Вы находитесь на странице: 1из 38

Here is a preprint (submitted manuscript) of the following paper:

Amicarelli A., B. Kocak, S. Sibilla, J. Grabe; 2017; A 3D Smoothed Particle Hydrodynamics model for erosional dam-
break floods; International Journal of Computational Fluid Dynamics, 31(10):413-434; DOI
10.1080/10618562.2017.1422731

Please refer to the published version of the manuscript (Version of Record -VoR-) to filter errors and typos, and for the
parts added during the revision process (with a full list of references):
“ The Version of Record of this manuscript has been published and is available in International Journal of
Computational Fluid Dynamics (2017) http://www.tandfonline.com/10.1080/10618562.2017.1422731 “

Notes not reported on the VoR.


Assuming both conservative particles and Weakly Compressible approach permits to use a reduced speed of sound and
is consistent with the momentum balance equation (which needs conservative particles).
We could not ask to modify a sketch (on the initial conditions; sketch not present on the preprint) to graphically
lengthen the gate/dam through the sediment layer, until the domain bottom.
A 3D SMOOTHED PARTICLE HYDRODYNAMICS MODEL
FOR EROSIONAL DAM-BREAK FLOODS

Andrea Amicarelli1,5, Bozhana Kocak2,3, Stefano Sibilla4, Jürgen Grabe2


1
Ricerca sul Sistema Energetico - RSE SpA, Department SFE, via Rubattino, 54, 20134, Milan, Italy
2
Hamburg University of Technology, Institute of Geotechnical Engineering and Construction Management,
Harburger Schlossstrasse 20, 21079, Hamburg, Germany
3
Hamburg Port Authority, Neuer Wandrahm 4, 20457, Hamburg
4
University of Pavia, Dept. of Civil Engineering and Architecture, via Ferrata, 3, 27100, Pavia, Italy
5
Corresponding author
andrea.amicarelli@rse-web.it
bozhana.kocak@tuhh.de
stefano.sibilla@unipv.it
grabe@tuhh.de

Abstract.
A mesh-less SPH model for erosional dam-break floods is outlined. The model formulation is

consistent with the “packing limit” of the Kinetic Theory of Granular Flow, despite the

approximations underlying a mixture model, which represents both the main liquid phase and the

solid, saturated granular material. The model is validated on the results from several experiments on

erosional dam breaks. A numerical comparison between the present mixture model and a 2-phase

SPH model for geotechnical applications (Gadget Soil; TUHH) is also performed on a 2D test case.

A demonstrative 3D erosional dam break on complex topography is investigated: the structural

failure of a dam triggers a violent water flow, which alternatively thrusts or erodes, and then

transports and deposits the granular material of the downstream mobile riverbed. The numerical

developments of this study are implemented in the SPH code SPHERA v.8.0 (RSE SpA),

distributed as FOSS on a GitHub public repository.

Keywords.
SPH; erosional dam breaks; mixture model; CFD; granular material; bed-load transport; safety of
hydroelectric plants; accidents; floods; particle methods.
1. INTRODUCTION
Hydroelectric plants play a key role in flood propagation as triggering elements, exposed and/or

damaged structures, or as flood control works.

Damage induced by partial or full dam breaks can be huge. The most devastating example is

provided by the 1975 dam break in Banqiao, China, triggered by the typhoon Nina, which caused

around 230’000 fatalities, representing the fourth deadliest flood ever.

The risk assessment of a dam break impact on human health and the surrounding environment is

crucial to secure the involved plants (e.g., [20]). In particular, the interaction of the dam break flow

with the underlying mobile granular bed plays a key role in the propagation of the dam break front

and the associated sediment transport phenomena. In this context, Computational Fluid Dynamics

(CFD) codes provide a relevant contribution, as discussed in the following.

Erosional dam breaks (i.e. dam breaks over granular beds) were experimentally studied by several

authors (e.g., [38], [30]). The associated experimental datasets have been also used to validate the

most popular numerical models used at present to simulate erosional dam breaks. These codes are

based on the 2D Shallow Water Equations (SWE ; e.g., [7], [14], [36]) and usually need an ad-hoc

tuning for the viscosity of the mixture of water and granular material.

A recent alternative to the solution of the governing equations with mesh-based Eulerian methods

such as Finite Volumes (FV) resides in the adoption of Lagrangian methods (e.g., [34], [46]).

Among the various numerical schemes, Smoothed Particle Hydrodynamics (SPH) has several

advantages: a direct estimation of free surface and phase/fluid interfaces; effective simulations of

multiple moving bodies and particulate matter within fluid flows; direct estimation of Lagrangian

derivatives (absence of non-linear advective terms in the balance equations); effective numerical

simulations of fast transient phenomena; no meshing; simple non-iterative algorithms (in case the

“Weakly Compressible” approach is adopted). On the other hand, SPH models are still affected by a

few shortcomings, if compared with mesh-based CFD tools, in terms of computational costs

(normally higher) and lower accuracy as a general purpose technique. Nevertheless, SPH models

2
are effective in several, but peculiar, application fields. Some of them are here briefly recalled:

flood propagation (e.g., [42], [17]); sloshing tanks (e.g., [18], [4]); gravitational surface waves (e.g.

[8]); hydraulic turbines; fast landslides; liquid jets (e.g., [12]); astrophysics and magneto-

hydrodynamics (e.g. [31]); body dynamics in free surface flows (e.g., [18], [3]); multi-phase and

multi-fluid flows; sediment removal from water reservoirs (e.g., [26]); landslides (e.g., [1]); natural

convection (e.g., [10]).

In this context, the present paper proposes a 3D SPH model for erosional dam breaks and is

organized as follows: the new numerical model is described in section 2, while the results of this

study are discussed in section 3. Validation is based on experimental datasets regarding three 2D

and one 3D erosional dam breaks ([13], [14], [38], [36]). An inter-comparison with a state-of-the-art

SPH model is also assessed, by taking into account a 2D erosional dam break with a very high

density ratio. Furthermore, a demonstrative 3D erosional dam break on complex topography is

reported. The overall conclusions are drawn in section 4.

3
2. THE NUMERICAL MODEL
After a brief, general introduction to Smoothed Particle Hydrodynamics (Section 2.1), this chapter

describes the proposed numerical model (Section 2.2) and briefly recalls the state-of-the-art model

(Section 2.3) here used for the inter-comparison of Section 3.4.

2.1. Smoothed Particle Hydrodynamics

SPH is a mesh-less CFD technique, whose computational nodes are represented by numerical fluid

particles (for a complete presentation of the method refer, for instance, to [27], [25] and [44]). In the

continuum, the functions and derivatives appearing in the governing balance equations of fluid

mechanics are approximated using convolution integrals over smoothing functions. Then the SPH

integral approximation < >I of a generic function f is:

  fWdx
3
f I , x0 (2.1)
Vh

where x0 is the position of a generic computational point and W is a weighting (or kernel) function,

which is non-zero only inside the integration volume Vh, also called kernel support, whose

characteristic (or smoothing) length is h (usually, the kernel support is a sphere of radius 2h in the

inner domain, far from boundaries). A generic first-order derivative along xi direction can be still

computed as in (2.1), replacing f by its derivative. After integration by parts, one obtains:

f W 3
  fWn dx f
2
dx (2.2)
xi xi
i
I , x0 Ah Vh

Here a surface integration is considered all over the surface Ah of the kernel support. This term

vanishes far from the boundaries, where the kernel function goes to zero, but is peculiar and

necessary at boundaries, where the kernel support is truncated by the frontiers of the fluid domain;

here SPH truncation errors are of major interest ([5]) and need particular boundary treatments. The

representation of the surface term in (2.2) can relevantly differentiate SPH models and is still object

of recent studies and algorithms for the treatment of solid boundaries (e.g., [2], [32], [11]).

4
In a discrete representation, the flowing mass is divided into elementary masses m, or particles. Far

from the boundaries, the particle SPH approximation of the generic first-order derivative becomes:

f W
  f b b (2.3)
xi x0 b xi b

where a summation over particle volumes m/ replaces the volume integral in (2.2). The

subscripts “0” and “b” refer to a generic computational particle and its neighbouring particles (i.e.

particles inside the kernel support centred on the particle located in x0), respectively. One may

notice that (2.3) is derived straightforwardly from (2.2) just for educational purpose, but this

approximation is normally corrected when implemented into the SPH balance equations (Section

2.2).

The generic nth derivative can be still approximated with the SPH technique, using the same

approach as for the first derivative. However, consistent schemes for the derivatives can be obtained

only if a suitable renormalization of the kernel gradient is adopted (see [35], [24] for a general

discussion on this issue).

The functions and derivatives comparing in the balance equations of fluid dynamics can be then

replaced by their corresponding SPH particle approximations, in the frame of a particle and

Lagrangian approach.

2.2. The SPH mixture model for erosional dam breaks

This SPH model represents the mixture of pure fluid and non-cohesive solid granular material,

under the “packing limit” of the Kinetic Theory of Granular Flow (KTGF, [6]). This limit refers to

the maximum values of the solid phase volume fraction and is peculiar of bed-load transport (e.g.,

erosional dam breaks) and fast landslides. The present model is integrated in the SPH code

SPHERA v.8.0 (RSE SpA, [37]).

The continuity equation for the fluid phase (“f”) reads ([6]):

5
d  f  f    f  f u f , j 
 (2.4)
dt x j

where  is density,  the phase volume fraction, u the velocity vector, t represents time and x the

position vector. Einstein notation applies to the subscript “j”, hereafter.

The continuity equation for the incompressible solid phase (“s”) reads ([6]):

d  s  s    s  s us , j 
 (2.5)
dt x j

The volume balance equation assumes the following form:

s  f 1 (2.6)

After defining the mixture density and velocity (the subscript “m” is always omitted):

 f  f u f ,i   s sus ,i
   f  f   s s , ui  (2.7)

the summation of (2.4) and (2.5) provides:

d  u j 
 (2.8)
dt x j

The model assumes that SPH particles are conservative (i.e. mixture particles do not exchange mass

with the surrounding ones), which is a reasonable hypothesis for high solid volume fractions in

saturated soils, according to the “packing limit” of the Kinetic Theory of Granular Flow (KTGF,

[6]):

d u
0 j 0 (2.9)
dt x j

Starting from (2.9), the model adopts a Weakly Compressible approach to obtain:

d u j
  (2.10)
dt x j

The SPH approximation of (2.10) reads:

d0 W W 3
 0  ub, j  u0, j  b  2 0  u w  u0   nn j dx (2.11)
dt b x j V'
x j
b h

6
where n is the unit vector normal to the frontier. The subscripts “0”, “b” and “w” refer to the

computational particle, a neighbouring particle and a wall frontier, respectively. The integral

boundary term is computed according to [11].

Considering the KTGF, the momentum equation for the fluid phase reads ([6]):

d  f  f u f ,i  p f  f ,ij
  i 3 g f  f   f   K gs u f ,i  us ,i  (2.12)
dt x j x j

where  is the deviatoric (or shear) stress tensor, g gravity acceleration and p pressure. The last term

depends on the relative velocity between the phases (filtration process) through the drag coefficient

Kgs.

The momentum equation for the solid phase reads ([6]):

d  s  s us ,i  ps  s ,ij p f
  i 3 g s  s   s  K gs u f ,i  us ,i  (2.13)
dt x j x j x j

Provided the volume equation and the definitions of the mixture velocity and density, the sum of

(2.12) and (2.13) provides:

d ui  p f ps  f ,ij  s ,ij


  i 3 g     (2.14)
dt x j x j x j x j

Considering the assumption on conservative SPH particles, the shear stress gradient term of the

fluid phase reads ([6]):

 f ,ij   f   u f , j   2 u f ,i 
f   f  (2.15)
x j  
 3 xi  x j  x 2j 

Under the hypothesis of plain strain, the shear stress gradient term is represented by [33] (plastic

model for dry granular material based on internal friction), by means of a parameter, which KTFG

names frictional viscosity, as described in the following.

The pressure of the solid phase is treated as follows:

ps ,m m' 3


 i'
  , m  
'
(2.16)
xi xi i 1 3

7
where  m' is the mean effective stress ([33] refers to Mohr-Coulomb criterion; the extension to

Mohr-Coulomb-Terzaghi criterion for saturated soils is straightforward). The shear stress gradient

term in the momentum equation reads ([33]):

 s ,ij
 k


 s' ,m  eij
1
 e , 1  u u 
eij   i  j 
x j x j 2  x j xi 
ij

(2.17)
1
 
eij    eij2   2 I 2 eij  , k  2 sin  
2

 i, j 

where  is the internal friction angle, eij is the strain-rate tensor and I2(eij) represents its second

invariant (formulation for incompressible fluids). One may notice that the term (2.17) is potentially

unstable at high internal friction angles and that, in the “packing limit” of the KTGF, the shear

stress terms of the collisional-kinetic regime are zeroed. Eq.(2.17) can be rearranged as follows:

 s ,ij    s ,m 
'

 2 sin   eij 
x j  2 I 2 eij  
(2.18)
x j
 

The KTGF introduces the following definition:

  ' sin   
 fr   m 
 2 I 2 eij  
(2.19)
 

to obtain:

 s ,ij 
 2 fr eij  (2.20)
x j x j

Eq.(2.20) is consistent with internal friction, as it does not depend on the magnitude of the strain-

rate tensor.

In analogy to the Navier-Stokes equation (for incompressible flows), under the strong but accepted

hypothesis of smooth spatial variations of fr, (2.20) provides:

 s ,ij  2 ui
  fr (2.21)
x j x 2j

8
which is valid in the “packing limit” of the KTGF (i.e. for s values of about 0.50-0.55). The model

then defines the mixture total pressure p as:

p  p f   m' (2.22)

One may consider that the mean effective stress can only be formulated under simplifying

assumptions (e.g., x, y and z need to be the principal axes). Thus,  m' is computed as the difference

between the total pressure and the fluid pressure:

 m'  max  p  max( p f ,0),0 (2.23)

Both fluid and solid pressures are limited to positive values as soils, which are either fully saturated

or dry, do not bear tension. Considering the continuity equation, the momentum equation for the

mixture can be rearranged as:

d ui  p f  m'  2 u f ,i  2 u s ,i
  i 3 g    f f  H  s   s , p  fr (2.24)
dt xi xi x 2j x 2j

where εs,p = 0.5 and H(x) is the Heaviside step function:

H x   max x,0
d
(2.25)
dx

Assuming SPH conservative particles implies that the velocity of each phase is basically equal to

the one of the mixture:

u f ,i  ui , us ,i  ui (2.26)

Considering (2.22) and the assumption of SPH conservative particles, (2.24) reduces to:

dui 1 p  2u
  i 3 g   2i (2.27)
dt  xi x j

where the mixture viscosity  is finally defined as:

   f  f  H  s   s, p  fr (2.28)

With the boundary treatment method proposed by [11], the SPH approximation of (2.27) becomes:

9
W W
u b  u 0 W
dui
p  p0 
1 p mb
  i 3 g  b  2 0  x dx 3  2  
0 xi b 0  0 r0b r
b
dt 0 b Vh' i b b b

 1
(2.29)
W W 3   1 W 3 
  M  b 2 u b  u 0    x b  x 0    M u SA  u 0     2 x  x 0  dx  2 uw,i  u0 ,i   
m
dx 
 r x  r x   r r 
b 0 0b i b  Vh 0 w
' i   Vh 0 w
'

where r the distance between two interacting particles, whereas M stands for the artificial viscosity

([28]). The boundary value uSA of the velocity in the external portion Vh’ of the kernel support is

assigned according to [11]. So far, the last term, representing the bottom drag, has been validated in

2D.

Despite its formulation as a mono-phase mixture, the model needs to adopt a simplified approach to

represent fluid pressure in the granular material. This parameter can be related to two different soil

conditions: uniform fully saturated soil and uniform dry soil (the first condition being applied to all

the test cases in this study):

 p f ,blttop   f g ( zblttop  z0 ) cos 2 TBT , fully saturated soil


pf   x0 , y0
(2.30)
0, dry soil

where the subscript “blt-top” refers to the top of the bed-load transport layer (or the layer of saturated

material). Eq.(2.30) assumes a 1D filtration flow parallel to the slope of the granular material. This

simplifying hypothesis is still consistent with SPH conservative particles; is the topographic

angle at the top of the bed-load transport layer and lies between the local interface normal and the

vertical:


TBT  max  arccos nblt top,3 ,  (2.31)
 2

The angle limiter in (2.31) allows one to assign null pf values in case of slope anomalies (very rare

and unstable).

The mixture pressure is computed by means of a barotropic equation of state (linearized around a

reference state indicated by subscript “ref”):

p  cref
2
  ref  (2.32)

10
It has been shown that the artificial speed of sound c in the Weakly-Compressible approach should

be at least 10 times greater than the maximum velocity to reduce the pressure error associated to

artificial compressibility effects below 1% ([27]). A unique speed of sound can be chosen (i.e. the

highest among the SPH particle values, no matter about their phase volume fractions).

The Leapfrog time integration scheme of the model is then constrained by the following stability

criteria:

 2h 2 2h 
dt  min 0 C ; CFL  (2.33)
  c u 

where, following [2], the Courant-Friedrichs-Lewy number (CFL) and the viscous term stability

parameter are CFL=0.1 and C=0.05, respectively. Owing to the high viscosities of the mixture, the

first term of (2.33) is usually the main constraint for dt in the modelling of bed-load transport and

fast landslides, at least at fine spatial resolutions. In this frame, it might be convenient to adopt a

multi-step approach, where the time integration of the equations of motion for the fluid particles

would be obtained with a longer time step than the one needed for the mixture particles.

In order to reduce the computational time and avoid the unbounded growth of (2.19), a threshold for

the mixture viscosity can be defined. Mixture particles with a higher viscosity are considered in the

elasto-plastic regime of soil deformation and are held fixed, whereas their pressure is derived from

the mixture particles flowing above them. The threshold value is deemed to be high enough not to

influence the simulation (Figure 2.1).

No-slip conditions were imposed on solid walls in the 2D simulations discussed in the following.

As a matter of fact, these boundaries were in general in contact with the bottom of the fixed bed, so

the choice of a no-slip rather than a slip condition did not play any role. On the other hand, 3D

simulations were obtained by imposing free-slip conditions in 3D. In fact, the depth of the granular

material layer in the 3D test case of Section 3.5 is high enough and the interactions with solid walls

quite exclusively concern either fluid particles at high Reynolds number or mixture particles with

null velocities, so that the choice of a slip condition everywhere appeared to be an appropriate

11
compromise. Nevertheless, no-slip conditions should be in general applied to those mobile mixture

particles which are interested by a locally laminar regime. This issue seems to play a minor role in

the simulations of this study (except for the “demonstrative” test case of Section 3.6, which only

proposes preliminary results) and represents a matter of on-going developments.

The model algorithm is synthesized in the following. At the beginning of each iteration, the time

step duration is assessed. The mixture viscosity and the momentum equation are computed and the

particle velocities are integrated in time to yield particle trajectories. Particles are then ordered on a

coarse Cartesian grid to efficiently compute, once per time step, all the quantities related to particle

distance needed by the SPH technique (relative distances, kernel functions and derivatives, Shepard

coefficient, the preconditioned dynamic vector for the neighbour search -[43]-). The solution of the

continuity equation yields then the density, which is then converted into pressure, through the state

equation (“Weakly Compressible” approach). Pressure is then regularized through a smoothing

procedure and converted again in density.

Figure 2.1. Effect of the maximum (threshold) mixture viscosity on the mixture rheology. Under the frictional regime
(||ij||>||ij||*), Schaeffer (1987) equation is applied (red horizontal line). Under the elasto-plastic regime (||ij||<||ij||*),
SPH mixture particles are held fixed (blue vertical line) to approximate negligible strains. Here the liquid viscosity
shear stress term is assumed to be negligible, as it usually occurs far from the top of the bed-load transport layer.
2.3. The geotechnical SPH model of Grabe & Stefanova (2014)

This section briefly introduces a state-of-the-art geotechnical SPH model (Gadget Soil, TUHH,

[41], [15], [16], [22], [39]), whose results permit a further assessment (Section 3.4) of the model

presented in this study (Section 2.2). Gadget Soil, which is based on Gadget H20 by Thomas Rung

and Christian Ulrich ([41]), aims at simulating geotechnical problems that include large

12
displacements and soil-water interaction. It is developed at the Institute of Geotechnical

Engineering and Construction Management at Hamburg University of Technology.

The model uses a weakly compressible approach with Tait’s equation for water as well as XSPH

velocity smoothing. The LES approach for turbulence is applied according to [41].

Three basic particle types are available in Gadget Soil: sand, water as well as liquefied sand

particles. The boundaries can be modelled using fixed boundary particles or ghost particles.

Saturated soil is described as a two phase material, consisting of soil and water particles.

The effective stress in the soil phase, by means of Terzaghi’s definition, is calculated using a

hypoplastic constitutive model, according to [45]. This is a rate type constitutive model that is able

to describe the nonlinear and anelastic behaviour of soil. The rate of change of the effective stress

rate tensor  ij'  is determined as follows:

 ij'  h1,ij  ij' , eij , ev  (2.34)

where h1,ij is a tensor valued function and ev the void ratio.

The constitutive law uses nine parameters in its conventional version ([45]) and five more

parameters in its extension for intergranular strains ([29]).

The interaction between the solid and the water phase is defined by means of an interaction force

for the laminar and the turbulent cases, thus representing the effects of the filtration process (flow

through porous media). The model provides alternative momentum equations for the water phase

du f ,i p p  W    W
  m f ,b  2f ,0  2f ,b    m f ,b  f 2,ij ,0  f 2,ij ,b  
   
dt 0 b   f ,0  f ,b  xi b b   f ,0  f ,b  xi b
(2.35)
g
A0,b u f ,i ,0  us ,i ,b W0,b   B0,b u f ,i ,0  us ,i ,b  W0,b  i 3
ms ,b ms ,b

2

b  s ,b  f ,0 b  s ,b  f ,0  f ,0
    
Rla min ar Rturbulent

and the solid phase (sand grains):

 '  '  W g
A0,k u f ,i ,b  us ,i ,0 W0,b   B0,b u f ,i ,b  us ,i ,0  W0,b  i 3
dus ,i m f ,b m f ,b
  mb  ij2,0  ij2,b  
2

dt 0 b   s ,0  s ,k  xi k 
b  s , 0  f ,b
  b  s ,0  f ,b
  
 s ,0 (2.36)
Rla min ar Rturbulent

13
where A and B here quantify the interaction factors for the laminar and the turbulent flow regimes,

respectively.

The continuity equations for both phases are discretized as follows:

d f W d s W
  f  u f ,b, j  u f ,0, j  b ,   s  us ,b, j  us ,0, j  b (2.37)
dt b x j dt b x j
b b

More details on the definition and determination of the interaction factors for the laminar and

turbulent cases are given in [39] and [16].

Since the constitutive model for soil would only work as long as no tension occurs, an additional

approach is needed, in case of effective significantly low stresses. Under these conditions, a soil

liquefaction threshold is introduced. The model considers the soil as liquefied when the average

normal stress is lower than the threshold value pliq=5Pa. The average normal stress (pav) is defined

depending on the number of spatial dimensions (nD) and the effective stress tensor  ij' :

 ij ij'
pav  (2.38)
nD

where Einstein notation works for the subscripts “j” and “i”. In case the threshold value has been

reached, the soil starts to behave as a viscous fluid with a variable viscosity, according to [41]:

pav sin   c

I 2 eij  (2.39)

The pressure for the liquefied soil is then calculated using a state equation analogous to that of water ([41],

[15]).

14
3. RESULTS

The validation of the model outlined in Section 2.2 involves comparison of numerical results with

experimental datasets on various flow configurations, namely three different 2D erosional dam

breaks ([13], [38], [14]) and a 3D erosional dam break ([36]). An inter-comparison with the state-

of-the-art SPH code Gadget Soil is also assessed by investigating a 2D bed-load transport

phenomenon with a very high density ratio s/f (2D erosional dam break with Karlsruher sand).

Finally, a “demonstrative” simulation of a 3D erosional dam break event on a complex topography

is shown to illustrate the capabilities of the proposed method.

3.1. 2D erosional dam break by Fraccarollo & Capart (2002)

The experiment is described by Fraccarollo & Capart ([13]) and was used as test for numerical

simulation by the same authors and, concerning SPH, by [23].

In the experiment, the length of the granular bed is Ls = 2m, with initial depth of Hs = 0.06m; the

water reservoir has length Lw = 1.000m and height Hw = 0.100 m; the solid grains are made of PVC

and have d50 = 0.00392m, s = 1’336kg/m3 and  = 38°; the phenomenon lasts for 1s.

The SPH simulation was performed with an initial particle spacing dx = 4mm, h/dx = 1.3and f=

0.37, resulting in a saturated mixture density = 1’212kg/m3.

The pressure coefficient is defined as C p  0.5U 2  1’923Pa; velocities and time are normalized by

U  2 gH w  1.961m/s and t0  H w  0.1s, respectively.


g

Figure 3.1 shows a sequence of snapshots with particles coloured according to their volume

fractions: pure fluid (blue) or mixture (red; saturated granular material). The dam is suddenly

removed at T=0. Due to a reduced effective mean stress, the granular material lifts up around the

dam break front (T=1.0s), noticeably alters the free surface evolution (T=2.5) and is transported

downstream, where it partially settles.

Figure 3.2 reports an example of the pressure and velocity field. Pressure is generally hydrostatic,

but in a wide region around the water front, where the mean effective stress relevantly deviates
15
from its equilibrium values. On the other hand, the velocity field is constrained by the positions of

the interfaces and the maximum values of velocity occur at higher positions than those found during

equivalent dam breaks events on fixed beds.

The free surface and interface profiles are compared with the experiments in Figure 3.3 at several

times T. Each image reports a couple of superior curves (free surface profiles), a couple of

intermediate curves (“mixture – pure fluid” interface) and a bottom curve representing the fixed bed

as defined in the experimental dataset.

On the other hand, the detection of the fixed bed is highly dependent on the choice of the velocity

threshold defining this interface, both numerically and experimentally. The fixed bed is numerically

represented by those particles interested by the elasto-plastic regime of soil deformation (Section

2.2). The procedure to detect this regime aims to make the model accurate at low strain-rates, but

cannot guarantee a reliable estimation of the fixed bed top as very low velocity values are

approximately zeroed. Nevertheless, the experimental definition of the fixed bed is also affected by

relevant uncertainties when a bed-load transport layer is evolving over it. In this context, only the

position of the experimental fixed bed top is reported (as provided by other authors), with no

numerical comparison for this curve, just to provide a very approximate quantification of the depth

of the bed-load transport layer.

The results show a good performance of the model, which can satisfactorily represent both the free

surface and the “mixture - pure fluid” interface. Nevertheless, the bed-load transport depth is

slightly underestimated, especially at the front of the water wave at later times (T=10), and the

numerical free surface is a little smoother than the experimental one. These minor shortcomings can

be related to several assumptions of the model (Section 2.2). Further, the size of the SPH mixture

particles is limited by the grain size, so that we cannot provide very fine resolution simulations with

a continuum, mono-resolution approach.

16
Figure 3.1. 2D erosional dam break: SPH simulation of the experiment by Fraccarollo & Capart (2002): sequence of
snapshots of the solid volume fraction field from T = 0 to T = 10. Colour scale ranging from s=0 (pure fluid, blue) to
s=1 (saturated granular material, red); z-scale amplified by 5 times.

a) b)
Figure 3.2. 2D erosional dam break: SPH simulation of the experiment by Fraccarollo & Capart (2002): normalized
pressure (a) and velocity (b) fields at T = 2.5; z-scale amplified by 5 times.

17
Figure 3.3. 2D erosional dam break: SPH simulation of the experiment by Fraccarollo & Capart (2002): comparison
between simulated and measured values of free surface and the top of the bed-load transport layer. The bottom curve
represents the fixed bed defined in the experimental dataset. Symbols: + computed free-surface; × measured free-
surface; * computed interface; □ measured interface; ■ measured fixed bed top.

3.2. 2D erosional dam break by Spinewine (2005)

This experiment by Spinewine ([38]) is also reported in [23]. In the experiment, the length of the

granular bed is Ls = 6m and its initial height is Hs = 0.100m; the water reservoir has length Lw =

3.000 m and height Hw = 0.350 m. The phenomenon lasts for 0.75s.

The granular material has the same properties of the one described in the erosional dam break by

[13]. The SPH simulation was performed with an initial particle spacing dx = 5mm. All the other

parameters are the same of the test case of Section 3.1.

Figure 3.4 reports a sequence of snapshots with the SPH pure fluid (blue) and mixture (red)

particles. Although the phenomenon is qualitatively similar to the previous test case, the velocity

and length scales are higher, the water front is faster and the interfaces are less regular.

18
Figure 3.4. 2D erosional dam break: SPH simulation of the experiment by Spinewine (2005): sequence of snapshots of
the solid volume fraction field. Colour scale as in Figure 3.1; z-scale amplified by 5 times.

Figure 3.5. 2D erosional dam break: SPH simulation of the experiment by Spinewine (2005): comparison between
simulated and measured values of free surface and top of the bed-load transport layer; symbols as in Figure 3.3.

The available measured profiles refer to the end of the simulation (T=3.968, t0=0.189s). The free

surface and water-mixture interface profile are compared with the experiments in Figure 3.5, and

show the reliability of the model.

3.3. 2D erosional dam break by Fraccarollo et al. (2003)

This third 2D erosional dam break test case refers to the Taipei experiment by Fraccarollo & Capart

(2002), with useful details also reported in Fraccarollo et al. ([14]).

19
Figure 3.6. 2D erosional dam break: SPH simulation of the experiment by Fraccarollo et al. (2003):
sequence of snapshots of the solid volume fraction field (colour scale as in Figure 3.1; z-scale amplified by 5 times) and
comparison between simulated and measured values of the position of free surface and bed-load transport layer top
(symbols as in Figure 3.3).

In the experiment, the length of the granular bed is Ls = 1.500m and its initial height is Hs = 0.050.

The water reservoir has length Lw = 0.750 m and height Hw = 0.100 m; the solid PVC grains have

d50 = 0.006m ([9]), s = 1’048kg/m3, f = 0.37 (resulting in a mixture density = 1’212kg/m3) and

= 26° (as assumed in the numerical simulations by [14]). The phenomenon lasts for 5t0 = 0.505s.

The SPH simulation was performed with an initial particle spacing dx = 0.010m and h/dx = 1.3.

Measured profiles are available at T=3,4,5.

20
Figure 3.6 shows a sequence of snapshots of the volume fraction field ranging from with pure fluid

(blue) to granular material (red).

Here the solid density is much lower than the one of the previous test cases, causing a stronger

reduction in the mean effective stress at the dam break front, a huge rise of the granular material

and a deeper bed load transport region, with respect to the previous test cases.

On the right column of Figure 3.6, the comparison of numerically simulated and measured positions

of the free surface and of the top of the bed-load region shows a good performance of the model in

reproducing both the interface profiles. Nevertheless, some minor discrepancies in the local patterns

of the free surface and the dam break front velocity could be possibly still improved if using a

multi-resolution approach.

3.4. 2D erosional dam break with Karlsruher sand

The mixture model of Section 2.2 here represents an erosional dam break with much higher values

of the ratios s/f and Hw/d50 than the previous test cases (z-axis has the same scale as x-axis in

Figure 3.7).

Experimental datasets seem not available under these conditions. Thus, inter-comparisons are

provided by applying the state-of-the-art SPH model Gadget Soil (Section 2.3) on the same test case

and considering its results as a reference.

The length of the granular bed is Ls = 3.000m, with initial depth of Hs = 0.200m; the water reservoir

has length Lw = 1.500m and height Hw = 1.300m; the solid grains represent Karlsruher sand, often

used for geotechnical experiments and simulations. Thus, d50 = 0.0007m, s = 2’650kg/m3 and  =

30°. Porosity is set to f= 0.41, so that the saturated mixture density is = 1’973kg/m3.

SPH simulations consider the initial particle spacing dx = 0.0125m, h/dx = 1.3 (SPHERA) or 2.4

(Gadget Soil) and time is normalized by t0  g 1H w  0.364s.

21
Figure 3.7. 2D erosional dam break with Karlsruher sand: sequence of snapshots for the field of the solid volume
fraction. SPH particles are coloured in blue (water) or red (saturated granular material).
T=0.00,0.69,1.10,1.37,1.65 (from the top to the bottom panel).
Left: SPHERA v.8.0 (model of Section 2.2). Right: Gadget Soil (state-of-the-art SPH code of Section 2.3).

Gadget Soil takes into account a hypoplastic constitutive model for sand. This is a comprehensive

model, which is able to describe the complex behavior of granular soils. A set of input parameter is

required, that are determined for the so called Karlsruher sand, a type of model sand (Table 3.1).

A sequence of snapshots for the field of the solid volume fraction is reported in Figure 3.7. The

final profiles (T=1.65) of the top of the bed-load transport layer show comparable shapes, despite a

22
Parameter Description Value
critical state friction angle
granular stiffness
exponent
minimal void ratio
critical void ratio
maximal void ratio
exponent
exponent
initial void ratio
Table 3.1. Hypoplastic constitutive model of Gadget Soil: soil parameters for the so called Karlsruher sand.

higher fragmentation of the interfaces for the 2-phase model. The two SPH models show a similar

behavior in terms of velocity propagation of the water front and provide comparable results in

estimating the position of the regions where the bed shapes are the highest. Nevertheless, one

notices appreciable differences between the simulations, mainly concerning the impact between the

water front and the soil grains. The local velocities of the bed shapes are also quite different. Gadget

Soil bed shapes deflect water upward more relevantly than SPHERA and causes a partial breaking

of the water front, which flows lifted from the soil (since T=0.69 to T=1.37 in Figure 3.7) and

ingurgitates air bubbles. The differences between the models mainly refer to the treatment of the

granular material: Gadget Soil represents a 2-phase model, whereas SPHERA a mixture model.

Gadget Soil should be much more precise than SPHERA for small displacements within the soil, as

it uses a hypoplastic constitutive model. Gadget Soil assumes a threshold value for the mean

effective stress to choose whether treating the soil as a Bingham fluid or a hypoplastic medium.

Nevertheless, the definition of this threshold may need particular attention. Finally, Gadget Soil

makes use of a LES scheme, whereas SPHERA does not use any turbulence scheme by simply

assuming this kind of erosional dam breaks to be almost gravity-driven inertial flows.

3.5. 3D erosional dam break by Pontillo (2010)

The 3D erosional dam break experiment by Pontillo ([30]) has been described and used for

validation of SWE numerical methods by the same author, [36] and [40].

23
The length of the granular bed is Ls = 6.000m and its initial height Hs = 0.100m. The water reservoir

has length Lw = 3.000 m, height Hw = 0.250 m and width WW = 0.250 m. A sharp enlargement is

located at x = 4 m, where the width of the domain doubles.

Density and porosity are s = 2’680kg/m3 and f = 0.42, respectively ([36]), while some

uncertainties remain in some key experimental values: the grain size is in the range 0.00165m < d50

< 0.00182m, while the internal friction angle was not measured: here a value = 32° is derived

from d50 and f, according to a reference plot of Kirkpatrick ([21]) for sand.

The model here considers only the first stages of the phenomenon, after the front crosses the

enlargement (until t=3s). The SPH simulation was performed with an initial particle spacing dx =

0.011m and h/dx = 1.3.

Figure 3.8. 3D erosional dam break SPH simulation: sequence of snapshots of the solid volume fraction field from T = 0
to T = 10. Colour scale ranging from s = 0 (pure fluid, blue) to s = 1 (saturated granular material, red).

Figure 3.9. 3D erosional dam break SPH simulation: time-history of the computed (+) and measured (×) water depth
before (probe U1, left) and after (probe U3, right) the enlargement.

24
Figure 3.10. 3D erosional dam break SPH simulation: sequence of snapshots of the non-dimensional velocity
magnitude.

a) b)
Figure 3.11. 3D erosional dam break SPH simulation: a) top view of the enlargement with the non-dimensional velocity
magnitude field; b) view of the pressure coefficient field.

Figure 3.8 shows a sequence of snapshots where the SPH particles range from pure fluid (blue) to

saturated mixture (red). The erosional dam break front is characterized by a strong decrease in the

mixture viscosity of the granular material, which lifts up and locally pushes upward a quite irregular

free surface. After the enlargement, part of the granular material seems to settle due to a local

velocity reduction, while an oblique hydraulic jump takes form on the left frontier. However, this

flow feature should not be accurately described by the model at the smallest scales.

Figure 3.9 reports the numerical and experimental free surface levels, registered before the

enlargement, on the left (probe “U1” at x = 3.750 m; y = 0.125 m), and after it, on the right (probe

“U3” at x = 4.200 m; y = 0.125). A noticeable delay of the front is recorded from the very first

stages of the simulation. However, an analogue shortcoming affects the results of the numerical

SWE model by Pontillo (2010, [30]). Apart from this discrepancy, the numerical free surface levels

seem to approach fairly well the experimental ones.

Figure 3.10 reports a sequence of the velocity magnitude field. One can notice the oblique hydraulic

jump (t=2s) and the peculiar position of the maximum values of velocity, at highest levels with

respect to the original bed top height.

25
The top view of the velocity magnitude field (Figure 3.11.a) highlights the 3D pattern of the

hydraulic jump, while the pressure field appears to be rather regular (Figure 3.11.b).

3.6. 3D erosional dam break on complex topography

Due to the lack of data on experimental 3D erosional dam breaks on complex topographies, a

demonstrative test case is set up to show the applicability of the model to real-life situations. The

topography and the reservoir are provided by a demonstrative ICOLD Benchmark, which

represented a dam breach ([19]). The 3D erosional dam break is triggered by assuming the

instantaneous collapse of a gravity dam, whose structure is not simulated. The water flow impacts,

erodes and transports a portion of the downstream mobile bed, which is composed by a deposit of

granular material (Figure 3.12). Its original sedimentation is assumed to be related to the presence

of a weir, whose structure deteriorated and was then removed before the dam building. The planar

domain dimension is 24.627×9.855 km, and the coordinates of the lower left corner of the domain

are (1.153; 0.071) km; the minimum elevation is zmin=67m. The water surface elevation in the

reservoir is Hw=272m. The dam centre is located at (4.500;6.682) km. The phenomenon lasts

around 25 minutes. Due to the demonstrative aim of this test, the resolution of the fictitious

topography of [19] is increased to dxtop=76m, with 21’823 topographic vertices. Both the

moderately coarse spatial resolution and the high velocity scale motivate the adoption of free-slip

wall boundary conditions. Six monitoring sections are selected to record flow rate and free surface

level time histories, as described in Figure 3.12 (top left panel). The granular material of the mobile

bed is characterized by d50=0.0001m, s=1’700kg/m3, =0.37 and =30°. The SPH simulation was

carried out with an initial particle spacing dx=4.0m, h/dx=1.2 and CFL=0.25.

Figure 3.12 and Figure 3.13 report top and valley views of the phenomenon, with SPH particles

representing pure water (in blue) or the saturated granular material (in green). The total number of

particles is 6.8×105. During the impact of the dam break front with the mobile bed (t=144s),

velocity reaches values up to 36m/s (Figure 3.13, top right panel).


26
Figure 3.14 reports a sequence of snapshots of the water reservoir and the mobile bed (bottom view

without topography).

Figure 3.12. 3D erosional dam break on complex topography. Sequence of snapshots (top views) with particles coloured
in blue (pure water: water reservoir) or green (mixture of saturated granular material: mobile bed). The top left panel
also shows the position of the monitoring sections (red dots).

The 2D fluid dynamics synthetic fields of the maximum water depth and the maximum specific

flow rate (Figure 3.15) can be used as input elements for those numerical tools which represent the

failure criteria of structures and infrastructures of the energy system (20), potential targets of floods

and landslides. These tools implement simplified failure criteria (e.g., sliding, buckling, floating,

toppling, impact of debris), based on the Limit State Equations. They could take advantage from

being fed by the synthetic fields derived from the 3D non-stationary fields of the present model.

Figure 3.16 describes the hydrographs in terms of total flow rate (top left panel), bed-load flow rate

(top right panel) and mixture depth (bottom panel), at the monitoring sections. Mixture depth

(Figure 3.16, left panel) is here defined as a generalization of the water depth and involves the

whole fluid domain (both the pure liquid and the mixture of granular material). The peak values of

the total flow rate almost occur instantaneously at the sections which lie upstream of the mobile bed

(“P1”, “P2” and “P3”). This feature is peculiar of small reservoirs and very fast dam breaks.

Downstream, the interaction of the dam break flow with the mobile bed further delays the

occurrence of the hydrograph peaks with respect to the first passage of the pure liquid. The total

flow rate peak decreases with the section barycentre height, but for “P4”, which is close to the initial

27
downstream end of the mobile bed (with a null slope at the beginning of the simulation). The time

history of the bed-load flow rates (Figure 3.16, top right panel) show peaks of one order of

magnitude lower than the hydrographs. Considering the monitoring sections with non-null values,

the peak value decreases with the section barycentre height, except for a local outlier in the plot of

“P6”. The monitoring section “P3” records a very low cumulated volume of granular material as this

monitor is located near the initial upstream edge of the mobile bed. The mixture depth (Figure 3.16,

bottom panel) quite monotonically decreases at the two monitoring sections which pass for the

reservoir (“P0”, “P1”).

Figure 3.13. 3D erosional dam break on complex topography: valley views (z-scale is amplified by a factor of 3). Top
right panel: example of the surface velocity field. Other panels: sequence of snapshots with SPH particles coloured in
blue (pure water) or green (mixture).

Figure 3.14. 3D erosional dam break on complex topography: bottom view without topography of the upstream portion
of the numerical domain. Particles coloured in blue (pure water) or green (mixture).

28
Figure 3.15. 3D erosional dam break on complex topography: maximum values of the mixture depth (left panel) and the
specific flow rate (right panel) after a post-processing procedure at the spatial resolution of 9.0m.

Figure 3.16. 3D erosional dam break on complex topography. Monitoring sections(hydrographs): pure water flow rate
(top left panel), bed-load flow rate (top right panel) and mixture depth (bottom panel).

The monitoring sections “P3” and “P4”, which lie within the initial volume of the mobile bed, record

of course a non-null initial mixture depth, which changes suddenly at the monitoring section “P3”,

due to the passage of the dam break front (t=40s-100s), whose huge impact with the mobile bed

locally deviates the flow upward. The strong increase at t≈400s at the same section can be related to

the passage of a temporary hydraulic jump, caused by a retrograde non-stationary steep-fronted

wave. This wave is generated by the presence of the mobile bed, which temporarily acts as a sharp

slope change. The peak of the mixture depth decreases with the section barycentre height, but for

“P4”, which is located close to the downstream end of the mobile bed, where the initial depth of

granular material is maximum.

Despite some simplifications reported above and the lack of available experiments for validation,

this test shows the capability of the present model in representing a 3D erosional dam break on a

complex topography, with high fluid volumes and over a long time scale: the model yields the

29
estimation of the major fluid dynamics parameters involved in flood risk management, and a direct

3D representation of both the water reservoir (pure liquid) and the mobile bed (mixture of saturated

granular material).

30
4. CONCLUSIONS

A SPH mixture model for erosional dam-break floods is presented. The validation against available

experimental data shows the good reliability of the model in reproducing 2D and 3D erosional dam

break events.

One of the advantages of the model is that mixture viscosity does not need to be tuned, as it occurs

in other numerical models commonly used for representing bed-load transport. The consistency of

the model formulation with the “packing limit” of the Kinetic Theory of Granular Flow (KTGF) is

discussed.

The performance of the mixture model is comparable with that of a state-of-the-art two-phase SPH

model for geotechnical applications (Gadget Soil, TUHH; [39], [41], [15], [16]) in simulating a 2D

erosional dam break, at very high values of the ratios s/f and Hw/d50.

The simulation of a 3D erosional dam break event on a complex topography proves the applicability

of the proposed method to practical problems in civil engineering.

The numerical developments of this study are implemented in the SPH code SPHERA v.8.0 (RSE

SpA), which is distributed as FOSS on a GitHub repository ([37]).

Acknowledgements.
The work of the RSE author has been financed by the Research Fund for the Italian Electrical
System (RdS) under the Contract Agreement between RSE SpA and the Italian Ministry of
Economic Development for the RdS periods 2012-2014 and 2015-2017, in compliance with the
Decrees of 9 November 2012 and 21 April 2016. Reference projects: “Sviluppo del Sistema e della
Rete Elettrica Nazionale”, Frigerio A. et al., 2012-2015; “A.5 - Sicurezza e vulnerabilità del sistema
elettrico”, Frigerio A. et al., 2015-2018.
We acknowledge the CINECA award under the ISCRA initiative, for the availability of High
Performance Computing resources and support. In fact, the HPC simulations on SPHERA refer to
the following HPC research projects:
a) HPCEFM7b – High Performance Computing for Environmental Fluid Mechanics 2017b (Italian
National HPC Research Project); instrumental funding based on competitive calls (ISCRA-C
project at CINECA, Italy); 2017-2018; Amicarelli A. (Principal Investigator, 160’000core-hours
equivalent to c.a.20keuro), G. Pirovano.
b) HPCEFM17 - High Performance Computing for Environmental Fluid Mechanics 2017 (Italian
National HPC Research Project); instrumental funding based on competitive calls (ISCRA-C
project at CINECA, Italy); 2016-2017; 200’000core-hours; Amicarelli A. (Principal
Investigator, 200’000core-hours equivalent to c.a.25keuro), G. Agate, N. Pepe, G. Pirovano, M.
Torresi, B. Fortunato, F. Fornarelli, F. Scarpetta, G. Viccione, E. Pugliese Carratelli, V.
Bovolin, S. Sibilla, E. Persi, G. Petaccia.

31
c) PANCIA - high PerformANce Computing for Interface and Atmospheric flows (Italian National
HPC Research Project); instrumental funding based on competitive calls (ISCRA-C project at
CINECA, Italy); 2016-2017; Vacondio R. (Principal Investigator), A. Amicarelli, G. Curci, A.
dal Palù, S. Falasca, A. Ferrari, L.M. Baldini; 110’000core-hours (equivalent to ca. 14keuro).
d) NMTFEPRA - Numerical Modelling of Turbulent Flows for Environment Protection and Risk
Assessment (Italian National HPC Research Project); instrumental funding based on
competitive calls (ISCRA-C project at CINECA, Italy); 2016-2017; Ferrero E. (Principal
Investigator), A. Bisignano, A. Amicarelli, G. Curci, S. Manenti, S. Todeschini, A. Bisignano,
S. Falasca. 40’000core-hours equivalent to c.a. 5keuro.
e) HPCEFM16 - High Performance Computing for Environmental Fluid Mechanics 2016 (Italian
National HPC Research Project); instrumental funding based on competitive calls (ISCRA-C
project at CINECA, Italy); 2016; Amicarelli A. (Principal Investigator, 146’000core-hours
equivalent to c.a. 18keuro), G. Curci, S. Falasca, E. Ferrero, A. Bisignano, G. Leuzzi, P. Monti,
F. Catalano, S. Sibilla, E. Persi, G. Petaccia.
f) HPCEFM15 - High Performance Computing for Environmental Fluid Mechanics 2015 (Italian
National HPC Research Project); instrumental funding based on competitive calls (ISCRA-C
project at CINECA, Italy); 2015; Amicarelli A. (Principal Investigator, 100’000core-hours
equivalent to c.a. 13keuro), A. Balzarini, S. Sibilla, G. Agate, G. Leuzzi, P. Monti, G. Pirovano,
G.M. Riva, A. Toppetti, E. Persi, G. Petaccia, L. Ziane, M.C. Khellaf.
g) HSPHMI14 - High performance computing for Lagrangian numerical models to simulate free
surface and multi-phase flows (SPH) and the scalar transport in turbulent flows (MIcromixing)
(Italian National HPC Research Project); instrumental funding based on competitive calls
(ISCRA-C project at CINECA, Italy); 2014-2015; Amicarelli A. (Principal Investigator;
71’000core-hours equivalent to c.a. 9keuros), G. Agate, G. Leuzzi, P. Monti, R. Guandalini, S.
Sibilla.

References.

32
1. Abdelrazek A.M., I. Kimura, Y. Shimizu; 2016; Simulation of three-dimensional rapid free-

surface granular flow past different types of obstructions using the SPH method; Journal of

Glaciology, 62(232):335-347.

2. Adami S., X. Y. Hu, N. A. Adams; 2012; A generalized wall boundary condition for smoothed

particle hydrodynamics; Journal of Computational Physics, 231, 7057–7075.

3. Amicarelli A., R. Albano, D. Mirauda, G. Agate, A. Sole, R. Guandalini; 2015; A Smoothed

Particle Hydrodynamics model for 3D solid body transport in free surface flows; Computers &

Fluids, 116:205–228.

4. Amicarelli A., G. Agate, R. Guandalini; 2013; A 3D Fully Lagrangian Smoothed Particle

Hydrodynamics model with both volume and surface discrete elements; International Journal for

Numerical Methods in Engineering, 95, 419–450, DOI: 10.1002/nme.4514.

5. Amicarelli A., J.-C. Marongiu, F. Leboeuf, J. Leduc, M. Neuhauser, Le Fang, J. Caro; 2011; SPH

truncation error in estimating a 3D derivative; International Journal for Numerical Methods in

Engineering, 87-7, 677-700.

6. Armstrong L.M., S. Gu, K.H. Luo; 2010; Study of wall-to-bed heat transfer in a bubbling

fluidised bed using the kinetic theory of granular flow; Internat. Journal of Heat and Mass Transfer,

53(21-22):4949-4959.

7. Chauchat J., M. Médale; 2010; A three-dimensional numerical model for incompressible two-

phase flow of a granular bed submitted to a laminar shearing flow; Comput. Methods Appl. Mech.

Engrg. 199: 439–449.

8. Colagrossi A., A. Souto-Iglesias, M. Antuono, S. Marrone; 2013; Smoothed-particle-

hydrodynamics modeling of dissipation mechanisms in gravity waves; Physical Review E -

Statistical, Nonlinear, and Soft Matter Physics, 87(2), art. no. 023302.

9. Costabile P.; 2006; Propagazione di onde di dam break su fondo erodibile; XXX Convegno di

Idraulica e Costruzioni Idrauliche (IDRA).

33
10. Danis M.E., M. Orhan, A. Ecder; 2013; ISPH modelling of transient natural convection;

International Journal of Computational Fluid Dynamics, 27(1):15-31.

11. Di Monaco A., Manenti S., Gallati M., Sibilla S., Agate G., Guandalini R., 2011; SPH modeling

of solid boundaries through a semi-analytic approach; Engineering Applications of Computational

Fluid Mechanics, 5, 1, 1–15.

12. Espa P., Sibilla S., Gallati M.; 2008; SPH simulations of a vertical 2-D liquid jet introduced

from the bottom of a free-surface rectangular tank, Adv. Appl. Fluid Mech., 3, 105-140.

13. Fraccarollo L., Capart H.; 2002; Riemann wave description of erosional dam-break flows; J.

Fluid Mech., 461, 183-228.

14. Fraccarollo L., H. Capart, Y. Zech; 2003; A Godunov method for the computation of erosional

shallow water transients; Int. J. Numer. Meth. Fluids, 41: 951–976.

15. Grabe, J., B. Stefanova; 2014; Numerical modeling of saturated soils, based on Smoothed

Particle Hydrodynamics (SPH). Part 1: Seepage analysis. geotechnik 37(3):191-197.

16. Grabe J., Stefanova B.; 2015; Numerical modeling of saturated soils based on Smoothed

Particle Hydrodynamics (SPH) - part 2: coupled analysis; Geotechnik, 38(3):218-229.

17. Gu S., X. Zheng, L. Ren, H. Xie, Y. Huang, J. Wei, S. Shao; 2017; SWE-SPHysics simulation

of dam break flows at South-Gate Gorges Reservoir; Water (Switzerland), 9(6), art. no. 387.

18. He J., N. Tofighi, M. Yildiz, J. Lei, A. Suleman A.; 2017; A coupled WC-TL SPH method for

simulation of hydroelastic problems; International Journal of Computational Fluid Dynamics,

31(3):174-187.

19. ICOLD - Committee on Computational Aspects of Analysis and Design of Dams; 2013; 12th

Benchmark Workshop on Numerical Analysis of Dams (Graz, Austria), Theme C.

20. Khakzad N., P. Van Gelder; 2017; Fragility assessment of chemical storage tanks subject to

floods; Process Safety and Environmental Protection, 111:75-84.

34
21. Kirkpatrick, W.M.; 1965; Effects of Grain Size and Grading on the Shearing Behaviour of

Granular Materials; Proceedings of the Sixth International Conference on Soil Mechanics and

Foundation Engineering, 1:273-277.

22. Kocak, B.; 2017; Zur numerischen Modellierung von hydraulisch-mechanisch gekoppelten

Prozessen in gesättigten granularen Böden mittels Smoothed Particle Hydrodynamics; PhD thesis,

Hamburg University of Technology, Serial of the Institute of Geotechnical Engineering and

Construction Managementm Nr. 39.

23. Leonardi M., T. Rung; 2013; SPH Modelling of Bed Erosion for Water/Soil-Interaction; 8th

SPHERIC ERCOFTAC International Workshop, 139-144, Trondheim (Norway).

24. Liu M.B., G.R. Liu; 2006; Restoring particle consistency in smoothed particle hydrodynamics;

Applied Numerical Mathematics, 56:19–36.

25. Liu M.B., G.R. Liu; 2010; Smoothed Particle Hydrodynamics (SPH): an Overview and Recent

Developments; Arch Comput Methods Eng, 17: 25–76.

26. Manenti S., S. Sibilla, M. Gallati, G. Agate, R. Guandalini; 2012; SPH Simulation of Sediment

Flushing Induced by a Rapid Water Flow; Journal of Hydraulic Engineering ASCE 138(3): 227-

311.

27. Monaghan J. J.; 1992; Smoothed Particle Hydrodynamics; Annu. Rev. Astron. Astrophys, 30,

543-74.

28. Monaghan J.J.; 2005; Smoothed particle hydrodynamics; Rep. Prog. Phys. 68, 1703–1759.

29. Niemunis, A., Herle I.; 1997; Hypoplastic model for cohesionless soils with elastic strain range,

Mechanics of Cohesive-Frictional Materials, 2 (4), pp. 279-299.

30. Pontillo M.; 2010; Trasporto ed “entrainment” di sedimenti in alvei mobili; PhD thesis,

Università degli studi di Napoli Federico II.

31. Price, D.J.; 2012; Smoothed Particle Hydrodynamics and Magnetohydrodynamics; J. Comp.

Phys., 231-3, 759-794.

35
32. Ren J., J. Ouyang, B. Yang, T. Jiang, H. Mai; 2011; Simulation of container filling process with

two inlets by improved smoothed particle hydrodynamics (SPH) method; International Journal of

Computational Fluid Dynamics, 25(7):365-386.

33. Schaeffer D.G.; 1987; Instability in the Evolution Equations Describing Incompressible

Granular Flow; Jou. of Differential Equations, 66:19-50.

34. Shadloo M.S., G. Oger, D. Le Touzé; 2016; Smoothed particle hydrodynamics method for fluid

flows, towards industrial applications: Motivations, Current state, And challenges; Computers and

Fluids, 136:11-34.

35. Sibilla S.; 2015; An algorithm to improve consistency in Smoothed Particle Hydrodynamics;

Computer and Fluids, 118:148-158.

36. Soares-Frazao S., Y. Zech; 2011; HLLC scheme with novel wave-speed estimators appropriate

for two-dimensional shallow-water flow on erodible bed; Int. J. Numer. Meth. Fluids, 66:1019–

1036.

37. SPHERA v.8.0 (RSE SpA), https://github.com/AndreaAmicarelliRSE/SPHERA

38. Spinewine B.; 2005; Two-layer flow behavior and the effects of granular dilatancy in dam-

break induced sheet-flow; PhD Thesis, Faculté des sciences appliquées, Université catholique de

Louvain.

39. Stefanova, B., J. Grabe; 2014; SPH model of water jet erosion in granular soils with a boundary

layer of liquefied soil; International Symposium on Geomechanics Micro to Macro, September 1-3,

2014, Cambridge, UK.

40. Swartenbroekx C., Y. Zech, S. Soares-Frazão; 2013; Two-dimensional two-layer shallow water

model for dam break flows with significant bed load transport; Int. J. Numer. Meth. Fluids,

73(5):477–508; DOI: 10.1002/fld.3809

41. Ulrich, C.; 2013; Smoothed-Particle-Hydrodynamics simulation of port hydrodynamics

problems; Schirftenreihe Schiffbau der Technischen Universität Hamburg-Harburg, Report Nr. 671.

PhD Thesis. Technische Universität Hamburg-Harburg, Hamburg. Schiffbau.

36
42. Vacondio, R., B.D. Rogers, P. Stansby, P. Mignosa; 2012; SPH Modeling of Shallow Flow with

Open Boundaries for Practical Flood Simulation; J. Hydraul. Eng. 138-6, 530–541.

43. Viccione G., V. Bovolin, E.P. Carratelli; 2008; Defining and optimizing algorithms for

neighbouring particle identification in SPH fluid simulations; International Journal for Numerical

Methods in Fluids, 58(6):625-638.

44. Violeau D.; 2012; Fluid mechanics and the SPH method: theory and applications. Oxford

University Press, Oxford.

45. von Wolffersdorff, P.A.; 1996; A hypoplastic relation for granular materials with a predefined

limit state surface; Mechanics of Cohesive-Frictional Materials, 1(1):251–271.

46. Zhang Y., D. Wan; in press; Numerical study of interactions between waves and free rolling

body by IMPS method; Computers and Fluids.

37

Вам также может понравиться