Proceedings of the 5th International Conference on

Advanced Concepts in Mechanical Engineering ACME 2012

14 -15 June 2012

ACME 3-04

A COMPARATIVE NUMERICAL ANALYSIS SPH-FEM

IN APPLIED MECHANICS
BY

VASILE NASTASESCU1
1
Military Technical Academy of Bucharest
Department of Integrated Aircraft Systems, Geomatic and Mechanics

Received:
Accepted for publication:

Abstract. The paper presents some results of my researching

regarding to modelling and simulation some problems by Smoothed
Particle Hydrodynamics (SPH) method. The SPH method results are
compared with Finite Element Method (FEM) results and also the
influence of some specific parameters of this method (smoothing length,
material constants etc) are also presented. Next to it, the presented
applications involve two important aspects: large deformations with high
strain rate, in the case of Taylor test and the effect of stress concentration,
in the case of a classical concentrator, a hole in a strip. The paper begins
with some theoretical fundamentals and it continues with two
applications, solved by FEM and SPH method, followed by conclusions.
Key words: numerical analysis, SPH method .

1. Introduction
Nowadays a lot of numerical methods for structural analysis exist and
among these, the Meshfree Methods (MM) in a large definition or the Meshfree
Particle Methods (MPMs) in a more accurate definition, Smoothed Particle
Hydrodynamics (SPH) Method is perhaps the most used method.
A meshfree particle method (MPM) in general refers to the class of
meshfree methods that employ a set of finite number of discrete particles to
represent the state of a system and to record the movement of the system. Each
particle can either be directly associated with one discrete physical object, or be


Corresponding author: [email protected]
 Vasile Nastasescu

generated to represent a part of the continuum problem domain. The particles

can range from very small (nano or micro) scale, to meso scale, to macro scale,
and even to astronomical scale.
In Computing Fluid Dynamics (CFD) and also in hydrodynamics with
material strength (a special case in applied mechanics) each particle possesses a
set of field variables such as mass, momentum, energy, position etc, and other
variables (e.g., charge, vorticity, etc.) related to the specific problem. The
evolution of the physical system is determined by the conservation of mass,
momentum and energy.

2. SPH Theoretical Fundamentals

The SPH method belongs to the meshless methods, so the investigated
domain is represented by a number of nodes, representing the particles of this
domain, having their material and mechanical (mass, position, velocity etc.)
characteristics. Each particle represents an interpolation point on which the
material properties are known.
Practically, depending on the used software, the boundary conditions
have to be imposed to some of particles, according to the problem analyzed, like
in the case of finite element method.
Theoretical fundamentals of this aspect are a bit different and require
more space to be presented here. The problem solution is given by the
computed results, on all the particles, using an interpolation function. We can
say that the fundamentals of SPH theory consist in interpolation theory; all the
behaviour laws are transformed into integral equations.
The kernel function, or smoothing function, often called smoothing
kernel function, or simply kernel, gives a weighted approximation of the field
variable (function) in a point (particle). Integral representation of a function
, used in the SPH method starts from the following identity:

(1)

where is a function of a position vector , which can be an one-, two- or

three-dimensional one; is a Dirac function, having the properties:

(2)

In equation (1), is the function domain, which can be a volume, that

contains the , and where is defined and continuous.
By replacing the Dirac function with a smoothing function
the integral representation of becomes:
 Proceedings of the 5th International Conference ACME 2012

(3)

where is the smoothing kernel function, or smoothing function, or kernel

function. The parameter , of the smoothing function , is the smoothing
length, by which the influence area (support domain) of the smoothing function
is defined.

Fig. 1 – The support domain of the kernel function.

As long as Dirac delta function is used, the integral representation,

described by equation (1), is an exact (rigorous) one, but using the smoothing
function instead of Dirac function, the integral representation can only be an
approximation. This is the reason for the name of kernel approximation.
Using the angle bracket, this aspect is underlined and the equation (3) can
be rewritten as:

(4)

The smoothing function is usually chosen to be an even one, which

has to satisfy some conditions, which are not presented here. The literature
presents different smoothing function. Theoretically, any function having the
properties presented above, can be employed as SPH smoothing function.
One of the most used smoothing function is a cubic B-spline kernel
function, in the form given by relation (5), where s = r/h, n is the number
representing the spatial dimension and is a constant which has the value: 2/3,
10/7 or respectivelly 1/ , depending on the space dimension.
 Vasile Nastasescu

(5)

Fig. 2 – The cubic B-spline kernel function.

The graphical representation of the smoothing function described by

relation (5) and its derivatives (first and second) can be seen in the Figure 2.

3. SPH Method in Applied Mechanics

Because some particle parameters are referring to material properties, we
can use SPH method in structural analysis, this aspect being only an
assumption, which along others considerations represents the theoretical
fundamentals. In some conditions, for enstance high velocity impact (HVI), the
mechanical structure material often behaves like a fluid. The SPH
method proved to be fully suitable for the study of fluid flow, so only the
material strength has to be added to equations of fluid mechanics.
The difference between modeling with SPH method of the fluid flow or
of the mechanical structure consist in the way in which the relationship of the
stress, strain and strain rate are expressed. So, we could say that SPH method
 Proceedings of the 5th International Conference ACME 2012

will be applied for hydrodynamics with material strength.

Because a detalied presentation of these theoretical aspects need much
more space, which exceed the space and just the aim of this paper, the equations
will be presented in a very concentrated form.
The equation of continuity, expressed by the continuity of density ( ) is:

(6)

The momentum conservation, in SPH formulation is written,

(7)

where the brackets may still contain a term representing the artificial viscosity
effect. The equation of energy ( ) conservation is represented by relation (8), in
which the term represents the artificial heating.
An other equation – relation (9) - representing the velocity, would have
to be added for an complet equation set.

(8)

(9)

In all the equations (6)…(9), the subscripts or show that the

parameter with this subscript, is referring to the particle or , respectively.
is the total particle number in the support domain centered on particle .
Others notations are: for isotropic pressure, for velocity, for
viscous shear stress and is the strain rate tensor.
The superscripts şi define the coordinate directions. The equations
(6)…(9) can be integrated by any standard methods, so all the parameters
attached to a particle are known at each time step.

4. FEM-SPH Numerical Comparative Results

This example is a comparative presentation of MEF-SPH numerical
analysis; a flat strip with a hole in its centre, is subjected to a tension load, in
dynamic regime. In the Figure 3, the finite element model is presented.
 Vasile Nastasescu

a) b)
Fig. 3 – The finite element model.

Fig. 3-a presents a detail of the model that shows a symmetric mesh, and
a no uniform one. The structure could be modelled using only a model of ¼, but
for a result control, I preferred to use entire structure, in this case the computer
time being not a restrictive condition. Therefore, a symmetric geometry and
loading will lead to symmetric results, if the calculus is right.

Fig. 4 – The particle model for SPH method.

Figure 4 presents the particle model adopted for using of SPH method. A
detail of the model is also presented in the figure 4-b. As we could see, the
particle distribution is an uniform one, this aspect being a requirement of a
proper using of SPH method.
 Proceedings of the 5th International Conference ACME 2012

Fig. 5 – Time evolution of the applied force

The structure consists in a strip (with a hole in its centre) with thickness
of 1 mm, having the dimensions 60 mm height and 20 mm width. The hole
diameter is 10 mm. The structure material is a high carbon steel.
The material model is a plastic-cinematic one, being suited for modeling
of the isotropic and kinematic hardening plasticity with the option of including
rate effects by Cowper-Symonds coefficients.
The finite element model has 1560 nodes and 1436 elements; the smallest
finite element dimensions being 0.25 mm and the biggest dimensions about 1.50
mm. The SPH model has 28412 particles, uniformly distributed, with an
internodal distance of 0.20 mm.
For both models (FEM and SPH), the maximum value of distributed load,
applied on the upper side of the strip, was of 8000 N; on the lower side, the
nodal displacement on the load direction was blocked.
In the figure 6, we can see the von Mises stress field at time of 20e-6 and
25e-6 seconds. Figures 6-a and 6-c are referring to FE model, and figures 6-b
and 6-d are referring to SPH model.
We could notice a good concordance between the stress field shapes and
between the values too. The errors, for maximum values of von Mises stress
calculated by SPH method, are -0.67% at the time of 20e-6 seconds and 0.32%
at the time of 25e-6 seconds. We also could notice the appearance of the stress
concentration effect, where from the crack started and grew to a total break.
Than, the symmetry of the stress field can be noticed, this aspect being a proof
of the correct calculation.
 Vasile Nastasescu

a) b)

c) d)
Fig. 6 – Von Mises stress field.

Correct results, by SPH method, are strongly affected by the right

choosing of the smoothing length ( ). Many programs use different criteria for
a proper choosing of the parameter, but often the programs allow users to
make their choice. I will not refer to the automatic choosing criteria, this aspect
being beyond the aim of this paper. I will refer to a practical method, which can
be used in connection with FEM.
The numerical analysis by FEM allowed me to know the total energy of
the structure under loading at any moment. The numerical analysis by SPH
allowed me to get the dependence between the total energy of the structure,
under loading at any moment, versus smoothing length ( ), as the figure 7
present this correlation.
 Proceedings of the 5th International Conference ACME 2012

Fig. 7 – Material total energy vs. smoothing length.

Using the graphical representation of the figure 7, the material total

energy (1639.70 Nmm) calculated by FEM, allow us to get the proper value of
smoothing length (1.35). Using this value of the smoothing length , all the
results by SPH method are in a very good agreement with the results by FEM.

Fig. 8 – Time evolution of the material total energy, by FEM.

In the figures 8 and 9, the time evolutions of the material total energy are
presented and as we can notice, the allure and the maximum value are in a very
good concordance.
 Vasile Nastasescu

Fig. 9 – Time evolution of the material total energy, by SPH method.

The maximum value of the material total energy by SPH method is

different by FEM with an error of -0.71%. Others results, SPH-FEM, can be
similarly compared and each time will find out a good agreement, which show
us that SPH method could be a fitted or better method in applied mechanics.

5. Conclusions
Using of the SPH method in applied mechanics are still under
researching, but until now it appears to be a fitted and a stronger one, specially
in impact problems with high velocity.
This paper demonstrates that the SPH method can be successfully used in
others dynamic problems in applied mechanics.

REFERENCES

Hallquist, J. O., (1998), LS-DYNA Theoretical Manual, Livermore Software

Technology Corporation, 2876 Waverley Way, Livermore, California 94550-1740.
Lacome J. L., (2004), Smoothed Particle Hydrodynamics Method in LS-DYNA, 3rd LS-
DYNA Forum, Bamberg, Germany
Liu, G. R., Liu, M. B., (2009), Smoothed Particle Hydrodynamics – a meshfree particle
method, World Scientific Publishing Co. Pte. Ltd., 5 Toh Tuck, Singapore 596224.
Nastasescu, V., Iliescu, N., (2009), Numerical Simulation of the Impact Problems by
SPH Method, Proceedings of 4th National Conference of Academy of Technical
Science in Romania, Iaşi, 1, 193-198.
 Proceedings of the 5th International Conference ACME 2012

O ANALIZĂ NUMERICĂ COMPARATIVĂ SPH-MEF IN MECANICA APLICATĂ

(Rezumat)
Alături de metodele numerice binecunoscute, precum Metoda elementelor
finite, Metoda fâşiilor finite, Metoda volumelor finite, Metoda elementelor de frontieră,
au apărut, începând cu anul 1977 şi s-au dezvoltat apoi metode numerice fără reţea
(meshless) sau altfel spus metode numerice fără conexiunea elementelor care compun
domeniul de analiză numerică. Aceste metode sunt adesea numite metode cu particule
libere. Metoda reprezentativă a acestor metode fără reţea este Metoda SPH (Smoothed
Particle Hydrodynamics). O traducere exactă a expresiei poate că nu ar fi suficient de
expresivă, încât consider preferabil să păstrăm prescurtarea din limba engleză.
Astfel de metode se bazează pe reprezentarea unui domeniu al analizei
numerice (1D, 2D sau 3D) constituit dintr-un set de particule fără conexiuni fizice între
ele, care descriu starea sistemului într-un echilibru dinamic, precum şi evoluţia sa.
Fiecare particulă poate fi asociată unei porţiuni fizice a domeniului şi la rândul ei îi sunt
asociate anumite proprietăţi, precum poziţie, masă, viteză, acceleraţie etc. Formele
particulelor pot fi diferite în cadrul metodei, dar cea mai utilizată este forma sferică, cea
de elipsoid sau de tipul unor capsule alungite. Cât priveşte mărimea particulelor, aceasta
cunoaşte o plajă extrem de largă, de la scara nano la macro, ajungând chiar la scări
astronomice. Acest lucru nu este de mirare dacă ne gândim că metoda SPH a apărut în
efortul astrologilor de a modela fenomenele astrologice.
La ora actuală, metoda SPH este validată ca metodă de mare eficienţă şi
folosită curent în mecanica fluidelor. Metoda permite modelarea cu uşurinţă a unor
probleme foarte dificile precum descrierea evoluţiei suprafeţei libere a fluidelor,
modelarea oricărui tip de curgere, formarea fenomenelor de tip tsunami etc., dar şi
rezolvarea directă a interacţiunii fluid structură.
Cât priveşte utilizarea metodei SPH în domeniul mecanicii aplicate, există deja
progrese şi realizări remarcabile dar, domeniul fiind extrem de vast, cercetările
ştiinţifice nu sunt încheiate. Acesta este contextul în care se încadrează şi prezenta
lucrare, izvorâtă din cercetările ştiinţifice proprii.
Mai trebuie reţinut faptul că s-au dezvoltat programe de calcul specializate,
deosebit de puternice, acompaniate de pre- şi post- procesoare grafice adecvate. Dar pe
lângă acest fapt, mai trebuie remarcat şi faptul că metoda SPH a fost deja implementată
în cele mai puternice programe de analiză numerică a mediilor continue; exemple
precum Ls-Dyna şi Autodyne sunt elocvente şi desigur nesingulare.
În această lucrare, după o sintetică prezentare a unor consideraţii teoretice
fundamentale, este analizat cazul unei solicitări dinamice la tracţiune a unei platbande
cu gaură. Folosindu-se şi un model de material adecvat, s-a reuşit simularea numerică a
apariţiei fenomenului de concentrare a tensiunilor, simularea apariţiei fisurii şi
dezvoltarea acesteia până la rupere. Toate acestea sunt prezentate comparativ cu
utilizarea metodei elementelor finite. Sunt analizate şi alte aspecte, precum influenţa
cantitativă a variaţiei lungimii de netezire, un criteriu de adoptare a mărimii acesteia etc.
De mare importanţă este şi descrierea modelului de calcul, care are deosebiri
fundamentale faţă de metoda cu elemente finite.
Prezenta lucrare poate fi deosebit de utilă studenţilor, cadrelor didactice şi
cercetătorilor care folosesc sau doresc să utilizeze metoda SPH.