MASTER OF SCIENCE
In
MECHANICAL ENGINEERING
Approved by:
________________________________ __________________
Nellore S. Venkataraman, PhD Date
Member, Graduate Committee
________________________________ __________________
Vikram Raj Pandya, PhD Date
Member, Graduate Committee
________________________________ __________________
Walter Silva Araya, PhD Date
President, Graduate Committee
________________________________ __________________
Andres Calderon, PhD Date
Representative of Graduate Studies
________________________________ __________________
Paul Sundaram, PhD Date
Chairperson of the Department
ABSTRACT
The objective of this study is to extend SilvaAraya and Chaudhry’s unsteady friction
model for water hammer analysis to series and branching pipe systems. The conduits at
these systems may have different diameters, wall thickness, wall material, and/or friction
factors.
For the computation of the initial steady state conditions a subroutine that solves the
network flow equations using the Hardy Cross linear analysis is used. The subroutine
allows for the input of minor losses, use of either the DarcyWeisbach or HazenWillians
ii
To create a “user friendly” Windows interface, the extended application was
developed using VBA (Visual Basic for Applications) in Excel for general program
applications.
The model was tested for three cases: single, series and branching piping systems.
The computed pressures are compared with the experimental data obtained for Silva
Araya (1993) for a single piping system case. For a series and a branching piping system
the computed results obtained by using unsteady friction are compared with the data
iii
RESUMEN
válvula. El modelo aproxima los perfiles de velocidad durante flujo transitorio y calcula
iv
de Ariete. Las tuberías en estos sistemas pueden tener diferentes diámetros, espesor de
Para el cómputo del estado permanente inicial se utiliza una subrutina que resuelve
las ecuaciones de flujo de la red, usando el análisis lineal de Hardy Cross. La subrutina
tiene en cuenta las pérdidas menores de entrada, el uso de las perdidas por fricción en
desarrolló utilizando VBA (Visual Basic for Applications) en Excel para las aplicaciones
El modelo fue probado para tres casos: sistemas de tubería sencillos, en serie y
obtenidos para Silva Araya (1993) para el caso de sistemas de tubería sencillos. Para
v
vi
ACKNOWLEDGEMENTS
I want to express a sincere acknowledgement to my advisor, Dr. Walter Silva for his
Rico, Mayagüez Campus for giving me the opportunity to carry out my master studies.
I would like to thank my mother and my wife, for their support, inspiration and love.
vii
Table of Contents
ABSTRACT .................................................................................................................................................II
RESUMEN ................................................................................................................................................. IV
ACKNOWLEDGEMENTS ....................................................................................................................VII
TABLE LIST ................................................................................................................................................X
FIGURE LIST ............................................................................................................................................ XI
1 INTRODUCTION .............................................................................................................................2
1.1 MOTIVATION .................................................................................................................................3
1.2 LITERATURE REVIEW ....................................................................................................................3
1.3 SUMMARY OF FOLLOWING CHAPTERS ..........................................................................................8
2 GOVERNING EQUATIONS...........................................................................................................9
2.1 CLOSED CONDUIT TRANSIENT FLOW GOVERNING EQUATIONS ....................................................9
2.1.1 Continuity Equation.................................................................................................................9
2.1.2 Momentum Equation ...............................................................................................................9
2.2 ENERGY DISSIPATION FUNCTION..............................................................................................11
2.3 MOMENTUM EQUATION FOR AXISYMMETRIC PIPE FLOW .......................................................14
2.4 METHOD OF CHARACTERISTICS ................................................................................................15
2.4.1 Characteristics Equations .....................................................................................................16
2.4.2 Boundary Conditions.............................................................................................................23
3 SYSTEM SPECIFICATION ...........................................................................................................37
3.1 SCHEMATIC DIAGRAM...............................................................................................................37
3.2 NUMBERING OF NODES .............................................................................................................38
3.3 NODE TYPES...............................................................................................................................40
3.4 NUMBERING OF CONDUITS .......................................................................................................41
4 METHODOLOGY ...........................................................................................................................42
4.1 COMPUTATION OF THE INITIAL STEADY STATE CONDITIONS ...................................................42
4.1.1 Generalized Network Equations............................................................................................43
4.1.2 Linearization of System Energy Equations............................................................................49
4.1.3 Hardy Cross Method .............................................................................................................50
4.2 GENERATION OF THE COMPUTATIONAL GRID ..........................................................................53
4.3 TURBULENCE MODELING ...........................................................................................................55
4.3.1 Turbulence Model for the Inner Region ................................................................................57
4.3.2 Outer Region Model and Blending Relation..........................................................................58
viii
4.4 INITIAL VELOCITY PROFILE ........................................................................................................61
4.5 COMPUTATION OF THE PRESSURE GRADIENT ...........................................................................63
4.6 SHEAR STRESS DISTRIBUTION .....................................................................................................64
4.7 COMPUTATION OF THE VELOCITY PROFILE ...............................................................................65
4.7.1 Predictor Step........................................................................................................................66
4.7.2 Corrector Step .......................................................................................................................67
5 VERIFICATION...............................................................................................................................69
5.1 SINGLE PIPING SYSTEM ..............................................................................................................69
5.2 SERIES PIPING SYSTEM ...............................................................................................................82
5.3 BRANCHING PIPING SYSTEM .....................................................................................................89
6 CONCLUSIONS ..............................................................................................................................94
APPENDIX A. VALUES OF T VS. TAU FOR DIFFERENT TYPES OF VALVES........................99
APPENDIX A1. BUTTERFLY VALVES...................................................................................................99
APPENDIX A2. CIRCULAR VALVES....................................................................................................101
APPENDIX A3. BIPLANES VALVES.....................................................................................................103
ix
Table List
Tables Page
x
Figure List
Figures Page
xii
1 INTRODUCTION
If sudden velocity and pressures variations are caused in a pipe system, for example
when pumps are shut off or valves are closed, a pressure wave develops which is
transmitted in the pipe at a certain velocity that is determined by fluid properties and the
pipe wall material. This phenomenon, called water hammer, can cause pipe and fittings
rupture. The intermediate stage flow, when the flow conditions are changed from one
steady state condition to another steady state, is called transient state flow or transient
flow; water hammer is a transient condition caused by sudden changes in flow velocity or
pressure.
Generally friction losses in the simulation of transient pipe flow are estimated by
using formulae derived for steady state flow conditions, this is known as the quasisteady
approximation. The head loss during transient flow is equal to the head loss obtained for
steady uniform flow with an average velocity equal to the instantaneous transient velocity.
absence of vapor cavitations or column separation, this is not accurate for the prediction
of the time history of pressure oscillations and the attenuation of pressure waves. An
accurate calculation of the damping effect due to unsteady friction losses is important for
2
1.1 Motivation
Water hammer analysis tools of the past have been noted for being difficult to use and
system design and operation has often been overlooked. Recently, pressure oscillations
caused serious damage in the important pipe conduction called the North Coast Super
The present research extended SilvaAraya and Chaudhry’s (1997) unsteady friction
model to series and branching pipe systems. The extended application was developed
using a Windows interface with VBA in Excel for general program applications. The
Piping systems engineers typically are not specialists in water hammer modeling.
With this work and fundamental knowledge they would be equipped with the proper tools
energy losses in unsteady pipe flows. Historically, the most common method of
representing frictional resistance in unsteady flows has been the use of DarcyWeisbach
steadyflow equation based on the mean velocity, considering that the shear stress at the
3
wall is the same in steadystate flow condition than in transient condition. These models
fail to correctly predict the transient pressure head field beyond the first wave cycle.
Ohmi et al. (1980) studied the flow behavior in turbulent oscillatory flow and found
that the shear stress depends on the ratio of frequency of oscillation and the time
averaged Reynolds number throw each cycle. When this ratio is small, the pressure
gradient and the wall shear stress are in phase with the mean velocity and the quasi
steady approximation is applicable. As this ratio increases, the mean velocity delays from
the shear stress and the pressure gradient. Then the friction factor is different from its
Experimental studies by Ramaprian and Tu (1979) showed that the turbulent structure
in oscillatory flow is different from that in equilibrium conditions, thereby increasing the
timemean shear stress and the friction losses. Therefore, the energy dissipation in
Vardy and Hwang (1991) adopted a fiveregion turbulence model to compute the
eddy viscosity distribution using empirical relations. In each region a different expression
was used to compute the eddy viscosity. This turbulence model is used to solve the
transient governing equations including axial and radial velocity components. The
4
The equations must be solved at every time step at every cross section to obtain the
pressure head and the axial velocity. It was reported that this model is very time
consuming.
axisymmetry unsteadyflow analysis in a pipe system. Taking into account the velocity
profile in the cross section, he used a two layer turbulence model, which divides the flow
into a viscous sublayer and a turbulent sublayer, based on the mixing length hypothesis
friction factor with the help of the turbulence model, developed by Nagano and Hishida
(1987), to compute the eddy viscosity and solve the transient pipe flow governing
equations.
SilvaAraya and Chaudhry (1993) developed an energy dissipation model for the
computation of laminar and turbulent unsteady friction losses and obtained instantaneous
velocity profiles to compute the Reynolds stresses in transient pipe flow. The energy
dissipation factor incorporated in the momentum equation is defined as the ratio of the
energy dissipated in transient flow to the energy dissipated according to the steadystate
5
friction term in the transient flow governing equations. For transient turbulent pipe flow a
mixing length eddy viscosity turbulence model proposed by Granville (1989) was chosen
for the inner region of the boundary layer and a constant eddy viscosity model proposed
Peralta Cooper (2001) replaced the mixing length model in SilvaAraya and
(1988a, 1993b). The results were very similar to those obtained by using Granville’s
mixing length model. As a general conclusion, for most transient conditions, the use of
simple turbulence models provides similar results to twoequation models. The use of
of pressure oscillation. However, unsteady turbulence models are still an area where
water hammer considering the energy dissipation. He studied the water hammer produced
by the instantaneous closure of a valve, located at the end of a pipe line connected to a
constant head reservoir. The model consists of linearizing the equations of transient flow,
6
Zhao and Ghidaoui (2004) formulated, applied, and analyzed first and second order
explicit finite volume Godunovtype schemes for water hammer problems. The finite
volume formulation ensures that both schemes conserve mass and momentum and
such as, valves, pipe junctions, and reservoirs, within the Godunov approach is similar to
that of the method of characteristics approach. The schemes were applied to a system
consisting of a reservoir, a pipe, and a valve and to a system consisting of a reservoir, two
transient laminar liquid flow by the method of characteristics. The procedure consists of
determining an approximate expression for frequencydependent friction such that the use
of this expression requires much less computer storage than the use of the exact
Schohl (1993) used a new approximation to the weighting function in Zileke’s (1967)
frequency dependent friction in transient laminar flow calculations. The new, fiveterm
approximation was fitted to the weighting function using a nonlinear least squares
approach.
7
Vardy and Brown (2004) presented a method for evaluating wall shear stress from
known flow histories in unsteady pipe flows. The method builds on previous work by
Trikha, but has two important differences. One of these enables the method to be uses
with much larger integration time steps than are acceptable with Trikha’s method. The
enables it to be used at indefinitely small times. The method is applicable to both laminar
Most of unsteady flow models with unsteady friction have been tested in simple pipe
systems, typically a valvereservoir system. Recently, some attempts have been done to
apply then in pipe networks. The present work has extended the application of a quasi
the governing equations of the phenomenon. Chapter 3 explains how to describe the
system you are analyzing, how many components the system has, and of what type and
size they are. Chapter 4 describes the procedure followed to obtain the energy dissipation
factor as an integrated part of the transient analysis. Chapter 5 shows the verification of
the generalized algorithm, with the results obtained by using the unsteady friction model
8
2 GOVERNING EQUATIONS
The dynamic equations in transientsstate flow in closed conduits are the equations
for the conservation of mass and the momentum equation. These equations are a set of
hyperbolic partial differential equations. Using the Reynolds Transport Theorem and
assuming onedimensional flow and elastic conduit with slightly compressible fluid,
which stretch or contracts with respect to time; these equations become (Chaudhry, 1987)
∂p ∂p ∂V
+ V + ρa 2
= 0 2.1
∂t ∂x ∂x
∂V ∂V 1 ∂p fV V
+V + + g sin θ + =0 2.2
∂t ∂x ρ ∂x 2D
Where: g = acceleration due to gravity, θ = pipe angle with respect to the horizontal,
9
D = pipe diameter, f = DarcyWeisbach friction factor.
( V∂V / ∂x ), are very small compared to the others terms and may be neglected. Therefore
∂p ∂V
+ ρa 2
= 0 2.3
∂t ∂x
∂V 1 ∂p fV V
+ + = 0 2.4
∂t ρ ∂x 2D
specified datum as, and using the discharge, Q = VA, instead of the flow velocity and
considering the energy dissipation factor in the equation of momentum. Equation 2.3
becomes:
∂H a 2 ∂Q
+ = 0 2.5
∂t gA ∂x
10
Equation 2.4 is modified as:
∂Q ∂H fQ Q
+ gA + ef =0 2.6
∂t ∂x 2 DA
friction losses in space and time. This parameter is set equal to one if the quasisteady
approach in transient pipe flow is used. The term in the energy equation that involves the
viscous and turbulent stresses is commonly called the dissipation function. A procedure
2.2 Energy Dissipation Function
For boundary layer planar or axisymmetric flow, the expression to compute the
∂u ⎛ ∂u ⎞
Φ = ⎜μ − ρ u 'v ' ⎟ 2.7
∂r ⎝ ∂r ⎠
Where: u = axial velocity of the fluid particle along the conduit length, r = radial
11
coordinate measured from the pipe centerline, ∂u/∂r = velocity gradient in the radial
direction, μ = dynamic viscosity of the fluid, ρ = fluid density, ρ u ' v ' = Reynolds
stresses. Equation 2.7 accounts the total shears stresses, viscous and turbulent stresses
and represents the energy dissipation per unit volume and per unit time.
Integration of Equation 2.7 across the conduit area from the wall to the pipe
centerline and multiplying by two can be estimated the dissipation integral. This result
accounts for the rate of energy dissipation per unit length in the conduit area.
Therefore:
R
D I = 2π ∫ 0
Φ rdr 2.8
t2 R
E t = 2π ∫ ∫ Φ rdrdt 2.9
t1 0
Silva Araya (1993) demonstrated that for steadystate the energy dissipation is given
by:
12
fV 3
D Is = ρA 2.10
2g
Where DLs = steady state dissipation integral, f = Darcy weisbach friction factor,
fV 3
E s = ρ A (t 2 − t 1 ) 2.11
2D
Equation 2.9 can be used to compute the unsteady energy dissipation integral if the
velocity profile at any time during the computation and the Reynolds stresses are known.
Division of Equation 2.9 by Equation 2.11 gives the ratio, called the energy
dissipation factor, as defined by Silva Araya and Chaudhry (1997), of the energy
Et
e f = 2.12
Es
The energy dissipation factor is a function timevarying velocity gradient and the
Reynolds stresses for turbulent flow. To solve Equation 2.12 requires a procedure to
compute the instantaneous velocity distribution in transient flow condition and also a
13
turbulence model. The next section describes the procedure implemented to find the
2.3 Momentum Equation for Axisymmetric Pipe Flow
∂u ∂u ∂u 1 ∂p ⎛ ∂ 2u 1 ∂u ⎞
+ ux + ur = − +ν ⎜⎜ + ⎟⎟ 2.13
∂t ∂x ∂r ρ ∂x ⎝ ∂ r 2
r ∂ r ⎠
Where ν = kinematics viscosity of the fluid. The coordinate directions are defined in
Figure 21.
14
Thus after time averaging the axisymmetric equation of motion, the following
∂u ∂ u 'v ' 1 ∂p ⎛ ∂ 2u 1 ∂u ⎞
+ = − + ν ⎜⎜ + ⎟ 2.14
∂t ∂r ρ ∂x ⎝ ∂r
2
r ∂ r ⎟⎠
Where u = time averaged velocity in the axial direction, ρ u ' v ' = Reynolds stress.
Equation 2.14 is the equation to compute the velocity distribution in turbulent transient
pipe flow.
2.4 Method of Characteristics
The equations describing the transientstate flow in closed conduits are quasilinear,
available. However, by neglecting or linearizing the nonlinear term, various graphical and
15
The method of characteristics has become quite popular and is extensively used. For
speed is constant), this method has proven to be superior to other methods in several
aspects, such as correct simulation of steep wave fronts, illustration of wave propagation,
Let us rewrite the momentum and continuity equations (Eqs. 2.5 and 2.6) as:
∂Q ∂H
L1 = + gA + RQ Q = 0 2.15
∂t ∂x
∂Q ∂H
L 2 = a 2
+ gA = 0 2.16
∂x ∂t
In which R = f/(2DA). Let us consider a linear combination of Eqs. 2.15 and 2.16:
L = L1 +λL2
Or
⎛ ∂Q ∂Q ⎞ ⎛ ∂H 1 ∂H ⎞
⎜ + λa 2
⎟ + λ gA ⎜ + ⎟ + RQ Q = 0 2.17
⎝ ∂t ∂x ⎠ ⎝ ∂t λ ∂x ⎠
16
If H = H(x,t) and Q = Q(x,t), then the total derivative
dQ ∂Q ∂Q dx
= + 2.18
dt ∂t ∂x dt
and
dH ∂H ∂H dx
= + 2.19
dt ∂t ∂x dt
1 dx
= = λa 2
2.20
λ dt
i.e.,
1
λ = ± 2.21
a
And by using Equations. 2.18 and 2.19, Equation 2.17 can be written as
dQ gA dH
+ + RQ Q = 0 2.22
dt a dx
If
dx
=a 2.23
dt
17
and
dQ gA dH
− + RQ Q = 0 2.24
dt a dx
If
dx
= −a 2.25
dt
Note that Equation 2.22 is valid if Equation 2.23 is satisfied and that Equation 2.24 is
In the x – t plane, Equations 2.23 and 2.25 represent two straight lines having slopes
± 1/a. These are called characteristic lines. Mathematically, these lines divide the x – t
plane in two regions, which may be dominated by two different kinds of solution, i.e., the
solution may be discontinuous along these lines. Physically, they represent the path
18
dt
Considering a single pipeline (Figure 23) Equations 2.22 and 2.24 are valid along the
pipe length (i.e., for 0 < x < L) and special boundary conditions are required at the ends
19
dt
Referring to Figure 22, knowing the values of Q and H at points A and B, their
values at point P can be determinates by solving Equations 2.22 and 2.24 as follows:
P gA P P
∫A
dQ +
a ∫
A
dH + R ∫ dQ Q dt = 0
A
2.26
The first two integral forms of Equation 2.26 are easily evaluated; however, the third
term is not, representing the friction losses, because the variation of Q with respect to t is
not explicitly known. By using a firstorder approximation, the integral of the third term
is evaluated as:
20
P
R ∫ dQ Q dt ≅ RQ A Q A ( t p − t A ) = RQ A Q A Δ t 2.27
A
In other words, we are saying that for the evaluation of this term, Q remains constant
gA
QP − QA + ( H P − H A ) + R Δ tQ A Q A = 0 2.28
a
Note that Equation 2.28 is exact except for the approximation of the friction term.
This firstorder approximation usually yields satisfactory results for typical engineering
applications.
gA
QP − QB − ( H P − H B ) + R Δ tQ B Q B = 0 2.29
a
By combining the known variables together, we may write Equation 2.28 as:
QP = CP − CaH P 2.30
21
And Equation 2.29 as:
QP = Cn − Ca H P 2.31
gA
CP = QA + H A − R Δ tQ A Q A 2.32
a
gA
Cn = QB − H B − R Δ tQ B Q B 2.33
a
And
gA
Ca = 2.34
a
Note that Equation 2.30 is valid along the positive characteristic line AP and Equation
2.31 along the negative characteristic line BP (Figure 22). The value of the constants CP
and Cn are known for each time step, and the constant Ca depends upon the conduit
properties. We will refer to Equation 2.30 as the positive characteristic equation and
Equation 2.31 as the negative characteristic equation. In Equations 2.30 and 2.31, we
have two unknowns, namely, HP and QP. The values of these unknowns can be
22
2.4.2 Boundary Conditions
Special boundary conditions are required to determine the transientstate head and
discharge at the boundaries. These are developed by solving Equations 2.22 or 2.24, or
both, simultaneously with the conditions imposed by the boundary. These conditions are
in terms of special relationships that define, at the boundary, the discharge, the head, or a
relation between them. Equation 2.22 is used for the downstream boundaries and
We shall designate the upstream and downstream ends with reference to the initial
flow direction even though the flow may reverse during the transient state. A section at
the upstream end of a conduit shall be numbered section 1 and the one at the downstream
different sections, we shall use two subscripts: The first subscript designates the conduit
number, and the second indicates the section number. For example, Qi,j represents flow at
the jth section of the ith conduit. For variables that have the same value at all sections of
a conduit, only one subscript will be used. For example, Cai refers to constant Ca
(Equation 2.34) for the ith conduit. Although CP and Cn may have different values at
different sections of a conduit, only one subscript will be used with them to indicate the
conduit number.
23
2.4.2.1 ConstantLevel Upstream Reservoir
In this case, it is assumed that the water level in the reservoir remains constant during
kQ P2i ,1
he = 2.35
2 gA i2
In which k is the coefficient of entrance loss. Then, referring to Figure 25, we obtain:
Q P2i ,1
H Pi ,1 = H res − (1 + k ) 2.36
2 gA i2
In which Hres = height of the reservoir water surface above the datum.
24
Figure 25. Constantlevel Upstream Reservoir.
with the negative characteristic equation (Equation 2.31). Elimination of HPi,1 from
Equations 2.31 and 2.36 and simplification of the resulting equation yield:
In which:
C a i (1 + k )
k1 = 2.38
2 gA i2
By solving Equation 2.37 and neglecting the negative sign with the radical term, we
obtain:
25
−1+ 1 + 4 k 1 ( C n1 + C a1 H res )
Q Pi ,1 = 2.39
2 k1
If the entrance losses as well as the velocity head are negligible, then:
H Pi ,1 = H res 2.40
In which Hres = height of the reservoir water surface above the datum. Equation 2.31
Q Pi ,1 = C ni + C a i H res 2.41
kQ P2i , n +1
he = 2.42
2 gA i2
Q P2i , n +1
H Pi , n +1 = H res − (1 − k ) 2.43
2 gA i2
26
Elimination of Hpi,n+1 from Equations 2.43 and 2.30 yields:
In which:
C a i (1 − k )
k2 = 2.45
2 gA i2
By solving equation 2.44 and neglecting the positive sing with the radical term:
1 − 1 − 4 k 2 ( C p1 − C a1 H res )
Q Pi , n +1 = 2.46
2k2
If the exit loss and the velocity head are negligible, then:
H Pi , n +1 = H res 2.47
Q Pi , n +1 = C p i − C a i H res 2.48
27
Figure 26. Constantlevel Downstream Reservoir.
At the dead end (Figure 27), QPi,n+1 = 0. Hence, from the positive characteristic
C pi
H Pi , n +1 = 2.49
C ai
28
Figure 27. Dead End.
Q oi , n +1 = ( C d Av ) o 2 gH oi , n +1 2.50
Hoi,n+1 = head upstream of the valve, and Av = area of the valve opening.
Q Pi , n + 1 = ( C d A v ) 2 gH Pi , n + 1 2.51
29
Dividing Equation 2.51 by Equation 2.50, taking square of both sides and defining the
Q 2
=
(Q oi , n +1 τ )
2
H Pi , n +1
Pi , n + 1 2.52
H oi , n +1
Substitution for Hp from the positive characteristic equation (Equation 2.30) into
Q P2i , n + 1 + C v Q Pi , n + 1 − C p i C v = 0 2.53
(
Q Pi , n +1 = 0 . 5 − C v + C v2 + 4 C p i C v ) 2.54
curves (Figure 28b and c) may be specified either in a tabular form or by an algebraic
expression. Note that τ = 1 corresponds to a valve opening at which the flow through the
valve is Qoi,n+1 under a head of Hoi,n+1. Values of t versus τ for different types of valves
30
are given in Appendix A.
(a)
t t
1 1
0 0
t
(b) (c)
31
2.4.2.5 Orifice at Lower End
For an orifice, the opening remains constant. Therefore, the above equations may be
used with τ = 1.
thicknesses, wall materials, and/or friction factors. If the difference in the velocity heads
at sections (I, n+1) and (i+1, 1) (Figure 29) and the head losses at the junction are
H Pi , n +1 = H Pi +1 ,1 2.55
The positive and negative characteristic equations for sections (I, n+1) and (i+1, 1)
are:
Q Pi , n +1 = C p i − C a i H Pi , n +1 2.56
Q Pi +1 ,1 = C n i +1 − C a i +1 H Pi +1 ,1 2.57
Q Pi , n +1 = Q Pi +1 ,1 2.58
32
It follows from Equations 2.55 through 2.55 that:
C p i − C ni +1
H Pi , n +1 = 2.59
C a i + C a i +1
Now Hpi+1, 1, Qpi, n+1, and Qpi+1, 1 can be determined from Equations 2.55 through
2.57.
For the branching junction shown in Figure 210, the following equations can be
written:
1. Continuity equation:
Q Pi , n +1 = Q Pi +1 ,1 + Q Pi + 2 ,1 2.60
33
2. Characteristic equations:
Q Pi , n +1 = C p i − C a i H Pi , n +1 2.61
Q Pi +1 ,1 = C ni +1 + C a i +1 H Pi +1 ,1 2.62
Q Pi + 2 ,1 = C n i + 2 + C a i + 2 H Pi + 2 ,1 2.63
If the head losses at the junction are neglected and it is assumed that the velocity
heads in all conduits are equal, then it follows from the energy equation that:
H Pi , n +1 = H Pi +1 ,1 = H Pi + 2 ,1 2.64
C p i − C ni +1 − C ni + 2
H Pi , n +1 = 2.65
C a i + C a i +1 + C a i + 2
Now Hpi+1, 1 and Hpi+2, 1 can be determined from Equation 2.64; Qpi, n+1, Qpi+1, 1, and
34
Figure 210. Branching Junction.
Figure 211 illustrates a valve positioned at the junction of two pipes that may have
35
The positive and negative characteristic equations for sections (I, n+1) and (i+1, 1)
are given for Eqs. 2.56 and 2.57 respectively. The continuity equation at the junction is
C pi − C ai H p i , n +1 − C ni +1
H p i +1 ,1 = 2.66
C a i +1
C p i − Q p i , n +1
H pi , n +1 = 2.67
C ai
And Qpi,n+1 is computed for Eq. 2.54 for a valve at downstream. Values of t versus τ
36
3 SYSTEM SPECIFICATION
This chapter describes how to specify the pipe system in the computer program.
3.1 Schematic Diagram
conduits or tunnels are conduits and the storage facilities, devices or appurtenances, such
as reservoirs, tanks, valves or pipe junctions are the boundaries. The conduit cross
A square is used for a node and a circle for a conduit. The node number is written
inside the square and the conduit number inside the circle. The type of each node will be
indicated by a numeral written inside a triangle (see Fig. 31). Each node type is
To identify the nodes in the system, they are numbered in a sequence starting at the
upstream end and, number the upstream node as one. Then, proceeding in the
downstream direction, the other nodes are numbered 2, 3, and so on. Fig. 32 illustrates
this procedure. The junction of two conduits is a node called a series junction. The
conduits joining at a series junction may have different diameters, wall thickness, wall
material, and/or friction factors. In a branching piping system having a treelike structure
(Fig. 33), designate one pipeline as the main line, and the remaining pipelines as the
branch lines. The branch lines may have subbranches and the subbranches may be
38
Nodes on the main line are numbered like a series system, i.e., we start at the
upstream end and then proceed in the downstream direction. The nodes on each branch
line are then numbered one branch at a time. The numbering on the branch line starts at
the upstream end of the branch and then proceeds in the initialsteadystateflow
direction. No particular sequence has to be followed to number the branches, i.e., you
decide which branch you want to number as branch no. 1, 2, 3, and so on.
Fig. 33 illustrates this numbering procedure for a branching system. We have
arbitrarily called one line as the main line and the remaining lines as branch lines. We
number the upstream node of the main line as 1 and number the remaining nodes on this
line proceeding in the downstream direction. The last node on the main line is 8. Then,
we select one of the branch lines as number 1 and number the nodes on this line as 9 to
11. Of the remaining two branch lines, we call one of them as branch number 2 and
number the nodes on this line as 12 to 14. Branch number 3 provides inflow to the main
line as compared to the other two lines in which flow is from the main line. We number
the upstream node on this line as 15 and the remaining two nodes on this line as 16 and
17.
39
Figure 33. Schematic Diagram of a Branching System.
3.3 Node Types
Assign numbers to designate the type of a node. For example, an upstream reservoir
is node type 1, downstream reservoir as node type 2, and so on. Table 31 lists various
types of nodes and numbers to designate their type. Write in a triangle the number
designating the type of a node, i.e., the number in the triangle refers to the type of the
node, and the number in the square represents the number of the node.
40
Node Type Node
1 Upstream Reservoir
2 Downstream Reservoir
3 Series Junction
4 Branching Junction
5 Dead End
6 Downstream Valve
7 Orifice at Lower End
8 Intermediate Valve
3.4 Numbering of Conduits
Different nodes are connected by conduits, and number these conduits to distinguish
them from another. It is not necessary to follow any particular sequence in numbering the
conduits as long as each conduit in the system in numbered once and the numbering of
the conduits starts at one. For example, the conduit between nodes 1 and 2 may be
numbered 1, 3, or any other number as long as it is not greater than the local number of
conduits in the system. The numbering of conduits should start at 1 and then continue
consecutively to the total number of conduits in the system. It does not matter if conduit
number 1 is located at the downstream end of the system, conduit no. 2 is located ate the
upstream end of the system, and conduit no. 3 is located as some intermediate location.
41
4 METHODOLOGY
This chapter describes the procedure followed to obtain the energy dissipation factor
includes the application of the Hardy Cross method for the computation of the initial
steady state conditions and the use of a technique to generate the computational grid
without having an excessive number of grid points. The turbulence model is described
and details on the computation of the initial steady state velocity profile. The
approximation of the time dependent pressure gradient and the numerical scheme to solve
4.1 Computation of the initial steady state conditions
For the computation of the initial steady state conditions a subroutine that solves the
network flow equations using the Hardy Cross linear analysis (adapted from [14]) is used.
The subroutine allows for the input of minor losses, use of either the DarcyWeisbach or
HazenWillians pipe frictional loss, and the provision of rated power for pumps. Either SI
It is necessary to know the flow direction in each pipe element at all times because
the required head summations depend on that flow direction. This process is handled by
42
initially designating an assumed flow direction in each pipe element. As the calculations
proceed, the direction of the flow rate is positive if it is in the initial assumed direction
and negative if it reverses direction. Input geometric data require signs based on the
initial assumed flow directions. The subroutine requires input data specifying assigned
pipe numbers for the paths of pipe sections making up primary loops and pipe sections
between fixedgrade nodes. The pipe numbers are input with a positive sign if the
assumed flow direction is in the path direction and with a negative sign if it is opposite.
Networks of piping such as those shown in Figures 4.1 and 4.2 can be represented by
∑ (± ) j Q j − Qe = 0 4.1
In which the subscript j refers to the pipes connected to a node, and Qe is the
external demand. The algebraic plus or minus sign convention pertains to the
assumed flow direction. Use the positive sign for flow into the junction, and the
43
Figure 41. Representative piping network: (a) assumed flow directions and
numbering scheme: (b) designated interior loops; (c) path between two fixed
grade nodes. (From [14])
∑ (± ) Wi i =0 4.2
In which the subscript i pertains to the pipes that make up the loop. There will be
44
a relation for each of the loops. Here it is assumed that there are not pumps
located in the interior of the network. The plus sign is uses if the flow in the
element is positive in the clockwise sense; otherwise, the minus sign is employed.
3. Energy balance along a unique path or pseudo loop connecting two fixedgrade
nodes:
∑ ( ± ) [W
i i − (H )
p i ]
+ ΔH = 0 4.3
Where: ΔH is the difference in magnitude of the two fixedgrade nodes in the path
ordered in a clockwise fashion across the imaginary pipe in the pseudo loop. The term
(Hp)i is the head across a pump that could exist in the ith pipe element. If F is the
number of fixedgrade nodes, there will be (F – 1) unique path equations. The plus and
minus signs in Eq. 4.3 follow the same argument given for Eq. 4.2.
Let P be the number of pipe elements in the network, J the number of interior nodes,
and L the number of interior loops. Then the following relation will hold if the network is
property defined:
P=J+L+F–1
45
each pipe; it is:
W = RQ x
+
∑k Q2 4.4
2
2 gA
8 fL
R = 4.5
gπ 2 D 5
frictional loss, and ∑ K are the minor loss components associated with each pipe
element.
If the minor losses can be defined in terms of an equivalent length, Eq. 4.4 can be
replaced by:
W = RQ x 4.6
8 f [L + (L e )]
R = 4.7
gπ 2 D 5
46
Where: Le is the equivalent length.
H p (Q ) = a o + a 1Q + a 2 Q 2 4.8
The coefficients ao, a1, and a2 are assumed known; typically, they can be found by
substituting three known data points from a specified pump curve and solving the three
an alternative means of including a pump in a line is to specify the useful power the
pump puts into the system. The useful, or actual, power W f is assumed to be constant
.
Wf
H (Q ) = 4.9
p
γQ
47
Figure 42. Two piping networks: (a) tree watering system: P = 17, L = 0, j =
8, F = 10; (b) water distribution system: P = 17, L = 4, J = 12, F = 2. (From
Wood, 1981.)
48
4.1.2 Linearization of System Energy Equations
Equation 4.3 is a general relation that can be applied to any path or closed loop in a
network. If it is applied to a closed loop, ΔH is set equal to zero, and if no pump exists in
Define the function Φ(Q) to contain the nonlinear terms W(Q) and HP(Q) in the form:
φ (Q ) = W (Q ) − H P (Q ) = RQ x − H P (Q ) 4.10
dφ d 2φ
φ (Q ) = φ (Q o ) + (Q − Q o ) + 2 (Q − Q o )2 + ... 4.11
dQ Qo
dQ Qo
chosen so that the difference (Q  Qo) is numerically small. Retaining the first two terms
on the righthand side of Eq. 4.11, and using Eq. 4.10, yields:
⎡ dH P ⎤
φ (Q ) ≅ RQ ox − H P (Q o ) + ⎢ x RQ ox −1 − ⎥ (Q − Q o ) 4.12
⎢⎣ dQ Qo ⎥
⎦
49
The parameter G is introduced as:
dH P
G = x RQ ox −1 − 4.13
dQ Qo
φ (Q ) = R Q ox − H P (Q o ) + (Q − Q o )G
= W o − H Po + (Q − Q o )G 4.14
In which Wo = W(Qo) and Hpo = Hp(Qo). Finally, Eq. 4.14 is substituted into Eq. 4.3 to
∑ ( ± ) [(W ) − (H ) + ∑ [Q
i o i po i i ]
− (Q o )i ]G i + Δ H = 0 4.15
The Hardy Cross method of analysis is a simplified version of the iterative linear
analysis. In Eq. 4.15, let (Qo)I be the estimates of discharge from the previous iteration,
and let Qi be the new estimates of the discharge. Define a flow adjustment ΔQ for each
50
loop to be:
Δ Q = Q i − (Q o )i 4.16
The adjustment is applied independently to all pipes in a given loop. Hence Eq. 4.15
can be written as
∑ ( ± ) [(W ) − (H )
i o i po i
+ Δ Q ∑ G i +Δ H = 0 ] 4.17
ΔQ =
− ∑ (± )[(W ) − (H ) ] − Δ H
o i po i
4.18
∑G i
For a closed loop in which no pumps or fixedgrade nodes are present, Eq. 4.18
ΔQ =
− ∑ (± )(W ) o i
4.19
∑G i
1. Assume an initial estimate of the flow distribution in the network that satisfies
continuity, Eq. 4.1. The closer the initial estimates are to the correct values, the
51
fewer will be the iterations required for convergence. One guideline to use is that
2. For each loop or path, evaluate ΔQ with Eq. 4.18 or 4.19. The numerators should
3. Update the flows in each pipe in all loops and paths, that is, from Eq. 4.16
Q i = (Q o )i + ∑ ΔQ 4.20
The term ∑ ΔQ is used, since a given pipe may belong to more than one loop;
hence the correction will be the sum of corrections from all loops to which the
pipe is common.
4. Repeat steps 2 and 3 until a desired accuracy is attained. One possible criterion to
use is
∑ Q i − (Q o )i
≤ε 4.21
∑ Qi
52
4.2 Generation of the computational grid
The skin friction in turbulent flow is greater than in laminar flow; therefore, the
velocity gradient is greater at the wall. To assure a good resolution of the velocity
gradient close to the wall it is very important to have small grid steps in this region. A
geometrically expanding grid spacing proposed by Cebeci and Smith (1974) was used in
this study. Figure 4.1 shows a typical grid. This was the same grid adopted by Silva
Araya and Chaudhry (1997) to compute the velocity distribution during the transient
flow.
r=0
Pipe Wall
53
This technique is a geometric progression, in the radial direction, which have the
property that the ratio of lengths of any two adjacent intervals is a constant; that is:
h j = Kh j −1 4.22
Where h j −1 the size of the jth 1 interval is, h j is the size of the jth interval, K is the
ratio of two successive intervals. The distance to the position in the radial direction is
Where: h1 the length of the first cell and K is equal to the ratio of two successive
steps. The total number of points J was calculated by the following formula:
ln [1 + ln( K − 1)( R / h1 ) ]
J = 4.24
ln K
Where: k = ratio of two successive steps, R = pipe radius, h1 = length of the first
interval. The implementation of this grid generation technique was based in knowing the
length of the first interval, the pipe radius and the number of grid points (N).
The constant expansion rate was computed using Newton’s iteration method, which
54
starts with an initial guess; the convergence is quadratic. Then the grid growth in the
radial direction until the grid points is equal to the pipe radius. In order to compare the
model results, it is important to have a similar grid; therefore this grid was similar to the
4.3 Turbulence modeling
The turbulent boundary layer can be considered as composed of an inner and outer
region. The inner region characteristics are strongly dependent on the conditions near the
wall and are affected by the pressure gradient and the shear stress distribution. This
region is subdivided into a very thin region near the wall, called the viscous sublayer,
where, where the flow is dominated by viscous effects. Farther from the wall the
turbulent stresses become more important. The transition between the viscous sublayer
and the fully turbulent region is called the buffer layer. Beyond this transition the
turbulent stresses dominate the flow and the average velocity profile can be estimated
using a logarithmic equation. In pipe flows the inner region includes most of the pipe
cross section. Figure 4.4 shows the regions of a turbulent boundary layer.
55
25
Viscous Buffer Logarithmic
Sublayer Region Region
20
15 Outer
Region
+
u
10
0
0.1 1 10 + 100 1000 10000
Y
As a general expression for the Reynolds stresses, Prandtl proposed the mixing length
theory (1925). According to this theory, the Reynolds shear stress is calculated as:
∂u ∂u
− ρ (u ' v ' ) = ρ l 2 4.25
∂r ∂r
Where: l is the mixing length. The “bar” symbol on the velocity was dropped in the
previous equation.
The Reynolds stresses can also be modeled by using the eddy viscosity concept
56
introduced by Boussinesq (1877). The basic assumption is that the turbulent stresses are
proportional to the velocity gradient. The same assumption is made in the mixing length
model. The coefficient of proportionality, ε, was called “eddy viscosity” and is defined as
⎛ ∂u ⎞
− ρ ( u ' v ' ) = ρε ⎜ ⎟ 4.26
⎝ ∂r ⎠
The close relation between the mixing length and the eddy viscosity concepts allows
converting mixing length into eddy viscosity in many cases. A mixinglength model was
used for the inner region of the boundary layer. An eddy viscosity model was selected for
the outer region. The mixinglength values are converted into eddy viscosity to make
The following mixing length model, proposed by Granville (1989), and normalized
by using the wall variables, is used for the inner region of the boundary layer.
l + = ky +
τ +
(1 − e − y + / λ1+
) 4.27
57
(for flows with zero pressure gradient), uτ = friction velocity given by uτ = τ w / ρ ,
Granville (1990) derived the relation between the mixing length model (Eq. 4.27) and
ε+ =
1⎛
( )
+ 2
⎜ 1 + 2l τ
2⎝
+
− 1 ⎞⎟ 4.28
⎠
Where ε + = ε / v is the ratio of the eddy viscosity and the kinematics viscosity.
For the outer region the eddy viscosity is computed using a model proposed by Kita
et al. (1980). This is a constant eddy viscosity values across the outer region given by
ε out
+
= CcR + 4.29
58
For the range 1 X 104 < Re < 2 X 106 the value of Cc is given by (Silva Araya and
Chaudhry, 1993).
According to the criterion given by Kita et al (1980). the outer region starts when the
0 . 154 Re + ⎛ Cc ⎞
y +
= ⎜1 + 1− ⎟ 4.31
lim
k ⎜ 0 . 077 ⎟
⎝ ⎠
A smoother transition between the inner and outer eddy viscosities can be achieved
given by:
e b (y )
+ +
ε + = ε out
+ − y out
4.32
59
Where the exponent b is computed as
b=
Ln ε out
+
( / ε c+ )
+
(
y out − y c+ ) 4.33
The value of ε c+ is linearly interpolated between the beginning of the falling limb of
the eddy viscosity distribution and ε out . y c+ is the value of y+ corresponding to ε c+ . Also
+
y out +
is linearly interpolated between y lim and R+ = uτR/v. The corresponding value of ε
is ε out
+
. Figure 4.5 shows a typical eddy viscosity distribution with the inner, transition and
outer regions.
+
Y
For laminar flow the velocity profile is computed using the following expression
(White, 1991)
1 ⎛ − ∂p ⎞
u = ⎜
4 μ ⎝ ∂x ⎠
⎟ R ( 2
− r2 ) 4.34
For turbulent flow, the initial mean velocity distribution can be written in the form:
u+ = f (y+) 4.35
Where f(y+) is the profile across the pipe, including the sublayer, buffer, logarithmic
If the shear stress, τ+, and the eddy viscosity distribution, ε+, are known, the velocity
profile can be obtained by the following equation (Silva Araya and Chaudhry, 1993).
y+ τ +
u+ = ∫
+
dy 4.36
0 1+ ε +
61
Where u+ = 0 at y+ = 0. For a fully develop steadystate flow, the shear stress τ+,
varies linearly across the pipe section from 0 at the wall to 1 at the pipe centerline.
The gradients of the velocity profile are computed by using a forwardtime, central
spaced explicit scheme modified for a geometrically expanded grid. The partial
∂u u
≈ i +1
( )
+ α 2 − 1 u i − α 2 u i −1
4.37
∂r α (α + 1 )Δ r −
Where α = Δri −1 / Δri is the ratio of two successive grid spaces and Δr − = ri −1 − ri .
In the previous expressions the subscript i refer to the ith position inside the pipe
cross section. A secondorder backward approximation is used for the gradient at the wall
∂u α 2 u N − 2 − (1 + α )2 u N − 1
≈ 4.38
∂r w α (1 + α ) (r N − 1 − r N − 2 )
Where N is the number of grid lines (the grid line at the wall is number N).
Prior to the initiation of the transient flow, the dissipation integral is computed
substituting the derivatives of the initial flow profile into Equation 2.8. The integrated
62
value is compared with the initial steadystate value given by Equation 2.10.
4.5 Computation of the pressure gradient
The computation of the pressure gradient needed in the Equation 3.14 is time
dependent and must be estimated from the local conditions. Given an approximation of
the velocity profile, for example the velocity profile at the previous time step, the
pressure gradient can be estimated from the momentum equation, expressed in a compact
r ⎛ ∂p ⎞ ∂u ∂ ⎛ ∂u ⎞
⎜ ⎟ = r − ⎜ rν ⎟ 4.39
ρ ⎝ ∂x ⎠
t
∂t ∂r ⎝ ∂r ⎠
form yields:
1 ∂p ΔQ 2ν ⎛ ∂u ⎞
= − t
⎜ ⎟ 4.40
ρ ∂x Δ tA R ⎝ ∂r ⎠ r=R
63
where ΔQ / Δt can be approximated by the change in discharge from the previous
time step. Adjustments in the value of the pressure gradient may be required if the flow is
reduced drastically; but Equation 3.51 is a good initial approximation (Silva Araya and
Chaudhry, 1997).
4.6 Shear stress distribution
Equation 2.14 can also be expressed in terms of the shear stress as:
∂u ∂p 1 ∂ (r τ )
ρ = − + 4.41
∂t ∂x r ∂r
r ∂u r 2 ∂p
ρ ∫
0 ∂t
rdr = −
2 ∂x
+ rτ 4.42
Dividing by the shear stress at the wall, the following nondimensional for results:
64
ρ r ∂u r ∂p
τ ∫
+
= rdr + 4.43
rτ w 0 ∂t 2τ w ∂ x
∂u u − ut
≈ t+Δt 4.44
∂t Δ tMOC
Where ΔtMOC is the time step chosen to solve the characteristic equations.
R ∂p
τ w = 4.45
2 ∂x
Equation 4.34 is used to update the shear stress distribution after every time step.
4.7 Computation of the velocity profile
equation for the velocity profile across the unequally spaced grid, developed by Silva
Araya and Chaudhry, (1997). This procedure is reference as the McCormack scheme
predictorcorrector where is second order accurate in space and time scheme (Anderson
et al., 1984). In the predictor a step forward finitedifferences approach is used for the
65
spatial derivatives, and a backward finite differences is used in the corrector step. This
sequence was alternated during the computations from one time step to the next; as
∂u u * − u in
= i 4.46
∂t Δt
∂u un − un
= i +1 + i 4.47
∂r Δr
∂ 2u 2 ( u in+ 1 − (α + 1) u in + α u in− 1 )
= 4.48
∂r 2 α (α + 1)( Δ r − ) 2
the time level, the subscript i refers to the grid point, and the asterisk indicates a predicted
value.
Thus, the momentum equation is solved for u i* (Silva Araya and Chaudhry, 1997) as
n +1
⎛ 1 ∂p ⎞ ⎛ε d ε ⎞ u in+ 1 − u in
u = u + ⎜⎜
* n
⎟⎟ Δt + ⎜ + ⎟ Δt
⎝ ρ ∂x ⎠
i i
⎝ r dr ⎠ Δr+
4.49
u in+ 1 − ( α + 1 ) u in + α u i − 1
+ 2εΔ t
α ( α + 1 )( Δ r − ) 2
66
4.7.2 Corrector Step
The predicted values are used in a backward finitedifference approximation for the
first spatial derivative and are also used for the second spatial derivative. The
approximations are
∂u u * * − u in
= i 4.50
∂t Δt
∂u u * − u i*− 1
= i 4.51
∂r Δr −
∂u 2 2 ( u i*+ 1 − (α + 1 ) u i* + α u i*− 1 )
= 4.52
∂r 2 α (α + 1 )( Δ r − ) 2
Where (**) indicates a corrected value. Therefore, the corrected velocity is:
n +1
⎛ 1 ∂p ⎞ ⎛ ε d ε ⎞ u i − u i −1
* *
u = u + ⎜⎜
** n
⎟⎟ Δ t + ⎜ + ⎟ Δt
⎝ ρ ∂x ⎠
i i
⎝r dr ⎠ Δ r +
4.53
u * − (α + 1) u i* + α u i*−1
+ 2 ε Δ t i +1
α (α + 1)( Δ r − ) 2
Finally, the velocity for the next time step is given by:
u i** − u i*
u in +1 = 4.54
2
The derivative of the eddy viscosity is obtained (Silva Araya, 1993) as:
67
dε ε + (α 2 + 1)ε i − α 2 ε i −1
= i +1 4.55
dr α (α + 1)( ri −1 − ri )
The velocity distribution is computed at every section of the pipe. At the end of the,
method of characteristic (MOC) and for every time step the velocity profile is integrated
across the pipe section to compute the flow discharge. If the integrated discharge differs
from the one predicted by the MOC, the pressure gradient is slightly adjusted to satisfy
the continuity equation. The adjustment is done until both discharges differ by less than
5%.
68
5 VERIFICATION
This chapter presents the results obtained by using the unsteady friction model for
piping systems. The generalized algorithm was analyzed for three cases: single, series
and branching piping systems. The computed pressures are compared with the
experimental data obtained for Silva Araya (1993) for a single piping system case. For a
series and a branching piping system the computed results obtained by using unsteady
friction are compared with the data obtained by using steady friction model.
5.1 Single Piping System
The experimental data obtained by Silva Araya (1993) was used for the verification
of the generalized algorithm in a single piping system. The experimental apparatus used
is shown in Figure 5.1. The system was built at the Albrook Hydraulics Laboratory at
A pump supplies water from the sump to the tank. At the valve, the water discharges into
another open tank and returns to the pump. Four pressure transducers were used to record
the pressure fluctuations for the valve closure in each section. These were located at X =
69
L / 4, X = L / 2, X = 3L / 4 and X = L from the reservoir.
1 2
1 6
The pipe material is commercial steel with an internal diameter of 0.101 m and a wall
thickness of 0.00635 m. The initial steadystate discharge was 0.0165 m3/s. This
corresponds to an average velocity of 20.02 m/s and an initial Reynolds number of 2.06
X 105. The relative roughness for this simulation was 0.001. Other parameters are listed
in Table 5.1. Table 5.2 shows the different types of nodes present in this system.
The transient conditions for this case were produced by a gradual closure of the
downstream valve. The valve closed completely in 1.085 sec. The curve of the effective
valve opening versus time curve used in the simulation is given in Figure 5.2
70
Parameter Value
Initial Discharge 0.0165 m3/s
Upstream Tank Head 2.65 m
Entrance Loss Coefficient 1.44
Friction Factor 0.0223
Initial Valve Position τ0 = 0.62
Wave Speed 1125 m/s
Density 999.1 Kg/m3
Dynamic Viscosity 1.06 X 103 m2/s
0.9
0.8
Effective Opening, TAU
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1 1.2
Time (s)
obtained by using steady friction. The maximum pressure is predicted accurately without
the unsteady friction model. The oscillations damp much more rapidly than the predicted
values. The shift in the frequency of the oscillations is due to the reduction of the wave
72
H Vs T (Valve)
Experimental Data
Steady Friction
60
50
40
Pressure Head (m)
30
20
10

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
(10)
(20)
Time (s)
50
40
Pressure Head (m)
30
20
10

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
(10)
(20)
Time (s)
(b) At x = 3L / 4
Figure 53. Pressure Head Histories for a Single Piping System Using Steady
Friction.
73
H Vs T (x = L/2)
Experimental Data
Steady Friction
60
50
40
Pressure Head (m)
30
20
10

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
(10)
(20)
Time (s)
(c) At x = L / 2
H Vs T (x = L/4)
Experimental Data
Steady Friction
60
50
40
Pressure Head (m)
30
20
10

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
(10)
Time (s)
(d) At x = L / 4
Figure 53. Pressure Head Histories for a Single Piping System Using Steady
Friction (Cont.)
74
Figures 5.4 (a) to 5.4 (d) compare the measured pressure heads with the pressure
heads obtained by the unsteady friction model. The maximum and the minimum
pressures as well as the damping rate are predicted accurately at the four locations along
the pipe.
The effect of the passage of a strong pressure wave created during a transient flow
Figure 5.5 (a) correspond to the initial steadystate velocity profile. Figure 5.5 (b)
through 5.5 (d) present several instantaneous velocities profiles. Figure 5.5 (e) shows that
during transient flow, the average velocity across the section could be almost zero;
however, local velocities could be significant. Figure 5.8 (f) presents a profile during a
The energy dissipation factor as a function of time for different points of the pipe is
75
H Vs T (Valve) Experimental Data
Unsteady Friction
60
50
40
Pressure Head (m)
30
20
10

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
(10)
(20)
Time (s)
50
40
Pressure Head (m)
30
20
10

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
(10)
(20)
Time (s)
(b) At x = 3L / 4
Figure 54. Pressure Head Histories for a Single Piping System Using
Unsteady Friction.
76
H Vs T (x = L/2)
Experimental Data
Unsteady Friction
60
50
40
Pressure Head (m)
30
20
10

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
(10)
Time (s)
(c) At the x = L / 2
H Vs T (x = L/4)
Experimental Data
Unsteady Friction
60
50
40
Pressure Head (m)
30
20
10

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
(10)
Time (s)
(d) At x = L / 4
Figure 54. Pressure Head Histories for a Single Piping System Using
Unsteady Friction (Cont.)
77
Velocity Profile (t = 0s)
0.06
0.05
Distance from Centerline (m)
0.04
0.03
0.02
0.01
0
0 0.5 1 1.5 2 2.5
Velocity (m/s)
(a) At t = 0 s
0.05
Distance from Centerline (m)
0.04
0.03
0.02
0.01
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Velocity (m/s)
(b) At t = 1 s
Figure 55. Velocity Profile in transient Flow at the Middle of the Pipe
78
Velocity Profile (t = 1.5s)
0.06
0.05
Distance from Centerline (m)
0.04
0.03
0.02
0.01
0
0.04 0.02 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
Velocity (m/s)
(c) At t = 1.5 s
0.05
Distance from Centerline (m)
0.04
0.03
0.02
0.01
0
0.04 0.02 0 0.02 0.04 0.06 0.08
Velocity (m/s)
(d) At t = 1.56 s
Figure 55. Velocity Profile in transient Flow at the Middle of the Pipe (Cont.)
79
Velocity Profile (t = 1.625s)
0.06
0.05
Distance from Centerline (m)
0.04
0.03
0.02
0.01
0
0.05 0.04 0.03 0.02 0.01 0 0.01 0.02 0.03 0.04 0.05 0.06
Velocity (m/s)
(e) At t = 1.625 s
0.05
Distance from Centerline (m)
0.04
0.03
0.02
0.01
0
0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0
Velocity (m/s)
(f) At t = 1.75 s
Figure 55. Velocity Profile in transient Flow at the Middle of the Pipe (Cont.)
80
Energy Dissipation Factor Vs Time (x = L / 2)
9000
8000
Energy Dissipation Factor (ef)
7000
6000
5000
4000
3000
2000
1000
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time (s)
(a) At x = L / 2
4500
4000
Energy Dissipation Factor (ef)
3500
3000
2500
2000
1500
1000
500
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time (s)
(a) At x = L / 4
81
5.2 Series Piping System
Hydraulics Transients” was used to apply the generalized algorithm in a series piping
system
Transient conditions in the piping system shown in Figure 5.7 were determined using
steady and unsteady friction. The arrangement consists of a constanthead tank open to
the atmosphere connected by two horizontal pipes located in series to a valve located
1000 m downstream of the tank. The conduits at these systems have different diameters,
wall material, and friction factors. Table 5.3 shows the different types of nodes present in
this system.
Reservoir
1 2 Valve
Qo
3
2
1
6
3
1
The initial steadystate discharge was 1 m3/s. This corresponds to an average velocity
of 2.26 m/s and an initial Reynolds number of 1.69 X 106 for pipe number 1 and an
average velocity of 3.53 m/s and an initial Reynolds number of 2.12 X 106 for pipe
number 2. Other parameters are listed in Table 5.4 and characteristics of the pipes are
Parameter Value
Initial Discharge 1 m3/s
Upstream Tank Head 67.7 m
Number of Pipes 2
Valve Operation Time 6s
Initial Valve Position τ0 = 1
Final Valve Position τf = 0
Density 1000 Kg/m3
Dynamic Viscosity 1.06 X 103 m2/s
83
Transient conditions were caused by closing the valve according to the τ – t curve
shown in Figure 5.8. Seven points on this curve were stored in the computer, and the τ
0.9
0.8
Effective Opening, TAU
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 1 2 3 4 5 6 7
Time (s)
Figure 5.9 shows the geometrically expanding grid spacing used for each pipe. Small
grid steps were used near the wall to assure a good resolution of the velocity gradient.
Figure 5.10 shows the initial eddy viscosity distribution in each pipe, and Figure 5.11
84
DISTANCE FROM CENTER LINE
0.4 Wall
0.35
Wall
0.3
0.25
ETA
0.2
0.15
0.1
PIPE1
0.05 PIPE2
0
0 10 20 30 40 50 60 70 80
GRID LINE NUMBER
Figure 59. Geometrically Expanding Grid Spacing for Series Piping System.
Eddy Viscosity Distribution
4500
4000 PIPE1
PIPE2
3500
3000
2500
ε+
2000
1500
1000
500
0
0 5000 10000 15000 20000 25000 30000 35000 40000
Y+
0.3
0.25
0.2
0.15
0.1
0.05
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Velocity (m/s)
Figures 5.12 (a) to 5.12 (d) compare the computed results obtained by using steady
and unsteady friction. The maximum pressure is predicted similarly by both models.
However the oscillations damp much more rapidly using unsteady friction than the results
86
H Vs T (Valve)
Steady Friction
Unsteady Friction
180
160
140
Pressure Head (m)
120
100
80
60
40
20
0
0 5 10 15 20 25 30
Time (s)
160
140
120
Pressure Head (m)
100
80
60
40
20
0
0 5 10 15 20 25 30
Time (s)
160
140
Pressure Head (m)
120
100
80
60
40
20
0
0 5 10 15 20 25 30
Time (s)
H Vs T (Middle Pipe 1)
Steady Friction
Unsteady Friction
180
160
140
Pressure Head (m)
120
100
80
60
40
20
0
0 5 10 15 20 25 30
Time (s)
The researcher modified the previous system to convert it into a branching system.
The modification, shown in Figure 5.13, consists of adding a new pipe exactly at the
junction of the previous case study, and thus to convert a series piping system in a
branching piping system. The boundary condition downstream of the new pipe is a node
Reservoir
3
Reservoir
1 4
2
2
Valve
2
1
3
4
1
6
Table 5.6 shows the different types of nodes presents in this system.
89
Node Number Node Type
1 (1) Upstream Reservoir
2 (4) Branching Junction
3 (6) Downstream Valve
4 (2) Downstream Reservoir
The diameter, length, wave speed and friction factor for pipes number 1 and 2 are the
same listed in Table 5.5. Characteristics for pipe number 3 are listed in Table 5.7.
Transient conditions were caused by closing the valve at the end of pipe number 2
Figure 5.14 shows the initial eddy viscosity distribution in each pipe, and Figure 5.15
shows the initial velocity profile. Figures 5.16 (a) to 5.16 (d) compare the computed
90
Eddy Viscosity Distribution
4500
PIPE1
4000 PIPE2
PIPE3
3500
3000
2500
ε+
2000
1500
1000
500
0
0 5000 10000 15000 20000 25000 30000 35000 40000
Y+
0.3
0.25
0.2
0.15
0.1
0.05
0
0 0.5 1 1.5 2 2.5 3 3.5
Velocity (m/s)
91
H Vs T (Valve) Steady Friction
Unsteady Friction
110
100
90
80
Pressure Head (m)
70
60
50
40
30
20
10
0
0 5 10 15 20 25 30
Time (s)
70
60
50
40
30
20
10
0
0 5 10 15 20 25 30
Time (s)
92
H Vs T (Middle Pipe No 1) Steady Friction
Unsteady Friction
110
100
90
80
Head Pressure (m)
70
60
50
40
30
20
10
0
0 5 10 15 20 25 30
Time (s)
100
90
80
Head Pressure(m)
70
60
50
40
30
20
10
0
0 5 10 15 20 25 30
Time (s)
1. A computer algorithm to extend Silva and Chaudhry unsteady friction model for
water hammer oscillations in piping systems to series and branching pipe systems
was developed. The conduits may have different diameters, wall thickness, wall
branching junction, dead end, downstream and intermediate valve, and orifice at
3. The generalized water hammer algorithm was developed using Visual Basic for
is an excellent tool to design and analyze fluid transients in liquid systems and no
compilers or interpreters are required because VBA is part of the Excel and MS
Office suite.
4. Commercial programs for transient analysis generally are very costly. This
94
REFERENCES
[2] Cebeci, T., Smith, A.M. 1974. Analysis of Turbulent Boundary Layers,
Academic Press, New york.
[4] Eichinger, P. and Lein G., (1992). “The Influence Of Friction Unsteady Pipe
Flow”, Unsteady Flow and transients, Betless and Watts. Balkema, Rotterdam.
[5] Hoffman, K.A. 1989. Computational Fluid Dynamics for Engineers, Texas:
Engineering Education System.
[6] Granville, P.S. 1990. A NearWall Eddy Viscosity Formula for Turbulent
Boundary Layers in Pressure Gradients Suitable for Momentum, Heat or Mass
Transfer, Journal of Fluids Engineering ASME. 112: 240 – 243.
[7] Granville, P.S. 1989. A modified Van Driest Formula for the Mixing Length of
Turbulent Boundary Layers in Pressure Gradients, Journal of Fluids Engineering
ASME. 111: 94 – 97.
95
[9] Larock Bruce E., Jeppson, Roland W. and Watters, Gary S. (2000). “Hydraulics
of Pipelines Systems”, CRC, Boca Raton, Florida, USA.
[11] Ohmi, M. Iguchi, M. and Hinami H., (1980). “Flow Pattern and Frictional
Losses in Pulsating Pipe Flow”. Part 1. Effect of Pulsating Frequency on the
Turbulent Flow Pattern, Bulletin of the JSME 23, No. 186, 2013 – 2020.
[12] Peralta Cooper, Eddy, (2001). “kω Twoequation turbulence model in the
Computation of the Energy Dissipation in Transient Flow”, M. Sc. Dissertation,
University of Puerto Rico, Mayagüez Campus.
[13] Pezzinga, G. (1999). “Quasi2D Model for Unsteady Flow in Pipe Networks.” J.
Hydraul. Div., Am. Soc. Civ. Eng., Vol. 125 (7), 676  685.
[14] Potter, Merle C. and Wiggert, David C. (1997). “Mechanics of Fluids”, Second
Edition, Prentive Hall, Inc. USA.
[15] Ramaprian, B.R. and Tu, S.W. (1979). “Experiments on Transitional Oscillatory
Pipe Flow”, Report No. 221, Iowa Institute of Hydraulic Research, August 1979,
115 p.
96
[17] Schetz, Joseph A. (1993). “Boundary Layer Analysis”, Prentice Hall,
Englewood Cliffs, New Jersey, USA.
[18] Shames, Irving H. (1992). “Mechanics of Fluids”, 3rd ed., New York, Mc Graw
Hill Co.
[22] Streeter, Wylie, Bedford (1998). “Fluid Mechanics”. Ninth edition, USA, WCB/
Mc Graw Hill Co.
[25] Swaffield J. A. and Boldy A. P. (1993). “Pressure Surge in Pipe and Duct
Systems”, Avebury Technical, England.
97
[26] Trikha, A. K. (1975). “An efficient method for simulating frequencydependent
friction in liquid flow” J. Fluids Eng., Vol 97(1), 97 – 105.
[29] White, F.M. (1991). “Viscous Fluid Flow”, 2nd ed., New York, Mc Graw Hill
Co.
[30] Wood, Don J., Lingireddy, Srinivasa and Boulos Paul F. (2005). “Pressure
Wave Analysis of Transient Flow in Pipe Distribution Systems”, MWH Soft,
Pasadena California, USA.
98
APPENDIX A. VALUES OF T VS. TAU FOR DIFFERENT
TYPES OF VALVES
99
BUTTERFLY VALVES
0.9
0.8
0.7
0.6
TAU
0.5
0.4
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
t/tc
(a)
(b)
Figure B 1. (a) t/tc Vs. TAU for Butterfly valves, (b) valve scheme.
100
APPENDIX A2. CIRCULAR VALVES
t/tc Close t/tc Opening (Z/D)·100 TAU
1.00 0.00 0 0.00
0.95 0.05 5 0.05
0.90 0.10 10 0.10
0.85 0.15 15 0.15
0.80 0.20 20 0.20
0.75 0.25 25 0.25
0.70 0.30 30 0.30
0.65 0.35 35 0.35
0.60 0.40 40 0.40
0.55 0.45 45 0.45
0.50 0.50 50 0.50
0.45 0.55 55 0.55
0.40 0.60 60 0.60
0.35 0.65 65 0.65
0.30 0.70 70 0.70
0.25 0.75 75 0.75
0.20 0.80 80 0.80
0.15 0.85 85 0.85
0.10 0.90 90 0.90
0.05 0.95 05 0.95
0.00 1.00 100 1.00
101
Circular Valves
0.9
0.8
0.7
0.6
TAU
0.5
0.4
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
t/tc
(a)
(b)
Figure B 2. (a) t/tc Vs. TAU Circular valves, (b) valve scheme.
102
APPENDIX A3. BIPLANES VALVES
103
Biplane Valves
0.9
0.8
0.7
0.6
TAU
0.5
0.4
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
t/tc
(b)
Figure B 3. (a) t/tc Vs. TAU for Biplane valves, (b) valve scheme.
104