Copyright © 2008 Tech Science Press CMES, vol.36, no.2, pp.

97-117, 2008

Large Deformation Analysis with Galerkin based

Smoothed Particle Hydrodynamics

S. Wong and Y. Shie

Abstract: In this paper, we propose a Galerkin-based smoothed particle hydro-

dynamics (SPH) formulation with moving least-squares meshless approximation,
applied to solid mechanics and large deformation. Our method is truly meshless
and based on Lagrangian kernel formulation and stabilized nodal integration. The
performance of the methodology proposed is tested through various simulations,
demonstrating the attractive ability of particle methods to handle severe distortions
and complex phenomena.

Keyword: SPH, Galerkin method, large deformation

1 Introduction

The problems of computational mechanics grow ever more challenging. For ex-
ample, in the simulation of manufacturing processes such as extrusion and mold-
ing, it is necessary to deal with extremely large deformations of the mesh while
in computations of castings the propagation of interfaces between solids and liq-
uids is crucial Onate, Idelson, Zienkievicz, and Taylor (1996); Idelsohn, Onate, and
Pin (2004); Bonet and Kulasegaram (2000); Feldman and Bonet (2007); Khayyer
and H. Gotoh (2008); Fang, Parriaux, Rentschler, and Ancey (in press); Antoci,
Gallati, and Sibilla (2007); Wang, Lu, Hao, and Chong (2005). In simulations
of failure processes, we need to model the propagation of cracks with arbitrary
and complex paths. In the development of advanced materials, methods which
can track the growth of phase boundaries and extensive microcracking are required
Belytschko, Lu, and Gu (1995); Belytschko and Tabbara (1996); Organ, Fleming,
Terry, and Belytschko (1996); Fleming, Chu, Moran, and Belytschko (1997); Hao,
Liu, and Chang (2000); Liu, Hao, and Belytschko (1999); Hao and Liu (2006);
Rabczuk and Belytschko (2006); Rabczuk and Zi (2007); Rabczuk, Areias, and
Belytschko (2007b); Zi, Rabczuk, and Wall (2007); Ventura, Xu, and Belytschko
(2002); Rabczuk and Belytschko (2004); Hao, Liu, Klein, and Rosakis (2004);
Chandra and Shet (2004); Maiti and Geubelle (2004); Rabczuk, Areias, and Be-
98 Copyright © 2008 Tech Science Press CMES, vol.36, no.2, pp.97-117, 2008

lytschko (2007a); Rabczuk and Belytschko (2007); Nishioka, Kobayashi, and Fuji-
moto (2007); Hagihara, Tsunori, and Ikeda (2007); Guz, Menshykov, and Zozulya
(2007); Guo and Nairn (2006); Rabczuk and Areias (2006); Gao, Liu, and Liu
(2006); Fujimoto and Nishioka (2006); Nishioka (2005); Chen, Gan, and Chen
(2008); Xu, Dong, and Zhang (2008); Wen, Aliabadi, and Lin (2008); Zhang and
Chen (2008). These problems are not well suited to conventional computational
methods such as finite element, finite volume or finite difference methods. The un-
derlying structure of these methods which originates from their reliance on a mesh
is not well suited to the treatment of discontinuities which do not coincide with the
original mesh lines. Thus, the most viable strategy for dealing with moving discon-
tinuities in methods based on meshes is to remesh in each step of the evolution so
that mesh lines remain coincident with the discontinuities throughout the evolution
of the problem. This can, of course, introduce numerous difficulties such as the
need to project between meshes in successive stages of the problem, which lead to
degradation of accuracy and complexity in the computer program, not to mention
the burden associated with a large number of remeshings.
The objective of meshless methods is to eliminate at least part of this structure by
constructing the approximation entirely in terms of nodes Atluri and Zhu (1998,
2000); Atluri and Shen (2002); Atluri (2002); Han and Atluri (2003); Tang, Shen,
and Atluri (2003); Liu, Han, Rajendran, and Atluri (2006); Liu, Jun, and Zhang
(1995a); Duarte and Oden (1996); Melenk and Babuska (1996); Belytschko, Lu,
and Gu (1994a, 1995); Rabczuk and Belytschko (2005); Idelsohn and Onate (2006).
Although in many meshless methods, recourse must be taken to meshes in at least
parts of the method, moving discontinuities can usually be treated without remesh-
ing with minor costs in accuracy degradation. Thus, it becomes possible to solve
large classes of problems which are very awkward with mesh-based methods.
Since the early 1990s, research focus was devoted to meshless methods. The oldest
meshless method is the Smoothed Particle Hydrodynamics (SPH) method devel-
oped in the late 1970s Lucy (1977). Improvements of this method flourished es-
pecially in the early 1990s Swegle, Hicks, and Attaway (1995); Monaghan (1994);
Vignjevic, Campbell, and Libersky (2000); Liu, Jun, and Zhang (1995b); Bonet
and Kulasegaram (2000); Krongauz and Belytschko (1997); Dilts (2000); Randles
and Libersky (1997, 2000). SPH is method based on the strong form. It employs
kernel function. Though many improvements were made, SPH and even many
corrected SPH versions suffer from instabilities. Belytschko, Krongauz, Organ,
Fleming, and Krysl (1996) showed that strong form based SPH can be regarded as
weak formulation SPH using nodal integration and they showed instability due to
under-integration. To overcome instability, stress-points were added to nodes Dyka
and Ingel (1995). The stress points have to be relocated when the body undergoes
Large Deformation Analysis 99

large deformations.
Another source of instability is connected to the kernel approximations. Eulerian
kernel functions are commonly used in SPH formulations since typical application
of SPH is for large deformations where Finite Element Method fails. However, Eu-
lerian kernel functions often result in instabilities as demonstrated by Belytschko,
Guo, Liu, and Xiao (2000). Lagrangian kernel functions can remove instabilities
but limit the application of the method to moderate deformations Rabczuk, Be-
lytschko, and Xiao (2004).
In this paper, we employ a SPH method to model problems with large deforma-
tions and damage in solid mechanics. Therefore, we interpret SPH as Galerkin
method and use a stabilized nodal integration that removes instabilities. We do
not need stress points that needs to be relocated when the body deforms. Updated
Lagrangian kernel functions are used to eliminate other instability. Therefore, La-
grangian kernel functions are consequently updated. Hence, the method can handle
large deformations. The method is applied to problems with severe distortions.
The paper is outlined as follows: We first state the governing equation. Then, we
develop the SPH formulation and apply the method to some problems. At the end,
we conclude our paper.

2 Governing Equations

In finite deformation analysis two possible co-ordinate systems can be chosen to de-
scribe the continuum under consideration [17, 18]. Lagrangian or material descrip-
tion is based on a certain reference configuration, that usually coincides with the
initial configuration. Then, all relevant quantities are referred to a initial problem
domain, Ω0 . The Eulerian or spatial description is based on the current continuum
configuration. Here, relevant quantities are referred to the current problem domain,
Ω. The former is most frequent in solid mechanics, whereas the latter is typical
in fluid mechanics. These two descriptions will lead, in general, to non-equivalent
discretizations in particle methods.
The governing equations solved in SPH method are the continuity equation, mo-
mentum equation and energy equation. Most SPH codes use an Eulerian rate form
for mass conservation
dρ
= −ρ div (v) (1)
dt

where d(·)/dt denotes the material time derivative and div(v) is computed in the
current configuration. For a Lagrangian description, there is no need to use the Eu-
lerian rate form, equation (1) and conservation of mass can be written in a material
100 Copyright © 2008 Tech Science Press CMES, vol.36, no.2, pp.97-117, 2008

form as an algebraic equation:

ρ J = ρ0 (2)
where ρ0 and ρ are, respectively, the initial and current densities and J is the de-
terminant of the deformation gradient, J = det(F), F = dXdx
= I − dX
du
where I is the
second order unity tensor and u denotes the displacement field. In the following, X,
x = x(X), ∇X and ∇x denote co-ordinates and gradient operators in the reference
and current configurations, respectively.
In a Lagrangian description, conservation of linear momentum can be written as

∇X · P + b = ρ0 ü, X ∈ Ω0 (3)

where b is the body force per unit volume, superimposed dots denote material time
derivatives and P is the first Piola-Kirchhoff stress tensor. Its Eulerian counterpart
is

∇x · σ + b = ρ v̇, x ∈ Ω (4)

where σ is the Cauchy stress tensor and ρ is the current density. The momentum
equation can also be written in an arbitrary Lagrangian-Eulerian (ALE) form

∇x · σ + b = ρ (v̇ + v∗ ∇x v) , x ∈ Ω (5)

where v∗ is the convective velocity. SPH-like particle methods follow the move-
ment of a set of particles, so v∗ = 0 and the convective term in equation (5) vanishes.
This is considered a Lagrangian description of the movement. When we say that
(4) is posed in Eulerian form we mean that relevant quantities are referred to the
current configuration, although the description is Lagrangian in the aforementioned
sense.
Conservation of energy may also be considered in processes involving heat trans-
fer or other related phenomena but we limit our studies to problems governed by
equations (2) to (5).
Above field equations are complemented with displacement and traction boundary
conditions that are given in the reference configuration:

u = ū, X ∈ Γ0u (6)

n0 · P = t0 = t̄0 , X ∈ Γ0t (7)

or in the current configuration

v = v̄, x ∈ Γv (8)
n·σ = t̄, x ∈ Γt (9)
Large Deformation Analysis 101

where the index t refers to traction boundaries, the index u refers to displacement
boundaries and index v to imposed velocities; n is the normal to the traction bound-
ary and the subscript 0 refers to quantities in the reference configuration.

3 SPH Method
The first SPH test and trial functions were given by

NJ (x) = VJWJ (x) (10)

where WJ (x) = W (x − xJ , h) is a kernel function with compact support centred

at particle J and VJ is the tributary or statistical volume associated to particle J.
The parameter h, usually called smoothing length or dilation parameter in the SPH
literature is a certain characteristic measure of the size of the support of WJ (x) (e.g.
the radius in circular supports). Exponential and spline functions are most frequent
kernels. We have not found a general criterion for an optimal choice and used the
quartic spline function that is commonly used in the literature:

1 − 6s2 + 8s3 − 3s4 s ≤ 1
W (x − xI , h) = w(s) = (11)
0 s>1

with s = x−xI
2h for circular support size.
The SPH approximation uh (x) of a given function u(x) can be posed in terms of
the shape functions, equation (10) and certain particle or nodal parameters uJ as
n
uh (x) = ∑ NJ (x)uJ (12)
J=1

with n numbers of neighboring particles with NJ (x) = 0. Using standard kernels,

the approximation given by equation (12) is poor near boundaries, and lacks even
zeroth-order completeness
n
∑ NJ (x) = 1 (13)
J=1

Thus, the set {NJ (x), J = 1, ..., n} does not constitute a signed partition of unity Me-
lenk and Babuska (1996); Duarte and Oden (1996). To eliminate this remedy, we
adopt a correction of the SPH approximation that is based on Moving Least Squares
(MLS) approximation. The MLS approximation is frequently used in meshfree
methods such as the famous Meshless Local Petrov Galerkin (MLPG) method
Atluri and Zhu (1998), the elementfree Galerkin (EFG)-method Belytschko, Lu,
and Gu (1994b) and Moving Least Squares Particle Hydrodynamics (MLSPH)
102 Copyright © 2008 Tech Science Press CMES, vol.36, no.2, pp.97-117, 2008

Dilts (2000). In fact, it can be shown that the latter two methods are specific cases
of the MLPG method.
The basic idea of the MLS approach is to approximate u(x), at a given point x,
through a polynomial least-squares fitting of u(x) in a neighbourhood of x as
m
uh (x) = ∑ pI (x)aI(z)|z=x = pT (x) a(z)|z=x (14)
I=1

where pT (x) is an m-dimensional polynomial basis and a(z)|z=x is a set of param-

eters to be determined, such that they minimize the following error functional:
  
J (a(z)|z=x) = W (z − y, h)|z=x u(y) − pT (y)a(z)|z=x dΩx (15)
y∈Ωx

being W (z − y, h)|z=x a symmetric kernel with compact support (denoted by Ωx ),

frequently chosen among the kernels used in standard SPH. The stationary condi-
tions of J with respect to a lead to

p(y)W (z − y, h)|z=xu(y) dΩx = M(x)a(z)|z=x (16)
y∈Ωx

with moment matrix


M(x) = p(y)W(z − y, h)|z=xpT (y) dΩx (17)
y∈Ωx

In numerical computations, the global domain Ω is discretized by a set of n par-

ticles. We can then evaluate the integrals in equations (16) and (17) using those
particles inside Ωx as quadrature points (nodal integration) to obtain, after rear-
ranging,

a(z)|z=x = M−1 (x)PΩx W(x)uΩx (18)

where the vector uΩx contains certain nodal parameters of those particles in Ωx, the
discrete version of M is M(x) = PΩx W(x)PTΩx , and matrices PΩx and W(x) can be
obtained as

PΩx = {p(x1 ), p(x1 ), ..., p(xn)} (19)

W(x) = diag {WI (x − xI , h)VI }, I = 1, ..., n (20)

where n denotes the total number of particles within the neighborhood of point x
and VI and xI are, respectively, the tributary volume (used as quadrature weight)
and co-ordinates associated to particle I.
Large Deformation Analysis 103

Introducing equation (18) in equation (12) the interpolation structure can be iden-
tified as

uh (x) = pT (x)M−1 (x)PΩx W(x)uΩx (21)

Henceforth, the MLS shape functions are

NT (x) = pT (x)M−1 (x)PΩx W(x) (22)

Instead of the global polynomial basis p(y), we use a scaled locally defined poly-
nomial basis p ((y − x)/h), that leads to better conditioned moment matrix M. In
this work, we use a linear polynomial basis

p ((y − x)/h) = {1, (y1 − x1 )/h, (y2 − x2 )/h} (23)

where (x1 , x2 ) and (y1 , y2 ) are, respectively, the Cartesian co-ordinates of x and
y. This guarantees linear completeness and constitutes a signed partition of unity
method.
Finally, we remark that all derivations above are done equivalently for a Lagrangian
description.

4 Discretization
The meshless discrete equations can be derived using a weighted residuals formu-
lation. In the MLPG method, different test and trial functions are employed and
a local weak form over local sub-domains is used. It is easy to show that due to
the choice of local weak form, MLPG leads to truly meshless method Atluri and
Zhu (1998), i.e. no background cells for integration are necessary as opposed to the
EFG method that is based on the global weak form and therefore does not result in
a truly meshless method since background mesh is needed to perform integration.
Therefore, it is commonly referred to as semi-meshless method.
The discrete counterpart of the Galerkin weak form is almost equivalent to that
obtained from kernel estimates Belytschko, Guo, Liu, and Xiao (2000); Beissel
and Belytschko (1996) such as classical SPH formulations. Furthermore, such an
equivalence indicates that SPH can be studied in the context of Galerkin methods.
We employ a global weak form and use nodal integration that maintains the truly
meshless character since no background mesh is needed for integration. However,
nodal integration leads to instabilities that need to be corrected (see below for more
details). The global weak (integral) form of the spatial momentum equation can be
stated in Lagrangian formulation as: Find u ∈ U and δ u ∈ U0 such that

δ W = δ Wint − δ Wext + δ Wkin = 0 (24)

104 Copyright © 2008 Tech Science Press CMES, vol.36, no.2, pp.97-117, 2008

with

δ Wint = ∇X δ u : P dΩ0
Ω0
 
δ Wext = δ u · t̄0 dΓ0 + δ u · b dΩ0
Γ0t Ω0

δ Wkin = ρ0 δ u · ü dΩ0 (25)
Ω0

with the approximation spaces U and U0 for the trial and test functions, respec-
tively,
 
U = u|u ∈ H 1 , u = ū on Γu
 
U0 = δ u|δ u ∈ H 1 , δ u = 0 on Γu (26)

For Eulerian description, we obtain similarly expressions


δ Wint = ∇x δ v : σ dΩ
Ω 
δ Wext = δ v · t̄ dΓ + δ v · b dΩ
Γt Ω

δ Wkin = ρδ v · v̇ dΩ (27)
Ω

with approximation spaces V and V0 for the trial and test functions, respectively:
 
V = v|v ∈ H 1 , v = v̄ on Γv
 
V0 = δ v|δ v ∈ H 1 , δ v = 0 on Γv (28)

The test and trial functions have the structure of equation (21). Introducing them
into the weak formulation with a Bubnov Galerkin method yields
   
n n
∑ δ uI ∑ − Ω0
∇X NI (X)P dΩ0 +
Ω0
NI (X)b dΩ0 +
Γ0t
NI (X)t̄0 dΓ0
I=1 J=1

+ ρ0 NI (X)NJ (X)u dΩ0 = 0 (29)
Ω0

Thus, for each particle I, the following identity must hold

n   
∑ ∇X NI (X)P dΩ0 = NI (X)b dΩ0 + NI (X)t̄0 dΓ0
J=1 Ω0 Ω0 Γ0t

+ ρ0 NI (X)NJ (X)u dΩ0 = 0 (30)
Ω0
Large Deformation Analysis 105

Figure 1: Voronoi cell diagram

The discrete equations for the Eulerian formulation look similar. These equations
can be recast into a matrix form

MIJ üJ = −fext

I + fI
int
(31)

with

MIJ = ρ NI (X) NTJ (X) dΩ0
Ω0

 
I =
fext NTI (X) t̄0 dΓ0 + NTI (X) bdΩ0 (32)
Γ0t Ω0


I =
fint ∇X NTI X) PdΩ0 (33)
Ω0

Nodal integration has been used, at least implicitly, in all SPH formulations, and
lies, indeed, in the basis of its early formulation. It is obviously the cheapest option
and the resulting scheme is truly meshless. The particles are used as quadrature
points and the corresponding integration weights are their tributary volumes:
n
MIJ = ∑ ρK NI (XK )NJ (XK )VK (34)
K=1

n n
I =
fext ∑ NTI (XK ) b̄K VK + ∑ NTI (XK ) t̄K0 AK (35)
K=1 K=1
106 Copyright © 2008 Tech Science Press CMES, vol.36, no.2, pp.97-117, 2008

n
I =
fint ∑ ∇X NTI XK ) PKVK (36)
K=1

In the above, VK represents the tributary volume associated to particle K. These

volumes are obtained from Voronoi diagrams, figure 1. Hence, the particles are set
up with certain initial densities, volumes and, therefore, masses. These physical
masses MK remain constant during the simulation and densities are field variables
updated through an algebraic equation (2).
Since nodal integration is the origin of well-known instabilities in meshless meth-
ods, we introduce a strain smoothing stabilization that avoids these instabilities
Chen, Wu, Yoon, and You (2001); You, Chen, and Voth (2002). Therefore, the
strain tensor is transformed by Gauss’ divergence theorem

1
ε̃ = (u ⊗ n + n ⊗ u dΓ) (37)
ΓVK VK

where VK is the volume of the Voronoi cell K, figure 1, and n is the normal to the
cell K.
Large deformations are accounted for by computing nodal deformation gradients
F. Then the first Piola-Kirchoff stress P is used in place of the Cauchy stress ten-
sor. For total Lagrangian, X refers to the undeformed reference configuration. For
updated Lagrangian, X refers to some intermediate reference configuration and the
first Piola-Kirchoff stress is computed accordingly. For total Lagrangian, the nodal
volume VK is the volume in the reference configuration X. For updated Lagrangian,
VK is updated by the volumetric strain (i.e. det(FK )VK ) when transitioning to a new
reference state.

5 Results
5.1 Billet
This academic example studies the behavior of our Galerkin based SPH method for
large deformations problems. A rubber billet with side length L = 2.54, Young’s
modulus E = 1 and Poisson’s ratio ν = 0.49 is compressed. We tested structured
and unstructured nodal arrangements and different methods:

• SPH with Eulerian kernel

• SPH with Lagrangian kernel

• SPH with updated Lagrangian kernel

Large Deformation Analysis 107

(a) initial configuration (b) Eulerian kernel at averaged strain of 0.4

(c) Eulerian kernel at averaged strain of 0.5 (d) Eulerian kernel at averaged strain of
0.5

(e) updated Lagrangian kernel at averaged strain (f) updated Lagrangian kernel at averaged strain
of 0.9 of 0.92
Figure 2: Displaced configuration of the billet with different methods

Figure 2 shows deformed configurations of this problems for different methods.

Figures 2b and c show the deformed configuration for the unstructured nodal ar-
rangement at average normal strain of the entire billet of 0.2 and 0.25, respectively,
for Eulerian kernel. Particle clumping occurs in figure 2c. It is well known that
nodal integration based SPH methods perform best when the initial configuration
is structured. Nevertheless, also for structured nodal arrangements, unphysical de-
formations occur for the Eulerian kernel as shown in figure 2d. Results with La-
grangian kernel got instable at averaged strain of the specimen around 0.2. The
deformed configuration with updated Lagrangian kernel formulation is shown in
figure 2e,f at averaged strain of 0.9 and 0.92. The simulation exhibits unphysical
deformation patterns first at averaged strains of 0.92. We also run computations
108 Copyright © 2008 Tech Science Press CMES, vol.36, no.2, pp.97-117, 2008

with updated Lagrangian kernel functions where we omitted stabilization and ob-
tained unphysical deformation patterns similar to the ones obtained by the Eulerian
kernel. Finite Element simulations terminated at about 0.5 average strain.

4500 4500

4000 4000
coarse
fine updated Lagrangian kernel
3500 3500
very fine Lagrangian kernel
Eulerian kernel
3000 3000
Reaction Force

Reaction Force
Finite Element

2500 2500

2000 2000

1500 1500

1000 1000

500 500

0 0
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
strain delta L/L [-] Strain Delta L/L [-]

(a) (b)
Figure 3: Reaction Force-strain curves from different billet simulations

Figure 3 shows the reaction force-strain curve for different methods and refine-
ments. The best results are obtained with stabilized SPH and updated Lagrangian
kernel functions.

5.2 Taylor bar impact

The next example to demonstrate the effectiveness of Galerkin based SPH is Tay-
lor bar impact. The following elastoplastic material properties were used: Young’s
modulus E = 117GPa, Poisson’s ratio ν = 0.35, yield strenght fy = 0.4GPa, lin-
ear hardening modulus ET = 0.1GPa and density ρ = 8930kg/m3. The initial
bar length was 32.4mm, the initial radius was 3.2mm and the initial velocity was
227m/s. Updates were made every 100 time steps. Figure 4 shows the displaced
Taylor bar for our method. The deformation pattern does not show any instable
behavior.

5.3 Aluminium perforation

The final example is perforation of aluminium target Borvik, Forrestal, Hopperstad,
Warren, and Langseth (in press). In ballistic experiments, aluminium plates with
varying thickness were subjected to impact load. The conical-nose steel projectiles
had nominal mass, diameter and length of 197g, 20mm and 98mm. The target
plates, having a free span diameter of 500mm and nominal thickness of 15, 20,
Large Deformation Analysis 109

(a) (b)
Figure 4: Displaced taylor bar

25 and 30mm, were cut to squares of 600×600mm2 from larger plates. Residual
velocities are reported in Borvik, Forrestal, Hopperstad, Warren, and Langseth (in
press).
We employed the Johnson-Cook material model given in Borvik, Forrestal, Hop-
perstad, Warren, and Langseth (in press) for the target. The von Mises equivalent
stress is expressed as
  ∗ C

σeq = A + Bεeq
n
1 + ε̇eq (1 − T ∗m ) (38)

where A = 59MPa, B = 511MPa, n = 0.285, C = 0.008 and m = 0.859 are material

constants identified for these tests in Borvik, Forrestal, Hopperstad, Warren, and
Langseth (in press) and literature therein, the dimensionless strain rate is given as
∗
ε̇eq = ε̇eq ε̇0 and the homologous temperature is given as T ∗ = (T − Tr )(Tm − Tr ).
Here, ε̇0 = 1s1 is the reference strain rate and Tr = 293K and Tm = 893K indi-
cate room and melting temperature, respectively. The temperature increase under
adiabatic conditions is calculated as
 εeq
σeq d εeq
ΔT = T − Tr = χ (39)
0 ρ Cp

where ρ = 2700.0kg/m3 is the density of the target material, Cp = 910J/kgK is the

specific heat and χ = 0.9 is the Taylor-Quinney coefficient that gives the proportion
of plastic work converted into heat.
110 Copyright © 2008 Tech Science Press CMES, vol.36, no.2, pp.97-117, 2008

(a) effective plastic strain (b) effective plastic strain

(c) von Mises stress

Figure 5: Displaced configuration of the aluminium impact

The projectile is modeled as a bilinear, elastic-plastic, von Mises material with

isotropic hardening

σ = E ε ε ≤ ε0 else, σ = σ0 + Et (ε − ε0 ) ε > ε0 (40)

where σ0 = 1190MPa is the yield strength and Et = 15, 000MPa is the tangent
modulus; E = 204, 000MPa is the elasticity modulus.
Displaced configurations of the 25mm thick aluminium target and the penetrat-
ing projectile at different time steps are illustrated in figure 5. Impact velocity
was 303m/s. Table 1 lists the residual velocities for this experiment and numerous
mesh refinement. Figure 6 shows the deceleration of the projectile during perfora-
tion. For a fine mesh, we run simulations for the other plate thickness and impact
velocity, i.e. t = 20mm and vi = 252m/s and t = 15mm and vi = 219m/s. The
results are shown in table 1 as well and compared to the experimental data. We
obtain good agreement that is independent of the mesh refinement. Also no diffi-
Large Deformation Analysis 111

Table 1: Results of the cantilever beam problem

Numb. thick. Exp. Imp. Exp. Res. Num. Res.
nodes [mm] vel. [m/s] vel. [m/s] vel. [m/s]
22132 25 303 161 165.6
49432 25 303 161 159.6
143425 25 303 161 158.9
324563 25 303 161 158.8
121453 20 252 33.9 36.3
103243 15 218.9 17.0 19.8

325

300
Projectile velocity [m/s]

275
22132 nodes
49432 nodes
250 143425 nodes
324563 nodes

225

200

175
test data (161m/s)

150
0 0.1 0.2 0.3 0.4 0.5
Time [ms]

Figure 6: Deceleration of the projectile at perforation

culties with mesh refinement are observed. We remark that we were not able to get
sufficient results with Eulerian kernel SPH method.

6 Conclusions

In this paper, we explored the application of a Galerkin based SPH formulation

to the simulation of problems involving large deformations in solid mechanics.
The Galerkin based SPH formulation is based on MLS-approximants and stabilized
nodal integration. It takes advantage of updated Lagrangian kernel functions that
guarantees stability of the meshless method while simultaneously maintaining the
applicability of the method to model large deformations.
We demonstrated the advantages of the method for three problems. The first prob-
lem is static compression of incompressible material and illustrates the advan-
112 Copyright © 2008 Tech Science Press CMES, vol.36, no.2, pp.97-117, 2008

tage of updated Lagrangian kernel formulation versus Eulerian kernel and Total
Lagrangian kernel formulation. The second problem is Taylor bar impact. It is
shown that stabilized nodal integration removes instabilities observed for colloca-
tion based SPH formulations. The final example is the application of the Galerkin
based SPH method to perforation of aluminium plates. It shows good agreement
to experimental data. For future research, we intend to develop more accurate con-
stitutive model for impact events. This includes the development of better failure
criterion.

References

Antoci, C.; Gallati, M.; Sibilla, S. (2007): Numerical simulation of fluid struc-
ture interaction by sph. Computers & Structures, vol. 85, no. 11-14, pp. 879–890.

Atluri, S. (2002): The Meshless Local Petrov-Galerkin (MLPG) Method. Tech

Science Press.

Atluri, S.; Shen, S. (2002): The meshless local petrov-galerkin method: a simple
and less-costly alternative to the finite element and boundary element methods.
Computations and Modelling in Engineering and Sciences, vol. 3, pp. 11–51.

Atluri, S.; Zhu, T. (1998): A new meshless local petrov-galerkin (mlpg) approach
in computational mechanics. Computational Mechanics, vol. 22, pp. 117–127.

Atluri, S.; Zhu, T. (2000): The meshless local petrov-galerkin (mlpg) approach
for solving problems in elasto-statics. Computational Mechanics, vol. 25, pp.
169–179.

Beissel, S.; Belytschko, T. (1996): Nodal integration of the element-free galerkin

method. Computer Methods in Applied Mechanics and Engineering, vol. 139, pp.
49–74.

Belytschko, T.; Guo, Y.; Liu, W.; Xiao, S. (2000): A unified stability analysis
of meshfree particle methods. International Journal for Numerical Methods in
Engineering, vol. 48, pp. 1359–1400.

Belytschko, T.; Krongauz, Y.; Organ, D.; Fleming, M.; Krysl, P. (1996): Mesh-
less methods: An overview and recent developments. Computer Methods in Ap-
plied Mechanics and Engineering, vol. 139, pp. 3–47.

Belytschko, T.; Lu, Y.; Gu, L. (1994): Element-free galerkin methods. Interna-
tional Journal for Numerical Methods in Engineering, vol. 37, pp. 229–256.
Large Deformation Analysis 113

Belytschko, T.; Lu, Y.; Gu, L. (1994): Element-free galerkin methods. Interna-
tional Journal for Numerical Methods in Engineering, vol. 37, pp. 229–256.

Belytschko, T.; Lu, Y.; Gu, L. (1995): Crack propagation by element-free

galerkin methods. Engineering Fracture Mechanics, vol. 51, no. 2, pp. 295–315.

Belytschko, T.; Tabbara, M. (1996): Dynamic fracture using element-free

galerkin methods. International Journal for Numerical Methods in Engineering,
vol. 39, no. 6, pp. 923–938.

Bonet, J.; Kulasegaram, S. (2000): Correction and stabilization of smooth parti-

cle hydrodynamics methods with application in metal forming simulations. Inter-
national Journal for Numerical Methods in Engineering, vol. 47, no. 6, pp. 1189–
1214.

Borvik, T.; Forrestal, M.; Hopperstad, O.; Warren, T.; Langseth, M. (in press):
Perforation of aa5083-h116 aluminium plates with conical-nose steel projectiles-
calculations. International Journal of Impact Engineering.

Chandra, N.; Shet, C. (2004): A micromechanistic perspective of cohesive zone

approach in modeling fracture. CMES: Computer Modeling in Engineering &
Sciences, vol. 5, no. 1, pp. 21–33.

Chen, J.-S.; Wu, C.-T.; Yoon, S.; You, Y. (2001): A stabilized conforming nodal
integration for galerkin mesh-free method. International Journal for Numerical
Methods in Engineering, vol. 50, pp. 435–466.

Chen, Z.; Gan, Y.; Chen, J. (2008): A coupled thermo-mechanical model for
simulating the material failure evolution due to localized heating. CMES: Com-
puter Modeling in Engineering & Sciences, vol. 26, no. 2, pp. 123–137.

Dilts, G. (2000): Moving least square particle hydrodynamics i: Consistency and

stability. International Journal for Numerical Methods in Engineering, vol. 44,
pp. 1115–1155.

Duarte, C.; Oden, J. (1996): An h-p adaptive method using clouds. Computer
Methods in Applied Mechanics and Engineering, vol. 139, pp. 237–262.

Dyka, C.; Ingel, R. (1995): An approach for tensile instability in SPH. Comput-
ers & Structures, vol. 57, pp. 573–580.

Fang, J.; Parriaux, A.; Rentschler, M.; Ancey, C. (in press): Improved sph
methods for simulating free surface flows of viscous fluids. Applied Numerical
Mathematics.
114 Copyright © 2008 Tech Science Press CMES, vol.36, no.2, pp.97-117, 2008

Feldman, J.; Bonet, J. (2007): Dynamic refinement and boundary contact forces
in sph with applications in fluid flow problems. International Journal for Numeri-
cal Methods in Engineering, vol. 72, no. 3, pp. 295–324.

Fleming, M.; Chu, Y.; Moran, B.; Belytschko, T. (1997): Enriched element-free
galerkin methods for crack tip fields. International Journal for Numerical Methods
in Engineering, vol. 40, pp. 1483–1504.

Fujimoto, T.; Nishioka, T. (2006): Numerical simulation of dynamic elasto

visco-plastic fracture using moving finite element method. CMES: Computer Mod-
eling in Engineering & Sciences, vol. 11, no. 2, pp. 91–101.

Gao, L.; Liu, K.; Liu, Y. (2006): Applications of mlpg method in dynamic
fracture problems. CMES: Computer Modeling in Engineering & Sciences, vol.
12, no. 3, pp. 181–195.

Guo, Y.; Nairn, J. (2006): Three-dimensional dynamic fracture analysis using the
material point method. CMES: Computer Modeling in Engineering & Sciences,
vol. 16, no. 3, pp. 141–155.

Guz, A.; Menshykov, O.; Zozulya, V. (2007): Contact problem for the flat
elliptical crack under normally incident shear wave. CMES: Computer Modeling
in Engineering & Sciences, vol. 17, no. 3, pp. 205–214.

Hagihara, S.; Tsunori, M.; Ikeda, T. (2007): Application of meshfree method to

elastic-plastic fracture mechanics parameter analysis. CMES: Computer Modeling
in Engineering & Sciences, vol. 17, no. 2, pp. 63–72.

Han, Z.; Atluri, S. (2003): Truly meshless local petrov-galerkin (mlpg) solutions
of traction and displacement bies. Computations and Modelling in Engineering
and Sciences, vol. 4, pp. 665–678.

Hao, S.; Liu, W. (2006): Moving particle finite element method with supercon-
vergence: Nodal integration formulation and applications. Computer Methods in
Applied Mechanics and Engineering, vol. 195, no. 44-47, pp. 6059–6072.

Hao, S.; Liu, W.; Chang, C. (2000): Computer implementation of damage

models by finite element and meshfree methods. Computer Methods in Applied
Mechanics and Engineering, vol. 187, no. 3-4, pp. 401–440.

Hao, s.; Liu, W.; Klein, P.; Rosakis, A. (2004): Modeling and simulation of
intersonic crack growth. International Journal of Solids and Structures, vol. 41,
no. 7, pp. 1773–1799.
Large Deformation Analysis 115

Idelsohn, S.; Onate, E. (2006): To mesh or not to mesh. that is the question...
Computer Methods in Applied Mechanics and Engineering, vol. 195, pp. 4681–
4696.

Idelsohn, S.; Onate, E.; Pin, F. D. (2004): The particle finite element method: a
powerful tool to solve incompressible flows with free surfaces and breaking waves.
International Journal for Numerical Methods in Engineering, vol. 61, pp. 964–989.

Khayyer, A.; H. Gotoh, S. S. (2008): Corrected incompressible sph method for

accurate water-surface tracking in breaking waves. Coastal Engineering, vol. 55,
no. 3, pp. 236–250.

Krongauz, Y.; Belytschko, T. (1997): Consistent pseudo derivatives in meshless

methods. Computer Methods in Applied Mechanics and Engineering, vol. 146, pp.
371–386.

Liu, H.; Han, Z.; Rajendran, A.; Atluri, S. (2006): Computational modeling of
impact response with the rg damage model and the meshless local petrov-galerkin
(mlpg) approaches. CMC: Computers Materials and Continua, vol. 4, pp. 43–53.

Liu, W.; Hao, S.; Belytschko, T. (1999): Multiple scale meshfree methods for
damage fracture and localization. Computational Material Science, vol. 16, no.
1-4, pp. 197–205.

Liu, W.; Jun, S.; Zhang, Y. (1995): Reproducing kernel particle methods. Inter-
national Journal for Numerical Methods in Engineering, vol. 20, pp. 1081–1106.

Liu, W.; Jun, S.; Zhang, Y. (1995): Reproducing kernel particle methods. Inter-
national Journal for Numerical Methods in Engineering, vol. 20, pp. 1081–1106.

Lucy, L. (1977): A numerical approach to the testing of the fission hypothesis.

Astronomical Journal, vol. 82, pp. 1013–1024.

Maiti, S.; Geubelle, P. (2004): Mesoscale modeling of dynamic fracture of

ceramic materials. CMES: Computer Modeling in Engineering & Sciences, vol. 5,
no. 2, pp. 91–101.

Melenk, J. M.; Babuska, I. (1996): The partition of unity finite element method:
basic theory and applications. Computer Methods in Applied Mechanics and En-
gineering, vol. 139, pp. 289–314.

Monaghan, J. (1994): Simulating free surface flows with sph. Journal of Com-
putational Physics, vol. 110, pp. 399–406.
116 Copyright © 2008 Tech Science Press CMES, vol.36, no.2, pp.97-117, 2008

Nishioka, T. (2005): Recent advances in numerical simulation technologies for

various dynamic fracture phenomena. CMES: Computer Modeling in Engineering
& Sciences, vol. 10, no. 3, pp. 209–215.

Nishioka, T.; Kobayashi, Y.; Fujimoto, T. (2007): The moving finite element
method based on delaunay automatic triangulation for fracture path prediction sim-
ulations in nonlinear elastic-plastic materials. CMES: Computer Modeling in En-
gineering & Sciences, vol. 17, no. 3, pp. 231–238.

Onate, E.; Idelson, S.; Zienkievicz, O.; Taylor, R. (1996): A finite point method
in computational mechanics: Applications to convective transport and fluid flow.
International Journal for Numerical Methods in Engineering, vol. 39, pp. 3839–
3866.

Organ, D.; Fleming, M.; Terry, T.; Belytschko, T. (1996): Continuous meshless
approximations for nonconvex bodies by diffraction and transparency. Computa-
tional Mechanics, vol. 18, pp. 225–235.

Rabczuk, T.; Areias, P. (2006): A meshfree thin shell for arbitrary evolving
cracks based on an extrinsic basis. CMES: Computer Modeling in Engineering &
Sciences, vol. 16, no. 2, pp. 115–130.

Rabczuk, T.; Areias, P.; Belytschko, T. (2007): A meshfree thin shell method
for non-linear dynamic fracture. International Journal for Numerical Methods in
Engineering, vol. 72, no. 5, pp. 524–548.

Rabczuk, T.; Areias, P.; Belytschko, T. (2007): A simplified mesh-free method

for shear bands with cohesive surfaces. International Journal for Numerical Meth-
ods in Engineering, vol. 69, no. 5, pp. 993–1021.

Rabczuk, T.; Belytschko, T. (2004): Cracking particles: A simplified mesh-

free method for arbitrary evolving cracks. International Journal for Numerical
Methods in Engineering, vol. 61, no. 13, pp. 2316–2343.

Rabczuk, T.; Belytschko, T. (2005): Adaptivity for structured meshfree particle

methods in 2d and 3d. International Journal for Numerical Methods in Engineer-
ing, vol. 63, no. 11, pp. 1559–1582.

Rabczuk, T.; Belytschko, T. (2006): Application of particle methods to static

fracture of reinforced concrete structures. International Journal of Fracture, vol.
137, no. 1-4, pp. 19–49.
Large Deformation Analysis 117

Rabczuk, T.; Belytschko, T. (2007): A three dimensional large deformation

meshfree method for arbitrary evolving cracks. Computer Methods in Applied
Mechanics and Engineering, vol. 196, pp. 2777–2799.

Rabczuk, T.; Belytschko, T.; Xiao, S. (2004): Stable particle methods based on
lagrangian kernels. Computer Methods in Applied Mechanics and Engineering,
vol. 193, pp. 1035–1063.

Rabczuk, T.; Zi, G. (2007): A meshfree method based on the local partition of
unity for cohesive cracks. Computational Mechanics, vol. 39, no. 6, pp. 743–760.

Randles, P.; Libersky, L. (1997): Recent improvements in sph modeling of

hypervelocity impact. International Journal of Impact Engineering, vol. 20, pp.
525–532.

Randles, P.; Libersky, L. (2000): Normalized sph with stress points. Interna-
tional Journal for Numerical Methods in Engineering, vol. 48, pp. 1445–1461.

Swegle, J.; Hicks, D.; Attaway, S. (1995): Smoothed particle hydrodynamics

stability analysis. Journal of Computational Physics, vol. 116, pp. 123–134.

Tang, Z.; Shen, S.; Atluri, N. (2003): Analysis of materials with strain-gradient
effects: A meshless local Petrov-Galerkin (MLPG) approach, with nodal displace-
ments only. CMES: Computer Modelling in Engineering and Sciences, vol. 4, pp.
177–196.

Ventura, G.; Xu, J.; Belytschko, T. (2002): A vector level set method and new
discontinuity approximation for crack growth by efg. International Journal for
Numerical Methods in Engineering, vol. 54, no. 6, pp. 923–944.

Vignjevic, R.; Campbell, J.; Libersky, L. (2000): A treatment of zero-energy

modes in the smoothed particle hydrodynamics method. Computer Methods in
Applied Mechanics and Engineering, vol. 184, pp. 67–85.

Wang, Z.; Lu, Y.; Hao, H.; Chong, K. (2005): A full coupled numerical anal-
ysis approach for buried structures subjected to subsurface blast. Computers &
Structures, vol. 83, no. 4-5, pp. 339–356.

Wen, P.; Aliabadi, M.; Lin, Y. (2008): Meshless method for crack analysis in
functionally graded materials with enriched radial base functions. CMES: Com-
puter Modeling in Engineering & Sciences, vol. 30, no. 3, pp. 133–147.

Xu, S.; Dong, Y.; Zhang, Y. (2008): An efficient model for crack propagation.
CMES: Computer Modeling in Engineering & Sciences, vol. 30, no. 1, pp. 17–26.
118 Copyright © 2008 Tech Science Press CMES, vol.36, no.2, pp.97-117, 2008

You, Y.; Chen, J.; Voth, T. (2002): Characteristics of semi- and full discretization
of stabilized galerkin meshfree method. Finite Elements in Analysis and Design,
vol. 38, pp. 999–1012.

Zhang, Y.; Chen, L. (2008): A simplified meshless method for dynamic crack
growth. CMES: Computer Modeling in Engineering & Sciences, vol. 31, no. 3, pp.
189–199.

Zi, G.; Rabczuk, T.; Wall, W. (2007): Extended meshfree methods without
branch enrichment for cohesive cracks. Computational Mechanics, vol. 40, no. 2,
pp. 367–382.