Вы находитесь на странице: 1из 2

KSCE Journal of Civil Engineering Water Engineering

Vol. 11, No. 4 / July 2007


pp. 209~214

Analysis of Transient Flow in a Piping System


By Hyuk Jae Kwon*

···································································································································································································································

Abstract

Transient flow in a piping system was studied using both experimental and computer models. Two different computer models, the
method of characteristics model and the axi-symmetrical model, are utilized and discussed. Experiments for transient flow in a piping
system were conducted to verify the results of the computer models. It was found that the energy decay is underestimated if the
Darcy-Weisbach friction coefficient f is used for the analysis of transient flow. Therefore, the equivalent head loss coefficient CL is
introduced and experimentally determined as 0.3 which is several times larger than the Darcy-Weisbach friction coefficient f for the
present experimental setup. The value of a parameter N in the axi-symmetrical model is calibrated using the results of experiments.
The parameter N is 0.22 for the present experimental setup and is smaller than the value used in previous studies.
Keywords: transient flow, piping system, head loss coefficient

···································································································································································································································

1. Introduction oscillations of the two-dimensional model damp out faster than


those of one-dimensional model since the steady friction term in
Chaudhry (1979), Watters (1979), and Wylie and Streeter one-dimensional model underestimates the head losses.
(1993) provided a general treatment of water hammer in piping Pezzinga (1999) explained that one-dimensional models in
systems. They specified boundary conditions associated with which the energy dissipation is computed by a relation between
different hydraulic devices in the distribution system under energy slope and mean velocity underestimate the friction forces
transient flow conditions. Islam and Chaudhry (1998), Karney and overestimate the persistence of oscillations. Pezzinga used
and McInnis (1992), and Wylie (1983) provided a systematic the implicit scheme for the momentum equation and the explicit
discussion of hydraulic transients and presented methods of scheme for the continuity equation to achieve numerical stability.
analysis. Kwon (2005) explored several numerical methods for This model assumed that the pressure gradient in the radial
the transient flow analysis in piping systems. These are the two- direction can be ignored and that the longitudinal velocity is a
dimensional model, the implicit method, and the explicit finite- function of radial direction, longitudinal distance, and time.
difference method. Results of a one-dimensional model and a quasi-2D model were
Karney (1990) and Karney and McInnis (1990) developed an compared.
alternative and simpler interpretation of transient conditions in a The main objective of the current study is to confirm and
pipeline. Karney (1990) also discussed the importance of energy compare two approaches of numerical analysis for the transient
relations, the proper expressions of internal energy, energy flow in a piping system. To accomplish this objective,
transformations, and compressibility. In 1995, McInnis and experiments for transient flow were conducted to verify the
Karney discussed the relatively unexplored area of transients in parameters and to confirm the accuracy of computer models.
complex pipe networks. The behavior predicted by a transient This study shows that the head loss coefficient used for the
model was compared with the results of field-testing. The method of characteristics model should be much larger than the
computer model was generally in agreement with the trend of Darcy-Weisbach friction coefficient and the parameter N used for
field data but the computer model poorly represented the long- the axi-symmetrical model should be smaller than the value used
term energy decay. In this analysis, many pipelines are neglected before. Furthermore, the strengths and weaknesses of the two
for the sake of simplicity as are many local losses and hydraulic computer models are summarized.
devices.
Brunone et al. (1995) discussed the rapid damping of water 2. Computer Models
hammer pressure after the sudden valve closure. Brunone
pointed out that the conventional one-dimensional MOC model 2.1 The Method of Characteristics Model
underestimates the energy decay and suggested an unsteady The two governing equations, the equation of motion and
friction term for a two-dimensional flow field. In his study, the continuity equation, can be simplified to form the two ordinary
computation was performed for a single pipe and the results from differential Eq. (1).
a two-dimensional unsteady friction term were compared with
dQ CL
those of one-dimensional model. It was shown that the pressure ------- r gA
------ dH
------- + ----------
-Q Q = 0 (1)
dt c dt 2DA

*Member, Lecturer, Department of Civil Engineering, Kangwon National University, Chuncheon 200-701, Korea (E-mail: hyukjae68@hotmail.com,
hjkwon@usc.edu)

Vol. 11, No. 4 / July 2007  209 


Hyuk Jae Kwon

where, Q represents the flow rate (m3/s), H is the pressure head u in j+ 1 – uin j Hin + 1 – Hin–+11 2S rj + 1 Win j + 1 – ri Win j
--------------------- + g --------------------------- + ------ --------------------------------------- = 0 (10)
(m), A is the cross-sectional area of a pipe (m2), c is the speed of 't 'x U 'A
the pressure wave (m/s) and CL is the equivalent head loss
The first subscript i is numbered along x direction and the
coefficient introduced and experimentally determined for the
superscript n represents time step. The second subscript j
present study.
represents r direction. The shear stress term is computed from
The method of characteristics in finite difference forms of (1)
the finite-difference form of (7) as follows:
can be written as follows:
uin j + 1 – uin j uin j + 1 – uin j
uin j + 1 – uin j ----------------------
gA C L 't n Win j+ 1 = –UQ ----------------------
- – Ul 2 ----------------------
- - (11)
Q n+1
i –Q n
i–1 + ------ Hin + 1 – Hin– 1 + ----------
-Q Qn = 0 (2) 'r 'r 'r
c 2DA i – 1 i – 1
gA C L 't n 3. Experiment Procedure
Qin + 1 – Qin+ 1 + ------ Hin + 1 – Hin+ 1 + ----------
-Q Qn = 0 (3)
c 2DA i + 1 i + 1
For measuring the pressure waves associated with the water
where, the superscript n+1 represents the unknowns and
hammer phenomena generated by sudden closure of control
subscript i is numbered along x direction in a pipe. The Courant
valves, a pipeline consisting 84.37 meter-long galvanized iron
stability condition must be satisfied as:
pipe was installed as shown in Fig. 1. The pipe diameter is
't 1 0.0508 m (2 inches). The upstream and downstream shut-off
------ d --- (4)
'x c valves are located at both ends of pipeline. The distance between
the upstream shut-off valve and the upstream pressure transducer
2.2 Axi-Symmetrical Model is 20.36 m and the distance from the upstream shut-off valve to
In the transient condition, the shear stress at the wall is not the downstream pressure transducer is 63.40 m. The sampling
wholly associated with the mean velocity since the velocity rate of pressure transducer was 4000 Hz for the present
profile is often different from the fully developed flow profile. experiment. The strain gauge input module was used for signal
Therefore, frictional head losses computed by using steady state conditioning. To control the upstream steady line pressures,
friction coefficient might be inaccurate in transient flow. The pressure regulator was installed. Once steady flow has been
unsteady flow could be modeled by means of the axi- established, two shut-off valves at the upstream and downstream
symmetrical model as a two-dimensional model by considering ends are manually closed. To compare the method of
the velocity profile in a cross section. This model is based on the characteristics model with the axi-symmetrical model, the line
continuity and momentum equations for the elastic pipe with pressure head of 70.26 m (100 psi) with the flow rate at 0.000757
circular cross section in cylindrical coordinates. The axi- m/s (12.0 gpm) is established. Steady pressures and flows were
symmetrical model is utilized to obtain the longitudinal velocity measured both upstream and downstream.
profile u as a function of r, x, and t as well as the average
pressure head H as a function of x and t. With similar 4. Results of Computer Models and Experiment
assumptions as in the method of characteristics model, the
continuity and momentum equations can be simplified as When the computations for transient flow are performed, if
follows: only the Darcy-Weisbach friction coefficient f is used for
representing the head loss, then the equivalent value of f must be
wH wQ-
c 2 ------
------- + -------- =0 (5) significantly greater. To overcome this over/underestimation of
wt gA0 wx
pressure fluctuations, a new coefficient is needed to accurately
wu wH 2S w rW estimate the head loss for transient flow. For this reason, the
------ + g ------- + ------ ------------ = 0 (6)
wt wx U wA equivalent head loss coefficient CL is introduced in this study. CL
is the combined head loss coefficient that includes various
For the shear stress, a mixing length solution can be used:
coefficients such as the friction coefficient and transient effects.
wu wu wu CL is calibrated using the experimental data sets of transient flow.
W = – UQ ------ – Ul2 ------ ------ (7)
wr wr wr Comparing the results of computer model with the experimental
data, a value of CL causing the smallest error is chosen for the
where, Q is the kinematic viscosity and l is the mixing length.
present experiment setup. Equivalent head loss coefficient CL for
The mixing length l (Marchi (1961) and Pezzinga (1999)) is
the present piping system is 0.3. This value is much bigger than
approximated as follows:
Darcy-Weisbach friction coefficient. However, equivalent head
loss coefficient CL should be used only for transient flow analysis
--l- = N --y- e – y e R (8)
R R to avoid the overestimations of head losses at steady state.
In an attempt to show the relative importance of local loss
where, y is the distance from wall to the pipe center, R is the pipe
effect, the values of f(L/D) in steady state are plotted along with
radius, and the coefficient N is to be determined experimentally.
the values of total head loss coefficient in the range of Reynolds
(5) and (6) are solved explicitly using a finite difference method
number covered. The Darcy-Weisbach friction coefficient f is
which is herein written as:
collected from the Moody’s diagram using a relative roughness
Hin + 1 – Hin c 2 Qin+ 1 – Qin kx/D of 0.0025~0.003 with galvanized iron pipe. It shows that the
- + -------- --------------------- = O
--------------------- (9)
't gA0 'x values of total head loss coefficient are much bigger than f(L/D)

 210  KSCE Journal of Civil Engineering

Вам также может понравиться