Vous êtes sur la page 1sur 5

Proceedings of the 19th IAHR-APD Congress 2014, Hanoi, Vietnam

ISBN 978604821338-1

A FLOW SIMULATION OF A BRIDGE-INDUCED FLOOD IN A RIVER USING SMOOTHED-


PARTICLE HYDRODYNAMICS (SPH) METHOD

N. KIMURA (1), & A. TAI (1)


(1)
Department of Urban and Environmental Engineering, Kyushu University, Fukuoka, Japan
e-mail nkimura3@civil.kyushu-u.ac.jp

ABSTRACT
Extreme weather has occurred more frequently worldwide over the last decade because of climate change. Therefore, at
times, very heavy rains have caused severe flooding in the world. There have been reports of uncommon floods (such as
flooding caused by the substantial accumulation of driftwood or debris) at a bridge in mountainous areas in southern
Japan. To reduce a flood-induced disaster of this kind, it is necessary to better understand the flood mechanism. In this
study, smoothed particle hydrodynamics method was employed to simulate flooding in the two-dimensional vertical
plane. The method could reveal the physical processes of a driftwood-induced flood event at a river bridge. As the first
step, we focused only on the effect of a bridge on water flow without driftwood. Open software, SPHysics
implementing smoothed particle hydrodynamics method, was calibrated with data of water level recorded at a
miniature bridge in our laboratory experiment. Like the laboratory experimental result, the calibrated simulation result
clearly showed that the flow was divided into a top surface flow above the bridge girder and an undersurface flow
below the bridge girder and the water level gradually was raised in an upstream direction from the bridge. In addition,
using the SPHysics, we assessed how the bridge characteristics (for example, the width and thickness of the bridge
girder and the spacing of cross sections) affect the intensity of flooding. A shorter thickness and lower position of the
bridge girder caused larger dam-up water levels. This suggests that there is potential flooding at a bridge even without
driftwood.

Keywords: Smoothed particle hydrodynamics, Bridge, Flooding, Extreme weather

1. INTRODUCTION simulation to gain a better understanding of flow-based


physical processes around the bridge.
The intergovernmental Panel on Climate Change (IPCC)
has reported that climate change has increased the Our purposes in this study are (1) to evaluate how
number of severe natural disasters in the past decade, well a particle-based hydrodynamics model works for a
and warning that more severe disasters may occur in the flood event caused by a bridge through a comparison
near future (IPCC 2012). In fact, most recently, there were between experiment and simulation data, and (2) to
catastrophic floods in July 2012 at several locations on assess the effects of bridge characteristics on water-level
northern Kyushu Island, Japan. One of the floods, in increase in front of the bridge and overflow patterns
Takeda City, was greatly abnormal because it was likely around the bridge.
caused by driftwood (Figure 1). This figure shows that
substantial driftwood was deposited in the river channel
and stuck at several bridges. This flood caused
widespread severe damage in the area. We chose this
area to better understand physical processes for a
driftwood-induced flood. However, because of
complicated physical processes in this actual flood event,
we firstly focused on only simplified, fundament tests of
bridge-induced flooding for laboratory experiment and
numerical simulation in this study. Those tests were
performed based upon two-dimensional (2D, i.e., one
horizontal and one vertical) flow without driftwood.
This kind of fundamental, experimental studies and
numerical studies with simplified-driftwood modeling
have been done (Adachi and Daido, 1957; Sakano, 2003;
Nakagawa et al., 1993; Gotoh et al. 2001; Shimizu and
Osada, 2007). However, it is harder to capture fluid Figure. 1 Evidence of flooding caused by driftwood after heavy
dynamics around the bridge using an Eulerian-based rainfall in Takeda: (a) map; (b) railroad bridge; (c) driftwood on
the river; (d) and (e) damaged bridge; (f) and (g) bridges blocked
numerical simulation because the flows should be
by wood.
separated by the bridge girder. Therefore, we
implemented a mesh-free particle method to this flood

1
2. METHOD

dva P P
2.1 Laboratory Experiment mb b2 a2 ab aWab g [3]
Flow measurement was conducted with a unidirectional dt b b a
channel, approximately 21 m long and 0.60 m wide and ,
tilted by 1/25 (Figure 2). The bridge-like wood (i.e., a
miniature bridge) with 0.03 m thickness and 0.15 m
d a
width was fixed at 3.5 m from the downstream end and mb vab aWab , [4]
at 0.08 m height from the channel bottom (Figure 2). dt b
Water flow was circulated in the channel from the

upstream top to the downstream end by a pump system. where a, b= particle numbers, mass, velocity

The flow was maintained uniformly through a vector, g vector of gravity acceleration, p
straightening plate. The water surface was kept at pressuredensity, viscosity term The viscosity
approximately 0.08 m height before and after the bridge term is computed by the artificial viscosity (Monaghan,
position by controlling discharge. Because of this 1992),
discharge control, a high flow condition that represents
heavy rainfall can be generated. The flow moved through

under the bridge, struck the bridge, or overflowed. This C ab ab ab vabrab 0
experiment setup is likely considered as a realistic flood
event in a mountainous steep stream without driftwood.
ab , [5]
0 vabrab 0
where
2

ab hvab rab rab 2 with the position

vector ( rab ra rb ), the velocity vector (

vab va vb ), C ab =0.5(Ca+Cb) , 2 0.01h2, and
= any coefficient. Note that 0.2 was set up for to
eliminate unstable computation in the surface treatment
in this study. We employed dynamic boundary
conditions, which force boundary particles to maintain a
fluid state (Crespo et al., 2007).
Following the setup of the laboratory experiment, our
numerical simulation was designed with a bridge and a
gate in a 2D long rectangular box with one horizontal
Figure 2. Setup of the laboratory experiment. direction (X) and one vertical direction (Z) (Figure 3). A
Step-shaped water-filled tank was initially located
upstream. The step shape of the water tank was created
2.2 Numerical Simulation to make the upstream water smoothly flow. A gate
We used well-accepted open software, SPHysics whose lower part was open was placed at the upstream
(https://wiki.manchester.ac.uk/sphysics), implementing location to control the flow from the water tank. Only
the smoothed particle hydrodynamics (SPH) method to one difference between laboratory experiment and
simulate a flood event in a river-like channel. The SPH numerical-simulation setups was the upstream channel
method, which works by dividing a fluid into a set of length. The length in the numerical setup was longer to
particles with a smoothing length (h) based on integral save expensive computational cost. For the
interpolants (e.g., Lucy, 1977; Monaghan, 2005), was used computational conditions, spacing for initial inter-
to simulate fluid flows, which are potentially separated particles was 0.01 m in X and Z and time step is 10-5 sec.
by an obstacle. In the SPH method, the fundamental Note that a dam that actually exists in the field was
principle is to approximate any function (e.g., velocity) as considered in the numerical simulation.


A r A r ' W r r ' , h dr ' , [1]

where h =smoothing length, given by c x 2 z 2


with coefficient (c) and small distance (), W = weighting

function, and r , r ' = position vector. A weighting
function expressed by a polynomial of the fifth degree is
given by

W q D 1 q 2 2q 1 0 q 2 [2]
4

where q = r h and D = coefficient. This weighting Figure 3. Initial conditions in the numerical simulation setup. Blue
function is applied to the NavierStokes equations. They color indicates water. The gradient of the slope is 1/25.
are given as
To clarify the effects of bridge characteristics on flow
dynamics around the bridge, we set up several test cases

2
with different width and thickness of a bridge girder and
the height of the bridge from the channel bottom. These
cases are listed in Table 1.

Table 1. Bridge characteristics (width, thickness, and height) for


four cases.

Case WIDTH THICKNESS HEIGHT


NOTE
name [m] [m] [m]
same as lab-
CASE 1 0.150 0.030 0.08
experiment
CASE 2 same 0.015 same ideal test
CASE 3 same 0.045 same same
CASE 4 0.075 0.030 same same Figure 4. Zoomed-in snapshot for the 2D simulated flow at 5 s
CASE 5 0.300 same same same near the bridge. Red triangles and blue particles showing
CASE 6 0.150 0.030 0.05 same observed water levels and simulated particle-based flow patterns,
respectively.
3. RESULTS AND DISCUSSION
As an ideal test, four cases (Case 2 to Case 5) were run
Laboratory measurements and numerical simulations
to evaluate the effects of bridge characteristics on flow
were conducted to clarify the raised water level due to a
dynamics. Water levels are approximately consistent
bridge girder. Table 2 shows the water level data,
except for Case 3, which had a larger thickness (Figure 5,
measured by the water-level indicator from the bridge
Table 3). The water level of Case 3 is higher as the flow is
front to the upstream position (i.e., 0.4 m from the bridge
stuck into the thicker bridge. That is, the water is piled up
front) with a 0.1 m interval for each cross section,
along the lateral-bride wall. Case 2 with half thickness of
perpendicular to the flow direction (Figure 2). The data
the bridge shows the upper water (surface flow)
were averaged over water levels on each cross section
separated by the bridge girder more splashes with a
laterally between both lateral walls. The measured water
larger thickness because of a smaller thickness. These
level increased gradually as flow was close to the bridge
results from the ideal test imply that only thickness of the
front (Table 2). This suggests that the risen water level
bridge affects the water level immediately in front of the
along the lateral wall of the bridge girder occurred near
bridge and thickness of the upper water over the bridge.
the bridge front because the bridge girder behaves like an
In addition, when the bridge was shifted to a lower
obstacle against flow.
potion, Case 6 indicated that the water level in front of
the bridge becomes higher, which is equivalent to that of
Table 2. Observed water levels (m) Case 3. This situation was likely caused by overflow due
DISTANCE FROM
to the obstacle (i.e., bridge) against the upstream flow
0.0 0.1 0.2 0.3 0.4 (Figure 6).
THE BRIDGE

WATER LEVEL 0.108 0.095 0.086 0.082 0.085

Note that water level is defined as height from the channel bottom
to water surface.

Test cases for the 2D flow simulation in one horizontal


and one vertical were conducted with the tank water
from an upstream location that potentially impacts a
bridge. Case 1 was run for a comparison with
observation (Table 1). Figure 4 shows that the water
level from the bridge to the upstream area was increased
by the flows impacting the bridge. In addition, the flow
was divided into a top surface flow above the bridge
girder and an undersurface flow below the bridge girder
because of particle-like movements. The simulated water
level at the front of bridge is overestimated slightly by
approximately 5% as compared with the observed water
level. This result suggests that the SPH method could Figure 5. 2D flow comparison among four zoomed-in snapshots at
capture the risen water level even without driftwood as 5 s near the bridge, simulated based upon different bridge
well as separated water. characteristics.

3
Figure 6. Zoomed-in snapshot for the 2D simulated flow at 5 s
near the bridge with a lower height of the bridge position.

For simulation results, we are interested in the


quantitative evaluation of how much the bridge
characteristics contribute to a raised water level. Note
that an increased height of water surface (hereafter, dam-
up) is defined as a height from the bridge top to water
level immediately in front of the bridge in this study. It
was calculated by simple average among neighborhood
particles which are in water surface. According to the
data in Table 3, the dam-up values for Case 2 and Case 6
were much larger because of shorter thickness with Figure 7. Zoomed-in snapshots for the 2D simulated pressure
strong overflow, and the obstacle under water surface distribution at 5 s near the bridge for Case 1 and Case 6. Colored
with strong overflow, respectively. Overall implication of scaling is given by log10. A small box figure shows a vertical
this ideal test is that the bridge height position and longer pressure profile in front of the bridge with log10 scaling. For
thickness are crucial to increase the water level pressure computation immediately in front of the bridge, the
immediately in front of the bridge and that a lower pressure values were averaged over from the bridge front to 0.05
m upstream position at each 0.01 m vertical interval because of
position of the bridge and a shorter thickness potentially
computational instability nearby the bridge.
cause a large dam-up. Like Case 6, a lower position of the
bridge than the upstream water level, in other word
smaller space underneath the bridge girder into which 4. CONCLUSIONS
the upstream flow hardly penetrates, may occur
A fundamental test for the high-intensity, rainfall-
potentially when the significant amount of driftwood is
induced flood at a river bridge was conducted with the
accumulated at the bridge.
SPH-used 2D simulation and the open channel-based
Finally, for a simple evaluation for the effect of flows laboratory experiment. This test was successfully
on the bridge, the pressure distributions at the bridge performed to gain a better understanding of physical
front for Case 1 and Case 6 are shown in Figure 7. Case 1 processes of this unique flood, caused by high flows that
suggests that the pressure at the bridge girder front was strike the bridge. The result even without driftwood
higher than that under the girder because of flow shows that water level rose slightly because of the bridge
collision. The pressure distribution at the upstream depth girder, and that the SPH reasonably captured the various
for Case 6 was much stronger due to the increased water flows separated by the bridge girder. In addition, the
level and the narrow space under the bridge girder. This evaluation for the effects of bridge characteristics on the
suggests the bridge may be damaged by high-pressure flow dynamics revealed that a lower position of the
flows bridge height and longer thickness of the bridge girder
raised water level immediately in front of the bridge and
that a lower position of the bridge height and a shorter
Table 3. Simulation results (m).
thickness of the bridge girder resulted in a larger dam-
Case name WATER LEVEL DAM-UP OVERFLOW up. In particular, a lower position of the bridge height
CASE 1 0.115 0.005 weak causes high-pressure-flow collision against the bridge
CASE 2 0.113 0.018 strong girder.
CASE 3 0.130 0.005 weak
CASE 4 0.116 0.006 weak
ACKNOWLEDGMENTS
CASE 5 0.116 0.006 weak Our study was supported by KAKENHI (Grant #
CASE 6 0.132 0.022 strong 24760398 and # 26420499).
Note that dam-up was measured immediately in front of the
bridge and that water level here is given by a height from water REFERENCES
surface to channel bottom. Adachi S. and Daido A. (1957). Experimental study on
washed timbers. Prev. Res. Inst. Annual (Kyoto Univ.),
No.1, 1-9. (Japanese)

4
Crespo, A.J.C., Gmez- Gesteira, M and Dalrymple, R.A.
(2007). Boundary conditions generated by dynamic
particles in SPH methods. CMC-Comput. Mater. Con.,
5(3), 173-184.
Gotoh H., Sakai T., Hayashi M. (2001). Lagrangian
particle method for analysis of dam-up process by drift
timbers. Annual J. Hydraul. Eng. JSCE, 45, 919-924.
(Japanese)
Intergovernmental Panel on Climate Change (IPCC)
(2012). Managing the risks of extreme events and
disasters to advance climate change adaptation. A
special report of working groups I and II of the IPCC
[Field, C.B., Barros V., Stocker T.F., Qin D., Dokken
D.J., Ebi K.L., Mastrandrea M.D., Mach K.J., Plattner
G.K., Allen S.K., Tignor M. and Midgley P.M. (eds.)].
Cambridge University Press, Cambridge, UK and New
York, USA, 582 pp.
Lucy, L.B. (1977). A numerical approach to the testing of
the fission hypothesis, Astron. J., 82, 10131024.
Monaghan, J. J. (1992). Smoothed particle
hydrodynamics. Annual Rev. Astron. Appl., 30, 543- 574.
Monaghan, J.J. (2005). Smoothed Particle
Hydrodynamics. Rep. Prog. Phys. 68, 17031759.
Nakagawa H., Inoue K., Ikeguchi M., Tsubota T. (1993)
Numerical simulation of drift wood behavior (2) dam
up of drift wood -. Annuals Disas. Prev. Res. Inst. Kyoto
Univ., 36, B-2, 487498. (Japanese)
Sakano A. (2003). Investigation of hydraulic features for
driftwood accumulation and water level increase to a
bridge. Japan National Institute for Land and
Infrastructure Management, Technical Report No.78, 87
pp. (Japanese)
Shimizu Y. and Osada K. (2007). Numerical experiments
on accumulation process of driftwoods around piers by
using a DEM-FLOW coupling model. Annual J. Hydraul.
Eng. JSCE, 51, 829-834. (Japanese)

Vous aimerez peut-être aussi