1 FEM AND FDM DIFFERENCES IN A TABULAR FORM

Finite Element Methods

Module 3

Part 1:

In the last class we discussed some of the model problems e.g., Electricity, Mass Transport
etc., which we can solve by using the FEM. Next we have discussed the difference between
FEM and FDM. We have gone through several procedures. First we have shown how to
define FEM and FDM, then we have discussed several advantages and disadvantages of
both FEM and FDM. Finally, we have shown the discretizations graphically for a model
rectangular domain and cultivated the differences between FEM and FDM.

1 FEM and FDM differences in a tabular form

The basic difference between the working rule of the two methods is:

1
 2 MATHEMATICAL PRELIMINARIES

FDM FEM

In FDM, the equations generated are In FEM, the interpolation functions

only for the nodes. This method (shape functions) connect the area
moves from one node to another between the two nodes.
node.
In FDM, pointwise approximations FEM makes a piecewise approxima-
are made to the governing equations. tion i.e., it ensures the continuity at
The continuity along the sides of the the nodes, sides, and other areas of
grid lines are not ensured. the elements.

FDM does not give any approxi- FEM can give the values at any
mating function to evaluate the ba- point. The values obtained at points
sic values (like deflection) using the other than nodes are by using suit-
nodal value. able interpolation formulae.

FDM makes a stair-type approxima- FEM can consider the slopes and
tion to slopes and curved bound- curves exactly.
aries.
FDM needs larger number of nodes FEM requires fewer nodes.
to get good results.

FDM FEM

FDM can handle only fairly compli- FEM can handle all complicated
cated problems. shapes/geometries.

High complexity level for FDM. FEM has very less complexity level
to get solutions.

2 Mathematical preliminaries

The study of finite element analysis requires the basics of elementary mathematics and
it stretches to slightly sophisticated levels. Most of the concepts used are vector analy-
sis, matrix theory and differential equations. The review of vector analysis ranges from

2
 2 MATHEMATICAL PRELIMINARIES

elementary definitions to more advanced integral theorems. The matrix theory that is
covered consists of elementary definitions, matrix manipulations, the solution of simul-
taneous equations and a briefing on differential equations. They are required for solving
boundary-value problems and to compare the solutions with the numerical solutions.
Lastly, Cartesian tensors are also considered in FEM.

Vectors

A vector is defined as a physical quantity that can be described by a single magnitude

and a direction that is related to a coordinate reference frame. Physical quantities that are
arbitrarily directed in space can be resolved into orthogonal components corresponding
to the reference frame. Once the components are found, they can be manipulated using
standard algebraic operations.

Matrices

A rectangular array of numbers with a definite number of rows and columns is a matrix.
Once an array has been defined as a matrix, it has certain mathematical properties that can
be classified within the context of matrix theory. Comprehensive knowledge of matrix
theory is not required for finite element analysis; however, certain fundamental concepts
are necessary for the study of finite element theory and its subsequent application. The
array of numbers can be written in the abstract as
 
 a11 a12 a13 ··· a1n 
 
 a21 a22 a23 ··· a2n 

[A] = 

a31 a32 a33 ··· a3n 
 
 ··· ··· ··· ··· · · · 
 
 am1 am2 am3 · · · amn 
 
The notation [ A] is used to indicate a matrix. The terms within the matrix are called
elements, and when an element or a group of elements is referred to, a subscript notation
will be used, such as ai j , where i indicates a row number and j indicates a column number.

3
 2 MATHEMATICAL PRELIMINARIES

The above matrix is called an m by n matrix or simply an m × n matrix, and m × n is referred

to as the order of the matrix. A row matrix is defined as a 1 × m matrix and, similarly,
a column matrix is defined as an m × 1 matrix. The column matrix is often written {A}.
A matrix [A] and a matrix [B] can be added or subtracted, element by element, as long
as both are m × n matrices. Proper addition and subtraction are not defined for matrices
of unequal order. Matrix multiplication is the process of multiplying one matrix by a
second matrix and is written [A][B]; in general, [ A][ B ] , [ B ][ A]. In matrix multiplication
[ A][ B ], [ B ] is said to be premultiplied by [ A] or [ A] is said to be postmultiplied by [ B ].

The division, element by element, of one matrix by a second matrix is not defined.
However, the inverse of a matrix, written as [A]−1 , serves a similar purpose and will
be discussed later. The transpose of a matrix is obtained by interchanging its rows and
columns. The transpose of [A] is written [A]T , and in subscript notation, the element
interchange is
[ ai j ]T = [ a ji ]

It follows that
[[A][B]]T = [B]T [ A]T

A symmetric matrix is defined as a square matrix with the property ai j = a ji for i , j.

A diagonal matrix has all elements of a square matrix equal to zero except those on the
principal diagonal, which is the diagonal from upper left to lower right. A unit matrix is a
special case of a diagonal matrix; all diagonal elements are equal to 1, and all off-diagonal
elements are equal to 0.

Part 2:

Determinants

An understanding of selected topics from the theory of determinants is necessary for the
successful solution of the simultaneous equations that result in finite element analysis.

4
 2 MATHEMATICAL PRELIMINARIES

Determinants of order 2 or 3 can generally be used to illustrate all the concepts required
for understanding the manipulation of determinants. A determinant is usually symbolized
by enclosing the array of numbers within vertical lines rather than brackets. For a matrix
denoted [ A], the notation for the determinant might be |A|, det[A], or | det A| and indicates
the determinant of the matrix [ A].

Every determinant has a determinantal equation. For higher-order determinants, it

can be quite formidable from a computational standpoint to obtain that equation. The
determinantal equation of a determinant of order 3 is obtained as follows:

a11 a12 a13

| det A| = a21 a22 a23

a31 a32 a33

= a11 a22 a33 + a12 a23 a31 + a13 a32 a21 − a11 a23 a32 − a21 a12 a33 − a31 a22 a13

and can be described as the product of the principal diagonal terms minus the product of
the secondary diagonal terms. This elementary concept can be applied to determinants of
order 2 or 3 but fails for higher-order determinants.

The minor of a determinant is the determinant that remains after a row and a column
is removed from the original determinant. The minor can be referenced to a particular
element of the determinant using the notation ai j . The minor of |a22 |

a11 a13
|M22 | =
a31 a33

a determinant of order 2. The cofactor of an element of a determinant is defined as

Ci j = (−1)i+ j Mi j , where Mi j is the minor of the element ai j .

Simultaneous Equations

A set of simultaneous equations can be written in matrix form as

[A] {x} = {f}

5
 3 DIFFERENTIAL EQUATIONS

The matrix [A] represents the matrix of coefficients that are multiplied by the unknown
quantities {x}. The column matrix on the right-hand side contains the known quantities
f . Multiplying by the inverse of [ A] gives

[ A]−1 [ A] {x} = [I] {X} = {x} = [A]−1 {f}

The use of the inverse for solving a set of simultaneous equations is inefficient for large
sets of equations.
Gaussian elimination: A method that is sometimes called Gaussian elimination is
faster and hence more efficient than the inverse method. Gaussian elimination is an
organized method of substituting each equation into the previous equation until the last
equation contains only one unknown. The unknowns are determined sequentially, starting
with the last equation and proceeding upward. The method is sometimes referred to as
upper triangulation.

Part 3:

3 Differential Equations

Finite element analysis is a method for the numerical solution of a differential equation.
It follows that without differential equations there would not be a finite element method.
Many practicing engineers and scientists learned the finite element method as an applica-
tion of structural analysis for civil engineering or aircraft structures. The state of the art of
finite element analysis several decades ago was responsible for that situation. The classical
stiffness method of structural analysis can be derived without mention of the governing
differential equations. That is, the fundamental relationships for deriving the stiffness
method are based upon solutions of differential equations, but the user can easily lose
sight of the origin of the analysis. Finite element analysis for beam and frame structures
can be based upon energy theorems without considering differential equations. Again,
the fault is not with the engineer or scientist, but historically the connection between
energy methods in structural analysis and the governing differential equation has not been

6
 4 STEPS INVOLVED IN FEM

emphasized. The differential equation is associated with the corresponding variational

function (energy theorem). The finite element method can be derived in a variety of ways,
but regardless of the derivation, the method is the numerical solution of a differential
equation.

4 Steps involved in FEM

Finite element analysis typically involves the following steps. We will cite stress analysis
and heat transfer as typical applications. Steps 1, 4 and 5 require decisions by the analyst
and provide input data for the computer program. Steps 2, 3, 6 and 7 are carried out
automatically by the computer program.

1. Divide the structure of the continuum into finite elements. Mesh generation program,
called preprocessors, which help the user in doing this work.

2. Formulate the properties of each element. In stress analysis, this means determining
nodal loads associated with all element deformation states that are allowed. In
heat transfer, it means determining nodal heat fluxes associated with all element
temperature fields that are allowed.

3. Assemble elements to obtain the finite element model of the structure.

4. Apply the known loads: nodal forces or moments in stress analysis, nodal heat fluxes
in heat transfer (Boundary conditions).

5. In stress analysis, specify how the structure is supported. This step involves setting
several nodal displacements to known values (which often are zero). In heat transfer,
where typically certain temperatures are known, impose all known values of nodal
temperature.

6. Solve simultaneous linear algebraic equations to determine nodal degrees of freedom

(d.o.f). These d.o.f are known nodal displacements in stress analysis and nodal
temperatures in heat transfer.

7
 4 STEPS INVOLVED IN FEM

7. In stress analysis, calculate element strains from the nodal d.o.f and in heat transfer,
calculate element heat fluxes from the nodal temperatures. Output interpretation
programs, called postprocessors, help the user sort the output and display it in
graphical form.

Remark 4.1. The power of the finite element method resides principally in its versatility.
The method can be applied to various physical problems. The body analyzed can have
arbitrary shape, loads, and support conditions. The mesh can mix elements of different
types, shapes and physical properties. This great versatility is contained within a single
computer program. User prepared input data controls the selection of problem type,
geometry, boundary conditions, element selection, and so on.