FEM Dr.

TN

FINITE ELEMENT METHOD

1. INTRODUCTION
FEM is anumerical technique, which can be used for any field of engineering problems such as
stress analysis, fluid flow, heat and mass transfer, lubrication, combustion, electromagnetic etc.
With the advances in computer technology any kind of complex problems can be made easy using
FEM. geometric
In FEM, a complex region defining a continuum or solution domain is divided into simple
at joints, which
shapes called finite elements as shown in Fig. 1. These elements are interconnected
are considered over these
are called nodes. The material property and governing equations assembly
elements and expressed in terms of unknown variables at element nodes. Then, an
equations
process results into a set of simultaneous equations. The solution of these simultaneous
gives the approximate behavior of the solution domain.
BOUNDARY OF REGION
INTEREST

TYPICAL NODE

TYPICAL ELEMENT

Fig. 1: Two-dimensional region divided into finite element

2. FEM Vs Classical Methods

1 In classical method exact equations are formed and exact solutions are obtained whereas in
FEM exact equations are formed but approximate solutions are obtained.
2. Solutions can be obtained only for few simple cases by classical method whereasit can be
obtained for all problems by FEM.
3. Whenever the complexities in shape, material, boundary conditions and loading as shown in
Fig.2 are occurs in the problems, classical method provides the solution with drastic
assumptions whereas in FEM no such assumptions are made.
4. The solutions for the problems with anisotropic material properties become very difficulty in
classical method. FEM can handle problems with anisotropic material properties without any
difficulty.
5. Ifthe structure consists of more than one material, it isdifficulty to use classical method but FEM
can be used with out such difficulty.
6. Problems with material andgeometric non-linearities cannot be handled by classical method but
there is no such difficulty in FEM.

Boundaries

(a) Irregular Shape (b) Irregular Boundary Condition (c) Irregular Loading
Fig. 2
 FEM Dr. TN
3. Finite Difference Method (FDM)
In finite difference method, the body or the system is discretized by a mesh of nodal poi
shown in Fig. 3. The field variable is represented by the discrete values of the variable at the no
For example, in case of plate bending problem the normal displacement wwill be taken as unknown
variable at each one of the nodes. The governing differential equation and the boundary conditions
are converted to finite difference form.
The finite difference form of governing differential equation and boundary conditions are then
applied to each of nodes. This will yield a set of linear simultaneous equations in the discrete field
variable w. The resulting simultaneous equations are solved for the nodal values of this variable.

Fig.3: Discretizationof region by mesh for finite difference method

4. FEM Vs FDM
1. FDM makes point wise approximation to the governing equations i.e. it ensures continuity only
at the node points but continuity along the sides of grid lines are not ensured.
FEM makes piece wise approximation i.e. it ensures continuity at note point as well as along the
sides of the element.
2. FDM does not give the values for the unknown variables at any point except at node points. It
does not give any approximating function to evaluate the basic values (deflection in case of
structural problems or temperature in case of heat transfer problems) using the nodal values.
FEM provides values for the unknown variables at node points and nodal variables can be
interpolated using suitable function to get the values at any point other than nodes.
3. irregular
FDM makes stair type approximation to irregular boundaries as shown in Fig.3 whereas in FEM
boundary can be discretized exactly using suitable elements.
4. FDM needs more number of nodes to get good results whereas FEM needs less number of
nodes.
5. Complicated problems such as irregular shapes, material properties, boundary conditions and
loading can't be handled by FDM but these problems can be handled by FEM.
5. Advantages and Disadvantages of FEM over other Numerical Techniques such as Finite
Difference Method (FDM):
Advantages
1. This method can be applied to any engineering problems, which are complex in nature.
2. This method can be systematically programmed to accommodate complex and difficult problems
having non-homogeneous materials, non-linear stress-strain behavior and complicated
boundary conditions.
3. This method posses variety of levels at which we may develop an understanding of the
technique i.e. the learner can understand this technique through simple one-dimensional
structure problem and can extend it to two- and three- dimensional problems.
4
The finite element mesh can mix elements of different type, shapes and physical properties.
5. element
User prepared input data controls the selection of problem type, geometry, boundary condition,
selection etc.
6. There is aclose physical resemblance between the actual structure and its finite element model.
Disadvantages
1. Acomputer., a reliable program and intelligent user are essential to get correct
result.
2. A general-purpose program has extensive documentations which cannot be ignored.
3. An experience and good engineering judgments are needed in order to define a good model.

2
 FEM Dr. TN

4. Many input data are required and voluminous output datamust be stored and understood.
6. Applications of FEM
from deformation and stress analysis of automotive,
Tne applications of FEM, in general, ranges analysis such as heat flux, fluid flow, magnetic flux,
field
aircraft, building and bridge structures to examples important
of specific applications of FEM for the
Seepage and other flow problems, Some
boundary value problems are as listed below:
engineering.
1. Structuralengineering, structural mechanics and aerospace
Ex.:(a) Analysis of beam, plate and shell structures.
(b) Analysis of complex structures. analysis.
(c) Two and three -dimensional stress
(d) Torsion of prismatic bar.
engineering and rock mechanics.
2. Soil mechanics, foundation stress analyses.
Ex.: (a) Two and three - dimensional problems.
(b)Construction and excavation
(c) Soil structure interaction.
bore wells, etc.
(d) Analysis of dams, tunnels, in soils and rocks.
seepage of fluids
(e) Steady state
3. Heat conduction.
distribution in solids and fluids.
Ex.: Steady state temperature engineering and water resources.
4. Hydrodynamic, hydraulic of fluids.
Ex.: (a) Solution for potential flow fluids.
(b) Solutions for viscous flow of and dams.
structures
(c) Analysis of hydraulic
5. Nuclear engineering,
containment structures.
Ex.: (a) Analysis of reactor distribution in reactors and reactor
structures.
(b) Steady state temperature

7. Some Definitions
domain ():
1. Continuum or Solution
being analysed.
Continuum is the physical body,structure or solid

x=0 x=l

Q0sxslh
Q0<x<l ’0sys,
Fig. (a): 1-D solution domain Fig. (b): 2-D solution domain

2. Field variable: Field variable is any variable such as

displacement, pressure, temperature,
velocity component etc.

3
 Dr. TN
FEM
called nodal variables. Generaly
3. Nodal variables: The values of field variables at nodes are
hence the analysis of FEM concerned
the nodal variables are the unknown of the problem and
with finding nodal variables.
the specified values of the field variables (or
4. Boundary conditions: Boundary conditions are
There are two types of boundary
their derivatives) on the boundaries of the problems.
boundary conditions and (2) Non
conditions. They are (1) Essential or geometric or kinematic
essentialor natural or forced boundary conditions.
boundary conditions are the prescribed values of
(1) Essential boundary conditions: The essential
boundary of the region.
the variables specified (or their first derivative) on the
shown in Fig. is used, the deflection, wand slope
For example, if a standard cantilever beam
dw
=0 are known as
dw
iszero at x=0(.e at fixed point). These
boundary conditions, w=0and dx
da

essential boundary conditions.

-ºX

boundary conditions are the prescribed

(2) Non-Essential boundary conditions: Non-essential
values of higher derivatives of the field variables.
non-essential bounday conditions are its
For example, in the cantilever beam shown, the
x= L)and are expressed as
bending moment and shear force at its free length (i.e. at
'w force EI d'w =0.
At x=L, BM, EIdu? 0 and shear

5. Boundary value problem:

obtained in the region of the body while on the
Boundary value problem is one in which a solution is
dependent variables (or their derivatives) are
boundaries (or edges) of the region the values of the structural
some times called ield problems.
prescribed. Boundary value problems are also
are the boundary value problems.
problems, soil mechanics problems, thermal problems, etc uniformly
shown in Fig. Here 1D bar is subjected to a
As an example consider a structural problem
concentrated load, F at its free end. The governing
distributed load, P, along its length and a
as
differential equation for this problem can be expressed
d ,du
AE-, =0 for 0<x<L --(a)
ds dx )
and the boundary conditions are
u=0 at x=0and AE=F at x=L --- (b)
dx
governing equation such as Eq. (a) and boundary
The problems which are represented by a
problems.
conditions given by Eq. (b) are referred as boundary value
4
 FEM
6. Initial value Dr. TN

problem:
Initial value problem is one in which the dependent variables (or their
derivatives) specifiedare
aly (1.e. at time t = 0). Initial value problems are
generally time dependent problems. Vibration
allaiysis of a body is an example for the initial value problem. As
an example the governing
almerentialequation of a spring-mass system subjected to excitation force, F) is
expressed as
m d'u -+ku = F()
--- (a)
with initial condition
du
U=u, and = V% when, time t =0. --- (b)
dt

i.e the displacement of mass from its equilibrium position and its velocity when time =0 are u,
and v, respectively. Those problems represented by such a governing Eq. (a) and such initial
conditions shown in eq. (b) are referred as initial value problems.

7. Degrees of free dom per node: It is the number of nodal variables required to determine per
node.

8. Degrees of freedom per element: It is the number of nodal variables required to determine per
element.
Let n =No. of d.o.f per node
nge =No. of d.o.f per element.
n, = No. of nodes per element.
Then, nd = n,* ndn
(1)For 1- Dproblem,
U3

1 2 3

Fig. 4(a): 2-noded bar element Fig.4(b): 3-noded bar element

1-D.0.F per node 1- D.0.F per node
2-D.O.F per element 3D.O.F per element
(2)For 2- D
problem,
V3
3 U3
2-D.O.F per node
V2
6-D.O.F per element
-W2
2

Fig.4(c):3 - nodded triangular element

5
 FEM
Dr, TN iWECI
nnovatNHOSion
9. Equilibrium equation of an elastic body:
the internal forces with that of external fore
The equilibrium equation of an elastic body relates equation is usually derived from the principle
under the equilibrium of the body, Equilibrium
minimum potentialenergy storer
potential energy of a loaded elastic body is given by the sum of the internal energy
The total
as result of deformation and potential energy of the external load. This column can also be
Consider a column loaded as shown in Fig. 5(a) and having deformation u. deformation is as shown
in Fig. 5(b), The spring force v/s
by a linear spring of stiffness K as shown
in Fig. 5(c).

PE=- Ku'
2
| Ku

(b) (c)
(a)
Fig. 5: Column and equivalent spring
TotalP.E, I =U+ Wp

Where, U= Ku-, Strain energy or internal potential energy

2
load
And Wp =-Pu,Potential energy due to external the capacity for
equation is because the force loses some of
Note that negative sign in the above
direction it acts.
doing work when it is displaces in the
Pu --- (a)
Therefore,total P.E, I=Ku-
2
the total P.E is to be minimum.
For the body to be in state of equilibrium,
=0
Therefore, for total P.E to be minimum, Ou
Differentiation of Egn. (a) w.r.t u yields,
Ku-P=0
or Ku=P is the equilibrium equation.
the spring and P= External force.
Where, Ku=Internal force exerted by having equilibrium
spring or body is displacement as shown in Fig. 4(d), the
If both end of the
equation becomes, in matrix form, as
K1 K2|
K21 K,Ju, P nodes iandj can be written as
for an e" body or element having
In general,the above equation Pi
P --- (b)

where,
K=Element stiffness matrix
g= Element nodal displacement vector and Fig. 4(d) PÊ

6
 FEM

Dr. TN
F=Element nodal force vector.
The elements of stiffness matrix in Eqn. (b) are
coefficient K,'is defined as the force reguired at called
as
influence coefficients. The influence
node to hold that node fixed when a unite
displacement at node j is given. On the other hand, the coefficient'K, is the force
Tto give unite displacement at same node '. required at node
These coefficients can be expressed in terms of area
OT Cls of the elastic body, length and material
given by property i.e. Young's modulus of the body and are
K =K = A,E, and K = K =
Then, Egn. (b) becomes as

,is the equilibrium equation for an e" element.

8. Finite Element Method Approaches

Majority of the application of FEM is in the area of solid mechanics and hence the problems in these
fields are usually tackled by one of the following three approaches:
1. Displacement method,
2. Equilibrium or Flexibility method and
3. Mixed method.

Displacements are assumed over the element as primary unknown quantities in displacement
method; stresses are assumed over the element as primary unknown quantities in equilibrium
method and some displacements and some stresses are assumed as primary unknown quantities in
the mixed method.
9. Steps Involved in FEM
The following are basic steps used in FEM. They are
1. Discretization of the continuum: Discretization is the process of subdividing a given
continuum into finite elements as shown in Fig. 1. The finite elements may be triangular or
quadrilateral etc. for a two-dimensional problems; they may be tetrahedron or rectangular prism etc.
for three-dimensional problems. The selection of finite elements in FEM is depends on the following
factors. They are:
1. Dimension of the problem.
2. Shape of the body or continuum.
3. Accuracy of the results required.
4. Cost and time for the analysis.
Fig. 7 shows some of typical finite elements used in FEM.

A
(a) Line element (b) 3-noded triangle (c) 4-noded quadrilateral
for 1- D problem for 2 D problem for 2 D problem

7
 Dr. TN

FEM

(d)4-noded tetrahedron (e)8-noded rectangular prism

for 3-D problem
for 3-Dproblem
Fig. 7: Typicalfinite elements
in Fig. 8 (a)
procedures, consider a stepped column shown
Toillustrate the concept of step-by-step the discretization of the stepped column into two elements.
as an example problem. Fig. 8 (b)shows
E =E, = 2*10'Nlm'.
Let h =, = lm; A= 20m, A, =10m and
A,E
100 N
It0-2

Fig. 8(a): Stepped column Fig.8(b): Discretization

2. Selection of displacement model: Simple functions are assumed to approximate the

displacements over each element. These functions are called displacement models or displacement
functions. The trigonometric or polynomial functions are assumed for these displacement models.
The polynomial functions are much easier for mathematical operation and hence they are widely
used in FEM.
The selection of displacement model is based on the following three factors;
1. The type and degree of the displacement model.
2. The particular displacement magnitude that describe the model.
3. The model should satisfy certain requirements.
3. Derivation of element stiffness matrix: The stiffness matrix consists of the coefficients of the
equilibrium equations derived from the material and geometric properties of an element. This matrix
relates the displacements at the nodal points (nodal displacement) to the nodal forces.
In general, using stiffness matrix the equilibrium equation for an e" element shown in Fig. 9 can be
written as

where,
element stiffness matrix

A,JE,
nodal displacement vector and u

(F=P P. nodal load vector.

Fig. 9: Typical element
The element stiffness matrices for the example problem considered are:

8
d 20

Dr. TN
FEM

For element-1: A, E 20* 2*10'! -4*102 Nm

1 2

10*2*10!! =2*10l2 Nm
For element-2

2 3

Therefore, K?=10 |-2 23

process includes the assembly of the overal
4. Assembly of algebraic equations: The assemblyindividual element stiffness matrices and the
or global stiffness matrix for the entire body from the
global load vector from the element nodal force vector.
stiffness matrix, global displacement
In general, the overall equilibrium equation between the global
vector and global load vector is expressed as
[KO}-(F} K can be assembled using the
For the example problem considered, the global stiffness matrix
following connectivity table.
Element No. Local node Nos. Global node Nos.
1 1
1

1
2
2 3

Write the global node numbers for the corresponding element stiffness matrices and assemble the
matrices we get global stiffness matrix as
1 2 3
[4 -4 01
[K]=10'2-4 4+2 -2|2
0 -2 23
The global nodal force vector becomes as
R Reaction at fixed node I
0 As there is no external load at node 2
100 As external load at node 3 acts in - ve x - direction
Globalnodal displacement vector becomes as

Using the above three matrices, the global equilibrium equation can be written as
 FEM Dr. TN

R
--- (A)
100

5. Boundary conditions: After assembling the elemental equilibrium equations, the boundary
conditions are to be applied and the overall equilibrium equation must be modified depending upon
the available boundary conditions. The overall equilibrium equation can be modified for given
boundary conditions using one of the following methods.
1. Elimination method.
2. Penalty method and
3. Multiconstraint method.
In our example problem, as node 1 is fixed, U,=0. Therefore, using elimination method the overall
system equation i.e. Eqn. (A) can be modified by eliminating 1 row and 1 column. It is given by
--- (B)

6. Solution of the unknown displacement: The simultaneous equations obtained in step 5 are
solved for the unknown displacements. Large sets of simultaneous equations are solved using
Gauss elimination method.
For the example problem, the solution of Eqn. (B) yields nodal displacements u2 and u3. They are
obtained as follows:
6*10u, -2 *10'Uz =0
-2*10u, +2 *10uz =-100
Adding two Eqs. 4*1o'u, =-100
Therefore, u, =-25 *107l m. Substituting uz into any one of the above Eqns. we get
W3 =-7.5*10-12 m.
Note: The reaction at fixed node 1 is obtained by substituting the values of nodal
displacements
uj,u2, Uy 0nto 1 equation of Eqn. (A).
ie. 4*10u -4*10u, =R
or R=-4*10(-25*10l2) = +100() since u, =0
7. Computation of element strains and stresses: Using the nodal displacements, the elemental
strains and stresses can be determined as follows:
In general, the strain in e" element can be expressed as

le
Then,the stress is expressed as o, = E,e
Therefore. the strains and stresses for the problem considered are:
-25 *10-12-0
For element 1:Strain. E. = "2 =
:-25*10(compressive)
Stress, a = E, *a =2*10'" *-25* 10-2=-5 Nim?
-75*10-12 -(-25*10)
For element 2: Strain, E = =-50 * 10- (compressive)
Stress, o¡, = E, *E, =2*10'*-50*10-12 =-|0 N/m²
10