Documente Academic
Documente Profesional
Documente Cultură
Abstract. Various types of behaviors of different soils have been predicted by us-
ing the finite element method (FEM) with comprehensive constitutive models de-
veloped in geomechanics. There are, however, still some problems for the large
deformation analyses within the framework of FEM. Numerical instabilities arise
due to the distortion of the FE mesh. In this work, deformation analyses of geoma-
terials using Smoothing Particle Hydrodynamics (SPH) method are carried out.
The SPH method belongs to the class of particle methods. In this paper, the ana-
lytical accuracy and the stability of SPH method are investigated for deformation
analyses of geomaterials which are assumed to be solid or fluid.
1 Introduction
The modeling of large deformations in geostructures within the framework of
FEM remains to be a major challenge, although there have been numerous com-
prehensive constitutive models developed in geomechanics. A common pathology
is the occurrence of numerical instabilities due to the distortion of the FE mesh.
On the other hand, some numerical methods have been proposed to solve large
deformation problems without FE mesh. The Eulerian method is one of the solu-
tions for large deformation problems because it is not necessary to take the defor-
mation of mesh into consideration. Simulations of large deformation problems of
geomaterials, for example, the lateral flow of liquefied ground [1, 2] and the large
deformation of slope failure [3] have been carried out using numerical schemes
based on the Eulerian method. Numerical results obtained in the previous studies
were in good agreement with theoretical solutions and experimental results. In the
previous studies, the deformation behavior of geomaterials was expressed under
the assumption that the geomaterials are a single-phase Bingham fluid with shear
strength of soils.
Smoothed Particle Hydrodynamics (SPH) [4, 5], a kind of particle method, is
also an effective method to solve large deformation problems because it does not
require a structured mesh system. Recently, SPH method has been widely used in
a variety of fields such as fluid dynamics [6] or solid mechanics [7]. The method
has also been applied to geotechnical engineering [8]. The objectives of this work
are to establish a computer program code for analyzing the deformation of geoma-
terials using SPH method. In order to use the nonstructural calculation points in
the SPH method, it is possible to express the complex configuration without
R. Wan et al. (Eds.): Bifurcations, Instabilities and Degradations in Geomaterials, SSGG, pp. 275–290.
springerlink.com © Springer-Verlag Berlin Heidelberg 2011
276 H. Nonoyama et al.
2 Numerical Method
f ( xα ) = ∫ f ( x β ) W (r , h)dx β (1)
Ω
where r = |xα-xβ|, h is a radius of the influence domain and Ω is the volume of the
integral that contains xα and xβ. Thus, the spatial derivative of Eq. (1) can be writ-
ten as
∂ f ( xα ) ∂ ∂W ( r , h )
∫ f ( x )W ( r, h ) dx = ∫ f ( x )
β β β
= dx β (2)
∂xi ∂xi Ω Ω ∂xi
The second step considers a discrete distribution of particles for which Eqs. (1)
and (2) are approximated by replacing the integral with the summation operation,
i.e.
mβ
f ( xα ) = ∑ β f ( x β )W αβ
N
(3)
β ρ
∂ f ( xα ) mβ ∂W αβ
f ( xβ )
N
=∑ (4)
∂xi β ρ β
∂xi
where mβ is the mass of a neighbor particle, ρ β its density, and N is the number of
particles in the influence domain.
Performance of the SPH Method for Deformation Analyses of Geomaterials 277
Many kinds of smoothing functions have been proposed in the literature. In this
work, the cubic B-spline function [9] is employed.
∂ρ ∂ (ρu i )
+ =0 (5)
∂t ∂xi
Dui 1 ∂σ ij
= fi + (6)
Dt ρ ∂x j
where ρ is the density, u is the velocity, σ is the stress and f is the external force.
The indices i and j denote the coordinate directions.
When applying the SPH interpolation theory to the gradients in Eq. (5), the
SPH equation of continuity at particle α is expressed as
dρ ∂W αβ
= ∑ m β ( uiα − uiβ )
N
(7)
dt β ∂xi
In place of Eq. (7), the density can also be directly determined from Eq. (3), i.e.
N
ρ α = ∑ m β W αβ (8)
β
When Eq. (8) is used, the density tends to be underestimated since there are a few
numbers of fixed particles in the vicinity of the free surface. This problem can be
avoided by using Eq. (7), the differential form. Alternatively, the next equation
can be used, in which the smoothing functions are summed for normalization [10]
and which allows for a solution even if the form is that of Eq. (8). Thus,
N N
⎛ mβ ⎞
ρ α = ∑ m β W αβ ∑β ⎜ ρ β ⎟W αβ (9)
β ⎝ ⎠
Applying the SPH interpolation theory to the gradients into the general equation of
motion, we get for particle α:
du i N ⎛ σ ijβ σ ijα ⎞ ∂W αβ
= ∑ mβ ⎜ + + δ Π ⎟ + fi (10)
dt β ⎝ ( ) ( )
⎜ ρβ 2 ρ α 2 ij ij
⎟ ∂x j
⎠
278 H. Nonoyama et al.
Π ij =
1
ρ ij
(− α vis
( ))
c φ αβ + β vis φ αβ
2
(11)
where α vis and β vis are the artificial viscosity parameters, c is the average of the
sound speed of each particles, ρ is the average of the density of each particle, uαβ
is the relative velocity and φ αβ is described as follows:
huiαβ ⋅ xiαβ
φ αβ = (12)
αβ 2
xi + 0.01h 2
σ ij = Dijkl
e
ε kle (13)
⎛ ∂f ∂f ⎞
⎜ ⎟
e
Dijmn D epqkl
∂σ ∂σ
dσ ij = ⎜⎜ Dijkl ⎟dε (14)
−
e mn pq
∂f ∂f ∂f ∂L ∂f ⎟ kl
⎜⎜
e
Dmnpq − ⎟⎟
⎝ ∂σ mn ∂σ pq ∂L ∂ε mn p
∂σ mn ⎠
where L is the hardening parameter and f is the yield function. In this study, the
yield function is based on the Cam-clay model [12] as follows:
λ −κ ⎛ pc 1 q ⎞
f = + ⎟ − εv = 0
p
⎜ ln (15)
1 + e0 ⎝ p0 M p ⎠
where λ is the compression index, κ is the swelling index, e0 is the initial void ra-
tio, pc is the consolidation yield stress, p0 is the initial mean stress, Μ is the stress
ratio at failure, q is the deviatoric stress, p is the mean principal stress and ε vp is
the plastic volumetric strain.
Performance of the SPH Method for Deformation Analyses of Geomaterials 279
where σ ij is the Cauchy stress rate tensor, σˆ ij is the Jaumann stress rate and ω ij is
the spin tensor.
Where τ is the shear stress, η0 is the viscosity after yield, γ is the shear strain
rate, c is the cohesion of soil, p is the hydraulic pressure and φ is the internal fric-
tion angle of soil. Because Eq. (17) cannot be directly calculated, we use an
equivalent viscosity η’ obtained for a Newtonian fluid as follows:
τ c + p tan φ
η′ = = η0 + (18)
γ γ
Figure 1 shows the equivalent viscosity of the Bingham fluid model. The value of
the equivalent viscosity is dependently on the shear strain rate. The constitutive
model used in this study can be obtained by introducing the equivalent viscosity
into the constitutive model of the Newtonian fluid.
τ
shear stress
yield
η0 : viscosity after yield
shear τy
stress η’ : equivalent viscosity
γ
.
shear strain rate
Dρ
=0 (19)
Dt
Dui 1 ∂p η ∂ ⎛ ∂ui ⎞
=− + ⎜ ⎟⎟ + fi (20)
Dt ρ ∂xi ρ ∂x j ⎜⎝ ∂x j ⎠
Dui 1 ∂p 1 ∂ ⎡ ⎛ ∂ui ∂u j ⎞⎤
=− + ⎢η ′ ⎜ + ⎟⎟ ⎥ + f i (21)
Dt ρ ∂xi ρ ∂x j ⎢⎣ ⎜⎝ ∂x j ∂xi ⎠ ⎥⎦
For the case of a Newtonian fluid, the viscosity coefficient is treated as a constant
value and its spatial derivatives are not considered. However, as we can see from
Eq. (21), the equivalent viscosity has a distribution in space. Therefore, the effect
of the spatial derivatives is taken into account in Eq. (21). Moreover, Eq. (21) can
be discretized as follows:
ui** − ui k
= fi k (22)
Δt
ui k +1 − ui* 1 ∂p k +1
=− (24)
Δt ρ ∂xi
Performance of the SPH Method for Deformation Analyses of Geomaterials 281
where Δt is the time increment, subscript k and k+1 indicate the quantities at each
calculation time step and * indicates the temporal quality. Equations (22), (23) and
(24) are the external force term, the viscous term and the pressure term of the equ-
ation of motion, respectively.
The following is an algorithm used in this study. A temporal value of the veloc-
ity ui** is obtained explicitly using the gravity fik and the velocity uik at the previ-
ous time from the Eq. (22):
A temporal value of the velocity ui* is obtained using the temporal value ui** and
the spatial derivative of ui* as
Δt ⎡ ∂ ⎛ ∂u i∗ ⎞ ∂
⎜η ′ ⎟+
⎛ ∂u ∗j ⎞ ⎤
⎜η ′ ⎟ (26)
u i∗ = u i∗∗ + ⎢
ρ ⎣⎢ ∂x j ⎜ ∂x ⎟ ∂x ⎜ ∂x ⎟ ⎥⎥
⎝ j ⎠ j ⎝ i ⎠⎦
According to previous studies [15], the right hand side of the temporal value ui* of
the spatial derivative in Eq. (26) can be discretized as follows:
∂W αβ
∂ ⎛ ∂u ∗ ⎞ β
4m η ′ η ′
N α β (u i
α ,∗
− uiβ ,∗ )( xiα − xiβ ) ⋅
∂xi
⎜⎜η ′ i
⎟⎟ = ∑ β α β
(27)
∂x j ⎝ ∂x j ⎠ β ρ η ′ +η ′
2
xiαβ
The temporal value of the position xi* is obtained by the temporal value ui* ob-
tained from Eq. (26), i.e.
The continuity equation requires that the density of fluid be constant. This is
equivalent to the particle number density being constant, n0. When the temporal
value of a particle number density n* is not n0, it is corrected to n0 as
n ∗ + n′ = n 0 (29)
1 ∂p k +1 (30)
ui′ = Δt
ρ ∂xi
282 H. Nonoyama et al.
There is a relationship between the correction value of the velocity ui’ and the cor-
rection value of the particle number density n’ from the equation of continuity, i.e.
1 n ′ ∂u i′
+ =0 (31)
n 0 Δt ∂x i
The Poisson equation is obtained from Eqs. (29), (30) and (31):
∂ ⎛ 1 ∂p k +1 ⎞ 1 n∗ − n 0
⎜⎜ ⎟⎟ = − 2 (32)
∂xi ⎝ ρ ∂xi ⎠ Δt n0
According to previous studies [15], the left hand side of Eq. (32) can be discre-
tized as follows:
∂W αβ
∂ ⎛ 1 ∂p k +1 ⎞ N 4m β ⎛ 1 ⎞
(p α
)(
− p β xiα − xiβ ⋅ ) ∂xi
⎜⎜ ⎟⎟ = ∑ β ⎜⎜ α ⎟
β ⎟
(33)
∂xi ⎝ ρ ∂xi ⎠ β ρ ⎝ρ +ρ ⎠
2
xiαβ
By solving Eq. (32), a pressure pk+1 at present time k+1 is obtained. In order to
prevent numerical instabilities due to a negative pressure, the negative value of the
pressure pk+1 set to zero. Using the pressure pk+1 from Eq. (30), the correction val-
ue of the velocity ui’ is obtained. Moreover, using the correction value of velocity
ui’, a velocity uik+1 and a position xik+1 at present time k+1 are obtained as
4 Solid Analysis
In order to validate the program for a solid, a simple shear test of elastic material
with Jaumann stress rate is carried out. Figure 2 shows the numerical model used
in this simulation. Here, the specimen is represented by a square object
(10cm×10cm) with a virtual area set around it. The number of particles including
the particles comprising the virtual area is 900. The initial interparticle distance is
1.0 cm and the radius of influence domain is 3.0 cm. The density is 1.5 g/cm3.
The time increment used in the calculations is 0.001s. The artificial viscosity
parameter is 10.0.
Performance of the SPH Method for Deformation Analyses of Geomaterials 283
L L L vx
L=10cm
L y γxy
L
x center axis
L
judgment of stress
Table 1 summarizes the numerical parameters used for this analysis. In the
simulation, the virtual area moves with constant displacement to describe simple
shear condition. The velocities of virtual area vx are calculated according to
vx = 0.10 y cm / s (36)
where G is the shear modulus, E is the Young's modulus and ν is the Poisson's ra-
tio. By using the concept of the Cauchy stress, the theoretical solution is given by:
E
τ xy = G sin γ xy = sin γ xy
2(1 + ν )
(38)
Figure 3 shows the relationship between shear stress and shear strain. The
obtained numerical results together with the theoretical solution are compared in
Fig. 3. It is confirmed that numerical result and theoretical solution are in good
agreement.
8
Shear stress τxy [Pa]
0
0 0.8 1.6 2.4 3.2
Shear strain γxy
Fig. 3. Relationship between shear stress and shear strain at the center of specimen
L=10cm vx
L
L
y
Table 2 summarizes the material parameters used for both FE and SPH analy-
ses. Two different values of the initial mean stress are used in this simulation.
Case 1 2
compression index λ 0.355
swelling index κ 0.0477
initial void ratio e0 2.0
stress ratio at failure Μ 1.45
consolidation yield stress pc[kPa] 98.0
initial mean stress p0[kPa] 68.6 98.0
Poisson’s ratio ν 0.33
Figure 5 shows stress paths of Cam-clay model at the center of specimen. Fig-
ure 6 gives the relationship between shear stress and shear strain. Numerical re-
sults from both SPH and FE analyses are described together in Figs. 5 and 6. The
critical state line (C.S.L.) obtained from the Cam-clay model is also plotted in Fig.
5. It is found that results from the SPH analysis are in good agreement with those
obtained from the FE analysis. Therefore, we conclude that the SPH model can
express quite well the behavior of soil.
150
Case1(FEM)
Case1(SPH)
Case2(FEM)
Case2(SPH)
100 Critical state line
q [kPa]
50
0
0 50 100 150
p [kPa]
40
35
Fig. 6. Relationship between shear stress and shear strain at the center of specimen
5 Fluid Analysis
EXP(Martin&Moyce(1952),1.125in)
EXP(Martin&Moyce(1952),2.25in)
EXP(Koshizuka et al.(1995))
SPH
4
L
3.5 2L
3
Xfront/L xfront
2.5
1.5
1
0 0.5 1 1.5 2 2.5 3 3.5
11/2
/2
t(2g/L)
0.2s
0.4s
0.6s
qu = (2 + π )c ≅ 5.14c (39)
where qu is the ultimate bearing capacity, and c is the cohesion of soil. The theo-
retical solution in Eq. (39) has been verified numerically in the literature; see e.g.
[19]. However, Eq. (39) involves some following assumptions:
• Pure cohesive material (c>0, φ 0)
• No friction between rigid body and ground
• Weightless material
Figure 10 shows the numerical model used in this simulation. A footing is placed
on the ground and is assumed to be a rigid body in this simulation. The footing
moves downward at a constant velocity. The constant value of 1.0×10-5 m/s is set
in this simulation. In order to prevent the penetration, the constant velocity is set
to a small value. The number of particles including the particles comprising the
wall is 5,146. The initial interparticle distance is 0.01 m and the radius of influ-
ence domain is 0.02 m. Table 3 summarizes the material parameters. Three differ-
ent values of cohesion are used in this simulation. No gravitational force is
included in the simulation. Moreover the boundary condition between the bottom
of the footing and the ground surface is described as a non-slip boundary. The ver-
tical stresses in the soils below the footing are calculated and the bearing capaci-
ties are determined with the average value of the vertical stresses in the particles
below the footing.
Case 1 2 3
viscosity after yield η0 [Pa·s] 1.0
cohesion c[Pa] 1.0 2.0 3.0
internal friction angle φ [deg] 0.0
Figure 11 shows the relationship between cohesion and bearing capacity at the
different cohesion. The obtained numerical results and the theoretical solution are
shown in Fig. 11. The numerical simulations give a bearing capacity value quite
close to the theoretical solution.
wall
footing(rigid body) 0.10 0.03
0.04
×
velocity 1.0 10-5m/s
0.45
cohesive ground
1.00 Unit:m
20
Prandtl
SPH
10
0
0 1 2 3 4
Cohesion [Pa]
6 Conclusions
In this study, the application of SPH method for deformation analyses of geomate-
rials is discussed. From the solid analysis, it is confirmed that the numerical re-
sults for simple shear test using an elastic model and an elasto-plastic model are in
good agreement with the theoretical solution as well as results of FE analysis. The
fluid analysis of a Dam-break (breach) problem involving a Newtonian fluid also
confirmed a good agreement between numerical results and existing experimental
data. Geomaterials are assumed to be a fluid by introducing a Bingham model
with shear strength of soils. It is confirmed that the bearing capacity of cohesive
ground is given as a factor of the cohesion of soil which is obtained within a cer-
tain level of accuracy when compared to Prandtl’s solution.
In the future, in order to take the influence of water into consideration, it is nec-
essary to introduce a scheme for soil-water coupled analysis.
References
[1] Uzuoka, R.: Analytical study on the mechanical behavior and prediction of soil lique-
faction and flow, Ph. D. Dissertation, Gifu University (2000) (in Japanese)
[2] Hadush, S.: Fluid dynamics based large deformation analysis in geomechanics with
emphasis in liquefaction induced lateral spread, Ph. D. Dissertation, Gifu University
(2002)
[3] Moriguchi, S.: CIP-based numerical analysis for large deformation of geomaterials,
Ph. D. Dissertation, Gifu University (2005)
[4] Lucy, L.B.: A numerical approach to the testing of the fission hypothesis. Astron.
J. 82, 1023–1024 (1977)
[5] Gingold, R.A., Monaghan, J.J.: Smoothed particle hydrodynamics: theory and appli-
cation to non-spherical stars. Monthly Notices of the Royal Astronomical Society 181,
375–389 (1977)
290 H. Nonoyama et al.
[6] Liu, G.R., Liu, M.B.: Smoothed Particle Hydrodynamics: A Meshfree Particle
Method, p. 449. World Scientific, Singapore (2003)
[7] Libersky, L.D., et al.: High Strain Lagrangian Hydrodynamics. J. Comput. Phys. 109,
67–75 (1993)
[8] Maeda, K., et al.: Development of seepage failure analysis method of ground with
smoothed particle hydrodynamics. Journal of structural and earthquake engineering,
JSCE 23(2), 307–319 (2006)
[9] Monaghan, J.J., Lattanzio, J.C.: A refined particle method for astrophysical problems.
Astronomy and Astrophysics 149, 135–143 (1985)
[10] Randles, P.W., Libersky, L.D.: Smoothed Particle Hydrodynamics: Some recent im-
provements and applications. Comput. Methods in Appl. Mech. Eng. 138, 375–408
(1996)
[11] Monaghan, J.J., Gingold, R.A.: Shock Simulation by the Particle Method SPH. J.
Comput. Phys. 52, 374–389 (1983)
[12] Schofield, A.N., Wroth, C.P.: Critical State Soil Mechanics, p. 310. Mcgraw-Hill,
New York (1968)
[13] Amsdam, A.A., Harlow, F.H.: The SMAC Method: A Numerical Technique for Cal-
culating Incompressible Fluid Flow. LA-4370. Los Alamos Scientic Laboratory
(1970)
[14] Sakai, Y., et al.: Incompressible viscous flow analysis by SPH. Journal of the Japan
Society of Mechanical Engineers. Series B 70(696), 1949–1956 (2004) (in Japanese)
[15] Cleary, W., Monaghan, J.J.: Conduction modeling using smoothed particle hydrody-
namics. J. Comput. Phys. 148, 227–264 (1999)
[16] Martin, J.C., Moyce, W.J.: An Experimental Study of the Collapse of Liquid Columns
on a Rigid Horizontal Plane. Philosophical Transactions of the Royal Society of Lon-
don, Ser. A 244, 312–324 (1952)
[17] Koshizuka, S., et al.: Particle Method for Incompressible Viscous Flow with Fluid
Fragmentation. Computational Fluid Dynamics Journal 4(1), 29–46 (1995)
[18] Prandtl, L.: Über die Härte plastischer Körper: Nachrichten von der Königlichen Ge-
sellschaft der Wissenschaften zu Göttingen. Math. Phys. KI. 12, 74–85 (1920)
[19] Tamura, T., et al.: Limit analysis of soil structure by rigid plastic finite element
method. Soils and Foundations 24(1), 34–42 (1984)