Hydraul.
v = velocity; Eng., 135(1), 45–56.

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / DECEMBER 2011 / 1657

J. Hydraul. Eng. 2011.137:1649-1658.


Prado, R., and Larreteguy, A. E. (2002). “A transient shear stress model shear stresses in highly transient pipe flows.” J. Hydraul. Eng., 133(11),
for the analysis of laminar water-hammer problems.” J. Hydraul. Res., 1219–1228; “Errata.” 135(1), 71.
40(1), 45–53. Vardy, A. E., and Brown, J. M. B. (2010). “Evaluation of unsteady wall
Ramos, H., Covas, D., Borga, A., and Loureiro, D. (2004). “Surge damping shear stress by Zielke’s method.” J. Hydraul. Eng., 136(7), 453–456.
in pipe systems: Modeling and experiments.” J. Hydraul. Res., 42(4), Vítkovský, J. P., Bergant, A., Simpson, A., and Lambert, M. (2006a). “Sys-
413–425. tematic evaluation of one-dimensional unsteady friction models in
Schohl, G. A. (1993). “Improved approximate method for simulating fre- simple pipelines.” J. Hydraul. Eng., 132(7), 696–708.
quency-dependent friction in transient laminar flow.” Trans. ASME, J. Vítkovský, J. P., Stephens, M., Bergant, A., Simpson, A., and Lambert, M.
Fluid Eng., 115(3), 420–424. (2006b). “Numerical error in weighting function-based unsteady fric-
Silva-Araya, W. F., and Chaudry, M. H. (1997). “Computation of tion models for pipe transients.” J. Hydraul. Eng., 132(7), 709–721.
energy dissipation in transient flow.” J. Hydraul. Eng., 123(2), Wood, D. J., Lingireddy, S., Boulos, P. F., Karney, B. W., and McPherson,
108–115. D. L. (2005). “Numerical methods for modeling transient flow in dis-
Suo, L., and Wylie, E. B. (1989). “Impulse response method for frequency- tribution system.” J. Am. Water Works Assoc., 97(7), 104–115.
dependent pipeline transients.” J. Fluids Engrg., Trans. ASME, 111(4), Wylie, E. B. (1996). “Frictional effects in unsteady turbulent pipe flow.”
478–483. Applied mechanics in Americas, M. Rysz, L. A. Godoy, and L. E.
Downloaded from ascelibrary.org by Louisville, University Of on 09/24/13. Copyright ASCE. For personal use only; all rights reserved.

Trikha, A. K. (1975). “An efficient method for simulating frequency- Suarez, eds., Vol. 5, University of Iowa, Iowa City, IA, 29–34.
dependent friction in liquid flow.” J. Fluid Eng., 97(1), 97–105. Wylie, E. B., and Streeter, V. L. (1993). Fluid transient in systems, Prentice
Vardy, A. E., and Brown, J. M. B. (2003). “Transient turbulent friction in Hall, Englewood Cliffs, New Jersey, 339.
smooth pipe flows.” J. Sound and Vibr., 259(5), 1011–1036. Zhao, M., and Ghidaoui, M. S. (2006). “Investigation of turbulence
Vardy, A. E., and Brown, J. M. B. (2004). “Efficient approximation behavior in pipe transients using κ-ε model.” J. Hydraul. Res., 44(5),
of unsteady friction weight functions.” J. Hydraul. Eng., 130(11), 682–692.
1097–1107. Zielke, W. (1968). “Frequency-dependent friction in transient pipe flow.”
Vardy, A. E., and Brown, J. M. B. (2007). “Approximation of turbulent wall Trans. ASME, J. Basic Eng., 90(1), 109–115.

1658 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / DECEMBER 2011

J. Hydraul. Eng. 2011.137:1649-1658.

You might also like