Sie sind auf Seite 1von 5

AE618A

(FINITE ELEMENT METHOD FOR FLUID DYNAMICS)


PROF. SANJAY MITTAL
TERM PAPER

ELEMENT FREE GALERKIN METHOD

SUBMITTED BY
MIT SHAH (12647)
SUYASH HARLALKA (12747)
AAKASH KARDAM (12004)
Abstract:

COMPUTERS have enhanced our abilities to simulate systems with great depth of sophistication and have indeed
contributed in the development of Scientific Computing Community. Problem Solving has mostly become
numerical dependent on simulation techniques in order to counter engineering problems. Albeit the techniques
like FDM (Finite Difference Method), CFD (Computational Fluid Dynamics) and FEM (Finite Element Method)
provide solutions to a plethora of problems, we are tormented by difficulties owing to their limitations in
application to complex problems and complex geometries. One of the integral part of all methods cited above is
the use of grids (mesh). Grid generation is itself a tedious and demanding job. To overcome such difficulties,
people have moved towards Mesh Free Methods. One such method is ELEMENT FREE GALERKIN (EFG) method.
EFG is a method that uses only a set of nodes and description of models boundary to generate discrete equations
and hence described as meshless. It can also be coupled with finite element method which allows the usage in
the study of crack region and FEM to model the remainder of the problem. Hence it provides us with a more
comprehensive study of 2-D and 3-D fracture mechanics. We aim to discuss EFG & its effective applications where
it has an upper hand as and when compared to mesh-methods.

Introduction:

The key idea of the mesh free methods is to provide accurate and stable numerical solutions for integral
equations or PDEs with all kinds of possible boundary conditions with a set of arbitrarily distributed nodes (or
particles) without using any mesh that provides the connectivity of these nodes or particles. One important goal
of the initial research is to modify the internal structure of the grid-based FDM and FEM to become more
adaptive, versatile and robust. Since there is no need to create a mesh, and the nodes can be created by a
computer in a fully automated manner, it saves a lot of human effort. The principal attraction of mesh free
methods is the possibility of simplifying adaptivity and problems with moving boundaries and discontinuities,
such as phase changes or cracks. In crack growth problems, for example, nodes can be added around a crack tip
to capture the stress intensity factors with the desired accuracy; this nodal refinement can be moved with a
propagating crack through a background arrangement of nodes associated with the global geometry. Mesh less
methods do not require us to define the elements and the connectivity. Only need to define is the node
distribution.

Support Domain

Now let us understand the concept of support domain. In Mesh less methods the displacement at a point x will
be an interpolation of the displacement of the nodes which lie in the support domain of the point.

Meshless method are divided into 4 categories:

Methods based on Global weak form

This method requires background cells for integration of the system matrices. These methods are therefore
pseudo mesh free methods. Example is Element Free Galerkin Method.

Methods based on Local weak form


This method requires background cells locally for integration of the system matrices. These methods are
therefore pseudo mesh free methods. Example is meshless local Petrov-Galerkin method

Methods based on Strong form

These methods do not require mesh at all. The examples are Local Point Collocation method and FDM.

Particle methods

These methods require predefinition of particles for their volume and mass. The example is Smoothed Particle
Hydrodynamics (SPH).

The Element Free Galerkin (EFG) Method was developed by Belytschko et al. based on Diffuse Elements Method
developed by Nayroles et al. EFG method has some major features which are as follows,

Moving Least Squares approximations is implemented to construct the shape functions


Galerkin Weak form with constraints to apply the essential Boundary conditions is used to develop the
discrete equation system equations.
EFG is a pseudo mesh free method as it requires background cells for performing numerical Integrations
to construct the system matrices

Moving Least Squares (MLS) Approximation:

MLS originated for data fitting and surface construction. The MLS method is used to create mesh free shape
functions for approximations. The two important features of MLS that makes it popular are

The approximated field solution is continuous and smooth in the entire problem domain (when sufficient
nodes are used)
It is capable of producing an approximation with a desired order of consistency.

Following is the procedure for constructing shape functions for mesh free methods using MLS approximation.

Let () the function of the field variable in the domain , the approximation of () is denoted by () . In
MLS we approximate the field function in the form of series representation as follows

() = () () ()()
=1

Where,
m is the number of monomials (polynomial basis)
() is the vector of coefficients given by
() = {0 () 1 () . . . . 1 () ()}
Here 0 () are functions of x.
Here () is a vector of basis functions that consists most often of monomial of the lowest order to ensure
minimum completeness. Considering pure polynomial basis function. In 1D complete polynomial basis of order
m is given by
() = {0 () 1 () . . . . 1 () ()} = {1, , 2 , . . , }
In 2-D space it is given by

() = (, ) = {1, , , , 2 , 2 . . , , }
And we can generalize it for 3-D.
Assuming that the support domain of x contains a set of n local nodes {1 , 2 . . . . 1 , }
The approximated value of the field function can be calculated as (the contribution of point in the support
domain of any point to the displacement field at the point )


(,
) = ( ) ()
=1

We should note that () is an arbitrary function of x. A functional of weighted residual is then created using the
approximated values of the values of the field function and the nodal parameters = ( )

= ( ) [ (, ) ( )]2
=1

= ( ) [ ( )() ]2
=1

Where ( ) is a weight function. Weight functions play their role effectively when sufficient nodes are used
i.e. (n (number of nodes) >>m (number of elements in the basis)). We now minimize the weight functions w.r.t
the co-efficient matrix


=0

This gives us the following expression of the coefficient vector

() = 1 ()()

Where A is called the MLS moment matrix given by

() = () ( )( )
=1

Where () ( ) and () has the form

() = {1 () 1 () . . . . 1 () ()}
= ()( )
Where is the vector of nodal parameters of the field variables for all nodes of the support domain.
Therefore the approximated displacement field can be computed as follows

()
= () (1 ()()
=1 =1

This can be written as



()
= ()
=1

Where () is the MLS shape function defined by


() = () (1 ()() = 1
=1

The shape functions obtained by MLS approximation do not satisfy the kronecker delta property meaning the
value of the displacement at node I is not equal to . This leads us to another important thing about applying
the essential boundary conditions on the weak form since we cannot apply boundary conditions directly to .
Therefore we use Lagrange multipliers to enforce the essential boundary condition.

Conclusions:

Element free Galerkin method is studied. The shape functions used in element free Galerkin method are studied
. Moving Least Square methods have been discussed and its application to get the shape function has been
shown. The boundary condition implementation using Lagrange multiplier has not been discussed.

The take home messages are :

Our ultimate aim is to be find () () = =1 ()


By minimizing the weighted residual ( = 0) we get our co-efficient matrix and then be again

resubstituting a(x) in the expression for u(x) we get ()

Shape functions in EFG do not satisfy kronecker delta property

This method takes more time and a lot of programming effort but it helps us to better analyse the crack
propagation problems

References:

Introduction of Mesh free Methods and Implementation of Element Free Galerkin (EFG) Method to Beam
Problem, Someshwar S. Pandey1, Paresh K. Kasundra2 & Sachin D. Daxini3
N. Sukumar, B. Moran, T. Black, T. Belytschko, An element-free Galerkin method for three-dimensional
fracture mechanics
Meshfree Methods: Moving Beyond the Finite Element Method, G.R. Liu

Das könnte Ihnen auch gefallen