Finite Element Methods

AE-6121

By
Maj. Dr. Riessom W/Giorgis

FEM Dr. Riessom 1

 Personal Expectation
Who am I?
Who are you?
Short presentation of yourself
What do you expect from this course?
How do we cover the entire course within the stipulated
time?

FEM Dr. Riessom 2

 Course Delivery
I would like to keep the lesson interactive
Everyone of us has his/her own experience
I highly value your inputs, I think that will
benefit everybody
The idea is to interchange opinions and points
of view
Mutual learning

FEM Dr. Riessom 3

 Student Evaluation Scheme
As per the course syllabus the mark distribution is
supposed to be:
Mid-semester Exam 30%
Assignment and Lab 30%
Final semester Exam 40%
However, it is modified to be:
30% tests, 30% project and 40% final exam to suit
block delivery

FEM Dr. Riessom 4

 Course Description
Review of Mathematical preliminaries and solid
mechanics concepts, variational and weighted integral
formulations, shape functions, natural coordinate
system, discretization, minimum potential energy
principle, compatibility conditions, assembling,
elimination and penalty approachs, higher order
elements, finite element modeling, aspect ratio,
convergence criteria, Types of symmetry, application of
FEM to 2-D and 3-d problems in solid mechanics, FEM
for transient heat flow problems and vibrations, use of
software in FEM.

FEM Dr. Riessom 5

 Course Content
The entire course consists of the following major
topics:
1. Overview of Finite Element Methods
2. Finite Element Formulation
3. One-dimensional Analysis
4. Two-dimensional Analysis
5. Three-dimensional Analysis
6. Dynamic Analysis
7. FEM Software
FEM Dr. Riessom 6
 Course Objective
At the end of the course, it is expected that
everyone of you will have:
Sufficient knowledge of mathematical foundation of
Finite Element Methods
Skill in developing FEM computer programs for
engineering problems
Skill in using commercial FEM packages to model
and solve practical mechanical engineering problems

FEM Dr. Riessom 7

 Course Project

In this project you will be required to

Choose an engineering system of your interest
Develop the finite element model using Matlab
Solve the problem using commercial software
Present and discuss the results

Project grade will depend on

1. Originality of the idea
2. Techniques used
3. Critical discussion

FEM Dr. Riessom 8

Don’t say you don’t have enough time.
You have exactly the same number of
hours per day that were given to Helen
Keller, Pasteur, Michelangelo, Mother
Teresa, Leonardo Davinci, Thomas
Jefferson and Albert Einstein!
(H. Jackson Brown)

FEM Dr. Riessom 9

 Lecture -1

Introduction to FEM

FEM Dr. Riessom 10

Methods of Engineering Analysis
There are three approaches usually followed to undertake
any engineering analysis.
 Experimental methods
Accurate but it needs man power and materials. So, it is
time consuming and costly process.
Analytical methods
Quick and closed form solution but for simple geometries
and simple loading conditions.
Numerical or approximate methods.
Approximate but acceptable solution for problems
involving complex material properties and loading.
FEM Dr. Riessom 11
 Numerical Methods
The common three numerical methods are:
Functional approximation
Rayleigh-Ritz (variation approach) for complex
structural
Weighted residual method for solving non-structural
Finite Difference Method
For heat transfer, fluid and structural mechanics
This method is difficult to use when regions have
curved or irregular boundaries
Finite Element Method
FEM Dr. Riessom 12
Modeling a physical problem
Physical Problem Change physical
problem

Mathematical Improve
mathematical
Model model

Numerical model

No!
Does answer
Refine analysis
make sense?

YES! Design improvements

Structural optimization
Happy Riessom
FEM Dr. 13
 Finite Element Method
Finite element method is a numerical method for
solving problems of Engineering and Mathematical
Physics.
So, Finite element method can be viewed simply as a
method of finding approximate solutions for partial
differential equations.
Or as a tool to transform partial differential equations
into algebraic equations, which are then easily solved.
In this method the body is considered as an assemblage
of elements connected at a finite number of Nodes.
FEM Dr. Riessom 14
 History of Finite Element Methods
1941 – Hrenikoff proposed framework method
1943 – Courant used principle of stationary potential energy
and piecewise function approximation
1953 – Stiffness equations were written and solved using digital
computers.
1960 – Clough made up the name “finite element method”
1970s – FEA carried on “mainframe” computers
1980s – FEM code run on PCs
2000s – Parallel implementation of FEM (large-scale analysis,
virtual design)

Clough
Courant

FEM Dr. Riessom 15

Applications of Finite Element Methods

Structural & Stress Analysis

Thermal Analysis
 Dynamic Analysis
Acoustic Analysis
Electro-Magnetic Analysis
Manufacturing Processes
Fluid Dynamics
Financial Analysis

FEM Dr. Riessom 16

Aerospace Engineering

FEM Dr. Riessom 17

Civil Engineering

FEM Dr. Riessom 18

Electrical Engineering

FEM Dr. Riessom 19

Biomedical Engineering

FEM Dr. Riessom 20

The Future – Virtual Engineering

FEM Dr. Riessom 21

FEM Dr. Riessom 22
 Major FEM Analysis Steps
Discretization of the domain into a finite number of
subdomains (elements).
Selection of interpolation functions.
Development of the element matrix for the
subdomain (element).
Assembly of the element matrices for each
subdomain to obtain the global matrix for the entire
domain,
Imposition of the boundary conditions.
Solution of equations.
Additional computations
FEM Dr. Riessom
(if desired). 23
 Simple Example

Approximation the area of a circle

Observation:
Complicated objects can be represented by geometrically simple
pieces
FEM Dr. Riessom 24
 Finite Element Method
Continued…
1. Assume a trial Solution that satisfies the boundary
condition
2. The domain residual or error is calculated while
satisfying the differential equation
3. The weighted sum of the domain residual
computed over the entire domain is rendered zero.
4. The accuracy of the assumed trial solution can be
improved by taking additional higher order terms
but computations becomes tedious.

FEM Dr. Riessom 25

 Finite Element Method
Continued…
Therefore, it is not a trivial task to choose a single trial
function over the entire domain satisfying the
boundary condition.
It is preferred to discretize the domain in to several
elements ( finite elements) and use several piece wise
continuous trial functions, each valid with in a segment
( finite element ).
The trial functions at element level are the element
shape function which interpolate the field variable with
in the element from its value at nodes.
FEM Dr. Riessom 26
 Common Types of Elements
One-Dimensional Elements Two-Dimensional Elements
Line Triangular, Quadrilateral
Rods, Beams, Trusses, Plates, Shells, 2-D Continua
Frames

Three-Dimensional Elements
Tetrahedral, Rectangular Prism
(Brick)
3-D Continua

FEM Dr. Riessom 27

 Discretization
Any continuous solution field such as stress, displacement, temperature,
pressure, etc. can be approximated by a discrete model composed of a set of
piecewise continuous functions defined over a finite number of sub-domains.

Three-Dimensional
Two-Dimensional Brick Elements
One-Dimensional
FEM Triangular
Dr. Riessom Elements 28
Frame Elements
 Two-Dimensional Discretization
Refinement
2

(Node
) 
-1

-2

-3
4
(Discretization with 228
3.5
3 
3
2.5
2
2.5
Elements)

1.5
2 1
0.5
(Triangular
1.5 0
-0.5
1 -1
y
Element)
x

Approximate
Piecewise Linear
Representation

(Discretization with 912

Elements)

FEM Dr. Riessom 29

 FE Equation Derivation
The Finite element equation can be derived by either of
the following methods:
1. Direct equilibrium method
2. Minimum potential method
3. Weighted Residual method
The Galerkin Weighted Residual formulation is the
most popular from the finite element point of view.
Piece-wise trial function approximation of the weak
form of the GWR technique forms the basis of FEM.

FEM Dr. Riessom 30

 Direct Equilibrium Method
u1 u2

F1 F2
1 2
k
Equilibrium at Node 1  F1  ku1  ku2
Equilibrium at Node 2  F2   ku1  ku2

or in Matrix Form

 k  k   u1   F1 
 k    
 k  u2   F2 
Stiffness Nodal Force
Matrix [ K ]{u}  {F } Vector
FEM Dr. Riessom 31
 Elastic Spring System
nodal displacement

compression tension

In equilibrium :
and

in matrix form :

Here, element stiffness :

We want to find {u} - unknown

From :

inverse [ke ]
Solution is :
FEM Dr. Riessom 32
 System assembly in global coordinates :
global nodal displacement
element

node
global nodal force
" separate elements "
" system of 2 - springs "

" separate nodes "

FEM Dr. Riessom 33
 Spring and Bar Element Stiffness Matrices :

For element 1 ;

For element 2 ;

Define :
U i - global nodal displ.

FEM Dr. Riessom 34

 Spring and Bar Element Stiffness Matrices :

For element 1 ;

For element 2 ;

combined

Global system :

FEM Dr. Riessom 35

 Spring and Bar Element Stiffness Matrices :

Global system :

System stiffness matrix :

FEM Dr. Riessom 36

Spring and Bar Element Stiffness Matrices :

find U 2 and U 3  ?
Boundary condition :
and the data :

from :

FEM Dr. Riessom 37

Spring and Bar Element Stiffness Matrices :

substitute values :

from 1st eq. :

from two remaining eqs., we obtain

solve for
and also

FEM Dr. Riessom 38

 Spring and Bar Element Stiffness Matrices :

Back substituting into each element,

equilibrium in each element,

FEM Dr. Riessom 39

 Linear Spring Element

FEM
Determine vertical displ. of each W

From :
Stiffness matrix for each element ;

actual

FEM Dr. Riessom 40

 Linear Spring Element

Suppose :

FE eq. for each element ;

FEM Dr. Riessom 41

 Linear Spring Element

Fill into global system ;

(BC.)

Solve for :

and we obtain from

FEM Dr. Riessom 42

 Linear Spring Element

Determine displacement of each node & force at node 3  ?

where the specified conditions :

System equilibrium for finite element equation :

FEM Dr. Riessom 43

 Linear Spring Element

Apply known values :

rearranged

FEM Dr. Riessom 44

 Linear Spring Element

Partition the stiffness matrix :

where :- and

FEM Dr. Riessom 45

 Linear Spring Element

From the 2nd -row :

Solve for unknown displacements

where

thus,

FEM Dr. Riessom 46

 Linear Spring Element

From :

and using we obtain

From :

we also find

FEM Dr. Riessom 47

47
 Linear Spring Element

For each element :

FEM Dr. Riessom 48

Calculate the end displacement of bar
(a) using 1 element (b) using 2 elements

actual structure :

FEM Dr. Riessom 49

 Minimum (Stationary) Potential
Energy
Of all displacement states of a body or structure, subjected
to external loading, that satisfy the geometric boundary
conditions (imposed displacements), the displacement state
that also satisfies the equilibrium equations is such that the
total potential energy is a minimum for stable equilibrium.

Total potential energy  Strain energy  Work done

and

FEM Dr. Riessom 50

 Minimum (Stationary) Potential Energy

For a system having

Equilibrium equations (conditions)

This leads to N algebraic equations

FEM Dr. Riessom 51

 Minimum (Stationary) Potential Energy

(a) Form the global stiffness matrix

(b) Find the displacements and reaction

Given data :

FEM Dr. Riessom 52

 Minimum (Stationary) Potential Energy

Strain energy :

Work Done :
Total potential energy :

FEM Dr. Riessom 53

 Minimum (Stationary) Potential Energy

Apply minimum potential energy :

FEM Dr. Riessom 54

 Minimum (Stationary) Potential Energy

Rewritten in matrix form :

Global stiffness matrix :

Global displacement matrix :

FEM Dr. Riessom 55

 Minimum (Stationary) Potential Energy

Substituting known values :

Eliminating :

Solving unknowns :

and

FEM Dr. Riessom 56

 Problems Assignment
Form the global stiffness matrix using min. potential energy

Given data :

Determine F3 and displacement at node 3

FEM Dr. Riessom 57

 Weighted Residual Method
For structural problems, potential energy function can be
easily formed, Rayleigh-Ritz method is used.
For any problem where the differential equation of the
phenomenon can be easily formulated, Weighted
residual becomes very useful.
There are many types of weighted residual methods, of
them three are very popular.
1. Point collocation method
2. Least square method
3. Galerkin’ method
Among these three, the Galerkin approach has the widest
choice and is usedFEM
in Dr.
FEMsRiessom 58
Weighted Residual Continued…
Weighted residual method is a powerful way of finding
approximate solutions to a differential equations
L

O
 W ( x) R ( x)dx  0
i d

In point collocation, residuals are set to zero at n

different locations,
In least square, the integral of the weighted square of
the residual over the domain is required to be
minimum.
In Galerkin’s, the trial function itself is considered as
the weighting function.
FEM Dr. Riessom 59
Weighted Residual Continued…
Our interest is to find the solution of the differential
equation y. we assume approximate function for y. when
we substitute the approximate solution to the differential
equation, we can get the residual.

Where are unknown parameters of the assumed trial

function. The assumed trial function can be expressed as
The trial function y(x) must exactly satisfy the boundary
conditions.
The method of weighted resident needs the parameters to
be determined by satisfying the following equation.

FEM Dr. Riessom 60

 Example 1
Consider a uniform rod subjected to a uniform axial load as
shown in figure. The governing differential equation is as
follows:

Assume trial solution

The domain residual

Minimizing the domain residual

Then the solution becomes

FEM Dr. Riessom 61
 Example 2
A cantilever beam under uniformly distributed load. The
governing differential equation is :

The boundary conditions are

Assume trial solution

The domain residual

Minimizing the domain residual

Then the solution becomes

FEM Dr. Riessom 62
 i. Point collection Method
In this technique, residential are set to zero at n different
locations and the weighting function is denoted as

The xi are referred as collocation pirates and are selected by

the choice of the analyst.
But the term
Therefore

FEM Dr. Riessom 63

 ii. Least Square Method

In this method, the integral of the weighted square of the

residual over the domain required to be minimum

Where

The requirement is

FEM Dr. Riessom 64

 iii. Galerkin’s Method
In this method, the trial function, itself is considered as
the weighted of function that is

Substituting value

Boris Grigorievich Galerkin

( 1871-1945)

FEM Dr. Riessom 65

 Use Galerkin's method to find solution
ODE. : -

Because there is a quadratic term in ODE.,

the trial functions in polynomial form are suitable
where

Noted that it satisfies homogeneous BCs.

FEM Dr. Riessom 66
 For this problem, one chooses

The approximate solution is

in which

substitute this into ODE.

FEM Dr. Riessom 67

 The residual is

Thus,

After integration gives c1  4

The approximate solution is
To compare the results, it must determine the exact solution

Starting from :

FEM Dr. Riessom 68

Applying the BCs.

The exact soltuion is

FEM Dr. Riessom 69
Reanalyze by using 2-term Galerkin solution with

The 2-term approximate solution is

and

The residual is

From :

FEM Dr. Riessom 70

 From :

then,

solved for

The 2-term approximate solution is

FEM Dr. Riessom 71

Comparison Plot

FEM Dr. Riessom 72