Vous êtes sur la page 1sur 9

Numerical Study of Dam-Break Flow

OMID SEYEDASHRAF 1, ALI AKBAR AKHTARI 2, HEJAR JEBELI AQDAM 3


M.Sc. Student of Hydraulic Engineering 1, Assistant Professor 2, M.Sc. Student of Hydraulic Structures 3

O.seyedashraf@gmail.com
Akhtari@razi.ac.ir
Hejar.jabaly@ut.ac.ir

Abstract
When a dam is breached, calamitous flash flooding occurs as the impounded water
flees through the opening into the downstream river, which is followed by massive live
and property lost. However, taking into account the large scale of the phenomenon, a
numerical scheme remains the preferred method for the provisional studies. Finite
Volume Method (FVM), Finite Difference Method (FDM) and Finite Element Method
(FEM) are the three main numerical methods used in simulating hydraulic incidents.
The use of non-linear Shallow Water Equations (SWEs) has become standard for dam-
break flow modeling and has been proved both useful and convenient, albeit the
existence of significant ambiguities due to the simplifying assumptions and
approximations used in the SWEs. In the present study attempts were made to
numerically investigate the fluid flow characteristics of dam-break phenomenon
through different governing equations and numerical methods. The resolution of 2D
SWE and 3D Navier-Stokes (NS) equations in the dam-break simulation, is put side by
side the experimental data, validated and compared the accuracy of both models. The
CFD analysis is executed by means of the commercial software package Fluent. The
volume of fluid (VOF) model is employed to depict the air-water interaction at the free
surface. It has been concluded that, both 2D SWE and 3D NS models are capable of
capturing the dam-break shocks reasonably accurate, yet, considering the CPU time
and overall evaluations, the 2D model with the SWE as the governing equations, is
preferable to the 3D NS model in numerical simulations of the dam-break phenomena.

Keywords: dam-break flow, shallow water equations, navier-stokes equations, open
channel bend, computational fluid dynamics.



2
Introduction
Dam-break flow is the immediate release of initially stationary water body by removing a
vertical obstacle, such as in case of a reservoir or a dam failure, the after effects transient
flow over the bed is termed as dam-break flow. When a dam is breached, calamitous flash
flooding occurs as the impounded water flees through the opening into the downstream
river. Generally, the response time available for warning is much shorter than that for
acceleration of runoff floods, called shocks. Consequently, the possibility of loss of lives is
much deplorable. The emphasis on dam failure hazards have become so well known that it
is protected by the rules of International Humanitarian Law (IHL) and dams shall not be
made the object of attack during armed quarrels, if that may cause brutal losses amongst
the civilian populations. There are 17 critical infrastructure areas like national monuments,
energy, chemical and agricultural systems, originally listed in the USA, Homeland Security
Presidential Directive (HSPD) 7 of 5/7/2007 obligation to ensure the protection of
resources including dams. However, there have been around 200 important dam and
reservoir failures in the world so far in the 20
th
century. The worlds most awful dam
failure occurred in Henan, a province in China, in August 1975, when the Banqiao Dam
and the Shimantan Dam failed horrendously due to the overtopping caused by torrential
rains. In this incident around 85,000 people died because of the flooding and many more
lost their lives during its consequent plague and starvation and millions of occupants lost
their houses [1]. This disastrous is analogous to the events of Chernobyl and Bhopal for the
nuclear and chemical industries [2]. Figure 1 shows the ten-year running average number
of dam failures as available in the digital library of National Performance of Dams
Program (NPDP) [3]. The running average is shown for the complete historic record (all
failures) and for the record excluding the many small dams that failed during the 1994
Georgia floods.

Figure1: Ten-year running average number of dam failures (excluding the many small dams that
failed during the 1994 Georgia floods)

In comparing with experimental investigations and because of the scale of the incident,
numerical methods could be more attention-getting to predict the flow behaviors,
hydrographs and routings. Numerical studies of currents in channels and rivers, mainly
driven by floods, have an interesting range of applications in environmental hydraulics,
navigation and providing safe drinking water to the rural masses and industry use.
However, the dynamics of the dam-break wave propagation is rather complex and its
behavior do not meet the terms of the regular assumptions of conventional steady and
gradually varied open-channel flows.


3
As a recent investigation, Soares Frazao et al. conducted a series of numerical modeling
and laboratory experiments of dam-break phenomenon on a 90
o
bend of rectangular cross
section with a straight outlet reach [4]. They solved 1D and 2D Shallow Water equations
(SWEs) in finite volume method and compared to the taken pictures of the water flow. He
measured both velocities in the bend and water depth profile along the channels, noticed
that the 2D model was in a good agreement with the experimental data. Ying et al. have
also developed numerical models for flows generated by a dam failure or levee breaching
process using a conservative form of SWEs, for more information refer to [5].
Numerical modeling of such phenomena using physical illustrations such as the Navier-
Stokes (NS) equations can frequently be awkward, due to the extent of the modeling
geometries as well as through resolving free surfaces. SWEs, of which there are a number
of demonstrations, provide an easier picture of such phenomena.
Due to Complexities of the nonlinear terms, computational domain and bottom
topography, numerical approximations are to be expected. Of these, finite difference
methods (FDMs), finite volume methods (FVMs) and finite element methods (FEMs) are
the most common. Normally, FEMs provide a better resolution of the flow domain but due
to their large matrix systems, are more expensive to implement and to execute, unless
modern methods e.g. MultiFrontal, are employed to solve the huge linear system [6]. In the
numerical studies, since spurious oscillation may occur near the discontinuities, in which it
is inevitable in a numerical dam-break model, and can substantially corrupt the results and
generate instabilities, it has been difficult to give such transcritical flows a clear
interpretation. Accordingly, the results obtained from nearly all classic numerical methods
suffer from these numerical fluctuations. To overcome this, a variety of high-order well-
balanced schemes for SWEs are developed and can be found in the literature (e.g. Toro E.F
[7], LeVeque [8], Kurganov and Levy [9]) these schemes construct excellent estimates of
the quasi-steady solutions and non-stationary steady states.
In the present study, attempts were made to numerically investigate the fluid flow
characteristics of dam-break phenomenon through different governing equations and
methodologies. The resolution of the 3D NS equations in FVM model is put side by side
the data obtained from a 2D SWE model and the experimental data by Soares Frazao et al.
[4]. To resolute the governing equations, the commercially available CFD package
FLUENT was utilized. Results were validated and compared to the ones obtained from the
2D model.

3D Reynold-Averaged Navier -Stokes
The governing equation for CFD is based on conservation of mass, momentum, and
energy. The used CFD package makes use of a FVM to solve the NS equations.
Advantages of using these types of computational tools are the possibility of evading the
SWEs restrictions, allowing a vertical description of the different variables of the flow e.g.
the velocity profile and the ability to model turbulent currents by various methodologies
such as the k- model. Making use of a Volume-of-Fluid (VOF) approach, we would be
able to resolve the free surface evolution. FVM involves discretization and integration of
the governing equation over the control volumes. The fundamental equations for transient
state laminar current are conservation of mass and momentum. In this research as the heat
transfer or compressibility is not involved, the energy equation is deleted from the equation
system. The governing equations are stated as:


4
( ) ( )
m
i
i
S
x
u
t
=
c
c
+
c
c

( )
( ) ( )
j
j i
i
j
j
i
j i j
j i
i
x
u u
x
u
x
u
x x
p
x
u u
t
u
c
' ' c
+
(
(

c
c
+
c
c
c
c
+
c
c
=
c
c
+
c
c


(1)



(2)
where t is time, u
i
is the i-th component of the Reynolds-averaged-velocity, x
i
the i-th axis
(with the axis x
3
vertical and oriented upward), is the water density, p is the Reynolds
averaged pressure, g is the acceleration due to the gravity, is the viscosity which is
equal to zero in this study and S
m
is the mass exchange between the two phases. It should
be noted that the unsteady solver will be used to get the velocities and other solution
variables now represent time-averaged values instead of instantaneous values. The term
appeared
( )
i j
u u ' ' is called Reynolds-stress. This term is obtainable from the Boussinesq
hypothesis which links Reynolds-stresses to the mean rate of deformation. However, in this
work due to the large scale of the phenomenon, this term is disregarded.

2D Shallow Water Numerical Model
Since the 17th century, Newton (16431727) and Leibniz (16461716) shaped the world
of modern mathematics by introducing calculus. With deferential in calculus, people
started to think about deferential equations. The SWEs also called Saint-Venant
equations in its one-dimensional form and after Adhmar Jean Claude Barr de Saint-
Venant, are a system of hyperbolic partial differential equations that depicts the flow below
a pressure surface in a fluid. The non-linear SWEs are regularly used for modeling flows in
which the depth D is much less than the wavelength L, like oceanographic or atmospheric
fluid flow. Models of such systems lead to the calculation of areas eventually affected by
pollution, coastal erosion, and polar ice-cap dissolving. The most frequently used SWE
form in dam-break investigations, which is derived from the Navier-Stokes equations, is as
follows:

0
t x y
U E F S + + + = (3)
h
U uh
vh
(
(
=
(
(



(4)
2
2
2
uh
gh
E u h
uvh
(
(
(
= +
(
(




(5)


5
2
2
2
vh
F uvh
gh
v h
(
(
(
=
(
(
+
(




(6)
0
bx
x
by
y
S ghb
ghb
t

(
(
(
(
= +
(
(
+ (





(7)

The vector E and F are the so-called ux vectors and S represents the topographical and
frictional source terms, where g is the gravitational acceleration. u and v are the depth-
integrated velocity components in the x and y directions, respectively, b is the bottom
elevation and h is the water depth.

GEOMETRY and MESH GENERATION
Different grid sizes were inspected to investigate the sensitivity and the precision of the
results. To obtain sound data of the shock propagation and the flow depth, the channel was
discretized by diminutive cells. Around 70 longitudinal, 20 latitudinal and 22 altitudinal
segments are created in the specified channel, as a result the three-dimensional flow
domain was splitted into a total number of 39048 hexahedral non-overlapping cells which
extent from 0.025m to 0.1m. Out of various possible meshing schemes, the chosen form is
suitable for the accuracy, computational costs and the CPU time of the convergence. Due
to the elimination of turbulence terms in equations (1) and (2) and neglecting the wall
functions, a regular unstructured grid is sufficed to model the fluid flow. Figure 1 and 2
respectively show the geometric layout and the plan view of meshing form.


Figure2: Schematic representation of the computational domain.



6



Figure 3: Meshing form used to perform the computations.

SOLUTION METHODOLOGY
The governing equations (1) and (2) are a set of convection equations with velocity and
pressure coupling based on the control volume technique. The general purpose code
FLUENT was employed for all the numerical simulations presented in this investigation.
The code employs the FVM in conjunction with a coupled technique, which solves all the
transport equations simultaneously in the whole domain through a false time-step
algorithm. Convection terms are discretized using the third order Monotone Upstream
Centered Scheme for Conservation Law (MUSCL). The linearized system of equations is
preconditioned in order to reduce all the eigen-values to the same order of degree.
Pressure-Implicit with Splitting of Operators (PISO) technique is engaged to deal with the
problem of velocity and pressure coupling. PISO methods incorporate pressure impact
through momentum equations into continuity equation to attain correction equations for
pressure. The Volume of Fluid (VOF) method was employed to simulate the air-water
interaction. The VOF method was developed by Hirts and Nicholas [10] and the
formulation relies on the fact that two or more phases are not interpenetrating. For each
extra phase that added to the system, a variable is introduced in the volume fraction of the
phase in the computational cell. In each control volume, the volume portions of all phases
sum to unity. Outflow boundary condition was chosen as the outflow basin with two
separate outlets with the same group ID. Symmetry boundary condition, in which all the
normal components and gradients are kept zero, was chosen as the upper surface boundary.
The sides and bottom surfaces are defined as wall boundary condition. A pressure based
solver is used to solve the equations since the flow is incompressible.
Regarding the CFL condition (which is a very limiting constraint on the time step t and
was named for its originators Courant, Friedrichs, and Lewy) with a value of 0.25, a set of
5600 time steps of 0.0025 seconds with the maximum 40 iterations per time step, was
conducted to solve the transient current. Using an ordinary unstructured grid has
considerably enhanced the acceleration of convergence.

Calibration and Validation
Although the CFD package FLUENT is widely used for engineering applications and
scientific investigations but validation of the numerical models is always essential. The


7
numerical results for the dam-break flow are validated by comparing thenumerical results
with the measurements made at laboratorial scale by Soares Frazao et al. Figure 2 depicts
the wave front tracking of a 2D model while Figure 3 shows results of the computed 3D
model.


Figure 4: Computed (2D Shallow-Water model) positions of dam-break front, at interval of 0.1 s

Figure 5: Computed (3D NS model) positions of dam-break front, at interval of 0.1 s

Due to elimination of the vertical coordinate from the flow equations in SWEs, The 2D
model couldnt model the secondary currents. Consequently, as depicted in the previous
figures, discrepancies in the numerical results between the 3D and 2D models are
inevitable.
In the following figures, water depth profiles are compared to the experimentally measured
ones.

Figure 6: Experimental and computed (2D and 3D numerical models) flow profiles, t=5(s).


8


Figure 7: Experimental and computed (2D and 3D numerical models) flow profiles, t=7(s).

Conclusions
The evaluation of the results with experimental data permits to illustrate a conclusion on
efficiency of a considered method. Discrepancies have been noted, between models using
different mathematical schemes and equations. As the gateway is detached instantaneously
a shock wave is made and propagates in through the downstream channel and a reflective
negative wave front is generated, which starts traveling upstream into the reservoir. Flow
regime transforms from subcritical to transcritical, and reaches to supercritical flow at
various section as the dam-break flow propagates downstream.
The comparison of the water depth profile of the experiments and the numerical results
show a good match. The 3D, NS model has captured the shocks outline precisely
(particularly at the time step t=7s); however, the 2D, SWE results seems more realistic and
depicts the shock propagation properly. Concerning Figures 6 and 7, a hydraulic jump is
noticeable at the inlet of the channel in both experimental and the 3D model, yet, due to
elimination of the vertical coordinate (x
3
) from the flow equations in SWEs and
consequently vertical description of the different variables of flow, this is omitted in the
2D model. Also, a temporal fluctuation of water depth at the tip of the 90
o
is shown at both
figures. At this location, there is no reflective wave front in the 3D model, but progressive
shock fronts from the upstream side originate the formation of peaks. The simulated 3D
model produced a good match with the photographed water depths at the time step t=7s;
though, the model underestimated less water than the reality passes the bend at the time
step t=5s and this led to lower depth before the bend.
Even if the water depth is to some extent underestimated, nevertheless, the time of
appearance of inundation peaks are depicted precisely. Eventually, it can be concluded that
both 2D SWE and 3D NS models are capable of capturing the dam break shocks
reasonably well but considering the accuracy and CPU time and over all evaluations, a 2D
model with the SWE as governing equations is more appropriate than a 3D NS model in
numerical simulations of the dam-break phenomena.





9

References
[1] Qing, D. : The River Dragon has come!: Three Gorges dam and the fate of Chinas Yangtze River and its
people, ME Sharpe, 1997.
[2] McCully, P. : Silenced rivers; The ecology and politics of large dams, Zed Books, London & New Jersey,
1996.
[3] The digital library of National Performance of Dams Program (NPDP). http://npdp.stanford.edu
[4] Soares Frazaol, S. and Zech, Y. : Dam Break in Channels with 90 Bend, JOURNAL OF HYDRAULIC
ENGINEERING / NOVEMBER 2002
[5] Ying, X. , Wangm SSY. and Khan, AK. : Numerical simulation of flood inundation due to dam and levee
breach, Proceedings of ASCE world water and environmental resources congress 2003, Philadelphia, USA.
[6] Duff , I. S. and Reid, J. K. : The Multifrontal Solution of Indefinite Sparse Symmetric Linear, ACM
Transactions on Mathematical Software (TOMS), v.9 n.3, p.302-325, Sept. 1983 DOI
10.1145/356044.356047
[7] Toro, EF. : Shock-capturing methods for free-surface shallow flows. New York, Wiley, 2001.
[8] LeVeque, RJ. : Balancing source terms and flux gradients in high resolution Godunov
methods: the quasi-steady wave-propagation algorithm, J Comput. Phys. 1998, 146:346-365.
[9] Kurganov, A. , Levy, D. : Central-upwind scheme for the saint-venant system. M2AN Math Model
Numer Anal 2002,36:397-425.
[10] Hirt, C.W. and Nicholas, B.D. : Volume of Fluid (VOF) method for the dynamics of free boundaries, J.
of Computational Physics, 39, pp. 201-225, 1982.

Vous aimerez peut-être aussi