Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Numerical and experimental study on dynamics of unsteady pipe flow involving backflow prevention assemblies
(USC Thesis Other)
Numerical and experimental study on dynamics of unsteady pipe flow involving backflow prevention assemblies
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
NUMERICAL AND EXPERIMENTAL STUDY ON
DYNAMICS OF UNSTEADY PIPE FLOW INVOLVING
BACKFLOW PREVENTION ASSEMBLIES
by
HYOUNG-JIN KIM
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulllment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(CIVIL ENGINEERING)
December 2012
Copyright 2012 HYOUNG-JIN KIM
Acknowledgements
First, I would like to thank my advisor, Dr. Jiin-Jen Lee, for his continuous supports and
eorts during my academic years at University of Southern California. Whenever I doubt
myself, he was there encouraging me not to loose goals academically and personally. I
have learned from him how to look a problem in a big picture, how to solve it eciently,
and how to express an idea in an eective manner.
I also would like thank the sta members (at the Foundation of cross connection con-
trol and hydraulic research) for giving the opportunity for testing utilities and providing
the valuable information of testing equipments and experimental data.
A special thank to my dad for his unconditional support and trust with his generous
mind. Without him, nothing would have be possible for my life.
Last, but not least, I would like to thank my beloved family, Jungwan, Noah, and
Sarah who have endured suering with patience and faith over last years of study.
ii
Contents
Acknowledgements ii
List of Tables vi
List of Figures vii
Abstract xi
Chapter 1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.1 Fundamental Partial Dierential Equations . . . . . . . . . . . . . 6
1.2.2 Classic Water Hammer Theory . . . . . . . . . . . . . . . . . . . . 6
1.2.3 Friction Models Used In Calculation . . . . . . . . . . . . . . . . . 11
1.2.4 Dynamics Involving Back
ow Prevention Assemblies . . . . . . . . 14
1.3 Objective and Scope of Present Study . . . . . . . . . . . . . . . . . . . . 16
iii
Chapter 2 Water Hammer Equations 17
2.1 Dierential Equations For Unsteady Flow . . . . . . . . . . . . . . . . . . 17
2.1.1 Continuity Equation . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.1.2 Momentum Equation . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.1.3 Moment-of-Momentum Equation of a Valve Disk . . . . . . . . . . 18
Chapter 3 Method of Characteristics 24
3.1 Solution by Method Of Characteristics . . . . . . . . . . . . . . . . . . . . 24
3.2 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2.1 Constant-Level Upstream Reservoir . . . . . . . . . . . . . . . . . 28
3.2.2 Series Junction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2.3 Valve At Downstream End . . . . . . . . . . . . . . . . . . . . . . 30
3.2.4 Back
ow Prevention Assembly . . . . . . . . . . . . . . . . . . . . 30
3.3 Coupling Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Chapter 4 Experiments 34
4.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2 Case 1: Without a Back
ow Prevention Assembly . . . . . . . . . . . . . 37
4.3 Case 2: With A Back
ow Prevention Assembly . . . . . . . . . . . . . . . 43
4.4 Comparison of Measured Data
Between Case 1 and Case 2 . . . . . . . . . . . . . . . . . . . . . . . . . . 49
iv
Chapter 5 Simulation Results and Discussion 53
5.1 Selection of F
T
and k
2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.2 Verication of Models: Case 1 . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.3 Implementation of A Coupling Model: Case 2 . . . . . . . . . . . . . . . . 63
5.3.1 Sensitivity of k
2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.3.2 Classication of Assemblies:
Ideal, Undamped, And Damped . . . . . . . . . . . . . . . . . . . 79
Chapter 6 Conclusion And Recommendation 81
6.1 Summary and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6.2 Recommendation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Bibliography 87
v
List of Tables
4.1 Physical Properties Of System . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2 Summary of Flow Conditions For Cases . . . . . . . . . . . . . . . . . . . 44
4.3 Summary of Pressure Reduction Rates by Back
ow Preventer . . . . . . . 52
5.1 Input Parameters Used In Simulation . . . . . . . . . . . . . . . . . . . . 55
vi
List of Figures
1.1 Dynamic Characteristics of A General Non-return Valve . . . . . . . . . . 15
2.1 Schematic Of A Check Valve . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2 Inherent Characteristics Of Valves . . . . . . . . . . . . . . . . . . . . . . 21
2.3 Flow Coecient Used For Present Model . . . . . . . . . . . . . . . . . . 22
2.4 Computer Algorithm For Coupling Model . . . . . . . . . . . . . . . . . . 23
3.1 Characteristic Lines In x-t Plane . . . . . . . . . . . . . . . . . . . . . . . 26
3.2 Constant-Level Upstream Reservoir . . . . . . . . . . . . . . . . . . . . . . 29
3.3 Series Junction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.4 A Schematic of Check Valve Between Two Pipes . . . . . . . . . . . . . . 31
4.1 Testing Loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2 Sequence of Water Hammers Occurred in System . . . . . . . . . . . . . . 38
4.3 Case 1 Pressure Time History At Upstream For 0t 10 . . . . . . . . 40
4.4 Case 1 Pressure Time History At Upstream For 0t 4 . . . . . . . . . 40
vii
4.5 Case 1 Pressure Time History At Upstream For 4t 8 . . . . . . . . . 41
4.6 Case 1 Pressure Time History At Downstream For 0t 10 . . . . . . . 41
4.7 Case 1 Pressure Time History At Downstream For 0t 4 . . . . . . . . 42
4.8 Case 1 Pressure Time History At Downstream For 4t 8 . . . . . . . . 42
4.9 Case 2 Pressure Time History At Upstream For 0t 8 . . . . . . . . . 45
4.10 Case 2 Pressure Time History At Upstream For 0t 2 . . . . . . . . . 45
4.11 Case 2 Pressure Time History At Upstream For 2t 4 . . . . . . . . . 46
4.12 Case 2 Pressure Time History At Downstream For 0t 8 . . . . . . . . 46
4.13 Case 2 Pressure Time History At Downstream For 0t 2 . . . . . . . . 47
4.14 Case 2 Pressure Time History At Downstream For 2t 4 . . . . . . . . 47
4.15 Case 2 Pressure At Downstream for 0:7t 0:9 . . . . . . . . . . . . . . 48
4.16 Comparison of Two Cases at Upstream for 0:6t 4:6 . . . . . . . . . . 50
4.17 Comparison of Two Cases at Upstream for 0:6t 1:6 . . . . . . . . . . 50
4.18 Comparison of Two Cases at Downstream for 0:6t 4:6 . . . . . . . . 51
4.19 Comparison of Two Cases at Downstream for 0:6t 1:6 . . . . . . . . 51
5.1 Quasi-steady Model with f = Darcy-Weisbach friction coecient for 0
t 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.2 Quasi-steady Model with f = 0.15 for 0t 8 . . . . . . . . . . . . . . . 56
5.3 Quasi-steady Model with f = 0.3 for 0t 8 . . . . . . . . . . . . . . . . 57
viii
5.4 Quasi-steady Model with f = Darcy-Weisbach friction coecient for 0
t 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.5 Quasi-steady Model with f = 0.15 for 0t 8 . . . . . . . . . . . . . . . 58
5.6 Quasi-steady Model with f = 0.3 for 0t 8 . . . . . . . . . . . . . . . . 58
5.7 Case 1 Simulated By Quasi-Steady 1D Model With F
T
= 0:3 . . . . . . . 59
5.8 Case 1 Simulated By Unsteady Friction 1D Model With F
T
= 0:3 and
k
2
= 0:045 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.9 Case 1 Comparison of Two Dierent Models At Upstream . . . . . . . . . 61
5.10 Case 1 Comparison of Two Dierent Models At Downstream . . . . . . . 62
5.11 Case 2 (Downstream) Present Model With F
T
= 0:3 and k
2
= 0:045 . . . 65
5.12 Case 2 (Upstream) Present Model With F
T
= 0:3 and k
2
= 0:045 . . . . . 66
5.13 Case 2 (Downstream) Present Model F
T
= 0:3 and k
2
= 0:045 for 0:7
t 1:5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.14 Case 2 (Upstream) Present ModelF
T
= 0:3 andk
2
= 0:045 for 0:7t 1:5 67
5.15 Case 2 Flow Rate Changes At Assembly for 0:7t 0:8 . . . . . . . . . 68
5.16 Case 2 Disk Angle Position . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.17 Case 2 (Downstream) Present Model With F
T
= 0:3 and k
2
= 0:08 . . . . 69
5.18 Case 2 (Upstream) Present Model With F
T
= 0:3 and k
2
= 0:08 . . . . . . 70
5.19 Case 2 (Downstream) Present Model F
T
= 0:3 and k
2
= 0:08 for 0:7
t 1:5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
ix
5.20 Case 2 (Upstream) Present Model F
T
= 0:3 and k
2
= 0:08 for 0:7t 1:5 71
5.21 Case 2 (Downstream) Present Model With F
T
= 0:3 and k
2
= 0:2 . . . . . 72
5.22 Case 2 (Upstream) Present Model With F
T
= 0:3 and k
2
= 0:2 . . . . . . 73
5.23 Case 2 (Downstream) Present ModelF
T
= 0:3 andk
2
= 0:2 for 0:7t 1:5 74
5.24 Case 2 (Upstream) Present Model F
T
= 0:3 and k
2
= 0:2 for 0:7t 1:5 74
5.25 Case 2 (Downstream) Present Model With F
T
= 0:3 and k
2
= 0:4 . . . . . 75
5.26 Case 2 (Upstream) Present Model With F
T
= 0:3 and k
2
= 0:4 . . . . . . 76
5.27 Case 2 (Downstream) Present ModelF
T
= 0:3 andk
2
= 0:4 for 0:7t 1:5 77
5.28 Case 2 (Upstream) Present Model F
T
= 0:3 and k
2
= 0:4 for 0:7t 1:5 77
5.29 Changes on Flow Rate At Assembly . . . . . . . . . . . . . . . . . . . . . 79
5.31 (Upstream) Comparison of Responses With Dierent Valve Types . . . . 80
5.30 (Downstream) Comparison of Responses With Dierent Valve Types . . . 80
x
Abstract
When control valves at the end of pipeline close simultaneously, two pressure waves are
generated at each end and the waves propagate toward to the other end. The pressure
waves continue to move back and forth along pipelines until they are damped out to next
steady states. This study provides information on the experimental data and the numer-
ical simulation of a rapid hydraulic transient event called water hammer. The energy
loss term due to friction in the present model consists of quasi-steady contribution and
unsteady contribution. For the present model, an equivalent friction coecient is used
to replace the quasi-steady friction coecient, inclusive unsteady friction loss and minor
energy loss factors. The unsteady component has been related to the combination of the
instant
ow acceleration and instant
ow convective acceleration. The numerical results
of the present model are compared with the experimental records. The computer results
by the present model which is based on the unsteady friction 1D model was successful to
follow general trends of water hammer phenomena, corresponding with sudden changes
in
ow. This study later extends to the dynamic characteristics of back
ow prevention
assemblies under a rapid transient condition. When numerous hydraulic devices are
xi
installed on water distribution systems, the dynamic characteristics of such hydraulic
devices have a signicant in
uence on the intensity of the associated water hammer
waves. A back
ow prevention assembly plays important roles as not only a safety hy-
draulic device but also an energy dissipater. A numerical program for rapid transient
pipe
ow interconnected with a back
ow prevention assembly has been developed using
the present coupling model and numerically solved by the method of characteristics.
xii
Chapter 1
Introduction
1.1 Background
Drinking water distribution systems are large networks of piping systems designed to
transport drinkable water from sources to consumers. A water distribution system con-
sists of storages (reservoirs or water tanks), pumps, networks of pipes, control valves
and other hydraulic appurtenances. The primary function of drinking water distribution
systems today is to deliver water of good quality eciently and safely to the end of water
users. Early water distribution systems were run solely by gravity at a relatively low
pressure. The supply of water was restricted low because of the limited water sources
reliable and the limited technologies available. For these reasons steady conditions were
prevalent throughout the systems. As a continuous demand for water has increased due
to fast growing population, there has been a growing need of improved water delivering
systems which are operated under a high pressure and a high
ow rate. Today water
1
distribution systems, therefore, have been enhanced by equipping with a series of pumps,
a intricate network of pipelines, and numerous hydraulic devices. Unavoidably, transient
conditions are of great possibility everywhere in modern water distribution systems. A
wide variation of water usages often causes extreme pressure
uctuations through the
system. For example, sudden stoppage of pumps (pump failure), immature operations
of valves and the in
uence of accidental events such as power outages and burst pipes,
all of these create transient
ow conditions. Many literature and the public press have
reported a large number of incidents of hydraulic transients in the water distribution
systems over the last century. Many safety devices have been developed and installed to
prevent water accidents and/or to minimize the further system damages in case of the
accidents.
Maintaining water distribution systems under normal operating condition as designed
is crucial to ensuring safe drinking water and its supply systems against contamination
and damage respectively. Thus, the importance of hydraulic transient analysis in water
distribution systems has been arisen and further research of transient pipe
ow should be
taken into account. Today, modeling of the systems is a fundamental part in providing
basis for planning and designing to engineers and implementing the right decisions to
operators.
Sudden momentum changes in
uid in motion creates excessive pressure changes
(positive and negative pressure surges) inducing back
ow into attached water piping sys-
tems. This rapid transient event is a phenomenon called water hammer. Traditionally
friction losses in the simulation of water hammer has been modeled using steady friction
2
approximation such as Darcy-Weisbach equation, which is known as quasi-steady approx-
imation. It is widely known that this assumption gives a satisfactory result only for slow
transients where the wall shear stress has a quasi-steady behaviour. This quasi-steady
friction approximation uses the friction coecient (named Darcy-Weisbach friction coef-
cient) depending on the state of the system at the previous time step. Even though the
friction coecient used in this model is estimated by means of an approximated formula
of the Moody diagram for each time step, it would be not valid for the simulation of tran-
sients because the value of the coecient is estimated on the basis of initial steady
ow
Q
0
. The experimental validation of this study shows that the quasi-steady approxima-
tion shows a poor agreement between the experimental data and computer calculation,
in magnitude and phase of pressure waves particularly for long-time-period records of
rapid transient events. Good estimation of friction losses under transient conditions is
one of the key issues encountered in the development of understanding and modeling
of rapid transient pipe
ows in the systems. Thus, further investigation of unsteady
friction eect in rapid transients of a simple reservoir-pipe-valve system are presented
in this study. Two one-dimensional unsteady friction models, the Zielke [29] and the
Brunone et al [1],[2] models, are reviewed in detail in the later chapter. Finally, the
Brunone unsteady friction model is incorporated with the pipe equations(water hammer
equations) and the valve equations for the analytic solution of the equations for unsteady
pipe
ow interacted with back
ow prevention assemblies. The derivation of this model
is presented in detail in chapter 3.
In order to prevent back
ow from occurring in distribution systems, back
ow preven-
3
ters are required and are installed between the delivery point of water mains and local
storage or use. The basic method of preventing back
ow is an air gap. This method
eliminates a direct cross-connection between a contaminated water source and a potable
source by providing an adequate space between them. However, an air gap loses the
pressure in the pipe system (drop to atmospheric pressure). Thus dierent back
ow
preventers are used to protect the water distribution system. Back
ow prevention as-
semblies are mechanical devices that provide a physical barrier to back
ow. In general, a
back
ow prevention assembly consists of a combinations of check valves, relief valves, air
inlet valves and/or shuto valves. The types of back
ow prevention assemblies would be
classied, according to the types of internal valves used. There are four types commonly
used.
Reduced Pressure Principle Assembly (RP)
Double Check Valve Assembly (DC)
Pressure Vacuum Breaker (PVB)
Atmospheric Vacuum Breaker (AVB)
The most important considerations in selecting a back
ow prevention assembly, are
the head loss and non-slamming characteristics. It should have an acceptable head
loss coecient for forward
ow under normal
ow condition, and not create excessive
transient pressures under the reverse
ow condition upon a sudden closing.
The dynamic characteristics of back
ow prevention assemblies depends mainly on the
internal check valves used. The valve position under the transient
ow condition is de-
4
termined by the
ow and valve dynamics. The dynamic behavior of back
ow prevention
assemblies is studied by using both experimental and theoretical approaches. Accurate
modeling of the dynamics of a back
ow prevention assembly requires experimental data
to quantify coecients for the hydraulic torque term in the moment-of-momentum equa-
tion. Based on the understanding of the dynamic interaction between the
ow and an
internal check, a nal numerical model is developed coupling those equations of motion
of checks into the unsteady model.
For the present study, a numerical solution for unsteady pipe
ow for the reservoir-
pipe-valve system is obtained by using the method of characteristics. A pair of partial
dierential equations for unsteady pipe
ow may be transformed into four ordinary
dierential equations by the method of characteristics. Those ordinary equations may
be integrated to achieve numerical solutions by nite dierential method.
1.2 Literature Review
Two dierential equations for unsteady pipe
ows are derived in this section based on
the classic water hammer theory. The governing equations are namely the conservation
of mass and the conservation of momentum. A set of water hammer equations can be
modied by making several assumptions on the
ow condition. Then, several unsteady
friction models are reviewed. The Quasi-steady model has been used to quantify the
system's friction coecient, F
T
for transient
ow. And unsteady friction models by
Zielke [29], Vardy [25], [26], and Brunone et al. [1], [2] are introduced and the Brunone
model has been selected for later use in this study.
5
1.2.1 Fundamental Partial Dierential Equations
The momentum and continuity equations are a set of non-linear, hyperbolic, partial
dierential equations that govern unsteady
ow in a closed conduit. Among the early
researchers of water hammer problems, Joukowsky (1898) [8] produced the well-known
equation that relates pressure changes, P , to velocity changes, V , according to Eqn
1.1. This equation was developed based on the rigid column theory in which the com-
pressibility of water and elasticity of pipe wall are ignored.
P =aV or H =
aV
g
(1.1)
where is the
uid density, H is the piezometric head, and a is the speed of sound.
Kortewegs (1878) formula denes the wave speed a for
uid in cylindrical pipes as
a =
p
K
0
=andK
0
=
s
K
(1 + (DK)=(eE))
(1.2)
whereD is the diameter of the pipe,e is the wall thickness,E is the modulus of elasticity
for the wall, andK is the bulk modulus of elasticity of
uid. Further investigation to the
governing equations of water hammer has continued by many researchers (Jaeger [6], [7],
Wood [27], Rich [16],[17], Parmakian [12], Streeter and Lai [18], and Streeter and Wylie
[19]), resulting in the following the classical water hammer equations for one-dimensional
unsteady pipe
ows.
1.2.2 Classic Water Hammer Theory
Fully developed by 1960s the classical water hammer equations may be able to describe
general physics necessary to model wave generation, propagation, and energy damping
6
in water distribution systems. However, it is widely known today that the model based
on quasi-steady friction losses hypothesis shows the basic reasons for dierences between
experimental and computational results obtained according to the classical water ham-
mer theory. Numerous researchers has developed miscellaneous unsteady friction models
that consider extra parameters aecting on the unsteady friction losses. The classical
Water hammer equations are presented in this chapter and several unsteady friction
models are derived, being modied from the classical water hammer equations by using
dierent approaches. Here, two governing equations based on the classic water hammer
theory may be applied for calculation of the liquid unsteady pipe
ow. The assumptions
made in the development of equations are
The
ow velocity and pressure at a cross-section are averaged and uniform (one-
dimensional).
The
uid is slightly compressible.
The pipe is full and remains full during the transient.
The pipe wall is linearly elastic and slightly deformable.
Free gas content of the liquid is small such that the wave speed can be regarded
as a constant.
These assumptions have been well adopted for many numerical analysis tools for unsteady
pipe
ows. The fundamental governing equations under the assumptions are described
by the continuity and motion equations [28], [4]. The conservation of mass for one-
7
dimensional unsteady pipe
ow is represented by
@P
@t
+V
@P
@x
gVsin +a
2
@V
@x
= 0 (1.3)
where
P = pressure
V = average velocity
a = wave speed
=
uid density
g = gravitational acceleration
= angle at which pipeline is inclined with the horizontal
t = time
x = distance
and wave speed is calculated using Eqn 1.2. The wave speed in a closed conduit can
be evaluated by investigating the
uid properties and conduit's elasticity as given by
A=pA in Eqn 1.2 depending on
1. Fluid Properties
Modulus of elasticity
Density
Amount of air, and so forth
2. Pipe properties
Modulus of elasticity
8
Diameter
Thickness
For very thick-walled pipe A=p is very small, and a
p
k= is the acoustic speed
of a small disturbance in an innite
uid. For very
exible pipe walls, the second term
in the denominator is relatively large and the wave speed becomes Eqn 1.4. For the
computer model presented in this study, a special formula for wave speed under the pipe
wall condition is used and it is expressed in Eqn 1.4 for calculation.
a =
s
K
w
=
w
1 + (1p
2
) (K
w
=E
p
) (D=e)
(1.4)
where
K
w
= Bulk modulus for water
w
= Density of water
p = Poisson's ratio for pipe material
E
p
= Young's modulus for pipe material
D =Pipe Diameter(m)
e = Pipe thickness(m)
For water under ordinary condition the wave speed can be calculated by
p
K= =
1440m=s. The speed of pressure waves is assumed to be maintained a constant value for
the present models. However, However, the measured data shows a signicant reduction
in the wave speed in transitional zone, causing an observable phase shift.
The piezometric head may replace pressure with the relation of
@P=@t =g(@H=@t) (1.5)
9
For most engineering applications, the convective term V (@H=@x) and Vsin are neg-
ligible compared to the other terms and may be neglected. Eqn 1.3 becomes
@H
@t
+
a
2
g
@V
@x
= 0 (1.6)
This is a simplied hydraulic-grade-line form of the continuity equation for unsteady
ow, Eqn 1.6.
The equation of motion for
uid
owing through a pipe is presented in Eqn 1.7. An
average cross-sectional pressure equal to the centerline pressure p(x;t) and an average
cross-sectional velocity V (x;t) are assumed in the one-dimensional equation derivation.
1
@P
@x
+V
@V
@x
+
@V
@t
+g sin +F = 0 (1.7)
where
= angle at which pipeline is inclined with the horizontal
F = head losses per unit length due to friction
When the convective term V (@V=@x) is assumed negligible, the pipeline is horizontal
(sin = 0), and pressure is replaced by the piezometric head in the same manner as in
the equation of continuity, the equation can be further reduced to
g
@H
@x
+
@V
@t
+gF = 0 (1.8)
which is the simplied hydraulic-grade-line form of the equation of motion. F in Eqn 1.8
is considered as a sum of the head loss due to steady and unsteady friction, respectively.
F =F
steady
+F
unsteady
(1.9)
10
The steady component of the friction may be based on the Darcy-Weisbach friction
relationship and is dened as
F
steady
=
f
D
VjVj
2g
(1.10)
in which
f = Darcy-Weisbach friction factor
D = Pipe diameter.
A common friction modeling according to the quasi-state
ow hypothesis assumes
that F
unsteady
equals zero. In addition, a miscellaneous unsteady friction models con-
sidering extra unsteady friction losses have been developed for many years in a large
number of literature. Two distinct modeling approaches compress several unsteady fric-
tion models into two main groups of research.
1.2.3 Friction Models Used In Calculation
After created in the pipe systems, pressure waves propagate along the pipeline and
dissipate after a short period of time, consequently reaching to another steady state. It
comes from the hydraulic resistance caused by both the internal friction of the
uid and
the friction at pipe walls. The quasi-steady friction model and two distinct unsteady
friction models are presented in this chapter.
The quasi-steady method assumes that unsteady friction factor has no contribution
to energy loss. Thus, the quasi-steady model usesF
steady
only. This quasi-steady friction
model is computationally economical and eective. However, its numerical results always
11
underestimate the energy losses due to friction during hydraulic transients.
F =F
steady
=
f
D
VjVj
2g
(1.11)
In the rst group of research the unsteady friction is calculated on the basis of the
past
ow acceleration (Zielke [29], Trikha [22], Vardy and Brown [23], [24], [25], [26]).
Zielke [29] developed an analytical model in which the unsteady head loss term,F
unsteady
,
is a function of
ow acceleration and weighted past velocity changes. The unsteady head
loss per unit length is expressed by
F
unsteady
(t) =
16
gD
2
Z
t
0
@V
@t
(u)W (tu)du (1.12)
where = kinematic viscosity; W = weighting function; and= convolution operator.
The convolution integral is approximated using the rectangular rule and the acceleration
term is approximated using a centered nite dierence as
F
unsteady
(t)
app
=
16
gD
2
M
X
j=1;3;5;
[V (tjt + t)V (tjt t)]W (jt) (1.13)
where M =t=t 1. Zielke [29] determined a weighting function applicable to laminar
ows and Trikha [22] developed a simplied model based on the Zielke's work, allowing
saving a lot of computing power and time. Whereas the previous models were only
applicable to laminar
ows, Vardy and Brown improved the weighting function applicable
for smooth-pipe turbulent
ows [23], [24] and rough-pipe turbulent
ows [26]. However,
Vardy and Brown showed the results of computation only in short duration of time. The
trend of energy decay in transient
ow cannot be fully observed in such a short time
duration. Also, this model is valid only for low Re number.
12
In the second group the unsteady friction is related to the instantaneous acceleration
(Daily et al. [9], Brunone et al. [1], [2]). The dependence of unsteady friction on
acceleration was initiated by Daily et al. [9]. In their model a result of experimental
work denes the unsteady friction term as
F =
f
D
V
2
2g
+
k
1
g
dV
dt
(1.14)
where the dimensionless coecient k
1
is equal to 0.01-0.015 for accelerating
ow and
0.62 for decelerating
ow. Carstens and Roller [3] suggested the value of k
1
is a function
of Reynolds number Re.
Brunone et al. [2] introduced an additional convective acceleration term to keep
unsteady friction when V@V=@t> 0 and to cancel it when V@V=@t< 0. The non-linear
term, @V=@x, that was ignored due to its small value in the process of derivation of
classical water hammer equations, is now multiplied by the wave speed, a, and added to
the instant acceleration term,@V=@t. The combination of two acceleration terms is then
multiplied by a constant coecient, k
2
for the unsteady friction model.
F =
f
D
V
2
2g
+
k
2
g
@V
@t
a
@V
@x
(1.15)
in which the coecient k
2
is originally evaluated by the experiment. The role of the
coecient, k
2
is very important for this model. This coecient may be assumed con-
stant or depending on the initial Re value. Since unsteady friction models based on
instantaneous values of the
ow are relatively simple and computationally eective, the
Brunone's model is selected and modied for later use in this study.
13
1.2.4 Dynamics Involving Back
ow Prevention Assemblies
A literature review on dynamics of check valves is presented for studying back
ow pre-
vention assembly's behavior during the transients in this section. In general, two lines
of research have been distinguished for many years. First is the dynamic characteristics
method and second is the moment-of-momentum equation method.
The dynamic characteristics method was initiated by Provoost [13], [14], [15] and
widely adopted in early 1980's among researchers and manufacturers. The dynamic
characteristic curve is determined experimentally. Given the
ow deceleration, the max-
imum reverse
ow velocity can be obtained from the curve. Assuming the valve is forced
to close by the reverse
ow, a maximum pressure peak is then calculated by using the
Joukowski formula. Thorley [21], [20] presented a plot of the dynamic characteristics
for various types of valves by collecting data from several sources. Koetizer et al. [5]
introduced a dimensionless form of dynamic characteristics curve. From Fig 1.1, basic
understandings of a non-return valve type may be obtained. An ideal back
ow pre-
vention assembly may be one which closes at the instant when the
ow velocity at the
assembly reaches to zero. However, limited reverse
ow will still occur in the system
due to the inertia and friction of the components. Some assemblies are spring-loaded or
power operated for more rapid closure so that to minimize the occurrence of back
ow.
For a system where the
ow reverses slowly, most assembles will close before any signi-
cant back
ow occurs. If
ow reversal occurs rapidly, a relative high reverse velocity may
occur before closure causing another water hammer in the system. For a given valve, the
maximum reverse velocity is a function of the rate change of back
ow. Most back
ow
14
Velocity
Valve
Position
0
1
Vmin
0
Vo
tc
Vr
Undamped damped
Time
(a)Vo: initial
ow Velocity (b)Vmin:minimum velocity to keep the valve fully open
(c)Vr: maximum velocity of reverse
ow (d)tc: time of closure
Figure 1.1: Dynamic Characteristics of A General Non-return Valve
prevention assemblies are types of undamped that they close in such a way that the
ow
changes from the reverse V
r
to zero rapidly (Fig 1.1).
All these attempts counted only for cases where
ow deceleration in a pipe system is
constant. Therefore, the dynamic characteristic method restricts to system-dependent
problems. The dynamic behaviors of back
ow prevention assemblies can be described
by using the moment-of-momentum equation of a valve disk which spin-moves around a
xed hinge. In this equation, the net torque applied to a moving check disk is equated
to the inertia torque. Among the net torque, a weight torque, a friction torque, and
external torque can be determined theoretically in a straightforward way. The hydraulic
15
torque, by contrast, is nearly impossible to quantify analytically or experimentally due
to the complex of transient
ow pattern. For the present study, the hydraulic torque
is estimated by the dierence of pressures calculated at two locations across the check
valve, proposed by Wylie [28].
1.3 Objective and Scope of Present Study
The main objective of this study is to develop a computer model to numerically solve
the problem of transient
ows in a simple water distribution system such as a reservoir-
pipe-valve system equipped with back
ow prevention assemblies.
This study is intended to facilitate calculation of transient conditions in water dis-
tribution systems more accurately and more eciently, considering unsteady friction
eects. A coupling model is then presented to investigate the relationship between the
system and the hydraulic devices. Although various boundary conditions are discussed
through the study, special attention is given to the dynamic behaviors of a back
ow
prevention assembly with a single internal check valve. The valve equation which relates
the system equations for transient
ows and the equation of moment-of-momentum for
the assembly is used to study the interactions between the assembly and the system.
In the present study, a set of computer algorithms for transient pipe
ows and the
dynamic equation of a back
ow prevention assembly have been developed. Numerical
simulations for transient pipe
ows are performed and compared with some experimental
results.
16
Chapter 2
Water Hammer Equations
In this chapter the one-dimensional dierential equations of motion and continuity for
unsteady pipe
ow are introduced. Also, the modied equation of motion is presented.
For the computer models, the method of characteristics, as a numerical scheme, has
been used in the present study. The water hammer eect can be simulated by solving
the following partial dierential equations.
2.1 Dierential Equations For Unsteady Flow
2.1.1 Continuity Equation
The one-dimensional conservation of mass equation(continuity) for slightly compressible
uids in cylindrical tube on any slope is rewritten as
@H
@t
+
a
2
g
@V
@x
= 0 (2.1)
17
with independent variables are x= distance and t= time. Other parameters are H=
piezometric head, V =
ow velocity, a= wave speed, and g= gravitational acceleration.
Wave speed is calculated using Eqn 1.4.
2.1.2 Momentum Equation
The unsteady friction model of Brunone et al. [1] assumes the friction term consists of
two components; a quasi-steady contribution, which is the traditional assumption, and
an unsteady contribution, which is related to the instantaneous acceleration @V=@t and
the instantaneous convective acceleration @V=@x.
F =
f
D
VjVj
2g
+
k
2
g
@V
@t
a
@V
@x
(2.2)
Incorporation of Equation (2.2) into the momentum equation, the basic equation of
motion is expressed as in Equation(2.3).
g
@H
@x
+
@V
@t
+
f
D
VjVj
2
+k
2
@V
@t
a
@V
@x
= 0 (2.3)
The parameter k
2
is evaluated by comparisons between experimental and numerical
results.
2.1.3 Moment-of-Momentum Equation of a Valve Disk
Although a single check valve cannot be a secure means of preventing back
ow, it is an
important component of back
ow prevention assemblies. A check valve is designed to
allow water to pass in only one desirable direction. The check valve used in this study
is of hinged swing check type with spring loading.
18
V
1
V
2
+ :clockwise
-
u
Figure 2.1: Schematic Of A Check Valve
The closing torque due to the internal spring holds the check valve closed. In order to
open the check valve, the water pressure in front of the check valve or upstream must be
greater than the closing torque. The dynamic behavior of a check valve can be described
by the moment-of-momentum equation of the check disk. Fig.2.1 provides a schematic
sketch in which is the disk angle, and with clockwise moments about the hinge point
considered positive. The moment-of-momentum equation yields
T
w
+T
e
+T
f
+T
h
=I
d
2
dt
2
(2.4)
The torque due to weight of the rotating disk is given by T
w
and is represented by
T
w
=W
s
r
c
sin (2.5)
in which W
s
= submerged weight of the disk assembly and r
c
= length from the hinge to
the mass center of the disk assembly.
Any external torque applied to the disk is included inT
e
. If a spring with a torsional
spring stiness of s is acting, T
e
is given by
T
e
=s (2.6)
19
The torque due to friction, T
f
, applied at the pin joint is likely to depend on the
angular velocity and it can be given by
T
f
=k
1
+k
2
d
dt
n
(2.7)
This T
f
is assumed negligibly small and left out from the equation.
The torque due to the hydrodynamic pressures, T
h
, is given by
T
h
=
Z
A
prdA (2.8)
in which r is the distance to the disk area where the pressure dierence across is p.
The pressure dierence is a function of the
ow, the angular position, the angular speed
of the disk. Since it is almost impossible to determine p across the disk analytically, a
valve equation relating
ow to pressure drop is used as
Q =C
d
A
o
p
2gH (2.9)
in which
H= the average pressure head drop across the valve
C
d
= the
ow coecient
A
o
= the open area through which the
ow passes
and C
d
is a function of the shape of
ow passage and Reynolds number. If Reynolds
eects are neglected, only a valve opening as a function of is needed.
All valves have an inherent
ow characteristic that denes the relationship between
valve opening and
ow rate under steady conditions. Dierent design of the plug and
20
Figure 2.2: Inherent Characteristics Of Valves
seat arrangement causes the dierence in valve opening between these valves. The most
common characteristics are shown in Fig 2.2 [11]. The percent of
ow through the valve
is plotted against valve opening, which in the present models, is assumed as a function
of disk angular position. The curve is based on constant pressure drop across the valve
and the inherent characteristic expressed by C
v
is shown in Fig 2.3 for a check valve
used in the present model. The maximum and minimum angular position of the disk is
80 and 0 in degree. Then, Eqn 2.8 can be rewritten as
T
h
=
HrA
v
(2.10)
in whichr is the distance from the hinge to the point of application of the average pressure
change across the valve, andA
v
is the disk area over which H acts. Substitution of all
21
0 10 20 30 40 50 60 70 80
0
0.2
0.4
0.6
0.8
1
Disk Angular Position (degree)
C
d
(flow coefficient)
Figure 2.3: Flow Coecient Used For Present Model
torques into Equation(2.4) gives
I
d
2
dt
2
=W
c
r
c
sin +s +
rA
v
QjQj
2g(C
d
A
o
)
2
(2.11)
Each torque of the equation as described above would be classied into two categories,
either the opening torque or closing torque. The opening torque includes the torque
due to weight of the rotating disk, T
w
, and the external torque due to spring, T
e
. The
hydraulic torque may be either opening or closing torque depending on the
ow direc-
tion. The values of each torque is calculated at each time step. Under the steady state
condition it is clear that the opening torque is greater than the closing torque. When
unsteady state is created in the system, if the opening torque is less than the opening
torque, the check valve may accelerate its move toward the close position. The charac-
teristics of valves may be determined, considering how the valves move along with the
22
Start
Get
, ,
v
Q
0
max
0
0
v
Q
C O
T T
max
C O
T T
Yes Yes
Yes
No
No
Get f C
d
Get new
0 max
Get
V O
A A ,
Return
Yes
Yes Yes
No
No
max 0
Figure 2.4: Computer Algorithm For Coupling Model
changes in
ow conditions. The parameters such as a response time(i.e. quick closing,
equal percentage), a closing time, linearity, and so on. The combination of the valve
equation and the pipe equations is essential to study the
ow-valve interaction. Fig 2.4
represents the
ow chart of programming algorithm for the coupling model used in the
present study. With given initial values of Q, , and
_
, new , and
_
can be updated
for each time step. Since the response time of the system is rapid, it is necessary to use
a higher-order integration scheme in handling the equation. Therefore, the fourth-order
Runge-Kutta method is used to solve the dierential equation.
23
Chapter 3
Method of Characteristics
A numerical solution of the governing equations for unsteady pipe
ow presented in the
previous chapter is introduced in this chapter. Also, various boundary conditions are
presented.
3.1 Solution by Method Of Characteristics
The popular method of characteristics is a simple and numerically ecient way of solving
the unsteady
ow equations. The continuity and momentum equations form a pair of
hyperbolic partial dierential equations in terms of dependent variables, V and H, and
independent variables,x andt. By using the method of characteristics, two partial dier-
ential equations can be transformed into the four ordinary dierential equations. These
equations are then integrated to obtain a nite dierence representation of the variables.
In the method of characteristics, each boundary and each conduit section are analyzed
separately during a time step. This advantage allows a powerful tool particularly for the
24
analysis of systems having complex boundary conditions. The characteristics method is
developed in more detail in this section.
The governing equations are summarized again as following:
L
1
=
@H
@t
+
a
2
g
@V
@x
= 0 (3.1)
L
2
=g
@H
@x
+
@V
@t
+
f
D
VjVj
2
+k
2
@V
@t
a
@V
@x
= 0 (3.2)
The momentum equation in Eqn 3.2 is rearranged as
L
2
= (1 +k
2
)
@V
@t
ak
2
@V
@x
+g
@H
@x
+
f
D
VjVj
2
= 0 (3.3)
The transformation of Equation(3.3) by the method of characteristics, multiplied by
a linear multiplier , gives two pairs of the ordinary dierential equations. Now, let's
consider the linear combination, L=(g=a)L
1
+L
2
.
(1 +k
2
)
@V
@t
+
(k
2
)a
1 +K
2
@V
@x
+
g
a
@H
@t
+
a
@H
@x
+
f
D
VjVj
2
= 0 (3.4)
With H =H(x;t) and V =V (x;t), the total derivatives may be written by chain rule,
Eqn 3.5 and 3.6, as:
dV
dt
=
@V
@t
+
@V
@x
dx
dt
(3.5)
dH
dt
=
@H
@t
+
@H
@x
dx
dt
(3.6)
Therefore, Eqn 3.4 can be written as:
(1 +k
2
)
dV
dt
+
g
a
dH
dt
+
fVjVj
2D
= 0 (3.7)
dx
dt
=
(k
2
)a
1 +k
2
=
a
(3.8)
25
-
x
6
t
t
6
?
x
t
t
t
t
-
@
@
@
@
@
A
P
R
B
C
+
C
Figure 3.1: Characteristic Lines In x-t Plane
The unknown multiplier can be determined from Eqn 3.8:
= 1 +k
2
or 1 (3.9)
Substitution of these values of into Eqn 3.7 leads to two pairs of ordinary dierential
equations, identied as C
+
and C
equations.
C
+
: (1 +k
2
)
dV
dt
+
g(1 +k
2
)
a
dH
dt
+
fVjVj
2D
= 0 (3.10)
dx
dt
=
a
1 +k
2
(3.11)
C
: (1 +k
2
)
dV
dt
g
a
dH
dt
fVjVj
2D
= 0 (3.12)
dx
dt
=a (3.13)
Eqn 3.10 and 3.12 are not valid everywhere in the x-t plane. The equations are valid
only along the straight lines(if the wave speed, a, is constant) given by Eqn 3.11 and
3.13, respectively. In Fig 3.1, Eqn 3.11 and 3.13 represent two straight lines having
26
slopes +
1+k
2
a
and
1
a
. To satisfy these characteristic relations, the x-t grid requires to
be chosen to ensure the Courant condition which is:
t
x
1
a
(3.14)
Once initial conditions and the grid are specied, Eqn 3.10 and 3.12 can be integrated
along the C
+
and C
characteristic lines.
C
+
:H
P
=C
P
B
P
Q
P
(3.15)
C
:H
P
=C
M
+B
M
Q
P
(3.16)
in which the coecients are all known constants when the equations are applied.
C
P
=H
R
+Q
R
(C
a
RjQ
R
j (1))
B
P
=C
a
+RjQ
R
j
C
M
=H
B
Q
B
((1 +k
2
)C
a
RjQ
B
j)
B
M
= (1 +k
2
)C
a
R =
fx
2gDA
2
C
a
=
a
gA
=a linearization constant
The weighting term in
uences the friction approximation on the third integral term
in Eqn 3.10 and 3.12 without changing the discretization terms such as t, x, or a.
Thus, it provides an excellent way of assessing the sensitivity of a transient simulation
to friction values. The term Q
R
and H
R
can be determined by using linear time line
27
interpolation such as
Q
R
=Q
A
I
P
(Q
A
Q
A
0) (3.17)
H
R
=H
A
I
P
(H
A
H
A
0) (3.18)
I
P
=
p
1 +k
2
1 (3.19)
and prime indicates the value at previous time step.
By eliminating Q
P
in the characteristic equations,
H
P
=
C
P
B
M
+C
M
B
P
B
P
+B
M
(3.20)
Q
P
=
C
P
C
M
B
P
+B
M
(3.21)
The values of H and Q are found at grid intersection points P
2
, P
3
, :::, P
N1
at jt.
Then time is incremented by t and the procedure is repeated for interior points.
3.2 Boundary Conditions
A number of simple boundary conditions used in the present model are introduced in
this section. Also, complex boundary condition such as for a check valve are developed.
3.2.1 Constant-Level Upstream Reservoir
If the volume of reservoir is considerably large, so that the changes in the reservoir level
may be small during the time period of interest, the water level in the reservoir can be
28
assumed constant. If the entrance losses as well as the velocity head are negligible, then
H
P
(1; 1) =H
res
Q
P
(1; 1) = (H
P
(1; 1)C
M
(1; 2))=B
M
(1; 2) (3.22)
For the present model, Q
P
(1; 1) is given as a known function of time (e.g., control valve
at upstream end). Hence, H
P
(1; 1) is calculated from Eqn 3.22.
H
res
Datum
Flow Pipe1
(1; 1)
-
6
?
Figure 3.2: Constant-Level Upstream Reservoir
3.2.2 Series Junction
A series junction is a junction of two conduits have dierent diameters, wall thicknesses,
wall materials, and/or friction factors. A simple junction connecting two pipes is shown
in Fig 3.3. If the dierence in the velocity heads at sections(i;N) and (i + 1; 1) and the
head losses at the junction are negligible, it can be written from the energy equation and
the continuity.
H
P
(i;N) =H
P
(i + 1; 1) (3.23)
Q
P
(i;N) =Q
P
(i + 1; 1) (3.24)
29
i
th
conduit (i + 1)
th
conduit
(i;N) (i + 1; 1)
Figure 3.3: Series Junction
From the characteristic equation for each conduit it is followed:
H
P
(i;N) =C
P
(i;N 1)B
P
(i;N 1)Q
P
(i;N) (3.25)
H
P
(i + 1; 1) =C
M
(i + 2; 2) +B
M
(i + 1; 2)Q
P
(i + 1; 1) (3.26)
It follows from Equation(3.23) through Equation(3.26) that
Q
P
(i;N) =
C
P
(i;N 1)C
M
(i + 1; 2)
B
M
(i + 1; 2) +B
P
(i;N 1)
(3.27)
Q
P
(i + 1; 1) =Q
P
(i;N) (3.28)
3.2.3 Valve At Downstream End
At the pipe outlet, the continuity condition at the downstream end requires that Q
P
=
Q
out
. If Q
P
is given as a function of time, from the C
+
equation, it follow that
H
P
=C
P
B
P
Q
P
(3.29)
3.2.4 Back
ow Prevention Assembly
A Back
ow prevention assembly is a mechanical device located within a given pipeline,
usually as close as possible to the drinking water service connection. Ideal back
ow
prevention assemblies close at an instant the
ow reverses. However, more realistically,
30
some level of reversal
ow usually occurs because of the inertia of the system. This
instantaneous stoppage of the reverse
ow causes the corresponding pressure rise called
valve slam. Fig (3.4) illustrates a schematic of a single check valve with nodes a and b
as the interconnecting junctions on both sides of the internal valve.
@
@
@
@
@
@
- -
(1;N)
node
(2; 1)
node
Conduit1 Conduit2 a b
C
+
C
Figure 3.4: A Schematic of Check Valve Between Two Pipes
For positive
ow with H
1;N
=H
a
and H
2;1
=H
b
, the valve equation is
Q
1;N
= Q
2;1
=Q
v
= C
d
A
o
p
2gH
= C
d
A
o
p
2g (H
a
H
b
) (3.30)
=)Q
2
v
= (C
d
A
o
)
2
(2g) (H
a
H
b
)
= (C
d
A
o
)
2
(2g) [(C
p1
C
M2
) (B
P1
+B
M2
)Q
v
]
Q
2
v
+ (C
d
A
o
)
2
(2g) (B
P1
+B
M2
)Q
v
(C
d
A
o
)
2
(2g) (C
p1
C
M2
) = 0 (3.31)
31
A quadratic equation to solve Equation(3.31) is given:
Q
v
=C
v
(B
P1
+B
M2
) +
q
C
2
v
(B
P1
+B
M2
)
2
+ 2C
v
(C
P1
C
M2
) (3.32)
where C
v
= C
2
d
A
2
o
g. Similarly for the negative
ow, Q
v
is calculated by using the
following equation.
Q
v
=C
v
(B
P1
+B
M2
)
q
C
2
v
(B
P1
+B
M2
)
2
2C
v
(C
P1
C
M2
) (3.33)
It is noted that a negative
ow is possible only ifC
P1
C
M2
< 0. Hence Equation(3.32)
is used if C
P1
C
M2
>= 0, and Eqn 3.33 is used if C
P1
C
M2
< 0. Once the
ow
is known, the characteristic equations are used to obtain the hydraulic heads for each
sectiona andb. The back
ow prevention assembly with a single check valve inside used
in the computer simulations is modeled for three dierent types:
An ideal valve type
undamped type
damped type
The model for an ideal valve does not allow a reversed
ow through the assembly (Q
v
is
always positive or zero) while other models may allow a back
ow in the system (Q
v
is
either positive or negative).
3.3 Coupling Modeling
The combination of the valve equations and the pipe equations is essential to study the
uid-valve interaction and to study the dynamic motion of check valves during transient
32
events. The pipe equations include the continuity and momentum equations introduced
in Eqn 2.1 and 2.3. The valve equations include the moment-of-momentum equation of
the check disk and the valve equation introduced in Eqn 2.11 and 3.30, respectively. At
each time step, the coupling model starts with initial disk angle positon, , and initial
disk angular velocity,
_
. The
ow coecient, C
d
, is given for a specic valve type and is
linearly interpolated. With Q
v
, , and
_
, the moment of momentum equation is solved
numerically by the Runge-Kutta 4
th
method to estimate new values of Q
v
0
,
0
, and
_
0
.
The prime represent the values of variables at the next time step. And coecients are
updated using new variables of Q
v
0
,
0
, and
_
0
.
33
Chapter 4
Experiments
Now two dierent sets of experimental records are presented in this chapter. The data
collected from the experiments will be used in the next chapter to compare with the
numerical simulations in an eort to determine varying coecients for a specic system.
Case 1 represents a simple valve-pipeline-valve system without a back
ow prevention
assembly installed. The comparisons are presented for case 1, one with quasi-steady fric-
tion only and the second with the unsteady friction coecient, K
2
, included to simulate
the unsteady friction eects during rapid transient events. Case 2 represents a system
within which a back
ow prevention assembly is installed. The comparison between two
dierent cases, one with a back
ow prevention assembly and the second without back
ow
prevention assembly.
34
4.1 Experimental Setup
As illustrated in Fig 4.1, a 84.73 m long galvanized iron steel pipe with 0.0525 m in
diameter was installed for experiments. In order to reduce the occupied space the pipeline
was set up bent using ten 90-degree-elbow ttings. The system was fed from upstream
main pipeline where the line pressure is normally maintained to 150 lbf=in
2
[psi]( 105
meter of water). Two control valves of ball valve type were installed at each end of
pipelines. Both control valves are then closed simultaneously by manual operation to
create the transient
ow. A positive pressure wave is created at downstream end of
pipeline and then it propagates to upstream. A negative pressure wave, by contrast,
is created at upstream end of pipeline, propagating to downstream. The line pressure
under the steady-state
ow condition was able to be controlled by the pressure-reducing
valve installed prior to the upstream control valve. In both Case 1 and Case 2, the initial
conditions are kept the same as soon as possible, except for the installation of a back
ow
prevention assembly for the system. Pressure time history data were measured by two
pressure sensors. Two pressure transducers were embedded at location of 20.4 m and
63.4 m measured from the upstream end. For the measurement of the
ow rate, a pair
of
ow meters was used. To regulate the line pressure of the system a pressure reducing
valve is installed at the upstream end just before the upstream control valve. Many tests
are performed varying the line pressure by adjusting the pressure valve. The physical
properties of the pipe material and water are summarized in Table 4.1.
35
(1) Upstream Shut-o Valve (2) Downstream Shut-o Valve (3) Transducer#1
(4) Transducer#2 (5) Back
ow Preventer (6) Flow Meter (7) Pressure Regulator
Figure 4.1: Testing Loop
According to the evaluation of total head losses by H.J. Kwon [10], the total head
loss of the system is a summation of the frictional loss and the minor losses for the
various items contained in the system. The eect of the minor losses to the system's
total head loss is signicant in the region of lower Reynolds number, used in the present
experiment.
Although many cases of experiments have been performed during the study, two cases
will be selected to represent two dierent pipeline system, equipped with/or without a
hydraulic device such as back
ow prevention assemblies. For case 2, the location of a
back
ow prevention assembly is measured 35.8 m from the upstream end.
36
Property Value
Total Length of Pipe 84:7344 m
Diameter of Pipe 5:25 10
2
m
Thickness of Pipe 3:912 10
3
m
Number of Elbow pipe ttings 10
Bulk Modulus of Pipe Material 2:07 Gpa
Young's Modulus of Pipe Material 206:86 Gpa
Density of Water 999:1845kg=m
3
Kinematic Viscosity of Water 1:13 10
6
m
2
=sec
Table 4.1: Physical Properties Of System
4.2 Case 1: Without a Back
ow Prevention Assembly
A set of experimental data for this case was obtained for a simple valve-pipeline-valve
system with the following characteristics: The line pressure at the upstream end under
the steady state condition was simply measured by a line pressure gauge. The line
pressure at upstream is maintained at 84:3683 m and the
ow rate before the control
valves are close is maintained at 7:88610
4
m
3
=sec. Since the shut-o valves at both end
were controlled manually, the detailed characteristics of closing or opening mechanism
is not available. Therefore, the changes in
ow rate at upstream and downstream are
given as a function of time, Q
up
=f
1
(t) and Q
down
=f
2
(t). It was estimated from the
measured data that the upstream shut-o valve started closing at 0.701 sec and completed
37
closing at 0.732 sec and the downstream shut-o valve started closing at 0.714 sec and
completely closed at 0.7355 sec. By linear interpolation, the changes in Q are given as
Q
up
=0:0254t + 0:0186 and Q
down
=0:0367t + 0:027. These linear equations for the
control valves will be used for boundary conditions of the computer simulation.
Hres
a
a
V2 = 0 V1 = 0
V0
H1
- H2
Datum
(a) 0tL=a
Hres a
V = 0
-H
H
Datum
a
a a
(b) L=at 2L=a
Figure 4.2: Sequence of Water Hammers Occurred in System
38
Fig 4.2 describes the sequence of water hammer event occurring in the system after
the shut-o valves at both ends close at the same time. At the instant of that the
downstream valve closes, the
uid nearest the upstream control valve is compressed
creating an extra pressure, H. The high pressure moves upstream at the wave speed
of a as a positive wave. Similarly, a negative wave that is created by the upstream
valve closure moves downstream at the same wave speed. When two opposite waves
rencounter in the middle of pipe, there may be extra energy losses due to friction. After
the instant of encounter between two waves, the velocity of
uid is everywhere zero. At
the instant ofL=a, the waves arrive at the ends of pipe and bound back to the backward
direction. Since the valves completely close, the pressure therefore drops to H
o
H
at the upstream end and jumps to H
o
+ H at the downstream end. The re
ected
waves reach at the ends of pipe at t = 2L=a. This process is repeated every 2L=a. The
action of friction between
uid and pipe wall damps out the oscillation of the waves and
eventually causes the waves to come to rest permanently.
Fig 4.3 and 4.6 are the pressure time history measured at the upstream and down-
stream locations, respectively. The gures clearly show that two pressure waves are
generated from upstream and downstream ends and they oscillate with the period of
2L=a. The magnitude of the pressure wave is quickly reduced due to great energy losses
mainly by friction. Evaluation of damping is discussed in detail in the next chapter 5.
39
0 1 2 3 4 5 6 7 8 9 10
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
Measured At Location of 20.4 m (Upstream)
Figure 4.3: Case 1 Pressure Time History At Upstream For 0t 10
0 0.5 1 1.5 2 2.5 3 3.5 4
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
Measured At Location of 20.4 m (Upstream)
Figure 4.4: Case 1 Pressure Time History At Upstream For 0t 4
40
4 4.5 5 5.5 6 6.5 7 7.5 8
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
Measured At Location of 20.4m (Upstream)
Figure 4.5: Case 1 Pressure Time History At Upstream For 4t 8
0 1 2 3 4 5 6 7 8 9 10
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
Measured At Location of 63.4 m (Downstream)
Figure 4.6: Case 1 Pressure Time History At Downstream For 0t 10
41
0 0.5 1 1.5 2 2.5 3 3.5 4
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
Measured At Location of 60.4 m (Downstream)
Figure 4.7: Case 1 Pressure Time History At Downstream For 0t 4
4 4.5 5 5.5 6 6.5 7 7.5 8
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
Measured At Location of 60.4 m (Downstream)
Figure 4.8: Case 1 Pressure Time History At Downstream For 4t 8
42
4.3 Case 2: With A Back
ow Prevention Assembly
This experiment is designed to nd the eects of water hammer on a back
ow prevention
assembly and the interactions between the transient
ow conditions and an assembly
installed in the system. A back
ow prevention assembly with a single internal check
valve is installed at location of 35.8 m away from the upstream end of pipeline. Similarly
to Case 1, pressure waves were created from both end by a sudden closure of both shut-o
valves. The line pressure at upstream is maintained at 84:2 m and the
ow rate before
the control valves are close is maintained at 7:886 10
4
m
3
=sec. It was shown that the
upstream shut-o valve started closing at 0.714 sec and completed closing at 0.737 sec
and the closing time of the valve is 0.023 second. The downstream shut-o valve started
closing at 0.712 sec and completed closing in 0.027 sec. The experiment conditions for
each case are summarized in Table 4.2 and the comparison plot of two cases is presented
in Fig 4.18. As seen in Fig 4.18, the response of the system with a back
ow prevention
assembly is more complex than one of just a simple system. The rst peak of two dierent
cases correspond very closely until the back
ow prevention assembly is closed. It is noted
that with the nearly same
ow condition and operation condition, same pressure waves
must be created and then moved in the same physical pattern until other hydraulic
devices like a back
ow prevention assembly in this study, interacts with the
ow in the
system. After the rst peak, the oscillation of the downstream transient continues with
a time increment of 4L=a, where L represents the distance from the downstream end
to the back
ow prevention assembly location and a represents the wave speed. Once
43
the internal check valve is closed, the pipeline is divided into two sections where are
distinguished from dynamic behavior. This experiment clearly shows the occurrence of a
check valve slam resulting from a sudden closure of the internal check disk and the
uid-
valve interaction during transient events. In Figs 4.9 and 4.12, experimental results for
a back
ow prevention assembly installed in the middle of the test pipeline are presented.
Information Case 1 Case 2
Line Pressure Head (m) 84.3683 84.2
Steady-State Discharge Rate (m
3
=s) 0.0007571 0.0007571
Wave Speed (m=s) 1367.2 1367.2
Closing Time of SOV Upstream (sec) 0.031 0.023
Closing Time of SOV Downstream (sec) 0.0245 0.027
Table 4.2: Summary of Flow Conditions For Cases
44
0 1 2 3 4 5 6 7 8
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
Measured at Upstream
Figure 4.9: Case 2 Pressure Time History At Upstream For 0t 8
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
Measured at Upstream
Figure 4.10: Case 2 Pressure Time History At Upstream For 0t 2
45
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
Measured At Upstream
Figure 4.11: Case 2 Pressure Time History At Upstream For 2t 4
0 1 2 3 4 5 6 7 8
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
Measured At Downstream
Figure 4.12: Case 2 Pressure Time History At Downstream For 0t 8
46
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
Measured at Downstream
Figure 4.13: Case 2 Pressure Time History At Downstream For 0t 2
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
CASE 1 with Check Valve at Downstream
Figure 4.14: Case 2 Pressure Time History At Downstream For 2t 4
47
0.7 0.72 0.74 0.76 0.78 0.8 0.82 0.84 0.86 0.88
20
40
60
80
100
120
140
160
Time (sec)
Piezometric Head (m)
Measured At Downstream ( 0.7 < t < 0.9 )
B
C
A
D
Figure 4.15: Case 2 Pressure At Downstream for 0:7t 0:9
As shown more in Fig 4.15, the check valve slam phenomenon is observed from this
experiment. Point A corresponds to the closing of the downstream control valve, and is
the start of the deceleration of the
ow at the back
ow prevention assembly. At point
B the control valve completes closing and then the internal check valve continues to
close from Point B to Point C. The maximum reverse
ow is established at Point C
and the valve disk strikes the seat causing slam and another water hammer within the
downstream region. Then the pressure wave created by a valve slam oscillates back and
forth. It is also noted that the energy losses for this case is much larger than those for
case 1. This damping eects may be resulted from the interaction between the motion
of an internal check valve and
ow conditions.
48
4.4 Comparison of Measured Data
Between Case 1 and Case 2
Fig 4.16 and Fig 4.17 present the pressure time histories at the upstream region and
Fig 4.18 and Fig 4.19 present the pressure time histories at the downstream region
after the shut-o valves close simultaneously. The oscillations in pressure occur quickly
being constrained within certain bounds, diminish rapidly with time, and nally reach
to another steady states. It is necessary to re-examine the importance of the role that
back
ow prevention assemblies play during the transient event. The pressure head at the
upstream in Case 1 is reduced by 50 m (from 84 to 34) in 0.02 seconds and bounds up to
130 m, the maximum. The range (maximum minus minimum) of pressure head is about
96 m. The maximum pressure head in Case 2, on the other hand, is 72 m, resulting
in the range of 38 m. Considering the pressure heads at the downstream region, Case
1 shows the pressure range of 109 m (from 130 to 21) while Case 2 shows the range
of 66 m (from 140 to 74). Local
ow separation caused by the motion mechanism of
back
ow prevention assemblies during the transients have the biggest in
uence on energy
dissipation particularly for the initial time of the transient event. It is clearly seen in
Figs 4.16and 4.18 that the energy losses in the upstream region is much greater than the
energy losses in the downstream region.
49
1 1.5 2 2.5 3 3.5 4 4.5
20
40
60
80
100
120
140
160
Time (sec)
Piezometric Head (m)
Comparison of Cases At Upstream
Case 1 (without BPA)
Case 2 (with BPA)
Figure 4.16: Comparison of Two Cases at Upstream for 0:6t 4:6
0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6
20
40
60
80
100
120
140
160
Time (sec)
Piezometric Head (m)
Comparison of Cases At Upstream
Case 1 (without BPA)
Case 2 (with BPA)
Figure 4.17: Comparison of Two Cases at Upstream for 0:6t 1:6
50
1 1.5 2 2.5 3 3.5 4 4.5
20
40
60
80
100
120
140
160
Time (sec)
Piezometric Head (m)
Comparison of Cases At Downstream
Case 1 (without BPA)
Case 2 (with BPA)
Figure 4.18: Comparison of Two Cases at Downstream for 0:6t 4:6
0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6
20
40
60
80
100
120
140
160
Time (sec)
Piezometric Head (m)
Comparison of Cases At Downstream
Case 1 (without BPA)
Case 2 (with BPA)
Figure 4.19: Comparison of Two Cases at Downstream for 0:6t 1:6
51
Information Upstream Region Downstream Region
Initial Pressure Head at t = 0 84 m 84 m
Range of Pressure dierence in Case 1 96 m 109 m
Range of Pressure dierence in Case 2 38 m 66 m
Percentage of Reduction 60.4 % 39.4 %
Table 4.3: Summary of Pressure Reduction Rates by Back
ow Preventer
As shown in Table 4.3, the back
ow prevention assembly reduces the pressure ranges
(which dened as maximum minus minimum value) by 60.4 percent and 39.4 percent for
the upstream and downstream wave respectively. The experimental result shows that
back
ow prevention assembly in general acts as a damper to the excessive pressure wave
created in the pipe system during the transient.
52
Chapter 5
Simulation Results and Discussion
The computer simulation results are presented in this chapter. First, the computer
simulation results are obtained by using both the quasi-steady 1D model and unsteady
friction 1D model for Case 1 without a back
ow prevention assembly. Verication of the
validity of the numerical models is then performed by comparing the numerical results
with experimental data. The unsteady friction 1D model is later expanded to incorporate
with the coupling model in order to solve for Case 2. The comparison of the computer
results with the experimental data for Case 2, verify the present model's validation for
the rapid transient
ow in the water distribution pipe system.
5.1 Selection of F
T
and k
2
The equivalent head loss coecient, F
T
has been used to replace with the Darcy-
Weisbach friction coecient for calculation to represent the total head losses due to
not only friction but minor losses. Because of the unique feature of the experimental
53
piping system, the energy losses due to minor-loss items could not be neglected. There-
fore, a new friction coecient must be chosen with much greater value than the Darcy-
Weisbach friction coecient. The value of F
T
was nalized to 0.3 by trial-and-error
until the agreement shown in the gures was obtained. It is veried from these results
that the agreement on the water hammer wave attenuation between the measured data
and the quasi-steady models gets better as the equivalent head loss coecient increases
as shown from Fig 5.1 through Fig 5.6. Also, the value of k
2
used for the unsteady
friction model was chosen by trial-and-error until the agreement shown in the gures
was obtained while the value of the steady friction coecient was xed to 0.3 for Case
1. The k
2
values used for Case 1 and Case 2 are 0.045 and 0.4 respectively. The value
selected fork
2
was held as a constant value during computing process. Parameters used
for computer simulations are summarized in Table 5.1.
5.2 Verication of Models: Case 1
The rst attempt with the quasi-steady model is made with dierent friction coecients
to simulate for Case 1. As seen in Figs 5.1,5.2, and 5.3, the greater the friction coecient
is, the greater the damping eect becomes. Fig 5.7 shows the experimental record at
the location of downstream transducer together with simulation results by the quasi-
steady friction model for Case 1. Fig 5.7a, shows a good agreement over the rst ten
oscillatory periods of time. The decay rate with F
T
= 0:3 is in good agreement with the
recorded data. However, for the long-term period of time after two second the computed
pressure oscillation appears to be slightly faster. This result shows the inconsistency
54
Parameters Quasi-Steady 1D Unsteady Friction 1D
Number of Pipes 1 2
Increment of x (x) 0.6725 m 0.6758 m, 0.6701 m
Increment of t (t) 0.0005 sec 0.00049
Number of Nodes 127 53, 75
Duration of Computation 10 sec 8 sec
Equivalent Friction Factor, F
T
0.3 0.3
k
2
NA 0k
2
< 1:0
NA 0 1:0
Table 5.1: Input Parameters Used In Simulation
in the assumption of the classical water hammer theory that the water hammer waves
propagate at a constant speed. A local vapour cavity formed at shut-o valves may
in
uence on the small changes in the wave speed.
Fig 5.8 shows the experimental record at the same location of downstream transducer
together with simulation results by the unsteady friction model with F
T
= 0:3 and
k
2
= 0:045 over 6 seconds. Both the oscillatory period and the decay rate agree quite
well with the experiment record. The numerical result by using the unsteady friction
model in the method of characteristics provides the reasonable result, especially for a
long-term oscillatory periods.
55
0 1 2 3 4 5 6 7 8
20
40
60
80
100
120
140
160
Quasi−steady Model (Upstream)
Piezometric Head (m)
Time(sec)
When f = Darcy Weisbach coefficient
Figure 5.1: Quasi-steady Model with f = Darcy-Weisbach friction coecient for 0t 8
0 1 2 3 4 5 6 7 8
20
40
60
80
100
120
140
Quasi−steady Model (Upstream)
Piezometric Head (m)
Time(sec)
When f = 0.15
Figure 5.2: Quasi-steady Model with f = 0.15 for 0t 8
56
0 1 2 3 4 5 6 7 8
20
40
60
80
100
120
140
160
Piezometric Head (m)
Time(sec)
Quasi−steady Model (Upstream)
When f = 0.3
Figure 5.3: Quasi-steady Model with f = 0.3 for 0t 8
0 1 2 3 4 5 6 7 8
20
40
60
80
100
120
140
160
Quasi−steady Model (Downstream)
Piezometric Head (m)
Time(sec)
When f = Darcy Weisbach Coeffient
Figure 5.4: Quasi-steady Model with f = Darcy-Weisbach friction coecient for 0t 8
57
0 1 2 3 4 5 6 7 8
20
40
60
80
100
120
140
160
Quasi−steady Model (Downstream)
Piezometric Head (m)
Time(sec)
Figure 5.5: Quasi-steady Model with f = 0.15 for 0t 8
0 1 2 3 4 5 6 7 8
20
40
60
80
100
120
140
160
Quasi−steady Model (Downstream)
Piezometric Head (m)
Time(sec)
Figure 5.6: Quasi-steady Model with f = 0.3 for 0t 8
58
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
Measured
Quasi 1D Model
(a) 0t 2
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
(b) 2t 4
4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 6
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
(c) 4t 6
Figure 5.7: Case 1 Simulated By Quasi-Steady 1D Model With F
T
= 0:3
59
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
At Location of Downstream
Measured
Computed by present model
(a) 0t 2
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
(b) 2t 4
4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 6
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
(c) 4t 6
Figure 5.8: Case 1 Simulated By Unsteady Friction 1D Model With F
T
= 0:3 and
k
2
= 0:045
60
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
Measured
Quasi(f = 0.3)
UNSTEADY (k
2
= 0.045)
(a) 0t 2
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
(b) 2t 4
4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 6
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
(c) 4t 6
Figure 5.9: Case 1 Comparison of Two Dierent Models At Upstream
61
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
Measured
Quasi (f=0.3)
Unsteady (k
2
= 0.045)
(a) 0t 2
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
(b) 2t 4
4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 6
20
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
(c) 4t 6
Figure 5.10: Case 1 Comparison of Two Dierent Models At Downstream
62
It is noted from Fig 5.7 and 5.8 that if one is only interested in the maximum
response and its decay ratio, there is no signicant dierence between two simulation
results. However, if the phase shift of pressure wave for a long-term period of time
needs to be simulated, it seems that the frequency-dependent analysis by the unsteady
friction model ts better with the experiment data. Since the understanding the
ow
and structure interactions during the transient requires more precise model, the unsteady
friction model is incorporated with a coupling model for the simulation of Case 2 where
the back
ow prevention assembly is set up.
The main point of the comparison here is to compare the dierence between re-
sults from the traditional analysis with quasi-steady assumption and results from the
frequency-dependent analysis with consideration of the additional energy loss due to un-
steady friction and the phase shift of pressure waves. The numerical result for Case 1
obtained from the unsteady friction model shows a better agreement with the recorded
data, hence the governing equations for the unsteady friction model are used later with
the coupling model for numerical analysis of the system with a back
ow prevention
assembly.
5.3 Implementation of A Coupling Model: Case 2
The present model is based on the unsteady friction 1D model coupling with the motion
equation of a back
ow prevention assembly. In Fig 5.25 and 5.26, the simulation results
by the present model are compared to the experimental record for Case 2. In the begin-
ning modeling a back
ow prevention assembly, the assembly is considered as an assembly
63
having an internal single check valve. The total head loss coecient (or equivalent head
loss coecient), F
T
, is 0.3 inclusive of minor losses and the constant value of k
2
equals
to 0.045 as same as the values for Case 1. The time histories of pressure head for the
present model are shown in Figs 5.11, 5.12, 5.13, and 5.14. The changes in
ow rate is
shown in Fig 5.15 and the angular position of check disk which is a function of
ow rate
and
ow acceleration is presented in Fig 5.16. In Fig 5.15, a reversal
ow at maximum
ow rate of3:514 10
4
m
3
=sec through the assembly is observed between 0.754 and
0.780 second. Most of back
ow prevention assemblies are designed as an fast responding
type to minimize allowable reversal
ows so as to considerably enhance the eectiveness
as a protective device.
5.3.1 Sensitivity of k
2
In order to look a degree of sensitivity of the parameter, k
2
, the present model is then
tested with dierent k
2
values. The computation results are shown in Figs 5.11, 5.12,
5.17, 5.18, 5.21, 5.22, 5.25, and 5.26.
64
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
At Location of Downstream
Measured
Present Model k
2
= 0.045
(a) 0t 2
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
(b) 2t 4
4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 6
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
(c) 4t 6
Figure 5.11: Case 2 (Downstream) Present Model With F
T
= 0:3 and k
2
= 0:045
65
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
20
40
60
80
100
120
Time (sec)
Piezometric Head (m)
At Location of Upstream
Measured
Present Model k
2
= 0.045
(a) 0t 2
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
20
40
60
80
100
120
Time (sec)
Piezometric Head (m)
(b) 2t 4
4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 6
20
40
60
80
100
120
Time (sec)
Piezometric Head (m)
(c) 4t 6
Figure 5.12: Case 2 (Upstream) Present Model With F
T
= 0:3 and k
2
= 0:045
66
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
40
50
60
70
80
90
100
110
120
130
140
Time (sec)
Piezometric Head (m)
At Location of Downstream
Measured
Present Model k
2
= 0.045
Figure 5.13: Case 2 (Downstream) Present ModelF
T
= 0:3 andk
2
= 0:045 for 0:7t
1:5
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
20
30
40
50
60
70
80
90
100
110
120
Time (sec)
Piezometric Head (m)
At Location of Upstream
Measured
Present Model k
2
= 0.045
Figure 5.14: Case 2 (Upstream) Present ModelF
T
= 0:3 andk
2
= 0:045 for 0:7t 1:5
67
0.7 0.71 0.72 0.73 0.74 0.75 0.76 0.77 0.78 0.79 0.8
−4
−2
0
2
4
6
8
x 10
−4
Time (sec)
Q (m
3
/sec)
At Location of Assembly
Figure 5.15: Case 2 Flow Rate Changes At Assembly for 0:7t 0:8
0.73 0.735 0.74 0.745 0.75 0.755 0.76 0.765 0.77
0
10
20
30
40
50
60
70
80
Time (sec)
Disk Angle Position
Fully open
Fully closed
Figure 5.16: Case 2 Disk Angle Position
68
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
At Location of Downstream
Measured
Present Model k
2
= 0.08
(a) 0t 2
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
(b) 2t 4
4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 6
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
(c) 4t 6
Figure 5.17: Case 2 (Downstream) Present Model With F
T
= 0:3 and k
2
= 0:08
69
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
20
40
60
80
100
120
Time (sec)
Piezometric Head (m)
At Location of Upstream
Measured
Present Model k
2
= 0.08
(a) 0t 2
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
20
40
60
80
100
120
Time (sec)
Piezometric Head (m)
(b) 2t 4
4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 6
20
40
60
80
100
120
Time (sec)
Piezometric Head (m)
(c) 4t 6
Figure 5.18: Case 2 (Upstream) Present Model With F
T
= 0:3 and k
2
= 0:08
70
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7
40
50
60
70
80
90
100
110
120
130
140
Time (sec)
Piezometric Head (m)
At Location of Downstream
Measured
Present Model k
2
= 0.08
Figure 5.19: Case 2 (Downstream) Present Model F
T
= 0:3 and k
2
= 0:08 for 0:7t
1:5
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
20
30
40
50
60
70
80
90
100
110
120
Time (sec)
Piezometric Head (m)
At Location of Upstream
Measured
Present Model k
2
= 0.08
Figure 5.20: Case 2 (Upstream) Present ModelF
T
= 0:3 andk
2
= 0:08 for 0:7t 1:5
71
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
At Location of Downstream
Measured
Present Model k
2
= 0.4
(a) 0t 2
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
(b) 2t 4
4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 6
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
(c) 4t 6
Figure 5.21: Case 2 (Downstream) Present Model With F
T
= 0:3 and k
2
= 0:2
72
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
20
40
60
80
100
120
Time (sec)
Piezometric Head (m)
At Location of Upstream
Measured
Present Model k
2
= 0.2
(a) 0t 2
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
20
40
60
80
100
120
Time (sec)
Piezometric Head (m)
(b) 2t 4
4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 6
20
40
60
80
100
120
Time (sec)
Piezometric Head (m)
(c) 4t 6
Figure 5.22: Case 2 (Upstream) Present Model With F
T
= 0:3 and k
2
= 0:2
73
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
40
50
60
70
80
90
100
110
120
130
140
Time (sec)
Piezometric Head (m)
At Location of Downstream
Measured
Present Model k
2
= 0.2
Figure 5.23: Case 2 (Downstream) Present ModelF
T
= 0:3 andk
2
= 0:2 for 0:7t 1:5
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
20
30
40
50
60
70
80
90
100
110
120
Time (sec)
Piezometric Head (m)
At Location of Upstream
Measured
Present Model k
2
= 0.2
Figure 5.24: Case 2 (Upstream) Present Model F
T
= 0:3 and k
2
= 0:2 for 0:7t 1:5
74
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
At Location of Downstream
Measured
Present Model k
2
= 0.4
(a) 0t 2
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
(b) 2t 4
4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 6
40
60
80
100
120
140
Time (sec)
Piezometric Head (m)
(c) 4t 6
Figure 5.25: Case 2 (Downstream) Present Model With F
T
= 0:3 and k
2
= 0:4
75
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
20
40
60
80
100
120
Time (sec)
Piezometric Head (m)
At Location of Upstream
Measured
Present Model k
2
= 0.4
(a) 0t 2
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 4
20
40
60
80
100
120
Time (sec)
Piezometric Head (m)
(b) 2t 4
4 4.2 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 6
20
40
60
80
100
120
Time (sec)
Piezometric Head (m)
(c) 4t 6
Figure 5.26: Case 2 (Upstream) Present Model With F
T
= 0:3 and k
2
= 0:4
76
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
40
50
60
70
80
90
100
110
120
130
140
Time (sec)
Piezometric Head (m)
At Location of Downstream
Measured
Present Model k
2
= 0.4
Figure 5.27: Case 2 (Downstream) Present ModelF
T
= 0:3 andk
2
= 0:4 for 0:7t 1:5
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
20
30
40
50
60
70
80
90
100
110
120
Time (sec)
Piezometric Head (m)
At Location of Upstream
Measured
Present Model k
2
= 0.4
Figure 5.28: Case 2 (Upstream) Present Model F
T
= 0:3 and k
2
= 0:4 for 0:7t 1:5
77
The discrepancies between the measured and computed results in Case 2 may be
caused by the dramatic changes in the wave speed. The repetitive opening and closing
movement of the check disk installed in the intermediate section of the pipe may cause
the formation of vapour cavitation and the turbulent
ows around the disk. As a result,
the wave speed reduces causing observable phase shifts at the upstream and downstream
regions. As previously observed in Figs 4.16 and 4.18, the wave speeds for Case 1 and
Case 2 are dierent. In Fig 4.16, there found a signicant energy damping at upstream
after 1 second. The
uctuation at the downstream continues even after 6 seconds while
the
uctuation at the upstream ends very quickly.
With the given assumption that the wave speed is held constant during oscillation,
a phase shift may also occur when the
ow is interfered in its direction (due to the
excessive bends in the experimental set-ups).
Although dierent in detail, the simulated result follows general trends of valve slam
phenomena. The pressure waves generated in the pipe system seems to experience a
phase-shift when it encounters sudden changes in
ow pattern, such as re
ection, trans-
mission, or turning direction. The phase shift seen in Fig 5.25 is much larger than that
for Case 1. Also, it is clear that the energy losses across the assembly become larger
than what these are expected. It is concluded that the delay of oscillation of the pres-
sure waves and the fast energy damping eect are caused largely by the formation of
cavitation and the occurrence of
ow separations around the assembly.
78
5.3.2 Classication of Assemblies:
Ideal, Undamped, And Damped
The present model is also implemented for the assemblies with three dierent types of the
assembly (ideal, undamped and damped). The shapes, decay rates, and periods of the
pressure waves created in the system are seriously aected by the assembly's behavior.
Depending on the time of valve closure, the path of moving disks, and the maximum
reverse
ow allowance, results may be widely dierent. For example, in Fig 5.30, the
pressure rise caused by the check valve slam is the greatest for the undamped valve. As
the time of closure increases, the intensity of the pressure wave weakens.
0.74 0.75 0.76 0.77 0.78 0.79 0.8 0.81 0.82 0.83
−5
−4
−3
−2
−1
0
1
2
3
x 10
−4
Time (sec)
Q (m
3
/sec)
At Location of Assembly
Ideal
Damped (Quick)
Damped (Slow)
Undamped
Figure 5.29: Changes on Flow Rate At Assembly
79
0.7 0.72 0.74 0.76 0.78 0.8 0.82 0.84 0.86 0.88
10
20
30
40
50
60
70
80
90
100
110
Upstream
Time (sec)
Piezometric Head (m)
Ideal
Damped (Quick)
Damped (Slow)
Undamped
Figure 5.31: (Upstream) Comparison of Responses With Dierent Valve Types
0.7 0.75 0.8 0.85 0.9 0.95 1
40
50
60
70
80
90
100
110
120
130
140
Downstream
Time (sec)
Piezometric Head (m)
Ideal
Damped (Quick)
Damped (Slow)
Undamped
Figure 5.30: (Downstream) Comparison of Responses With Dierent Valve Types
80
Chapter 6
Conclusion And Recommendation
6.1 Summary and Conclusion
The present study contains experimental and numerical analysis for rapid transients in
water distribution systems involving back
ow prevention assemblies. Two sets of funda-
mental equations for unsteady pipe
ows are examined and its numerical results solved
by the method of characteristics are compared with a set of experimental data. Also, the
moment-of-momentum equation developed for a back
ow prevention assembly with an
internal check valve has been incorporated into a coupling model for the simulation of the
valve-
uid interaction, on basis of the unsteady friction 1D model. The hydraulic torque
in the moment-of-momentum equation is estimated by using a valve equation relating
ow to pressure drop across the assembly with the
ow rate. Fourth-order Runge-Kutta
integration scheme is used to solve the moment-of-momentum equation for angle incre-
ments for each time step. Pure water hammer waves are observed in Case 1 and check
81
valve slam phenomena is observed in Case 2. Case 2 is simulated by the present model.
The assembly internal valve used in the computer simulations is modeled as an ideal,
undamped, damped fast-responding, and damped slow-responding type. Ideal back
ow
prevention assemblies close in the instant of the
ow is zero, preventing back
ow. In re-
ality, however, a back
ow prevention assembly allows a certain level of back
ow through
it. The sudden stoppage of reverse
ows creates another pressure rises called check valve
slam, propagating in the system.
The key equations used for the coupling model in the present study is summarized
as follows:
1. Quasi-steady friction model
@H
@t
+
a
2
g
@V
@x
= 0
g
@H
@x
+
@V
@t
+
f
D
VjVj
2g
= 0
2. Unsteady friction model
@H
@t
+
a
2
g
@V
@x
= 0
g
@H
@x
+
@V
@t
+
f
D
VjVj
2
+K
2
@V
@t
a
@V
@x
= 0
3. Wave Speed
a =
s
K
w
=
w
1 + (1p
2
) (K
w
=E
p
) (D=e)
4. Moment-of-momentum equation
T
w
+T
e
+T
f
+T
h
=I
d
2
dt
2
82
5. Hydraulic torque, T
h
T
h
=
HrA
v
=
rA
v
QjQj
2g(C
d
A
o
)
2
6. Valve discharge, Q
v
Positive
ow:
Q
v
= C
v
(B
P1
+B
M2
) +
q
C
2
v
(B
P1
+B
M2
)
2
+ 2C
v
(C
P1
C
M2
)
when C
P1
C
M2
0
Negative
ow:
Q
v
= C
v
(B
P1
+B
M2
)
q
C
2
v
(B
P1
+B
M2
)
2
2C
v
(C
P1
C
M2
)
when C
P1
C
M2
0
or
= 0
for ideal valve
The conclusions made from this present study are as follows:
1. The sudden changes in
ow rate at both ends of pipeline create excessive pressure
rises in the distribution system.
2. The pressure waves created in the distribution system propagate at the speed of
wave and decay with time.
83
3. The speed of pressure waves is assumed to be a constant for the present models.
However, the measurement shows a signicant reduction in the wave speed and an
observable phase shift.
4. The damping of the pressure
uctuations in the system under transient conditions
is considerably greater than that estimated by the steady state friction relationship.
5. The quasi-steady approximation clearly exhibits the discrepancies between the ex-
perimental data and computer calculation, in magnitude and phase of pressure
waves particularly for long-time-period records of rapid transient events.
6. The present model with the equivalent loss coecient, F
T
, using the method of
characteristics is successfully implemented to simulate additional energy losses due
to the unsteady friction eect during water hammer events.
7. To simulate the dynamic behavior of a back
ow prevention assembly when pres-
sure waves pass through the assembly, the water hammer equations (continuity
and momentum equations) are solved for each time step simultaneously with the
moment-of-momentum equation of a swinging check disk installed inside the back-
ow prevention assembly.
8. A review of time history curve indicates that when a rapid transient occurs, the
oscillatory waves are decayed quickly and delayed due to the phase shift.
9. The signicant damping and phase shift of the pressure
uctuations when a back-
ow prevention assembly is installed within the system, are observed in the mea-
84
surement. The changes in
uid properties due to air entrainment and cavity for-
mation in the assembly may result in signicant reduction in the wave speed.
10. Also, the excessive bends in the experimental set-up is a major feature that may
in
uence on the wave speed.
11. The property of pipe wall materials and the rigidity of systems may in
uence the
intensity of water hammer waves.
12. Cavity due to the repetitive opening and closing movement of a check valve disk,
turbulent eects, uncertainties of measurement and input data, approximate de-
scription of boundary conditions, and the systematic errors in the numerical model
cause the deformation of the water hammer waves.
13. Local
ow separation caused by the motion mechanism of back
ow prevention
assemblies during the transients have the biggest in
uence on energy dissipation
and wave speeds for the initial time of the transient event.
14. The energy losses due to a back
ow prevention assembly itself (not caused by
friction) in the upstream region are much greater than the energy losses in the
downstream region, resulting in the dierent damping ratio.
15. Precise modeling of a back
ow prevention assembly is essential to obtain good
computer results of transient modeling.
85
6.2 Recommendation
This research represents a study of the numerical modeling for the unsteady pipe
ow
and for the dynamic behavior of a back
ow prevention assembly with a single internal
valve when it interacts with changes of
ows in a closed conduit system. The present
model is not complete but eective for the rapid transient modeling. The future study
will extend to the study for the additional aecting factors on the water hammer wave
formation, propagation, and energy dissipation. There are many problems involved in
modeling dierent types of assemblies to deal with the unsteady friction term. This does
seem to be a very fruitful area for further research.
86
Bibliography
[1] B. Brunone, U. Golia, and M. Greco. Modelling of fast transients by numerical
methods. In International meeting on hydraulic transients with water column sepa-
ration, 9th Round Table, IAHR, pages 273{282, Valencia, Spain, 1991.
[2] B. Brunone, U. Golia, and M. Greco. Some remarks on the momentum equations for
fast transients. In International meeting on hydraulic transients with water column
separation, 9th Round Table, IAHR, pages 201{209, Valencia, Spain, 1991.
[3] M. Carstens and J. Roller. Boundary-shear stress unsteady turbulent pipe
ow. Journal of the Hydraulics Division, American Society of Civil Engineers.,
85(HY2):67{81, 1959.
[4] M. Chaudhry. Applied Hydraulic Transients. Van Nostrand Reinhold in New York,
2nd edition, 1987.
[5] A. K. H. Koetizer and C. Lavooij. Dynamic behavior of large non-return valves. In
Procs. 5th Int. Conf. on Pressure Surges, Hannover, 1986.
[6] C. Jaeger. Theorie Generale du Coup de Belier. Dunod, Paris, 1933.
87
[7] C. Jaeger. Engineering Fluid Mechanics. Blackie, London, 1956.
[8] N. Joukowski. ber den hydraulischen stoss in wasserleitungsrhren (in german), on
the hydraulic hammer in water supply pipes. Proceedings of the Annual Convention:
American Water Works Association. Translated from German By O Simin 1904.
[9] R. O. J.W. Daily, W.L. Hankey and J. Jordan. Resistance coecient for accelerated
and decelerated
ows through smooth tubes and orices. Trans. ASME, 78(7):1070{
1077, 1956.
[10] H. J. Kwon. Transient Flow In Water Distribution System. PhD thesis, University
of Southern California, 2005.
[11] E. P. Management. Control Valve Handbook. Fisher Controls International LLC,
4th edition, 2005.
[12] J. Parmakian. Water-HammerAnalysis. Prentice-Hall Englewood, 1st edition, 1955.
[13] G. Provoost. The dynamic behavior of non-return valves. In Procs. 3rd Int. Conf.
on Pressure Surges, Canterbury, 1980.
[14] G. Provoost. The dynamic characteristic of non-return valves. In 11th IAHR Symp.
of the Section on Hydraulic Machinery, Equipment and Cavitation, Operation Prob-
lems of Pump Stations and Power Plants, Amsterdam, 1982.
[15] G. Provoost. A critical analysis to determine the dynamic characteristic of non-
return valves. In Procs. 4th Int. Conf. on Pressure Surges, Bath, 1983.
88
[16] G. Rich. Waterhammer analysis by the laplace-mellin transformations. Transaci-
tions of ASME, pages 1944{1945, 1944.
[17] G. Rich. Hydraulic Transients. McGraw-Hill, New York, 1st edition, 1951.
[18] V. Streeter and C. Lai. Waterhammer analysis including
uid friction. Transactions
of American Society of Civil Engineers, (128):1491{1524, 1963.
[19] V. Streeter and E. Wylie. Hydraulic Transients. McGraw-Hill, New York, 1967.
[20] A. Thorley. Dynamic response of check valves. In Procs. 4th Int. Conf. on Pressure
Surges, Bath, 1983.
[21] A. Thorley and J. Oei. Dynamic behavior of a swing check valve. In 11th IAHR
Int. Symp. on Water Column Separation, Obernach, 1981.
[22] A. Trikha. An ecient method for simulating frequency-dependent friction in tran-
sient liquid
ow. Journal of Fluids Engineering, 97(1):97{105, 1975.
[23] A. Vardy and J. Brown. Transient, turbulent, smooth pipe friction. Journal of
Hydraulic Research., 33(4):435{456, 1995.
[24] A. Vardy and J. Brown. Transient turbulent friction in smooth pipe
ows. Journal
of Sound and Vibration., 259(5):1011{1036, 2003.
[25] A. Vardy and J. Brown. Ecient approximation of unsteady friction weighting
functions. Journal of Hydraulic Engineering., 130(11):1097{1107, 2004.
89
[26] A. Vardy and J. Brown. Transient turbulent friction in fully-rough pipe
ows.
Journal of Sound and Vibration., 270(1-2):233{257, 2004.
[27] D. J. Wood. Waterhammer analysis-essential and easy (and ecient). Journal of
Environmental Eng., 131(8):1123{1131, 2005.
[28] E. Wylie and V. Streeter. Fluid transients in systems. Prentice-Hall, 2nd edition,
1993.
[29] W. Zielke. Frequency-dependent friction in transient pipe
ow. Journal of Basic
Eng., 90(1):109{115, 1968.
90
Abstract (if available)
Abstract
When control valves at the end of pipeline close simultaneously, two pressure waves are generated at each end and the waves propagate toward to the other end. The pressure waves continue to move back and forth along pipelines until they are damped out to next steady states. This study provides information on the experimental data and the numerical simulation of a rapid hydraulic transient event called water hammer. The energy loss term due to friction in the present model consists of quasi-steady contribution and unsteady contribution. For the present model, an equivalent friction coefficient is used to replace the quasi-steady friction coefficient, inclusive unsteady friction loss and minor energy loss factors. The unsteady component has been related to the combination of the instant flow acceleration and instant flow convective acceleration. The numerical results of the present model are compared with the experimental records. The computer results by the present model which is based on the unsteady friction 1D model was successful to follow general trends of water hammer phenomena, corresponding with sudden changes in flow. This study later extends to the dynamic characteristics of backflow prevention assemblies under a rapid transient condition. When numerous hydraulic devices are installed on water distribution systems, the dynamic characteristics of such hydraulic devices have a significant influence on the intensity of the associated water hammer waves. A backflow prevention assembly plays important roles as not only a safety hydraulic device but also an energy dissipater. A numerical program for rapid transient pipe flow interconnected with a backflow prevention assembly has been developed using the present coupling model and numerically solved by the method of characteristics.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Open channel flow instabilities: modeling the spatial evolution of roll waves
PDF
Numerical analysis of harbor oscillation under effect of fluctuating tidal level and varying harbor layout
PDF
Numerical study of flow characteristics of controlled vortex induced vibrations in cylinders
PDF
Wave induced hydrodynamic complexity and transport in the nearshore
PDF
Computational fluid dynamic analysis of highway bridge superstructures exposed to hurricane waves
PDF
A coastal development idea for Gulf of Thailand to improve global trades
PDF
The development of a hydraulic-control wave-maker (HCW) for the study of non-linear surface waves
PDF
Vision-based and data-driven analytical and experimental studies into condition assessment and change detection of evolving civil, mechanical and aerospace infrastructures
PDF
Novel techniques for analysis and control of traffic flow in urban traffic networks
PDF
Experimental and kinetic modeling studies of flames of H₂, CO, and C₁-C₄ hydrocarbons
Asset Metadata
Creator
Kim, Hyoung-Jin
(author)
Core Title
Numerical and experimental study on dynamics of unsteady pipe flow involving backflow prevention assemblies
School
Viterbi School of Engineering
Degree
Doctor of Philosophy
Degree Program
Civil Engineering
Publication Date
11/20/2012
Defense Date
10/24/2012
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
backflow prevention assemblies,OAI-PMH Harvest,unsteady pipe flow,water hammer
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Lee, Jiin-Jen (
committee chair
), Masri, Sami F. (
committee member
), Moore, James Elliott, II (
committee member
), Wellford, L. Carter (
committee member
)
Creator Email
hyoungjk@gmail.com,hyoungjk@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-116209
Unique identifier
UC11292211
Identifier
usctheses-c3-116209 (legacy record id)
Legacy Identifier
etd-KimHyoungJ-1313.pdf
Dmrecord
116209
Document Type
Dissertation
Rights
Kim, Hyoung-Jin
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
backflow prevention assemblies
unsteady pipe flow
water hammer