Você está na página 1de 12

Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 4, pp.

541552 (2011)

DAM-BREAK FLOW IN THE PRESENCE OF OBSTACLE:


EXPERIMENT AND CFD SIMULATION
Hatice Ozmen-Cagatay * and Selahattin Kocaman #

* Civil Engineering Department, Cukurova University, 01330 Adana, Turkey


E-Mail: hatmen@cu.edu.tr (Corresponding Author)
Civil Engineering Department University of Mustafa Kemal, skenderun, Turkey

ABSTRACT: The aim of this paper is to present an experimental and numerical investigation of dam-break flow
over initially dry bed with a bottom obstacle. This test case highlights not only the bottom slope effects but also
those of abrupt change in channel topography. Dam-break flow was applied in a smooth prismatic channel of
rectangular cross-section over a trapezoidal bottom sill on the downstream bed. The present study scrutinized the
formation and propagation of negative bore behind the sill. The flow was numerically simulated by the VOF-based
commercially available CFD program, Flow-3D, solving the Reynolds Averaged Navier Stokes equations with the
k- turbulence model (RANS) and the Shallow Water Equations (SWE). To validate CFD models an experiment was
carried out. Using an advanced measuring technique, digital image processing, the flow was recorded simply
through the glass walls of channel; thus, continuous free surface profiles were acquired synchronously with three
cameras along the channel. The adopted measuring technique is non-intrusive and yields accurate and valuable
results without flow disturbances. Comparison of the computed results with experimental data shows that RANS
model reproduces the flow under investigation with reasonable accuracy while simple SWE model indicates some
discrepancies particularly in predicting the negative wave propagation.
Keywords:

dam-break, digital image processing, obstacle, CFD, VOF, k- turbulence model

numerical studies exist. Bed slope effect has been


investigated experimentally in a sloping smooth
channel with dry bed (Lauber and Hager, 1998;
Nsom et al., 2000). Stage hydrographs induced
dam-break flow over two trapezoidal bumps was
studied experimentally and compared with
numerical results (Aureli et al., 1999). Some
previous researchers carried out experiments of
dam-break flow over triangular bottom sill by
digital imaging measurements (Soarez-Frazao,
2007; Aureli et al., 2008). Furthermore, a similar
problem was numerically solved using SWEs
(Brufau et al., 2002; Quecedo et al., 2005,
Kesserwani and Liang, 2010). Most of the
previous numerical studies on dam-break
described the flow by shallow water equations.
Mohapatra and Bhallamudi (1996) numerically
solved the conservative form of the SWEs by the
MacCormack finite-difference scheme for effects
of the contractions and expansions on the
dambreak flow depth at the dam site. GarciaNavarro et al. (1999) compared SWE simulations
provided by three different schemes against exact
solution for dambreak problem over presence of
strong changes in geometry. Zhou et al. (2004)
presented well-balanced schemes for applications
involving wetting and drying. Liang et al. (2006)
compared the TVD-MacCormack solver with the

1. INTRODUCTION
The flood wave resulting from sudden dam failure
causes catastrophic events. The estimatation of
inundated areas is necessary for mitigation of
hazards.
Dam-break
flood
waves
are
hydrodynamically the most complex type of
rapidly varied unsteady flows. Since the flow
usually propagates over complex domain
topography, it becomes even more challenging.
The irregular topography of downstream channel
plays an important role in determining the
maximum flow depth, propagation velocities of
the flood wave and regime of flow. Moreover, it
causes negative wave formation in the upstream
direction. Local sills, debris in natural river bed as
well as buildings, bridges in urban areas act as
obstacles to the flow. They may cause the rising
of flow depth and variation of the flow pattern.
Furthermore, the presence of an obstacle like a
sill on the river bed accentuates the bottom slope
effect. The propagation of dam break flow in the
presence of an obstacle can be considered
assuming initially dry downstream bed despite the
existing of very low water level.
Analytical solution of dam break flow in sloping
channel is available (Dressler, 1958). In literature,
there are limited experimental studies while many
Received: 16 Feb. 2011; Revised: 8 Jun. 2011; Accepted: 27 Jun. 2011
541

Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 4 (2011)

software for dam-break problem (Vasquez and


Roncal, 2009; Biscarini et al., 2010; OzmenCagatay and Kocaman, 2010).
The present paper aims to present numerical
(CFD) and experimental investigation of dambreak flow in rectangular channel with initially
dry bed over trapezoidal bottom sill (bump).
Hence, an experiment has been carried out to
acquire new data for validation of CFD
simulations. The study reveals the effect of abrupt
channel bottom variations on flood wave
propagation. This study focuses on scrutinizing
the formation and time evolution of negative bore
in the upstream direction behind bump as well as
description of the flow regime. Dam-break flow
was simulated numerically by VOF-based CFD
modelling. A commercially available software
package, Flow-3D (Flow Science 2007), was used
in numerical simulation. It enables solving the
Reynolds Averaged Navier Stokes (RANS)
equations with the k- turbulence model, and the
Shallow Water Equations (SWE). The description
of unsteady flow behind bump was a great
challenge, however, the adopted advanced
measuring technique, digital image processing,
overcame these difficulties during experiments.
Unlike most of previous experimental works, in
this experiment synchronous filmed image of the
flow was obtained with three CCD cameras along
the channel without tests repetition and then, free
surface profiles were determined. The main
advantages of this measuring technique are
simplicity, cheapness, and reasonable sensitivity
through non-intrusive measurement. The results
of the experimental test were compared with
numerical results. Due to the difficulty of
obtaining field data for such flows, the highquality experimental data may be useful for other
numerical researchers to validate their models in
future practical applications especially to irregular
topography.

ADI type solvers for dambreak flow over uneven


bottom topographies and they reported that TVDMacCormack model reproduces the flow in
question. Aureli et al. (2008a) solved classical
SWEs with finite volume for the dambreak flow
over bottom sill. A weighted surface-depth
gradient method was proposed for dambreak flow
over non-horizontal beds as well as in the
presence of wet/dry fronts and it iwas applied to
real case study (Aureli et al., 2008b). Ying et al.
(2009) developed a robust and accurate numerical
model which is based on conservative form of
SWEs by using finite volume method on an
unstructured triangular grids for dambreak flows
that may occur over complicated terrain and
involve subcritical flows, supercritical flows and
transcritical flows. Liang and Borthwick (2009)
investigated dambreak wave propagation with an
adaptive quadtree simulation of shallow flows
over an initially dry floodplain containing three
humps. Liang and Marche, (2009) presented a
well-balanced numerical scheme for simulating
frictional shallow flows for dambreak flows over
a domain with irregular bed profile. Liang (2010)
extended this model with finite volume Godunovtype scheme which provides shock-capturing
capability for wave propagation over complex
domains. An unstructured Godunov-type finite
volume model was developed by Song et al.
(2011) for the numerical simulation of
geometrically
challenging
two-dimensional
shallow water flows with wetting and drying over
convoluted topography. However, there are less
studies related to using RANS equations
involving the turbulence modelling (Abdolmaleki
et al., 2004; Shigematsu et al., 2004; Quecedo et
al., 2005). Turbulence effects can be considered
in certain SWE models (Brufau and GarciaNavarro, 2000; Liang and Marche, 2009; Liang
and Borthwick, 2009; Erpicum et al., 2010).
Recently, with the great developments in
computer technology, 3D VOF based CFD
modelling has become popular (Flow-3D, Fluent,
CFX, Star CD, OpenCFD, PHOENICS etc.).
These programs allow the utilization of various
turbulence models (k-, k-, LES, RNG, etc). In
recent years, CFD has been used to investigate the
flow in several real case studies (Hargreaves et al.,
2007; Bridgeman et al., 2009; Wu, 2010). It is
difficult to obtain real case data in dam-break
events. At this point, CFD can be an alternative
tool in understanding dam-break mechanism.
However, CFD solutions should be validated by
comparing with experimental data to show their
capabilities in simulating the topographic effects.
Recent researchers have adopted CFD modelling

2. EXPERIMENTAL DATA
2.1

Experimental set-up

The experiment was carried out in the Civil


Engineering Hydraulics Laboratory of the
Cukurova University in Turkey using a
rectangular horizontal channel 8.9 m long, 0.30 m
wide, and 0.34 m high (Kocaman, 2007). The
channel bottom and walls are made of 9 mm thick
glass (Fig.1). The plate (dam) is located 4.65 m
from the channel entrance that separates the
upstream part of the channel, representing the
reservoir (Fig.2). It was initially filled with 0.25
m of water at rest. Downstream part of the
542

Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 4 (2011)


890
Plate

(a)

Reservoir

h0=25

465

35 30 35

(D

34

7.5
153

100

185

Plate

(b) A

grabber card. Once the flow images were


transferred simultaneously to the computer via the
frame grabber card, synchronous images of the
three camera records were combined, allowing for
a panoramic flow view without changing the
cameras positions (Cagatay and Kocaman, 2008).
The major difference from previous experimental
works is the synchronous recording of free
surface profiles along the downstream channel
entirely without tests repetition thanks to using
the frame grabber card.
The test area was illuminated by fluorescent
spotlight located above the channel. The
background of the opposite glass wall was coated
with light coloured paper to exclude objects
behind the channel. To prevent shadows and
reflections caused by uncontrolled ambient light,
the working area was isolated using thick whitecoloured curtains.
The recorded video images were digitized as
384288 pixels at 50 frames/s. Due to wide-angle
lens usage distorted images were corrected using
a calibration plate containing 42 uniform black
and white coloured square meshes (100 mm100
mm) (Fig.3). The spatial calibration parameters
were separately estimated for each camera by
matching the known coordinates of the corners on
video images, which were recorded from different
viewpoints of the calibration board. 25 images
were used for each camera. The software
Camera Calibration Toolbox for Matlab was
used to correct distortions (Bouguet, 2004). Fig.
3a shows original image including barrel
distortion. After radial calibration process, Fig. 3a
converted to plane image (Fig. 3b). To convert the
pixels into length, a predetermined reference
length on the calibrated image (Fig. 3b) like a
map scale was used. Bands with black and white
coloured grids 100 mm long were attached to the
upper and lower sides of the channel walls. The
intersection points were also used as references to
combine images. From the video images, the
corresponding reference length was determined.
The software ImageJ was used to obtain the
free surface profiles from the digitized images
(Rasband, 2005). To define free surface profiles,
the images were resized to 768576 pixels to
select free surface levels more accurately.
Afterwards, to reduce unwanted stain-like traces,
a filter was applied for smoothing the resized
image. The water-air interface was then sharpened
by contrast enhancement. The threshold value as
used by previous researchers was retained
(Bateman et al. 2006; Soares-Frazao, 2007). The
free surface levels were obtained from recorded

425

A 30

Frame grabber card


Cameras
Computer

Connection cables

Fig. 1

Test arrangement and initial conditions (a)


section A-A; (b) plan, lengths in cm.

Fig. 2

Photograph of channel and dam break


mechanism.

channel was dried carefully against possible


seepage from the reservoir. A symmetrical
trapezoidal shaped bottom sill, 0.075 m high and
1m base length, was located 1.53 m downstream
from the plate (Fig.1). It was made of 4 mm thick
Plexiglas. For dam-break modelling, a mechanism
allowed for the instantaneous removal of the
vertical plate (Fig.2). The 4 mm thick plate,
coated with aluminium, was made of rigid plastic.
A steel rope was attached to the plate top; the
rope was drawn over a pulley of 15 kg weight
attached at the other end. By releasing a weight
from 1.50 m above the floor, the plate was
removed instantaneously. The plate removal time
was between 0.06 and 0.08 s from video records.
From earlier studies, it should be shorter than
1.25(h0/g)1/2
for
a
sudden
removal,
corresponding to 0.2 s in the present tests (Lauber
and Hager, 1998). To better identify the free
surface levels and behaviour of dam-break flows
with video cameras, the reservoir water was
coloured with dye.
2.2

Measuring technique and calibration

Using digital image processing, the free surface


profiles were determined satisfactorily. The
components of the digital imaging system are
three CCD cameras, a computer, and a frame
543

Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 4 (2011)

(a)

Fig. 3

(b)

Calibration of radial distortions: (a) Original image; (b) Calibrated image.

images by clicking manually on the interface


between the water surface and air at various
points. The estimated error of the video images
for the water level was 1 to 2 mm by comparison
with water surface levels measured by calliper at
still water condition.

direction, p = pressure, = fluid density, t = time,


gi = gravitational force in subscript direction and
fi = diffusion transport term .VF and Aj (cell face
areas) = 1, thereby reducing the equations to the
basic incompressible RANS equations. Sij = strain
rate tensor, b,i = wall shear stress and =
kinematic viscosity, t= kinematic eddy viscosity.
Herein, the equations of motion are closed with
the standard k- turbulence closure scheme given
by

3. CFD SIMULATION: FORMULATION


AND NUMERICAL SOLUTION
The numerical solutions of RANS, SWE and the
computations of the free surface by VOF involved
the software Flow-3D that contains a CFD
package based on the Finite Volume Method in a
Cartesian, staggered grid. The program evaluates
the location of the flow obstacles by
implementing a cell porosity technique called the
fractional area/volume obstacle representation of
FAVOR method (Flow Science, 2007).

The governing continuity and RANS equations


for Newtonian, incompressible fluid flow are
(1)

u
u i
1

u j Aj i

x j
t V F

1 p g i f i

xi

3.2

(2)

1
VF

b ,i

A j Si j

x j

(4)

Shallow water model in Flow-3D (SWE)

The shallow-water equations result from a vertical


integration of the NavierStokes equations under
the assumption that the fluid is incompressible
and the pressure distribution is hydrostatic.
Vertical accelerations can be neglected if a flow is
shallow enough.
The Volume-of-Fluid (VOF) method for tracking
fluid interfaces and the FAVOR(TM) method for
the description of geometry can be combined in a
shallow water model of Flow-3D (Flow Science,
2007). The standard one-dimensional SWEs are

where

fi

where T is the turbulent viscosity, k is the


turbulence kinetic energy, C is the closure
coefficient and is the turbulent dissipation rate.
The standard k- turbulence closure model was
used to determine turbulent viscosity (Launder
and Spalding, 1974). The governing equations for
k and were modelled by the conventional
advection-diffusion equations with appropriate
production and dissipation terms. Previous
researchers also used the k- model to simulate
simple laboratory experiments; particularly for
flows at high Reynolds numbers (Wilcox, 2000).
Shigematsu et al. (2004) used the k- model in
dam-break application.

3.1 Reynolds-averaged Navier-Stokes


equations (RANS)

u A 0
xi i i

C k 2

(3)

u u j
Si j t i

x j xi
where ui = velocity component in subscript
direction, VF = volume fraction of fluid in each
cell, Ai = fractional area open to flow in subscript

h hB u h hB

0
t
x
544

(5)

Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 4 (2011)

1 p
u
u
u

gx fx
t
x
x

for the atmospheric pressure on the free surface.


In Flow 3-D, if the upper boundary is taken as
symmetry, atmospheric pressure can be effective
on the free surface of water. Since the water
surface is defined by VOF, zero shear stress and
constant atmospheric pressure were applied as
boundary conditions over the air-water interface
(Hirt and Nichols, 1981). The channel sidewalls
were chosen as symmetry, implying no flux and
shear of any property across it. The k- turbulence
model provided logarithmic velocity distributions
in the boundary layer, called the wall function.
The no slip condition was defined as zero
tangential and normal velocities.

(6)

where x = space coordinate, t = time, h = height of


fluid, hB = height of bottom contour, u =
horizontal velocity component, gx = gravity
acceleration in x direction, p = pressure, = fluid
density and fx = viscous stress. The volume
fraction variable, VF and F(=h-hB),is used to
define a variable bottom contour and the depth of
fluid, respectively. Both VF and F are necessary in
the shallow-water application of VOF and
FAVOR. The fluid height in terms of VOF
representation at the centres of control volumes is
h-hB= FVF . At the edge of elements, where Ax =
open area fraction at the x face, the fluid height is
h-hB= FAx (VF=Ax). Substituting VF, F and
fractional area Ax into Eq. (5)

V F F uAx F

0
t
x

3.4

The computational domain was subdivided into a


mesh of fixed rectangular cells using Cartesian
coordinates. The computational domain was
described by means of a Cartesian cell-centred
discretisation with square cells of size x=z=5
mm. Mesh sensitivity was analyzed for four grid
sizes of 2, 5, 10, and 20 mm, indicating that the
mesh size was important only in the vicinity of
wave front. Because of insignificant differences
between the wave fronts using 2 and 5 mm, a grid
size of 5 mm in both directions was adopted to
reduce computational efforts. In the SWE solution
via Flow-3D, x=5 mm was used. A minimum of
two real cells must be defined in each mesh block
in the z-direction to apply VOF to SWE. Fluid
was allowed to exist only in the lower layer of
cells; thus, the cell size in the z-direction should
be sufficient to contain all fluid in that layer. The
z-axis was horizontally divided into two layers,
with a lower layer of 0.27 m and an upper layer of
0.23 m. Therefore, the solution domain height
was 0.50 m in SWE, and 0.30 m in RANS. The
mesh systems consist of totally 88,300 cells for
RANS and 10,704 for SWE. The computation
durations for 10 s were 35 min and 1.4 min for
RANS and SWE simulations, respectively.

(7)

fx is included by defining a laminar viscosity and


bottom roughness:

fx

u
u 2

Fdz
dz

Computational meshes

(8)

where = fluid molecular viscosity,


=component bottom roughness and dz = half of
the cell height. Since the bottom roughness is
assumed as smooth, the second term of Eq.(8) is
zero.
3.3 Solution domain, boundary, and initial
conditions
The dimensions of the solution domains were 8.9
m long and 0.30 m high for RANS model and 8.9
m long and 0.50 m high for SWE model. As
initial conditions, a 4.65 m long and 0.25 m high
volume of fluid was defined as a reservoir and no
volume of water was described at downstream
from dam, due to dry bed. A symmetrical
trapezoidal shaped obstacle, 0.075 m high and 1m
base length, was defined and located 1.53 m
downstream from the plate. All channel and
obstacle surfaces were assumed smooth. The time
steps t were determined according to the
Courant-Friedrichs-Lewy condition.
In the numerical computations, the upstream
boundary was set as wall due to no flow into the
reservoir and constant reservoir length. The
downstream boundary was outflow since the
downstream end of channel was open over drybed. The lower boundary was set as wall. The
upper boundary was set as symmetry to account

4. RESULTS AND DISCUSSION


Figure 4 displays the flow picture frames captured
for the time interval of 2.5 s-3.8 s were produced
in CFD simulation of RANS (Fig.4a) and
experiment (Fig.4b). Fig.4 focuses on the
evolution of the negative wave.
After the plate was lifted to simulate the sudden
dam-break, a very complex flow pattern
developed due to the presence of the obstacle.
Once the flow reaches the obstacle, a part of the
wave is reflected and forms a bore travelling
545

Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 4 (2011)


(a)

(b)

t=2.50s

t=2.50s

t=3.00s

t=3.00s

t=3.26s

t=3.26s

t=3.54s

t=3.54s

t=3.66s

t=3.66s

t=3.80s

t=3.80s

t=5.00s

Fig. 4

Evolution of reflected wave: (a) with computed velocity vectors and contours with magnitudes; (b) measured,
lengths in m.

negative wave front travelling towards upstream


direction increase. Turbulence effects are
significant particularly on the reflected wave front.
In dam break flood waves, turbulence is usually
generated through either bottom friction or wave
breaking (Shigematsu et al., 2004).
Prior to presenting the comparison of the
numerical free surface profiles with the measured
one over channel with trapezoidal obstacle, free
surface profiles considering the dam-break flow
without this obstacle are given in Fig.5. The flow
depths h and horizontal distances x are nondimensionalized by the initial water depth h0.
Similarly, time t is multiplied by (g/h0)1/2 to obtain
dimensionless time T=t (g/h0)1/2. Computed free
surface profiles are in a good agreement with
experimental data, however, SWE shows slight
disagreement especially at positive and negative
fronts (Fig.5).

towards upstream direction while the other part


moves up the bump. The major difference of the
Fig. 4a and b is the observation of a spilling type
wave breaking in the experiment while a plunging
type wave breaking in the computed results.
Previously, the same problem was solved
numerically using RANS equations without
comparison with experimental data (Quecedo et
al., 2005). Present experimental data may be
useful for numerical researchers to validate their
numerical models. Furthermore, Fig. 4a shows the
time evolution of reflected wave (negative bore)
with computed velocity vectors and contours with
magnitudes. This is the superiority of the CFD
model in the computation of the velocity field.
Herein, reference vector represents the velocity of
1.78 m/s for t=2.5s-3.8s and 1.65 m/s for t=5s.
Velocities become minimum just behind the
bump. As the time progresses, velocities of the
546

Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 4 (2011)

surface profiles in front of the bump (x/h0= 9-14)


are always in good agreement with the
experimental data. It can be said that CFD
simulations of dam-break flow in the presence of
obstacle accurately predict the free surface
profiles. However, there are significant deviations
on reflected wave profile for RANS at T=20.67
and T=23.05, which can be attributed to the from
formation of plunging type wave breaking in the
computed results, while a spilling type wave
breaking occurs in the experiment. Herein,
different turbulence models k-, RNG etc,
especially LES (Large Eddy Simulation) can be
tested to capture spilling type wave breaking.
Despite the disagreement in predicting the
negative wave front, the simple SWE model is
capable of reproducing free surface profiles
except near negative wave. The disagreement may
be caused by using the non-conservative form of
governing equations in SWE model of Flow-3D.
In literature, certain studies have overcome the
problem by using the conservative form of SWEs
for dam-break flows over complex topography
(Aureli 2008b, Liang and Marche, 2009; Liang
and Borthwick, 2009; Ying et al., 2009).
Fig.7 shows the distance versus computed (RANS)
local Froude numbers at T= 25.68 in the absence
and presence of bottom sill located at dry channel
downstream. For the classical dam break problem
over dry bed without sill, Froude numbers
immediately upstream of plate are below unity,
namely, the flow is subcritical. On the other hand,
presence of sill causes mixed flow conditions
along the channel. As shown in Fig.7,
immediately downstream of the plate they are
above unity, thus, the flow is supercritical. While
the flow remains supercritical after a distance of
x0/h0 = 3.6 from the plate in the absence of bump,
the flow transcends abruptly to subcritical for
presence of bump, with the influence of hydraulic
jump- up to the distance x0/h0=8. After that point,
it transcends abruptly to supercritical again.
Accordingly, there are abrupt transitions from
supercritical to subcritical flow and from
subcritical to supercritical behind bump. Mixed
flow regimes require special attention when
modelling this type of flow due to the different
directions of wave propagation in subcritical and
supercritical flows.
Fig.8 shows the variations of computed (RANS)
local Froude numbers with non-dimensional
distance at corresponding non-dimensional point
of time of Fig. 6. The flow is always subcritical
(Fr<1) over reservoir. For T=11.9, the flow
transcends from subcritical to supercritical and
remains as supercritical downstream. For

1.0
RANS
SWE

0.8

Experiment

h/h0

0.6
0.4
T = 6.64

0.2
0.0
-8

Fig. 5

-4

x/h0

12

Comparison of computed free surface profiles


with experimental data over channel without
bottom sill at T= 6.64.

Comparisons of computed free surface profiles


with experimental data of dam-break flow over
dry bed with trapezoidal bottom sill at various
points of time are shown in Fig.6. With passing of
flood wave over the bump, a negative wave (bore)
is observed on free water surface of upstream face
of (behind) bump. Free surface profiles are
accurately predicted by both numerical
simulations for T=11.90, and 17.54 except the
appearance of a discontinuity on SWE at T=17.54.
While negative surge begins to form at T=20.67
behind the bump on measured profile, water
levels continue to rise and maximum water level
just downstream of x/h0=6 is overestimated in
RANS model. The rising limb of the free surface
is steep and convex in RANS model while it is
concave in experimental profile and speed of
reflected wave front is lower in RANS than that
of measured. This difference can arise from
formations of hydraulic jump on the free surface
and spilling type wave breaking towards upstream
in experiment. On the other hand, in RANS
solution, plunging type wave breaking is formed
resulted from reflection of the flood wave against
bump. As for SWE model at T=20.67, a
discontinuity is observed on negative wave front
resulted from early wave breaking. The
discontinuity of negative wave front in SWE
remains up to T=29.69 and it moves towards
upstream direction with a speed significantly
slower, compared to those of measurement and
RANS. With early wave breaking in both
numerical solutions at T=23.05, velocities of
negative wave front in SWE and RANS are
significantly lower than measured one. There is
good agreement between RANS solution and
measured free surface profiles for last two points
of time (T=29.69 and T=41.84). However, SWE
model underestimates the negative wave front
speed at T=29.69 while it sligtly overestimates the
maximum water levels. Both of the computed free
547

Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 4 (2011)

1.0

RANS

h/h0

SWE
Experiment

0.5

Obstacle

T = 11.9

0.0
0

x/h0

10

12

14

1.0
RANS

h/h0

SWE
Experiment

0.5

Obstacle

T = 17.54

0.0
0

x/h0

10

12

14

1.0
RANS

h/h0

SWE
Experiment

0.5

Obstacle

T = 20.67

0.0
0

x/h0

10

12

14

1.0
RANS

h/h0

SWE
Experiment

0.5

Obstacle

T = 23.05

0.0
0

x/h0

10

12

14

1.0
RANS

h/h0

SWE
Experiment

0.5

Obstacle

T = 29.69

0.0
0

x/h0

10

12

14

1.0
RANS

h/h0

SWE
Experiment

0.5

Obstacle

T = 41.84

0.0
0

Fig. 6

x/h0

10

12

14

Comparison of computed free surface profiles with experimental data over channel with bottom sill at various
points of time.
548

Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 4 (2011)

2
Fr

without sill
with sill

Fr=1

supercritical

subcritical

subcritical

supercritical

0
-2

0
Fig. 7

4
x/h0

10

Distance versus computed local Froude numbers at T = 25.68.

T=11.90
T=17.54

T=20.67
T=23.05
T=29.69

Fr

T=41.84

Fr=1

0
-20

-15

-10

-5

10

x/h0

Fig. 8

Variations of computed (RANS) Froude numbers over distance at various points of time.

T=17.54-23.05, with the transition of flow


through the condition (Fr=1) on the plate (dam)
axis (x/h0=0), flows are supercritical behind bump
and then they transcend abruptly to subcritical
with formation of hydraulic jump between x/h0=5
and 6. At the middle crest of the bump, flows
transcend from critical depth to supercritical. As
demonstrated in Fig.8, change in the channel
longitudinal profile is the indicator of the
unsteady mixed flow regimes. The unsteady flows
return to steady and all the profiles coincide after
passing bump. Actually, the flow behaves as
steady flow over broad-crested weir. As time
progresses (T=29.69-41.84), the flow remains
subcritical behind bump. Although the flow
reaches Fr=1 on x/h0=0 at T=29.69, it remains
subcritical behind bump. For T=17.54-23.05,
flows transcend three times through the critical
condition (Fr=1). It is observed from Figs. 7 and 8

that Flow-3D can handle change from subcritical


flow to supercritical flow and reverse transitions.
5. CONCLUSIONS
This paper reveals the effect of abrupt channel
bottom variations on unsteady flows. Dam-break
flow in smooth prismatic channel with rectangular
cross-section and horizontal bed is studied. The
problem was handled experimentally and
numerically in the presence of trapezoidal bottom
sill, located initially dry downstream. The results
show that the presence of an obstacle causes
reflection of the flood wave and formation of a
negative bore which propagates in the upstream
direction, while the other part moves up the bump.
Thus, mixed flow regime occurs behind the bump.
The flow under investigation was simulated
numerically using VOF-based CFD commercial
549

Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 4 (2011)

software package, Flow-3D, solving the RANS


equations with the k- turbulence model and the
SWEs. In RANS solution, at most stages, good
agreement was achieved not only in speed of
negative wave front but also in prediction of free
surface profiles. However, evident disagreements
on maximum water depth at certain points of time
can arise from formation of plunging type wave
breaking in RANS solution. Despite not
accurately predicting negative bore propagation
and underestimating negative wave front speed,
the simple SWE model is capable of reproducing
the free surface profiles. Moreover, the SWE
approach may be attractive due to the advantage
of requiring considerably less CPU time for large
computational domains. Flow-3D enables the
regime of dam-break flow to be described by
means of determining Froude numbers, which are
not available experimentally in this work due to
the difficulty of velocity measurements. It seems
suitable for analyzing the dam-break flow in the
presence of bottom obstacle. The simulation
performance of the CFD was validated with the
new high-quality laboratory data. The behaviour
of unsteady flow was well-detected by digital
image processing without flow disturbances.
Continuous free surface profiles could be
acquired synchronously along the entire channel
without repetition of test by changing the camera
positions. Main advantages of this measuring
technique are yielding of free surface profiles in
reasonable accuracy without using any physical
instrument and its non-intrusive character.
In the future, current experimental data can be
useful for other researchers to validate their
numerical models over topography with abrupt
bottom variations. Numerical parts of this study
can be extended to different turbulence models,
e.g. LES, in order to capture spilling breaking
profile observed experimentally. Furthermore,
unlike the simplified SWE model, more robust
SWE solutions involving conservative forms of
equations developed by other researchers can be
compared with current measurements.

C
dz
F
fi
fx
gi
gx
h
h0
hB
k
p
u
Sij
t
T
v
VF
x, z

b,I

t
T

REFERENCES
1. Abdolmaleki K, Thiagarajan P, MorrisThomas MT (2004). Simulation of the dam
break problem and impact flows using a
Navier-Stokes Solver. Proceedings of 15th
Australasian Fluid Mechanic Conference,
Sydney, Australia, 13-17.
2. Aureli F, Maranzoni A, Mignosa P, Ziveri C
(2008a). Dam-break flows: Acquisition of
experimental data through an imaging
technique and 2D numerical modelling.
Journal of Hydraulic Engineering ASCE
134(8):1089-1101.
3. Aureli F, Maranzoni A, Mignosa P, Ziveri C
(2008b). A weighted surface depth gradient
method for the numerical integration of the
shallow water equations with topography.
Advances in Water Resources 31(7):962-974.
4. Aureli F, Mignosa P, Tomirotti M (1999).
Dam-break flows in presence of abrupt
bottom variations. Proceedings of XXVIII
IAHR Cong 1999, Graz, Australia, 163-171.

ACKNOWLEDGEMENT
This work was partly supported by Cukurova
University Research Fund under project no:
MMF2007BAP6. This support is gratefully
acknowledged.
NOMENCLATURE
Ai

directions
closure coefficient
half of the cell height
fraction of fluid
diffusion transport term
viscous stress
gravitational force in subscribe
direction
gravity acceleration in x direction
flow depth
initial upstream water level
height of bottom contour
turbulent kinetic energy
pressure
horizontal velocity component
strain rate tensor
time
dimensionless time t= t(g/h0)1/2
vertical velocity component
volumetric fluid fraction in each cell
cartesian coordinates in horizontal and
vertical directions
dissipation rate of turbulent kinetic
energy
fluid density
bottom roughness component
fluid molecular viscosity
wall shear stress
kinematic viscosity
eddy viscosity
turbulent viscosity

fractional flow area to in subscript


550

Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 4 (2011)

5. Bateman A, Granados A, Medina V, Velasco


D, Nalesso M (2006). Experimental
procedure to obtain 2D time-space high-speed
water surfaces. Proceedings of River Flow
2006, Lisbon-Portugal, 3:1879-1888.
6. Biscarini C, Francesco DS, Manciola P
(2010). CFD modelling approach for dam
break flow studies. Hydrology and Earth
System Sciences 14(4):705-718.
7. Bouguet J-Y (2004). Camera calibration
toolbox for Matlab.
8. Bridgeman J, Jefferson B, Parsons SA (2009).
Computational fluid dynamics modeling of
flocculation in water treatment: A review.
Engineering Applications of Computational
Fluid Mechanics 3(2):220-241.
9. Brufau P, Garcia-Navarro P (2000). Two
dimensional dam break flow simulation.
International Journal for Numerical Methods
in Fluids 33(1):35-57.
10. Brufau P, Vazquez-Cendon ME, GarciaNavarro P (2002). A numerical model for the
flooding and drying of irregular domains.
International Journal of Numerical Methods
in Fluids 39(3):247-275.
11. Cagatay H, Kocaman S (2008). Experimental
study of tail water level effects on dam-break
flood wave propagation. Proceedings of River
Flow 2008, Cesme-Turkey, 1:635-644.
12. Ozmen-Cagatay H, Kocaman S (2010). Dambreak flows during initial stage using SWE
and RANS approaches. Journal of Hydraulic
Research 48(5):603-611.
13. Dressler RF (1958). Unsteady non-linear
waves in sloping channel. Proc. R. Soc.
London, Ser. A, 247(1249):186-198.
14. Erpicum S, Dewals BJ, Archambeau P,
Pirotton M (2010). Dam-break flow
computation based on an efficient flux vector
splitting. Journal of Computational and
Applied Mathematics 234(2010):2143-2151.
15. Flow Science Inc (2007). Flow-3D Users
Manuals. Santa Fe NM.
16. Garcia-Navarro P, Fras A ,Villanueva I
(1999). Dam-break flow simulation: Some
results for one-dimensional models of real
cases. Journal of Hydrology 216(3-4):227247.
17. Hargreaves DM, Morvan HP, Wright NG
(2007). Validation of volume of fluid method
for free surface calculation: The broad-crested
weir.
Engineering
Applications
of
Computational Fluid Mechanics 1(2):136-146.
18. Hirt CW, Nichols BD (1981). Volume of
fluid method for the dynamics of free

19.

20.

21.
22.

23.

24.

25.

26.

27.

28.
29.

30.
31.

32.

551

boundaries. Journal of Computational


Physics 39(1):201-225.
Kesserwani G, Liang Q (2010). Wellbalanced RKDG2 solutions to the shallow
water equations over irregular domains with
wetting and drying. Computers and Fluids
39(10):2040-50.
Kocaman S (2007) Experimental and
Theoretical Investigation of Dam-Break
Problem. PhD thesis. University of Cukurova,
Adana, Turkey (in Turkish) .
Lauber G, Hager WH (1998). Experiments to
dam-break wave: Sloping channel. Journal of
Hydraulic Research 36(5):761-773.
Launder BE, Spalding DB (1974). The
numerical computation of turbulent flows.
Computer Methods in Applied Mechanics and
Engineering 3(2):269-289.
Liang Q (2010). Flood simulation using a
well-balanced shallow flow model. ASCE
Journal
of
Hydraulic
Engineering
136(9):66975.
Liang Q, Borthwick AGL (2009). Adaptive
quadtree simulation of shallow flows with
wet-dry fronts ove complex topography.
Computers and Fluids 38(2):221-34.
Liang D, Falconer RA, Lin B (2006).
Comparison between TVD-MacCormack and
ADI-type solvers of the shallow water
equations. Advances in Water Resources
29(12):1833-1845.
Liang Q, Marche F (2009). Numerical
resolution of well-balanced shallow water
equations with complex source terms.
Advances in Water Resources 32(6):873-884.
Mohapatra PK, Bhallamudi SM (1996).
Computation of a dam-break flood wave in
channel transitions. Advances in Water
Resources 19(3):181-187.
Nsom B, Debiane K, Piau JM (2000). Bed
slope effect on the dam-break problem.
Journal of Hydraulic Research 38(6):459-464.
Quecedo M, Pastor M, Herreros MI, Merodo
JAF, Zhang Q (2005). Comparison of two
mathematical models for solving the dam
break problem using the FEM method.
Computer Methods in Applied Mechanics and
Engineering 194(9):3984-4005.
Rasband WS (2005). ImageJ. US National
Institutes of Health, Bethesda M.D.
Shigematsu T, Liu PLF, Oda K (2004).
Numerical modelling of the initial stages of
dam-break waves. Journal of Hydraulic
Research 42(2):183-195.
Soares-Frazao S (2007). Experiments of dambreak wave over a triangular bottom sill.

Engineering Applications of Computational Fluid Mechanics Vol. 5, No. 4 (2011)

33.

34.

35.
36.

37.

38.

Journal of Hydraulic Research 45(Extra


Issue):19-26.
Song L, Zhou J, Guo J, Zou Q, Liu Y (2011).
A robust well-balanced finite volume model
for shallow water flows with wetting and
drying over irregular terrain. Advances in
Water Resources 34(7):915-932.
Vasquez JA, Roncal JJ (2009). Testing
river2D and flow-3D for sudden dam-break
flow simulations. Proceedings of CDA
Annual Conference, Whistler, BC Canada 4455.
Wilcox DC (2000). Turbulence modeling for
CFD. La Canada CA:DCW Industries, Inc.
Wu B (2010). CFD analysis of mixing in
large
aerated
lagoons.
Engineering
Applications
of
Computational
Fluid
Mechanics 4(1):127-138.
Ying X, Jorgeson J, Wang SSY (2009).
Modeling dam-break flows using finite
volume method on unstructured grid.
Engineering Applications of Computational
Fluid Mechanics 3(2):184-194.
Zhou JG, Causon DM, Mingham CG, Ingram
DM (2004). Numerical prediction of
dambreak flows in general geometries with
complex bed topography. Journal of
Hydraulic Engineering 130(4):332-40.

552

Você também pode gostar