11

Meshfree Methods
Saeid Zahiri
Shiraz University,
Islamic Republic Of Iran
1. Introduction
In this chapter we will describe about numerical simulation with meshfree methods. We
know; phenomena in nature, whether physical, geological, mechanical, electrical, or
biological, can often be describe by means of algebraic, differential, or integral equations.
Obtaining exact solutions for these equations is ideal. Unfortunately, we can only obtain
exact ones for limited practical problems because most of these problems are complex.
Therefore using numerical procedure to obtain approximate solutions is inevitable. One of
the most important tools in the field of numerical methods that has been developed newly is
meshfree or meshless methods.
A meshfree method is a method used to establish system algebraic equations for the whole
domain of problem without using a predefined mesh for the domain discretization. This
infant method uses a set of scattered nodes, called field nodes, to establish the problem
domain and boundaries, which do not require any priori information on the relationship
between the nodes for the interpolation or approximation of the unknown functions of field
variables. In the FEM, a continuum with a complicated shape is divided into elements, finite
elements. The individual elements are connected together by a topological map called a mesh.
Meshfree methods have been proposed and achieved remarkable progress over the past few
years. According to the formulation procedure, meshfree methods fall into three categories:
meshfree weak form methods (like: EFG, MLPG, LRPIM,), meshfree strong form methods
(like: SPH, Collocation method,) and meshfree weak-strong form methods based on the
combination of both weak form and strong form (like: MWS method). These three categories
and their limitations, applications, advantages and other descriptions will be introduced. In
seeking for an approximate solution to the problem governed by PDEs and boundary
conditions, one first needs to approximate the unknown field function using shape (trial or
base) functions before any formulation procedure can be applied to establish the discretized
system equations. In this chapter definition of base and shape functions and various
techniques for meshfree shape function constructions are discussed. These shape functions
are locally supported, because only a set of field nodes in a small local domain are used in
the construction and the shape function is not used or regarded as zero outside the local
domain. Such a local domain is termed the support domain or influence domain. The
concept and kinds of support domain and determination of the dimension of the support
domain will be described.
After introducing the concept of support domain, the point interpolation method (PIM) in
detail will be discussed. Point interpolation method is one of the series representation

www.intechopen.com

232

Numerical Analysis Theory and Application

methods for the function approximation, and useful for creating meshfree shape functions.
A scalar function defined in the problem domain that is represented by a set of scattered
nodes will be shown. There are two types of PIM shape functions have been developed so
far using different forms of basis functions Polynomial basis functions and radial basis
functions (RBF) have often been used in meshfree methods. These two types of PIMs will be
discussed in the following chapter.
For satisfying the boundary conditions, penalty method, direct method, lagrange multiplier
method and direct interpolation method can be used to enforce essential boundary
conditions. One of these methods due to using meshfree method can be elected and they
will be explained and compared.
To simulate some problems, the partial differential equations and boundary conditions for
two dimensional solid mechanics and fluid mechanics problem and heat transfer problem
especially thermodynamics of plates and shells will be given in sub-sections. These
problems are solved with meshfree methods.

2. Meshfree methods categories

According to the formulation procedure, meshfree methods fall into three categories:
meshfree weak form methods (like: EFG, MLPG, LRPIM,), meshfree strong form methods
(like: Collocation method, SPH,) and meshfree weak-strong form methods based on the
combinations of both weak forms and strong forms (like: MWS method).
2.1 Strong form methods
Regarding to formulate the governing equations, the direct approximate solution from the
differential equations is used. It means the strong form of governing equations for boundary
conditions are directly discretized at the field nodes to obtain a set of discretized system
equations. If Taylor series is used and the differentiations are replaced, the method is the
strong form method. The strong form method does not need the numerical integration. Thus
The background mesh even locally is not needed for the strong form methods.
Meshfree strong form methods have some attractive advantages: a simple algorithm,
computational efficiency, and truly meshfree. However, Meshfree strong form methods are
often unstable, not robust, and inaccurate, especially for problems with derivative boundary
conditions. Several strategies may be used to impose the derivative (Neumann) boundary
conditions in the strong form methods, such as the use of fictitious nodes, the use of the
Hermite-type Meshfree shape functions, the use of a regular grid on the derivative
boundary.
2.2 Weak form methods
In Meshfree weak form methods, the governing partial differential equations (PDEs) with
derivative boundary conditions are first transformed to a set of so called weak form integral
equations using different techniques. The weak forms are then used to derive a set of
algebraic system equations through a numerical integration process using sets of
background cells that may be constructed globally or locally in the problem domain.
Meshfree weak form methods were relatively under developed before 1990, but there has
been a substantial increase in research effort since then.
There are now many different versions of Meshfree weak form methods. Meshfree weak form
methods based on the global weak forms are called Meshfree global weak form methods, and

www.intechopen.com

Meshfree Methods

233

those based on local weak forms are called Meshfree local weak form methods. Meshfree
global weak form methods are based on the global Galerkin weak form for equations of
problems and the Meshfree shape functions. Two typical Meshfree global weak form
methods: the element free Galerkin (EFG) method (Belytschko et al., 1994a) and the radial
point interpolation method (RPIM) (GR Liu and Gu, 2001c; Wang and GR Liu, 2000; 2002a).
Another typical Meshfree global weak form method is the reproducing kernel particle
method (RKPM) proposed by Liu and coworkers in 1995 (Liu et al., 1995). The main idea of
RKPM is to improve the SPH approximation to satisfy consistency requirements using a
correction function. RKPM has been used in nonlinear and large deformation problems
(Chen et al., 1996; Chen et al., 1998; Liu and Jun, 1998), inelastic structures (Chen et al., 1997),
structural acoustics (Uras et al., 1997), fluid dynamics (Liu and Jun et al., 1997), et cetera.
Meshfree local weak form methods were developed by Atluri and coworkers based on the
local Petrov-Galerkin weak form , and the Meshfree shape functions. Some other Meshfree
weak form methods have also been developed, such as the hp-cloud method (Armando and
Oden, 1995), the partition of unity finite element method (PUFEM) (Melenk and Babuska,
1996; Babuska and Melenk, 1997), the finite spheres method (De and Bathe, 2000), the free
mesh method (Yagawa and Yamada, 1996), et cetera.
2.3 Weak-strong form methods
These Meshfree methods are called Meshfree weak-strong (MWS) form methods in this
book because are based on the combination of weak and strong form methods. The MWS
method was developed by GR Liu and Gu (2002d, 2003b). The key idea of the MWS method
is that in establishing the discretized system equations, both the strong form and the local
weak form are used for the same problem, but for different groups of nodes that carries
different types of equations/conditions. The local weak form is used for all the nodes that
are on or near boundaries with derivative (Neumann) boundary conditions. The strong form
is used for all the other nodes. The MWS method uses least background cells for the
integration, and it is currently the almost ideal Meshfree method that can provide stable and
accurate solutions for mechanics problems.
There are also Meshfree methods based on the integral representation method for function
approximations, such as the Smooth Particle Hydrodynamics (SPH) methods (Lucy, 1977;
Gingold and Monaghan, 1977; GR Liu and Liu, 2003, etc.). In the standard SPH method, the
function approximation is performed in a weak (integral) form, but strong form equations
are directly discretized at the particles.
2.4 Comparisons between three meshfree categories
Each meshfree method has features with advantages and defects. With these properties, the
appropriate method can be selected to solve the problem. The features of methods are
presented in sub-sections.
2.4.1 Advantages and disadvantages
Convergence rate and highest accuracy are important properties in numerical methods.
When the problems include Dirichlet boundary conditions, the strong form methods are the
best but in case of Neumann boundary conditions, weak form methods are optimum and
when the both of Dirichlet and Neumann boundaries are used in problems, the weak-strong
form methods are useful.

www.intechopen.com

234

Numerical Analysis Theory and Application

The strong form methods are with good convergence rate and they are truly meshless. The
procedure is straightforward, and the algorithms and coding are simple. They are
computationally efficient, and the solution is accurate when there are only Dirichlet
boundary conditions.
However, Meshfree strong form methods have disadvantages: they are often unstable and
less accurate, especially for problems governed by PDEs1 with derivative boundary
conditions. Derivative boundary conditions (DBCs) involve a set of separate differential
equations defined on the boundary; these are different from the governing equations
defined in the problem domain. These DBCs require special treatments. Unlike integration,
which is a smoothing operator, differentiation is a roughening operator; it magnifies errors
in an approximation. This magnified error is partially responsible for the instability of the
solution of PDEs. Hence, Meshfree strong form methods are often unstable. Special
treatments are employed to implement the derivative boundary conditions in Meshfree
strong form methods. However, such treatments cannot always control the error. A
technique suitable for one problem may not work for another, even one of the same types. A
set of parameters tuned for one problem may not work for another.
The common feature of Meshfree weak form methods is that the PDE of a problem is first
replaced by or converted into an integral equation (global or local) based on a principle
(weighted residual methods, energy principle). Weak form system equations can then be
derived by integration by parts. A set of system equations of Meshfree weak form methods
can be obtained from the discretization of the weak form using meshfree interpolation
techniques. There are four features of the local weak form . The integral operation can smear
the error over the integral domain and, therefore improve the accuracy in the solution. It
acts like some kind of regularization to stabilize the solution. The requirement of the
continuity for the trial function is reduced or weakened, due to the order reduction of the
differential operation resulting from the integration by parts. The force (derivative)
boundary conditions can be naturally implemented using the boundary integral term
resulting from the integration by parts. The system equations in the domain and the
derivative boundary conditions are conveniently combined into one single equation.
These features give Meshfree weak form methods the following advantages. They exhibit
good stability and excellent accuracy for many problems. The derivative (Neumann)
boundary conditions can be naturally and conveniently incorporated into the same weak
form equation. No additional equations or treatments are needed and no errors are
introduced in the enforcement of traction boundary conditions. A method developed
properly using a weak form formulation is applicable to many other problems. A set of
parameters tuned for one method for a problem can be used for a wide range of problems.
This robustness of the weak form methods have been demonstrated through many practical
problems. It is this robustness that makes the weak form methods applicable to many
practical engineering problems.
However, Meshfree global weak form methods are meshfree only in terms of the interpolation
of the field variables. Background cells have to be used to integrate a weak form over the
global domain. The numerical integration makes them computationally expensive, and the
background mesh for the integration means that the method is not truly meshless.
In the Meshfree local weak form methods, the local integral domain in the interior of the
problem domain is usually of a regular shape. It can be as simple as possible and can be
1

Partial Differential Problems

www.intechopen.com

Meshfree Methods

235

automatically constructed in the process of computation. The Meshfree local weak form
methods have obtained satisfactory results in solid mechanics and fluid mechanics (Atluri
and Shen, 2002; GR Liu, 2002).
Although the Meshfree local weak form methods made a significant step in developing
ideal meshfree methods, the numerical integration is still burdensome, especially for nodes
on or near boundaries with complex shape. The local integration can still be
computationally expensive for some practical problems. It is therefore desirable to minimize
the need for numerical integrations.
The Mesh Weak-Strong method is designed to combine the advantages of strong form and
weak form methods and to avoid their shortcomings. This can be performed only after a
thorough examination of the features of both types of methods, presented in the above
sentences. An Meshfree weak-strong (MWS) form method was proposed recently by GR Liu
and Gu (2002d); it aimed to remove the background mesh for integration as much as
possible, and yet to obtain stable and accurate solutions even for PDEs with derivative
boundary conditions. The MWS method has been successfully developed and used in solid
mechanics (Gu and GR Liu, 2005; GR Liu and Gu, 2003b) and fluid mechanics (GR Liu and
Wu et al., 2004; GR Liu and Gu et al., 2003c).
The convergence of the MWS method is studied numerically by comparison with other
methods. The weak form method treats the Neumann boundary condition naturally and
easily. In addition, the accuracy achieved by meshfree methods based on the weak form
equations are generally much better than those based on strong form equations. However,
the efficiency is a big problem for the weak form methods because of the need for weak
form integration.
The MWS method proposed by Liu and Gu was based on both collocation and local radial
point interpolation formulation. In the present MWS method, the strong form of meshfree
collocation method is applied to the internal nodes and the nodes on the essential
boundaries, while the local radial point interpolation weak form is applied to the nodes on
the natural boundaries. The advantages of this MWS method are:
1. The Neumann boundary condition can be imposed straightforwardly and accurately
with arbitrary nodal distributions.
2. Stable and accurate solution can be obtained with high efficiency.
2.4.2 Applications of each category
Strong form methods are suitable for Dirichlet boundary conditions problems and weak
form methods are used more with problems that have Neumann boundary conditions.
Weak-strong form methods are appropriate for problems with both of Dirichlet and
Neumann boundary conditions.

3. Shape functions
In seeking for an approximate solution to a problem governed by PDEs and boundary
conditions, one first needs to approximate the equation variables using shape functions,
befor any formulation procedure can be applied to establish the descretized system
equations.
This section discusses various techniques for MFree shape function constructions. These
shape functions are locally supported, because only a set of field nodes in a small local
domain are used in the construction and the shape function is not used or regarded as zero

www.intechopen.com

236

Numerical Analysis Theory and Application

outside the local domain. Such a local domain is termed the support domain or influence
domain or smoothing domain.
3.1 Point interpolation methods shape functions
The point interpolation method (PIM) is one of the series representation methods for the
function approximation, and is useful for creating Meshfree shape functions. Consider a
scalar function T(x) defined in the problem domain that is represented by a set of
scattered nodes. The PIM approximates T(x) at a point of interest x in the form of
T(x)=

(1)

where the Bi(x) are the basis function defined in the space Cartesian coordinates XT=[x, y] , m
is the number of basis functions, and the ai are the coefficients.
For function approximation, a local support domain is first formed for the point of interest
at x which includes a total of n field nodes. For the conventional point interpolation method
(PIM), n=m is used that results in the conventional PIM shape functions that pass through
the function values at methods. The RPIM interpolation augmented with each scattered
node within the defined support domain.
For the weighted least square (WLS) approximation or the moving least squares (MLS)
approximation, n is always larger than m. There are two types of PIM shape functions have
been developed so far using different forms of basis functions. Polynomial basis functions
(GR Liu and Gu, 1999; 2001a) and radial basis functions (RBF) (Wang and GR Liu, 2000; GR
Liu, 2002) have often been used in Meshfree methods.
3.1.1 Conventional polynomial PIM
Using polynomials as the basis functions in the interpolation is one of the earliest
interpolation schemes. It has been widely used in establishing numerical methods, such as
the FEM. Consider a continuous function u(x) defined in a domain , which is represented
by a set of field nodes. The u(x) at a point of interest x is approximated in the form of

u(x)=m
i=1

={

PT(x)=(1 x y x2 xy y2)

x . . }

.
.
.

for m=6, p=2 (2-D)

(2)

(3)

where pi(x) is a given monomial in the polynomial basis function in the space coordinates
xT=[x,y], m is the number of monomials, and ai is the coefficient for pi(x) which is yet to be
determined. The pi(x) in Equation is built using Pascal's triangles, and a complete basis is
usually (but not always)
3.1.2 Radial point interpolation shape functions
In order to avoid the singularity problem in the polynomial PIM, the radial basis function
(RBF) is used to develop the radial point interpolation method (RPIM) shape functions for
Meshfree weak form methods (GR Liu and Gu, 2001c; Wang and Liu, 2000; 2002a,c). The
RPIM shape functions will be used for both Meshfree weak form and strong form polynomials
can be written as

www.intechopen.com

237

Meshfree Methods

Fig. 1. Pascal-khayyam triangle of monomials for two dimensional domain

u(x,y)=nj=1

x,y

m
i=1

x,y

(4)

Where Rj(x) is a radial basis function (RBF), n is the number of RBFs, pi(x) is monomial in the
space coordinates xT=[x, y], and m is the number of polynomial basis functions. When m=0,
pure RBFs are used. Otherwise, the RBF is augmented with m polynomial basis functions.
Coefficients ai and bj are constants. r is the distance between the point of interest (x,y) and a
node (xi,yi) at
r=

(5)

There are a number of types of radial basis functions (RBF), and the characteristics of RBFs
have been widely investigated (Kansa,1990; Sharan et al.,1997; Franke and Schaback, 1997;
etc). Four often used RBFs, the multi-quadrics (MQ) function, the Gaussian (Exp) function,
the thin plate spline (TPS) function, and the Logarithmic radial basis function, are listed in
Table.1.

Table 1. Typical radial basis functions with dimensionless shape parameters

Note: dc is a characteristic length that relates to the nodal spacing in the local support
domain of the point of interest x, and it is usually the average nodal spacing for all the nodes
in the local support domain.
R1(r1 ) R2 (r1 ) Rn (r1 )
R (r ) R (r )
Rn (r2 )
1 2
2 2
R0

R
r
R
r
R
(
)
(
)
2 n
n ( rn )
1 n
the polynomial moment matrix is

www.intechopen.com

(6)

238

Numerical Analysis Theory and Application

1
1
1
x

x2
xn
PmT 1

P
x
P
x
P
(
)
(
)
(
m 2
m xn )
m 1

rk ( xk xi )2 ( y k yi )2

U R0
U S S T
0 Pm

Pm a
G a0
0 b

T ( x )U
u( x ) RT ( x ) PT ( x ) G 1 U S
S

T ( x ) RT ( x ) PT ( x ) G 1

{1( x ) 2 ( x )n ( x ) n 1( x ) n m ( x ))

(7)

(8)
(9)
(10)

(11)

The above equations are brought to show the procedure of shape function produce. The
shape functions are obtained and then the discretized derivatives can be used to
governing equations and the parameters are shown with the equation

u( x ) T ( x )US i ui
n

i 1

(12)

The derivatives of u(x) are easily obtained as

u, l ( x ) T, l ( x )US

(13)

where l denotes either the coordinates x or y.

3.2 Support domain
The accuracy of interpolation for the point of interest depends on the nodes in the support
domain as shown in Fig.2. Therefore, a suitable support domain should be chosen to ensure
an efficient and accurate approximation. For a point of interest at , the dimension of the
support domain
is determined by

rsx s .dcx
rsy s .dcy

(14)

is the dimensionless size of the support domain, and dc is the nodal spacing near the point
at
. If the nodes are uniformly distributed, dc is simply the distance between two
neighboring nodes. When nodes are non uniform and where is a constant of shape
parameter, dc can be defined as an average nodal spacing in the support domain of . The
exponential function of the support domain
controls the actual dimension of the support
domain.
Rectangular support domains ( and : dimensions of the support domain in x and y
directions). The support domain is centred by .

www.intechopen.com

239

Meshfree Methods

Fig. 2. Support domains of points of interest at

in Meshfree models.

The actual number of nodes, n, can be determined by counting all the nodes included in the
support domain. Generally, an =2.0~3.0 leads to good results for many problems that we
have studied. Note that the support domain is usually centered by a point of interest at .

4. Satisfying boundary conditions

For the Dirichlet boundary condition, the essential boundary conditions for u can be simply
given as follows: (when node is on the boundary)
(15)

u=

The essential boundary condition can be directly imposed using the direct interpolation
method. another method is the Penalty method has been used to enforce essential boundary
conditions in the MLPG and LRPIM Methods. Since RPIM shape functions possess the
Kronecker delta function property, the essential boundary conditions can be easily enforced
as in the FEM (see, e.g., GR Liu and Quek, 2003).
The natural boundary conditions can be satisfy automatically when we use weak-strong
form method and no additional equation or treatment is needed.

4.1 Direct method

The ith component is prescribed by setting
(16)

ui=

Such an essential boundary condition can then be enforced directly into the system Equation
through the following modifications to the global matrix and the global right vector. The
global matrix, K, is changed to

k11 k1( i 1)

k
( i 1)1
0
0
K 0

k
( i 1)1

kn1 kn( i 1)

www.intechopen.com

0
0
0
1
0

k1n

k( i 1)n

0
0

k( i 1)n

knn

k1( i 1)

0
0 kn( i 1)

(17)

240

Numerical Analysis Theory and Application

The components in the global right vector are changed to

Ti
Fi
Fj k jiTi

i j
i j

(18)

The direct method can exactly enforce essential boundary conditions, but changing matrices
and vectors needs additional computational operations. In addition, the algorithm of the
direct method is also complicated.
4.2 Penalty method
The penalty method is a convenient alternative for enforcing the essential boundary
conditions, in which the diagonal entry kii in the stiffness matrix, is changed to

kii .kii

(19)

where is the penalty coefficient that is the much larger number than the components of
the global matrix K. In the global right vector F, only the component Fi is changed as
follows
.k .T
Fi ii i
Fj

i j
i j

(20)

The penalty method has some advantages: there are only two changes of matrices, and the
algorithm is very simple. However, the penalty method can only approximately satisfy the
essential boundary conditions. In addition, the accuracy is affected by selection of the
penalty coefficient.
the global matrix, K, is then changed to
... k1( i 1)
k11

k
( i 1)1
K ki1 ki( i 1)

k( i 1)1

kn1
... kn( i 1)

k1i

k( i 1)i

kii

k( i 1)i

kni

k1( i 1)

ki ( i 1)

kn( i 1)

k1n

k( i 1)n

kin

k( i 1)n

knn
...
...

(21)

5. Examples for numerical simulations

In this section, some problems are brought to show the abilities of meshfree methods for
solving the heat, solid and fluid mechanics problems.
5.1 Heat conduction
Meshfre methods are used to solve the heat transfer problem. For example, heat conduction
in the plate is solved.

www.intechopen.com

241

Meshfree Methods

5.1.1 Formulation of heat transfer in the plate

Formulation of heat transfer in Cartesian coordinate is

T
T
T
( k ) ( k ) q C p ( )
t
dx x
dy y
If k(conductivity coefficient) is constant and for steady state
generation we have:
(

2T
2T

)
(
)0
x 2
y 2

T is the Temperature and is the rate of energy generation and

the specific heat in the formula.

(22)

and without any energy

(23)

is the density and Cp is

5.1.2 Numerical results and discussion Domain representation for heat transfer
First, the temperature distribution in square plate is obtained. In problem 1, the bottom wall
is in temperature T0 and other walls are in temperature 0 and in problem 2, the up wall has
Neumann boundary condition. To check the validity of the method, three different
problems are considered. Fig.3 shows the domain representation for problems 1 and 2 by
the scattered nodes. The essential and natural boundary conditions should be satisfied on
the boundary nodes.

Fig. 3. The problem is represented by 256(16x16) regular nodes

5.1.3 Problem 1 with essential boundary conditions
Fig.4 shows the problem 1 and its boundary conditions. The temperature distribution in the
plate obtained by MWS method presented in this chapter is given in Fig. 5.
The constant temperature lines in this figure are shown by solid lines. Table.2 compares the
results obtained by collocation method and LRPIM method with those obtained by the
analytical method.

www.intechopen.com

242

Numerical Analysis Theory and Application

Fig. 4. Problem 1 and its boundary conditions

Fig. 5. Temperature distribution with essential boundary conditions (problem1 solved with
MWS method)
the analytical solution of the problem can be written as

T(x, y )

1n 1 1
sin(n x )( tanh(n ).cosh(n y ) sinh(n y ))
n
n 1

(24)

(x,y) are the coordinate of points in the plate. T is the temperature. We showed the
difference between three meshfree methods and the difference between using different
number of nodes to give the better results.
We used the error norm

Err( j )

www.intechopen.com

T ( j ) Tanalytic ( j )
Tanalytic ( j )

(25)

243

Meshfree Methods

and Total Error defined as

( Err( j ))
n

TotalErr

j 1

(26)

256 nodes

x=0.5

1156 nodes

Analytical

Collocation

LRPIM

MWS

LRPIM

MWS

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1

1
0.8017
0.6208
0.4679
0.3449
0.2500
0.1765
0.1194
0.0737
0.0351
0

1
0.7978
0.6137
0.4624
0.3390
0.2462
0.1728
0.1174
0.0720
0.0345
0

1
0.7980
0.6142
0.4632
0.3401
0.2473
0.1737
0.1181
0.0725
0.0419
0

1
0.7976
0.6134
0.4620
0.3386
0.2458
0.1724
0.1171
0.0717
0.0342
0

1
0.8010
0.6198
0.4667
0.3445
0.2493
0.1760
0.1190
0.0735
0.0350
0

1
0.8013
0.6203
0.4671
0.3448
0.2495
0.1761
0.1191
0.0736
0.0350
0

Table 2. Comparison between Mfree methods and analytical method in problem 1

x=0.5
256 nodes
Analytical Collocation LRPIM
y
Err
Err
Err
0.1
0
0.0051
0.0046
0.2
0
0.0119
0.0106
0.3
0
0.0126
0.0100
0.4
0
0.0183
0.0139
0.5
0
0.0168
0.0108
0.6
0
0.0232
0.0159
0.7
0
0.0193
0.0109
0.8
0
0.0271
0.0163
0.9
0
0.0256
0.1937
Total Err
0.018
0.0319

MWS
Err
0.0051
0.0119
0.0126
0.0183
0.0168
0.0232
0.0193
0.0271
0.0256
0.0178

1156 nodes
Collocation
LRPIM
Err
Err
0.0015
0.0009
0.0029
0.0016
0.0045
0.0026
0.0038
0.0012
0.0056
0.0028
0.0062
0.0028
0.0067
0.0034
0.0068
0.0027
0.0085
0.0028
0.005
0.0023

MWS
Err
0.0004
0.0008
0.0009
0.0009
0.0008
0.0006
0.0008
0.0014
0.0001
0.0007

Table 3. Errors in problem 1

5.1.4 Problem 2 with natural boundary condition
The MWS method is used to solve the same problem with both essential and natural
boundary conditions by 256(16x16) and 1156(34x34) nodes. The temperature distributions
for those problems are given in Fig.7. It should be noted that the essential boundary
conditions are satisfied exactly whereas the natural (Neumann) boundary conditions are
satisfied in the weak form formulation.

www.intechopen.com

244

Numerical Analysis Theory and Application

Fig. 6. Problem 2 and its boundary conditions

Fig. 7. Temperature distribution with 1156 nodes (Problem2 solved with MWS method)
In Tables 4 and 5 the LRPIM and MWS methods are compared with the analytical method.
The numerical values for the temperature distributions with 256 and 1156 nodes are also
given in Tables 4 and 5. The defined error equations (25 and 26) are used to show the
accuracy of MWS and LRPIM.

5.2 Lid driven cavity problem

In this section, the lid driven cavity problem is solved by meshfree method. In this problem,
the cavity is full of fluid and the upper plate in the cavity drive horizontally. It is shown that
the moving boundary conditions in the top wall are easily applied and natural boundary
condition can be satisfied.

www.intechopen.com

245

Meshfree Methods

x=0.5

256 nodes

1156 nodes

analytical

MWS

LRPIM

MWS

LRPIM

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1

1
0.8047
0.6271
0.4782
0.3606
0.2718
0.2071
0.1617
0.1320
0.1152
0.1098

1
0.8006
0.6196
0.4723
0.3538
0.2675
0.2026
0.1591
0.1293
0.1137
0.1081

1
0.8009
0.6202
0.4732
0.3550
0.2688
0.2039
0.1603
0.1304
0.1147
0.1089

1
0.8035
0.6254
0.4762
0.3589
0.2705
0.2059
0.1608
0.1314
0.1147
0.1093

1
0.8041
0.6262
0.4771
0.3598
0.2712
0.2066
0.1614
0.1318
0.1151
0.1097

Table 4. Comparison between Meshfree methods and analytical method in problem 2

x=0.5
y
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1

Analytical
Err
0
0
0
0
0
0
0
0
0
0
Total Err

256 nodes
MWS
LRPIM
Err
Err
0.0051
0.0047
0.0120
0.0110
0.0123
0.0105
0.0189
0.0155
0.0158
0.0110
0.0217
0.0155
0.0161
0.0087
0.0205
0.0121
0.0130
0.0043
0.0158
0.0082
0.0162
0.0108

1156 nodes
MWS
LRPIM
Err
Err
0.0015
0.0007
0.0027
0.0014
0.0042
0.0023
0.0047
0.0022
0.0048
0.0022
0.0058
0.0024
0.0056
0.0019
0.0045
0.0015
0.0043
0.0009
0.0046
0.0009
0.0046
0.0017

Table 5. Errors in problem 2

5.2.1 Formulation and boundary conditions of driven cavity problem
The application of Navier-Stokes equation in solving fluid flow has evolved in the past few
decades with meshfree method as one of the most adopted techniques. In this section the
Navier-Stockes equation is solved:

u
u
P 1
2u 2u
( 2 2)
v

Re x
x
y
x
y

(27)

The boundary conditions are shown in fig.8 and it is shown that three walls are without
motion and the upper wall move with the fix speed.

www.intechopen.com

246

Numerical Analysis Theory and Application

Fig. 8. Boundary conditions for driven cavity problem

5.2.2 Numerical results
The results from the solution of driven cavity problem when Reynolds number is 100 are
shown below:

Fig. 9. Stream line contours for Reynolds 100

The vortex on the corner is created and is related to Reynolds number and the corner vortex
will grow if Reynolds number increases. The vorticity and velocity contours are shown:
Location of large vortex
(0.6172,0.7344)
(0.6196,0.7373)
(0.6231,0.7460)
(0.5547,0.6055)
(0.5500,0.6125)
(0.5411,0.6005)

Minimum stream function

-0.0103
-0.0103
-0.0129
-0.114
-0.113
-0.114

Reynolds
100
100
100
400
400
400

Table 6. Minimum stream function and location of large vortex

www.intechopen.com

Reference
Ghia and Shin
Hou and Doolen
Present
Ghia
Gupta and Kalita
Present

247

Meshfree Methods

Fig. 10. Vorticity contours for driven cavity problem for Reynolds 100

Fig. 11. Horizontal velocity contours for driven cavity problem for Reynolds 100
In Table 6 the results are compared with the minimum stream function and the location of
large vortex. The results show the ability of meshfree methods to simulate the fluid
mechanics problems.
5.3 Cantilever beam problem
Numerical studies are conducted for a cantilever beam that is often used for benchmarking
numerical methods because the analytic solution for this problem is known. This problem is
a sample of solid mechanics.
5.3.1 Formulation of cantilever beam problem
The equilibrium equation is used with the formula:

ij , j bi 0

(28)

is the stress vector and is the body force vector components. The strain-displacement
relations are another formula that are brought in two directions:

www.intechopen.com

248

Numerical Analysis Theory and Application

Fig. 12. The beam problem

xx

u
x

(29)

is the strain component and u is the displacement in the horizontal direction.

yy

v
x

(30)

is the strain component and v is the displacement in the vertical direction.

The last equation is Hooks law:

De

(31)

De is the matrix of elastic constant.

5.3.2 Numerical results
The analytical solution is obtained for displacement of points of beam:

Py
D2
2
)
( 6L 3x )x ( 2 )( y
6EI
4

(32)

u is the displacement of points in horizontal direction and P is the force at the end of the
beam. E is the elasticity modulus and is the poisson ratio and moment of inertia is I, D is
the height and L is the length

Fig. 13. The beam after effect of force

www.intechopen.com

249

Meshfree Methods

The energy norm is defined to compare the results.

en

1
( ( numer exact )De ( numer exact )d
2

is the problem domain and

analytical solutions.

and

Energy norm

(33)

are the strain vector with numerical and

Solution method

No.

0.0258

MWS

0.026

LRPIM

Table 7. Energy norm for beam problem

These results show the capacity of MWS and LRPIM methods to solve the solid mechanics
problems. The Table 7 shows the errors are minimums with comparison with the analytic
solution.

6. Acknowledgement
This research has been mainly financed by Shiraz university. The author appreciate the
support from Shiraz university.

7. Conclusion
The meshfree methods are numerical methods that can be used to solve the many different
and complicated problems. The heat transfer problems, solid and fluid mechanics problems
have been solved with meshfree methods.
Three categories are used to solve the problems. Strong form methods, weak form methods
and weak-strong form methods(MWS) are meshfree categories. They can be used to solve
the problems with Dirichlet and Neumann boundary conditions. For examples the heat
conduction problem and lid driven cavity and cantilever beam are solved that they have
different type of boundary conditions. Solutions are related to many parameters: the
selected meshfree method, number of nodes, shape function parameters et cetera.
Nowadays, many changes are employed to different types of meshfree methods. The
advantages are improved and the high convergence rate and high accuracy are accessible.

8. References
Hou, S. , Doolen, G. & Cogley, A. (1995). Simulation of cavity flows by lattice boltzmann
method, Journal of computational physics, Vol.118, pp. 329-347
Gu, Y.T. & Liu, G.R. (2005). A meshfree weak-strong (MWS) form method for time
dependent problems, Computational Mechanics, Vol. 35, No.2 , pp. 134-145
Gu, Y.T & Liu, G.R. (2001). A Local Point Interpolation Method (LPIM) For Static And
Dynamic Analysis Of Thin Beams, Computer Methods in Applied Mechanics and
Engineering. Vol. 190

www.intechopen.com

250

Numerical Analysis Theory and Application

Hong, W.u. & Quan, W. (2007) Meshless method based on local weak form s for steady-state
heat conduction problems, International Journal of Heat and Mass Transfer Vol.51,
(2008) pp. 3103-3112
Incropera, Frank. & Witt, David.P. (2002). Introduction to heat transfer, 4th edition, springer
Liu, G.R. & Gu, Y.T. (2005). An Introduction to Meshfree Methods and Their Programming,
springer
Liu, G.R. , Wu, Y.L. & Ding, H. (2005). Meshfree weak-strong(MWS) form method and its
application to imcompressible flow problems, International Journal for Numerical
Methods in Fluids, Vol.46, pp. 1025-1047
Liu, G.R. & Gu, Y.T. (2003). A meshfree method: Meshfree Weak-Strong (MWS) form
method, for 2-D solids, Computational Mechanics, Vol.33, No.1, pp. 2-14
Liu, G.R. , Yan, L. , Wang, J.G. & Gu, Y.T. (2002). Point Interpolation Method Based On Local
Residual Formulation Using Radial Basis Functions, Structure Engineering Mechanic,
Vol.14, No.6, pp. 713-732
Liu, G.R. & Gu, Y.T. (2001). A local radial point interpolation method (LRPIM) for free
vibration analyses of 2-D solids, Journal of Sound and Vibration, Vol.246, No.1, pp.
2946
Liu, G.R. & Gu, Y.T. (2003). A meshfree formulation of local radial point interpolation
method (LRPIM) for incompressible flow simulation, Computational Mechanics
Vol.30, pp. 355365
Rao, S. (2004), The Finite Element Method in Engineering, Elsevier Science & Technology
Books.
Reddy, J.N. (2006), An introduction to finite element method, Third edition, Springer, McGrawHill Publishing Corporation
Zahiri, S. , Daneshmand, F. and Akbari, M.H. (2009). Using meshfree weak-strong form
method for 2-D heat transfer problem, Proceedings of ASME 2009 International
Mechanical Engineering Congress and Exposition, IMECE2009-12525, Lake Buena
Vista, Florida, USA, November 13-19, 2009

www.intechopen.com

Numerical Analysis - Theory and Application

Edited by Prof. Jan Awrejcewicz

ISBN 978-953-307-389-7
Hard cover, 626 pages
Publisher InTech

Published online 09, September, 2011

Published in print edition September, 2011

Numerical Analysis Theory and Application is an edited book divided into two parts: Part I devoted to
Theory, and Part II dealing with Application. The presented book is focused on introducing theoretical
approaches of numerical analysis as well as applications of various numerical methods to either study or
solving numerous theoretical and engineering problems. Since a large number of pure theoretical research is
proposed as well as a large amount of applications oriented numerical simulation results are given, the book
can be useful for both theoretical and applied research aimed on numerical simulations. In addition, in many
cases the presented approaches can be applied directly either by theoreticians or engineers.

How to reference

In order to correctly reference this scholarly work, feel free to copy and paste the following:
Saeid Zahiri (2011). Meshfree Methods, Numerical Analysis - Theory and Application, Prof. Jan Awrejcewicz
(Ed.), ISBN: 978-953-307-389-7, InTech, Available from: http://www.intechopen.com/books/numericalanalysis-theory-and-application/meshfree-methods

InTech Europe

University Campus STeP Ri

Slavka Krautzeka 83/A
51000 Rijeka, Croatia
Phone: +385 (51) 770 447
Fax: +385 (51) 686 166
www.intechopen.com

InTech China

Unit 405, Office Block, Hotel Equatorial Shanghai

No.65, Yan An Road (West), Shanghai, 200040, China
Phone: +86-21-62489820
Fax: +86-21-62489821