Valladares TFM 2018 Final

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

Computational Methods for Geomorphic Flows: Application to

River Dike Failure Due to Overtopping Flow

José Manuel Valladares Bustillo


Master’s Final Paper

ABSTRACT
The failure of river dikes can produce considerable damages to private property,
cropping fields and even cause loss of human lives. One of the reasons for dike failure is
the overtopping of water above the crest level. This occurs when the approach water
level exceeds the crest elevation of the structure, resulting a transient flow that erodes
the dike body. The purpose of this work is to compare two one-dimensional numerical
schemes for the solution of the Shallow Water Equations describing geomorphic flows.
The one-sided upwind finite volume scheme and the MacCormack finite difference
scheme were considered with different equilibrium sediment transport equations for
the bed-load, given its dominant role. Results were compared with laboratory
observations. Although both schemes produce reasonable results during initiation of
motion, they diverge from experiments at large times, thereby indicating that the full
physical process is still not well understood. The finite volume scheme was simple to
implement, it is easy to extrapolate to two-dimensional simulations on unstructured
meshes, and is free from spurious oscillations.

1. Introduction

Fluvial dikes are hydraulic structures used for flood control, to channelize flow,
and to prevent riverbanks deformation [Rifal et al., 2017]. Failure of earth dikes can
have catastrophic effects on cropping fields, residential zones and on the
environment due to the rapidly increasing water level and the sudden change in flow
speed [Alkema and Middelkoop, 2005]. A frequent and dangerous cause of dike
breaching is flow overtopping, which occurs when an inflow discharge pulse produce
water levels above the dike crest [Vorogushyn et al., 2010]. Despite the importance
of dike breaching, most computational studies are centered on dam-break flow
waves. Overtopping flow and its effects on the morphology of the dike are little
understood due to the difficulty to obtain precise measurements on lab experiments
and field investigations[He et al., 2015].

Numerical simulation of transient flows due to dike overtopping is useful to


quantify flood damages and to have a better understanding of hydraulic structures
design. Thus, it is important to study computational methods applied to predict the

1
morphology and erosion of hydraulic structures and river beds. For these reasons
dike overtopping has a significant importance in both research and practice.

In the past decades there have been a significant number of studies on dam break
processes solving the nonlinear shallow-water equations coupled with sediment
transport equations [He et al., 2015]. Simulation of movable beds is more difficult
than fixed beds due to the unknown boundary changing in time and space as function
of sediment transport. Tingsanchali and Chinnarasrin [1999] simulated dam erosion
due to overtopping flow but neglected the effect of sediment transport on the
shallow water equations, while Wu and Wang[2008] simulated dam-break flow over
movable beds using explicit algorithms that considered the effect of sediment on the
flow density and bed deformation.

The prediction and simulation of dike breach due to overtopping is conducted


coupling the shallow-water equations with sediment transport conservation laws
[Alhasan et al., 2016], taking into consideration the concentration of sediments, the
bed deformation, and non-equilibrium conditions [Wu, 2008]. In this work two
different numerical schemes were used to solve the shallow water equations for
geomorphic flows, thereby simulating the erosion-deposition processes due to
overtopping flow above a fluvial dike. The first one is the one-sided upwind finite
volume scheme [Wu and Wang, 2007, 2008], while the other one is the MacCormack
finite difference scheme [Tingsanchali and Chinnarasrin, 1999]. The sediment
transport capacity must be determined resorting to empirical equations [Wu, 2008],
and it plays a major role in sediment transport models. In dike overtopping flows
bed-load is the dominant sediment transport mode. Thus, different bed-load
transport equations such as those by Wu et al. [2000], Meyer-Peter & Muller [1948],
Wong and Parker [2006], and Nielsen [1992] were used in a non-equilibrium
sediment transport model. Simulations were conducted using both numerical
schemes for the different bed-load equations and compared with laboratory
experiments.

2. Materials and methods

Fluvial dikes are man-made structures constructed to contain, channelize and


divert flow from a river in order to provide protection to an area of interest. A dike
cross section consists of a body, a crest and the berms on one or both sides of the
structure (Fig. 1). The body has a trapezoidal shape, with the slopes playing a
substantial role in dike geotechnical stability. Its height depends on the river
maximum level.

2
Figure1. Typical dike cross section and its elements

The dike material is important for its stability and the sediment particle size is a
crucial magnitude in its erosion resistance. In this work only non-cohesive,
homogenous material is considered. According to Chinnarasri et al. [2003] four stages
are observed during the dike’s erosion process due to overtopping: Firstly the outer
side of the crest starts to erode almost instantly, setting the initial failing point of the
structure. Afterwards the outside slope starts to budge as water flows over it. Then,
the dike's cross section acquires a wave-shape profile and finally a large sediment
wedge deposits at both ends of it.

Transient geomorphic flows in rivers are described by shallow water equations


accounting for the sediment transport. In the case of geomorphic flows, the
interaction of the flow with the sediment transport and the river morphology
becomes of great significance. To account for this interaction Wu and Wang [2007]
proposed the following generalized shallow water equations(Fig. 2)

(1)

( )

(2)

where t is the time, x the longitudinal coordinate, A the flow cross section, Q the flow
discharge, the density of the water-sediment mixture, the change in bed area, g

3
is the gravitational acceleration, the water surface elevation, the local flow
depth, n the Manning coefficient, and R the hydraulic radius.

Figure 2: Flow and riverbed cross section

Sediment transported by rivers is divided into suspended and bed load transport
rates. The sediment transport capacity can be computed separately for both
transport modes; however, another approach is to calculate the total sediment being
transported without making any distinction between bed load and suspended load
[Wu, 2008]. For this work the latter approach will be used, as it simplifies the
calculations while still producing the information necessary for engineering purposes.
The transport equation for the total sediment load is as follows [Wu et al., 2004]

(3)

where is the volumetric concentration of sediments, is the total sediment flux, B


is the channel width (taken here as unity considering a rectangular cross-section),
and E and D are the sediment entrainment and deposition rates, respectively. For the
bed deformation Wu and Yang [2007] proposed

( )( )

(4)

4
where is the porosity of the bed material calculated using the Komura [1963]
formula modified by Wu and Wang [2005].Using Wu et al. [2004] formula, the net
exchange rate B(E-D) is obtained by

(5)

with L as the non-equilibrium adaptation length of sediment transport. The sediment


transport capacity is given by

(6)

Here and are the bed and suspended load transport rates per unit width.
Erosion initiates once the shear stress of the flow exceeds the critical shear stress of
the bed material. For geomorphic flows and bed deformation due to dike
overtopping the bed load transport rate becomes the main form of transport and its
characterization becomes the basic problem regarding transport formulas
[Schmocker, 2011]. There are several empirical formulas based on flume
experiments, field investigation and a combination of both. As a starting point, Wu et
al. [2000] formulas for suspended load and bed load transport rate were used to
calculate and in the numerical models as

(7)

5
( )

(8)

Then, three different formulas of bed load transport rate obtained previously were
used to produce alternative simulations. The alternative equations used are listed in
table 1.

Name Formula Reference

Meyer-Peter Meyer-Peter, E. and Müller, R., 1948, Formulas for


Bed-Load Transport, Proceedings, 2nd Congress,
& Muller International Association of Hydraulic Research,
Stockholm: 39-64.

Wong and Parker Wong, M., and G. Parker (2006), Reanalysis and
correction of bed load relation of Meyer-Peter and
Muller using their own database, J. Hydraul. Eng.,
132, 1159-1168

Nielsen √ Nielsen, P. (1992). Coastal bottom boundary layers


and sediment transport. World

Scientific, River Edge, N.J.

Table 1.Bed load transport rates used in the numerical models. Cq is the transport rate non-
dimensional parameter√( ) .

The shallow water equations are a complex of nonlinear hyperbolic differential


equations to be solved using numerical methods [Wu 2008]. In this work two
techniques are used, namely the finite volume method and the finite difference
method. In the finite volume method an integral form of the conservations laws is
solved, whereas in the finite difference method the original set of PDEs is considered
for numerical solution. The first-order upwind finite volume scheme is tested in this
work, given its simplicity and good performance. Experimental work on dike
breaching suggests that these flows are free from shocks. Therefore, a first-order
scheme seems adequate, given that a sharp resolution of bores is not necessary.
However, a second-order scheme will be considered to test this assumption. In

6
particular, MacCormack second-order finite difference method was used
[MacCormack, 1969]. Finite difference methods are difficult to extrapolate to two-
dimensional flow on irregular domains, whereas the generalization of the first-order
upwind method to two-dimensional flows on unstructured meshes is
straightforward. Comparison of both methods for the 1D case will depict if finite-
difference method shows any advantage to its consideration for the 2D case. The
spatial mesh for the 1D case is shown in Fig. 3, where the 1D finite-volumes are
sketched. The finite-difference mesh nodes are coincident with the gravity centers of
the finite volumes for a coherent comparison of both computational schemes.

Figure 3. Representation of the spatial cell grid

Wu and Wang [2007] reformulated equations (1) and (2) eliminating the flow
density in the left sides resulting in the conservative equation

(9)

where U, F(U), and S(U) are the vectors of conservative variables, fluxes and sources
respectively defined as

[ ] [ ]

[ ]

(10)

7
For the finite volume method the following discretized equation was obtained by Wu
and Wang [2007]

( )

(11)

where i+1/2 refers to the interface between adjacent finite volumes. There are
several approaches to determine the numerical flux. In this paper the one-sided
upwind method was used. The topographic source term is computed with an average
of up- and downwind free surface slopes, and the sediment transport source term
with a central discretization (Appendix A).

At the inflow boundary a transmissive condition was implemented to compute the


water depth using a gosh cell. In this case the information propagates backward in
the next time step as shown in figure (4). The inflow discharge was introduced as an
instantaneous pulse directly at the gosh cell i=1 and kept constant during all the
simulation. At the downstream boundary section both depth and discharge were
computed at a gosh cell using transmissive boundary conditions.

Figure 4. Water depth transmissive boundary condition

For the finite difference method the two step predictor-corrector MacCormack
scheme was used. In the predictor step the fluxes are evaluated at the time level n
using backward finite differences. This results in an estimation of the flow conditions
at the new time level n+1 as denoted in equation (11)

8
( )

(12)

For the corrector step the fluxes are evaluated with the estimated values of U in the
predictor step using forward finite differences. This step provides a new estimation of
the flow condition at the new time level n+1 as denoted in equation (12)

( )

(13)

The final step is conducted doing an average of the predictor step and the corrector
step results, resulting

(14)

According to Godunov’s theorem any scheme of order higher than one will suffer
spurious oscillations near steep gradients, like at the vicinity of a shock wave. These
dispersive errors can be suppressed by introducing artificial viscosity into the scheme.
In this work the method of Jameson et al. [1981] was used. First, compute the
parameter

(15)

9
(16)

Here k is the artificial viscosity coefficient, which was settled to 1.5. Finally the
numerical smoothing is computed as follows

(17)

In the finite difference scheme the boundary conditions are implemented by using
the method of characteristics. The upstream section is subcritical, and, therefore, we
set the constant inflow discharge as the known variable on the boundary and
calculate the flow depth variable resorting to the backward characteristic.
Meanwhile, the downstream section is supercritical; consequently, both depth and
discharge are computed by simultaneous solving the backward and forward
characteristics using information from the previous time step.

The results obtained using both schemes were compared with hydraulic model
experiments conducted by Schmocker [2011]. The trapezoidal model dike was
formed with homogenous, non-cohesive sediments, and the tests were conducted on
a rectangular flume. All breach processes were recorded with a high-speed camera
and the images were analyzed using a graphical program to correct any distortion.

3. Results

Two different computational models were developed to simulate overtopping


flow over a non-cohesive granular dike. For both cases the bed load transport rates
formulas of Wu, Meyer-Peter & Muller, Wong & Parker and Nielsen were used to
investigate their impact on the simulation of the eroded dike and the water surface
profile. To observe the performance of the models, experimental data at times 4s
and 10s since initiation of the experiments were considered.

10
The dike had a height of 0.2m, crest length of 0.1m, upstream discharge of
, a Manning coefficient of 0.018 and slope 2:1 at both sides. The sediment
diameter was 0.002m, with a density of , a settling velocity of and a
porosity of 0.3115. For the computational models a mesh with Δx= 0.025 m and a
Courant–Friedrichs–Lewy number equal to 0.25 was used.

The results obtained with the one-sided first-order upwind finite volume scheme
and MacCormack finite difference scheme at times 4s and 10s are presented in
figures (5-8). Figure (5) shows that at the initial stages of motion the finite volume
scheme approximate quite well the laboratory experiments. The bed load transport
rate of Wu and Meyer-Peter & Muller gave the best results, while Nielsen formula is
not accurate. At t =10s(figure 6) the numerical results diverge from the experimental
observations. Wu and Meyer-Peter & Muller formulas gave the best results, while
Nielsen formula and Wong and Parker produce poor results. In the finite difference
scheme it was difficult to fully suppress the oscillations at both times of study (figure
7 and 8). Both schemes produced similar results, but the finite volume method had
the advantage of being free from oscillations. Figure (9) represents the dike erosion
evolution throughout time for the finite volume scheme.

11
Figure 5. Comparison of one-sided first-order upwind finite volume scheme using Wu,
Meyer-Peter & Muller, Wong & Parker and Nielson bed load transport rate formulas with
flume tests at time study 4s.

12
Figure 6. Comparison of one-sided first-order upwind finite volume scheme using Wu, Meyer-
Peter & Muller, Wong & Parker and Nielson bed load transport rate formulas with flume tests
at time study 10s.

13
Figure 7. Comparison of MacCormack finite difference scheme using Wu, Meyer-Peter &
Muller, Wong & Parker and Nielson bed load transport rate formulas with flume tests at time
study 4s.

14
Figure 8. Comparison of MacCormack finite difference scheme using Wu, Meyer-Peter &
Muller, Wong & Parker and Nielson bed load transport rate formulas with flume tests at time
study 10s.

15
Wu

Meyer-Peter & Muller

Wong and Parker

Nielsen

Figure 9. Comparison of dike erosion for the of one-sided first-order upwind finite volume
scheme using Wu, Meyer-Peter & Muller, Wong & Parker and Nielson bed load transport rate
formulas for several times of study.

16
4. Conclusions

Two different numerical schemes were applied to simulate dike erosion due to
overtopping flow over a granular fluvial dike. Both gave similar results, but the finite volume
scheme was free from oscillations, given that it is a first-order scheme. In contrast, in the
second-order finite different scheme it was difficult to remove the spurious oscillations.
Therefore, it is preferable to use a 2d extended version of the finite volume first-order one-
sided upwind scheme for future generalizations to complex domains on unstructured
meshes.

Dike erosion sediment transport is dominated by bed load. Several empirical formulas
were tested, and Wu’s equation was found to produce the best results.

Discrepancies between numerical simulations and experiments suggest that there are
important elements not considered in the physics of current shallow water flow models for
dike erosion. Two important physical features not accounted for in the present work are the
dynamic pressures on the dike surface as the water flows over the dike, and the body dike
deformation due to the seepage pore pressures. Further research on the physics of these
processes is therefore needed in parallel to advances in the practical numerical
implementation.

17
References

Alkema, D., and H. Middelkoop (2005). The Influence of Flood Plain


Compartmentalization on Flood Risk Within the Rhine‐Meuse Delta.Nat. Hazards, 36,
125–145.

Alhasan, Z., Jandora, J. andRiha, J. (2016). Comparison of Specific Sediment Transport


Rates Obtained from Empirical Formulae and Dam Breaching Experiments.
Environmental Fluid Mech,16: 997-1019.

Chinnarasri, C., Tingsanchali, T., Weesakul, S. and Wongwises, S. (2003).Flow Patterns


and Damage of Dike Overtopping. Int. J. Sediment Res., 18(4), 301-309.

He, Z.; Hu, P.; Zhao, L.; Wu, G.; Pähtz, T. (2015).Modeling of Breaching Due to
Overtopping Flow and Waves Based on Coupled Flow and Sediment
Transport.Water, 7(8), 4283-4304.

Jameson, A., Schmidt, W., Turkel, E. (1981).Numerical Solution of the Euler Equations
by Finite Volume Methods Using Runge–Kutta Time-Stepping Schemes. AIAA Paper,
14th Fluid and Plasma Dynamics Conference, Fluid Dynamics and Co-located
Conferences, 1981.

Komura, S. (1963).Discussion of Sediment Transportation Mechanics: Introduction


and Properties of Sediment. J. Hydraul. Div., ASCE, 89, HY1, 236-266.

Meyer-Peter, E. and Müller, R.(1948).Formulas for Bed-Load Transport. International


Association of Hydraulic Research, IAHSR 2nd Congress, Stockholm, 39-64.

Nielsen, P. (1992).Coastal Bottom Boundary Layer and Sediment Transport. Advanced


Series on Ocean Engineering, World Scientific Edition, vol. 4.

Prosser, Ian and Rustomji, Paul (2000).Sediment Transport Capacity Relations for
Overland Flow. Progress in Physical Geography: Earth and Environment, Vol 24, Issue
2, pp. 179 – 193.

Rifal, I.,Erpicum, S., Archambeau, P., Violeau, D., Pirotton, M., El KadiAbderrezzak, K.,
Dewals, B. (2017).Overtopping induced failure of noncohesive, homogeneous fluvial
dikes. Water Resour. Res., 53, 3373–3386.

MacCormack, R.W. (1969).The Effect of Viscosity in Hypervelocity Impact Cratering.


American Institute of Aeronautics and Astrophysics, pp. 69-354

18
Schmocker, Lukas (2011). Hydraulics of Dike Breaching (doctoral dissertation).
Retrieved from ETH-Zurich Research Collection, p. 31.

Tingsanchali, T. and Chinnarasri, C. (1999).Mathematical Model of Dam Surface


Evolution Due to Flow Overtopping. Proc. of 1999 Int. Water Resources Engineering
Conf., Seattle, USA

Vorogushyn, S., B. Merz, K.‐E. Lindenschmidt, and H. Apel (2010).A New Methodology
for Flood Hazard Assessment Considering Dike Breaches. Water Resour. Res., 46,8.

Wong, M., and G. Parker (2006).Reanalysis and Correction of Bed Load Relation of
Meyer-Peter and Muller Using Their Own Database. J. Hydraul. Eng., 132, 1159-1168.

Weiming Wu & Sam S.Y. Wang, S.S.Y. (2008).One-Dimensional Explicit Finite-Volume


Model for Sediment Transport. Journal of Hydraulic Research, 46, 1, pp. 87-98.

Wu, W. (2004).Depth-Averaged 2-D Numerical Modeling of Unsteady Flow and


Nonuniform Sediment Transport in Open Channels. J. Hydraul. Eng., ASCE 130(10),
1013-1024

Wu, W., S.S.Y. Wang,and Y. Jia.(2000). Nonuniform Sediment Transport in Alluvial


Rivers. J. Hydr. Res., IAHR, 38(6), 427–434.

Wu, W. and Wang, S.S.Y. (2005).On Sediment Deposit Porosity and Settling Velocity.
Under review by J. Hydraul, Eng. ASCE.

Wu, W., and Wang, S.S.Y. (2007). One-Dimensional Modeling of Dam-Break Flow
Over Movable Beds. J. Hydraul, Eng. ASCE 133(1), 48-58.

Wu, W. (2007). Computational River Dynamics, Numerical methods, (pp. 113),


Leiden, Netherlands: Taylor & Francis/Balkema.

19
Appendix A

Topographic and Sediment Transport Source Terms

Topographic source term:

If Qi> 0 and Qi+1 > 0:

If Qi< 0 and Qi-1 < 0:

20
In any other case:

The variables w1 and w2 are the weighting factors of the downwind and upwind free
surface slope and are based on the Courant number Crdown and Crup respectively.

Sediment transport source term:

( )

where Ct is the sediment mixture concentration.

21

You might also like