|
|
The
simulation of distributed free gas in liquids using Verwey-Yu scheme
Sameh
Mansour
Researcher, HRI, Egypt
Currently, research fellow, TU
Braunschweig, Germany
Abstract
Describing fluid transients in
which free air is present has attracted a lot of research effort in the last
two decades. The numerical modeling of
these flows poses special problems due to numerical dispersion and
attenuation. The available methods are
the method of characteristics using several kinds of interpolations, implicit
methods and discrete methods. In 1993 a
numerical scheme was suggested by A.Verwey and J.H. Yu. This scheme was tested to be applicable in
water hammer and cavitation as well as fluid-structure interaction
simulations. This scheme has the
distinct advantage of being stable over wide range of courant
applications. It is intended in this
paper to use this scheme in the simulation of transients in liquids with free
gas. The inherent appeal of this scheme
is that it will be possible to carry out the simulation even when the existence
of free gas in the liquid leads to reducing the wave propagation velocity,
benefitting from the numerical stability of the scheme. In this way it would be possible to avoid
interpolation, which introduces numerical damping. Special attention has to be paid to the numerical procedure in
order to reach a stable application of this technique. The results show good accordance with those
produced by delft hydraulic which were used as a benchmark in order to assess
the applicability of the Yerwey-Yu scheme in this type of simulation.
Introduction
In liquid in which free gas is
present the propagation wave velocity becomes a strong function of the pressure
level, and the simulation of the transient response of a system becomes a
highly nonlinear problem. Figure 1
shows the relation between the wave propagation celerity and the percentage of
released air in a pipe (m) together with the pressure.

The available simulation
methods are the method of characteristics, implicit methods and discrete
methods. When the method of
characteristics is used with a specified grid in x-t plane, interpolations are
necessary either in the spatial or time directions, or both. The numerical modeling of all these methods
poses special problems due to numerical dispersion and attenuation. Using Verwey-Yu scheme it will be possible
to allow for propagation wave speed changes due to the released air in the
fluid and pressure changes. As the
scheme allows for courant changes in the range of 0.1 to 1 and gives good and
stable results. First the transient
equations will be numerically treated, using Taylor's series expansion in order
to subtract the third order terms from the truncation error. Then the simulation will be carried out and
using an iterative technique, it would be possible to determine the pressure
and the amount of released air in the fluid and then calculate its effect on
the wave propagation velocity.
Governing
equations
In a discrete air model
isothermal expansion and compression of the air volume is assumed. A mass conservation is used at the
computational sections. Together with
the momentum equation it would be possible to simulate the transient flow in a
fluid filled pipe system. The governing
equations can be written as follow:
|
|
|
|
in which
|
|
a propagation wave velocity m/s
A cross sectional area m2
D pipe diameter m
e pipe wall thickness m
E Young's modules of elasticity Pascal
g gravitational acceleration m/s2
H pressure head m
k fluid bulk modules Pascal
m percentage of dissolved gas -
p pressure Pascal
Q discharge m3/s
R gas solubility coefficient Pascal.Kelvin
T temperature Kelvin
ρ fluid density kg/m3
Methodology
Equation 3 shows that the
propagation wave celerity is a function of the pressure in the pipe and the
percentage of air released. Due to this
fact it would not be possible to maintain a unique value of courant number during
the simulation. Instead of using
interpolation, as in the method of characteristic, the Verwey-Yu scheme
provides a practical solution. Using
this scheme it will be possible to use a unique grid of space and time. During a simulation courant number will vary
between 1 and 0.1 depending on the amount of released air and the pressure
inside the pipe. The scheme can be
implemented after treating equations 1 and 2 according to Taylor's series
analysis to describe the third order terms from the truncation error and deduct
them from the original equations. Then
it will be possible to carry out a stable simulation even under a very low
courant conditions.
Numerical
implementation
Using
Taylor's series expansion to analyze the continuity equation, the following
equation can be written:
|
|
Compare equation 4 with the original continuity equation, and using θ, ψ value equal to 0.5, the second order differential terms vanish. The truncation error remains with the third order differential terms, which are equal to :
|
|
It is clear that this
important term will vanish when courant number is equal to unity or when it is
possible to deduce this term from the original continuity equation. This is possible when the Verwey-Yu scheme
is used as it will be possible to describe this term numerically. As this scheme consists of two steps in
space and 3 steps in time.
Then the continuity equation
and momentum equations can be written in the following form:
|
|
In which:
|
|
and
|
|
In which:
|
|
After subtracting these terms
from the simulated equations a stable simulation can be carried out even under
a very low courant condition.
Test and results
A numerical model based on
this approach is built and tested against the benchmarks defined by delft
hydraulic. This test involves a single
horizontal, frictionless pipeline. The
data of this test is shown in figure 2.
The transient action is introduced by lowering instantaneously the
upstream head from 60 m to 10.3 m, being the level of the pipe.

At the first instance the
sharp negative pressure pulse disperses.
Due to reflection at downstream boundary a positive pressure wave
appears. This wave will steepen until a
shock appears. The shock wave will
amplify and be transferred to the upstream end of the pipeline etc. The response of the system at a distance x =
1200 m from the LHS boundary is shown in figure 3 . This figure presents the results given by delft hydraulics. Figure 4 presents the results at the same
point given by the produced code, based on Verwey-Yu scheme and the
concentrated cavity model. The
following numerical data were used in both simulations :Δx = 200 m
and Δt = 0.2 s. These data means
that the simulations will be carried out under courant number equal to unity at
the beginning of the simulation.
Comparing these figures, it can be seen from the results that good accordance is reached. The results given by delft hydraulics show numerical oscillations or a numerical damping of the physical oscillation as they concluded in their research. They have recommended to decrease the time step in order to decrease the numerical damping. The implementation of the Verwey-Yu scheme in the simulation of this type of flow eliminates the necessity of interpolation. This result directly in the elimination of numerical damping and keeps the physical oscillations. Using this new technique, it was possible to maintain a predefined computational grid of space and time. This was done by computing a time step from the initial steady state computations and allows the system to oscillate according to the variation in courant number, due to the change in wave speed, tacking advantage of the fact that this new technique maintains its applicability over a wide range of courant numbers.

Another test was performed in
order to obtain further validation of the produced model.
The second test was performed
also by delft hydraulics. This test has
the same features as the first test but the under pressure wave was introduced
by lowering the pressure head at the LHS boundary from 60 mwc to 30 mwc
instantaneously. The results of this
test are introduced in the form of time series of pressure head at a point
X=1200m from the LHS boundary. Figure 5
presents the results given by the produced model. Some oscillations can be noticed in the time series. It is also possible to conclude that the air
entrained in the system acts as damper to the pressure wave. The wave amplitude decays until the system
reaches the steady state again, in this case it is 30 mwc along the pipeline.
Figure 6 presents the results given by delft hydraulics. Comparing both figures, it is possible to
notice some phase difference between the two solutions and to conclude that the
two models perform in more or less the same manner. The phase difference can be elaborated due to the fact that in
the case of modeling this test using the produced model, the under pressure wave
was introduced to the system over 5 time steps. While in the case of the results presented by delft hydraulics,
the under pressure wave was introduced over two time steps. This results in an initial phase difference
between the two solutions. It is not
possible to simulate the under pressure wave, using the produced model, with
less than these 5 time steps. As the
instantaneous pressure drop simulation will produce numerical oscillation due
to the fact of lacking the resolution of the under pressure wave.

Conclusion and
recommendation
The Verwey-Yu scheme can be
used successfully for water hammer computations as well as cavitation
computations. It is very important to
understand the numerical behavior of the scheme before applying it to practical
computations. The scheme can be used
throughout a wide range of Courant Numbers (0.1<Cr<1.0). It is very important to introduce the
changes in boundary conditions through enough point per wave length. This is essential to resolve the changes in
boundary conditions. If the under
pressure wave, which will produce cavitation, is introduced through less than 20 points per wave length, an
instability problem can be produced.
This instability is detected through the oscillations that show up in
the time series of the pressure head.
These oscillations are due to the phase error which is produced during
simulations carried out under Courant conditions less than unity. The phase error produces short waves. This short waves are the oscillations that
appears in the time series of pressure head.
To eliminate these oscillations, two different methods can be applied.
-More weight can be added to
the calculated time step. This can be
applied by introducing larger values of (θ) than 0.5. The drawback of this method is that more
values of (θ) than 0.505 means numerical diffusion to the solution. This method is not recommended due to the
produced numerical diffusion.
-The changes in the boundary
conditions should be introduced through more points per wave length than 20
. This will eliminate the production of
short waves and oscillates. This method
is recommended and essential to be noted and respected in all simulations that
are carried out under courant number less than unity.
Reference
[1]
M.B.ABBOTT,"Computational
Hydraulics", Ashgate Publishing LTD,
England, 1992
[2]
M.B.ABBOTT
and D.R.BASCO,"Computational Fluid Dynamics, An Introduction for Engineers",
Longman Group UK, LTD, 1989
[3]
E.S.CLYDE
and J.P.TULLIS,"Aeration Scale Effects", ASCE Spcialty Conference, Boston, MA,
ASCE, New York, August 1983
[4]
DELFT
HYDRAULICS,"Courses in the Hydrodynamics of Pipeline Systems", Delft, The
Netherlands, November 1991
[5]
A.HEINSBROEK,"Cavitation
Conceptual Model Aspects", Delft Hydraulics, Delft, The Netherlands, 1993
[6]
S.G.S.Mansour,"The
Application of a Space-Compact High-Order Implicit Scheme for Cavitation
Computation in Waterhammer Simulation",Hydroinformatics'96, ETH Zürich,
Switzerland
[7]
S.G.S.Mansour,"The
Application Verwey-Yu Scheme in FSI simulations", Hydroinformatics'98, DHI
Copenhagen, Denmark
[8]
C.A.PROVOOST
and E.B.WYLIE,"Discrete Gas Model to Represent Distributed Free Gas in
Liquids", Delft Hydraulics, delft, The Netherlands, April 1982
[9]
J.PAUL
TULLIS,"Hydraulics of Pipelines", Wiley-Interscience Publications, New York
1989
[10]
A.Verwey
and J.H.Yu,"A Space-Compact High-Order Implicit Scheme for Water Hammer
Simulations",XXV IAHR, Tokyo, Japan, August 1993