FINITE ELEMENT METHOD

Dr. NGUYỄN HỒNG ÂN

Head
Department of Mechanics of Materials & Structures
Tel: 0909.48.58.38
Email: anhnguyen@hcmut.edu.vn
 Chapter 1: INTRODUCTION

• Mathematic Model
• Finite Element Method
• Historical Background
• Analytical Process of FEM
• Applications of FEM
• Computer Programs for FEM
1. Mathematical Model
(1) Modeling
Physical Mathematica Solution
Problems l Model

Identify control variables

Assumptions (empirical law)

(2) Types of solution

Sol. Eq. Exact Eq. Approx. Eq.

Exact Sol. ◎ ◎
Approx. Sol. ◎ ◎
(3) Methods of Solution
(3) Method of Solution

A. Classical methods
They offer a high degree of insight, but the problems are
difficult or impossible to solve for anything but simple
geometries and loadings.
B. Numerical methods
(I) Energy: Minimize an expression for the potential energy of the
structure over the whole domain.
(II) Boundary element: Approximates functions satisfying the
governing differential equations not the boundary
conditions.
(III) Finite difference: Replaces governing differential equations
and boundary conditions with algebraic finite difference
equations.
(IV) Finite element: Approximates the behavior of an irregular,
continuous structure under general loadings and constraints
with an assembly of discrete elements.
2. Finite Element Method

(1) Definition

FEM is a numerical method for solving a system of

governing equations over the domain of a
continuous physical system, which is discretized
into simple geometric shapes called finite element.

Continuous system Discrete system

Time-independent PDE Linear algebraic eq.
Time-dependent PDE ODE
(2) Discretization
Modeling a body by dividing it into an equivalent
system of finite elements interconnected at a finite
number of points on each element called nodes.
 uniform loading
Approximate method
Geometric model
Fixed boundary

Element Node
Finite element Element
Cantilever plate
model Mesh
in plane strain
Discretization

Node
Problem: Obtain the
stresses/strains in the
plate
 Fundamental Concept of FEM
• A continuous field of a certain domain
having infinite degrees of freedom is
approximated by a set of piecewise
continuous models with a number of finite
regions called elements. The number of
unknowns defined as nodes are determined
using a given relationship
i.e.{F}=[K]*{U}.
 General Steps
1) Discretize the domain
a) Divide domain into finite elements using appropriate
element types (1-D, 2-D, 3-D, or Axisymmetric)
2) Select a Displacement Function
a) Define a function within each element using the nodal
values
3) Define the Strain/Displacement and Stress/strain
Relationships
4) Derive the Element Stiffness Matrix and
Equations
a)Derive the equations within each element
 General Steps
5) Assemble the Element Equations to Obtain the
Global or Total Equations and Introduce
Boundary Conditions
a)Add element equations by method of superposition to
obtain global equation
6) Solve for the Unknown Degrees of Freedom (i.e
primary unknowns)
7) Solve for the Element Strains and Stresses
8) Interpret the Results
3. Historical Background
Chronicle of Finite Element Method
Year Scholar Theory

1941 Hrennikoff Presented a solution of elasticity problem using one-dimensional elements.

1943 McHenry Same as above.

1943 Courant Introduced shape functions over triangular subregions to model the whole
region.
1947 Levy Developed the force (flexibility) method for structure problem.

1953 Levy Developed the displacement (stiffness) method for structure problem.

1954 Argyris & Kelsey Developed matrix structural analysis methods using energy principles.

1956 Turner, Clough, Derived stiffness matrices for truss, beam and 2D plane stress elements. Direct
Martin, Topp stiffness method.

1960 Clough Introduced the phrase finite element .

1960 Turner et. al Large deflection and thermal analysis.

1961 Melosh Developed plate bending element stiffness matrix.

1961 Martin Developed the tetrahedral stiffness matrix for 3D problems.

1962 Gallagher et al Material nonlinearity.

Chronicle of Finite Element Method
Year Scholar Theory

1963 Grafton, Strome Developed curved-shell bending element stiffness matrix.

1963 Melosh Applied variational formulation to solve nonstructural problems.

1965 Clough et. al 3D elements of axisymmetric solids.

1967 Zienkiewicz et. Published the first book on finite element.

1968 Zienkiewicz et. Visco-elasticity problems.

1969 Szabo & Lee Adapted weighted residual methods in structural analysis.

1972 Oden Book on nonlinear continua.

1976 Belytschko Large-displacement nonlinear dynamic behavior.

~1997 New element development, convergence studies, the developments of

supercomputers, the availability of powerful microcomputers, the development
of user-friendly general-purpose finite element software packages.
 Advantages of FEM
• Model irregularly shaped bodies
• Compute General load conditions
• Model bodies composed of different materials
• Solve unlimited numbers and kinds of boundary
conditions
• Able to use different element sizes in places where
loads or stresses are concentrated
• Handle non-linear behavior using linear
approximations
• Reduce System Cost
Disadvantages of FEM
Errors Inherent in FEM Formulation
Mistakes by Users
4. Analytical Processes of Finite Element Method
(1) Structural stress analysis problem
A. Conditions that solution must satisfy
a. Equilibrium
b. Compatibility
c. Constitutive law
d. Boundary conditions
Above conditions are used to generate a system of equations
representing system behavior.

B. Approach
a. Force (flexibility) method: internal forces as unknowns.
b. Displacement (stiffness) method: nodal disp. As unknowns.
For computational purpose, the displacement method is more
desirable because its formulation is simple. A vast majority of
general purpose FE softwares have incorporated the displacement
method for solving structural problems.
(2) Analysis procedures of linear static structural analysis

A. Build up geometric model

a. 1D problem
line
b. 2D problem
surface
c. 3D problem
solid
B. Construct the finite element model
a. Discretize and select the element types
(a) element type
1D line element
2D element
3D brick element

(b) total number of element (mesh)

1D:
2D:
3D:
b. Select a shape function
1D line element: u=ax+b
c. Define the compatibility and constitutive law

d. Form the element stiffness matrix and equations

(a) Direct equilibrium method
(b) Work or energy method
(c) Method of weight Residuals

e. Form the system equation

Assemble the element equations to obtain global system
equation and introduce boundary conditions
C. Solve the system equations
a. elimination method
Gauss’s method (Nastran)
b. iteration method
Gauss Seidel’s method

Displacement field strain field stress field

D. Interpret the results (postprocessing)

a. deformation plot b. stress contour
5. Applications of Finite Element Method

Structural Problem Non-structural Problem

Stress Analysis Heat Transfer
- truss & frame analysis Fluid Mechanics
- stress concentrated problem Electric or Magnetic
Buckling problem Potential
Vibration Analysis
Impact Problem
 FEM Packages
• Large Commercial Programs
– Designed to solve many types of problems
– Can be upgraded fairly easily
– Initial Cost is high
– Less efficient
• Special-purpose programs
– Relatively short, low development costs
– Additions can be made quickly
– Efficient in solving their specific types of problems
– Can’t solve different types of problems
 Computer Programs for FEM
• Algor • IMAGES-3D
• ANSYS • MSC/NASTRAN
• COSMOS/M • SAP2000
• STARDYNE • GT-STRUDL
• ETABS • ABAQUS
 Note on Stiffness matrix
For a 1-D bar, the stiffness matrix is derived from the
stress/strain relationship in Hooke’s law and the definitions
of stress and strain.

σx = Eεx. ; σx = P/A ; εx = du/dx = (d2x – d1x)/L

By substitution: -f1x = EA (d2x – d1x) L E

L A
f1x = EA (d1x – d2x) f1 f2
L d1x d2x
Similarly for f2x: f2x = EA (d2x – d1x)
L
Combining into matrix form, the stiffness matrix is defined
as
[k] = EA 1-1
L
-1 1
 Note on the displacement
function
•For a given set of nodes there exists a function that approximates
the displacement at any position along the bar.
•This function, called the displacement function, is derived from
Pascal’s Triangle.
•A new constant is introduced into the function for every node in
the discretized domain.

For 1-D u(x) = a1 + a2x + a3x2 + …

 Note on the displacement function
If a 1-D bar is broken into 2 elements, the displacement function
would be u(x) = a + a x + a x2.
1 2 3

Putting it into matrix notation: u(x) = [1 x x2] a1

a2
By knowing the distances to the nodes a3
and the displacements at those nodes, the
equation becomes: u1 1 0 0 a1
u2 2
= 1 x2 x2 a2 , where
u3 1 x3 x32 a3

x1 = 0, x2 and x3 are the distances to the nodes and u1, u2, and u3 are the
displacements.

The coefficients are found by solving the equation.

Linear Algebra Recap
 What is a matrix?
A rectangular array of numbers (we will concentrate on
real numbers). A nxm matrix has ‘n’ rows and ‘m’
columns

M11 M12 M13 M14  First row

M 3x4 
= M 21 M 22 M 23 M 24 
Second row
M 31 M 32 M 33 M 34  Third row

First Second Third Fourth

column column column column
Row number
M12 Column number
What is a vector?

A vector is an array of ‘n’ numbers

A row vector of length ‘n’ is a 1xn matrix

a 1 a2 a3 a4
A column vector of length ‘m’ is a mx1 matrix

 a1 
a 
 2
a3 
Special matrices

Zero matrix: A matrix all of whose entries are zero

0 0 0 0 
03x 4 
= 0 0 0 0  
0 0 0 0 
Identity matrix: A square matrix which has ‘1’ s on
the diagonal and zeros everywhere else.

1 0 0 
I3x 3 
= 0 1 0 
0 0 1
Matrix operations
Equality of matrices

If A and B are two matrices of the same size,

then they are “equal” if each and every entry of one
matrix equals the corresponding entry of the other.

 1 2 4 a b c 

A = − 3 0 7  
B = d e f  
 9 1 5   g h i 
a = 1, b = 2, c = 4,
A = B  d = −3, e = 0, f = 7,
g = 9, h = 1, i = 5.
Matrix operations Addition of two
matrices

If A and B are two matrices of the same size,

then the sum of the matrices is a matrix C=A+B whose
entries are the sums of the corresponding entries of A and
B

 1 2 4  − 1 3 10 
  
A =  − 3 0 7 B =  − 3 1 0  
 9 1 5  1 0 6 
 0 5 14 

C = A + B = − 6 1 7  
 10 1 11 
 Addition of matrices
Matrix operations
Properties

Properties of matrix addition:

1.Matrix addition is commutative (order of addition does
not matter)
A+B =B+A
2.Matrix addition is associative

A + (B + C ) = ( A + B ) + C
3.Addition of the zero matrix
A+0=0+A = A
Matrix operations Multiplication by a
scalar

If A is a matrix and c is a scalar, then the product cA is a

matrix whose entries are obtained by multiplying each of
the entries of A by c

1 2 4
A =  − 3 0 7 c = 3
 9 1 5
 3 6 12 

cA =  − 9 0 21 
 27 3 15 
 Multiplication by a
scalar
Matrix operations
Special case

If A is a matrix and c =-1 is a scalar, then the product

(-1)A =-A is a matrix whose entries are obtained by
multiplying each of the entries of A by -1

 1 2 4

A =  − 3 0 7  c = −1
 9 1 5
−1 −2 − 4

cA = -A =  3 0 − 7
 − 9 −1 − 5
Matrix operations Subtraction

If A and B are two square matrices of the same size, then

A-B is defined as the sum A+(-1)B

 1 2 4  − 1 3 10 
  
A =  − 3 0 7 B =  − 3 1 0  
 9 1 5  1 0 6 
 2 − 1 − 6

C = A − B = 0 − 1 7  
8 1 − 1
Note that A - A = 0 and 0 - A = -A
 Transpose
Special
operations
If A is a mxn matrix, then the transpose of A is
the nxm matrix whose first column is the first
row of A, whose second column is the second
column of A and so on.

1 2 4 1 − 3 9

A = − 3 0  
7   A = 2
T
0 1 
 9 1 5   4 7 5 
 Transpose
Special
operations

If A is a square matrix (mxm), it is called

symmetric if

A =A T
Matrix operations Scalar (dot) product of
two vectors
If a and b are two vectors of the same size

 a1   b1 
  
a = a 2 ; b = b 2  
 a 3  b 3 
The scalar (dot) product of a and b is a scalar obtained by
adding the products of corresponding entries of the two
vectors

a b = ( a 1b1 + a 2 b 2 + a 3 b 3 )
T
Matrix operations Matrix multiplication

For a product to be defined, the number of columns of A

must be equal to the number of rows of B.

A B = AB
mxr rxn mxn
inside

outside
 Matrix operations Matrix multiplication

If A is a mxr matrix and B is a rxn matrix, then the product

C=AB is a mxn matrix whose entries are obtained as follows.
The entry corresponding to row ‘i’ and column ‘j’ of C is the dot
product of the vectors formed by the row ‘i’ of A and column ‘j’
of B

1 2 4  −1 3
A 3x3 
=  −3 0 7  B3x2 
=  −3 1 
 9 1 5   1 0 
 −3  −1
T
5  1
C3x2 
= AB = 10  
−9  notice  2    −3 = −3
 
 −7 28   4   1 
 Multiplication of
Matrix operations matrices
Properties

Properties of matrix multiplication:

1.Matrix multiplication is noncommutative (order of
addition does matter)
A B  B A in g e n e ra l
•It may be that the product AB exists but BA does not
(e.g. in the previous example C=AB is a 3x2 matrix,
but BA does not exist)
•Even if the product exists, the products AB and BA
are not generally the same
Matrix operations matrices
Properties

2. Matrix multiplication is associative

A (B C ) = ( A B )C
3. Distributive law
A (B + C ) = AB + AC
(B + C )A = BA + CA
4. Multiplication by identity matrix
A I = A ; IA = A
5. Multiplication by zero matrix A0 = 0 ;0 A = 0
6.
(A B ) T
=B A
T T
 Miscellaneous
Matrix operations properties

1.If A , B and C are square matrices of the same

size, and A  0 then A B = A C does not
necessarily mean that B =C
2. A B = 0 does not necessarily imply that
either A or B is zero
 Definition
Inverse of a
matrix

If A is any square matrix and B is another

square matrix satisfying the conditions

A B = BA = I
Then
(a)The matrix A is called invertible, and
(b) the matrix B is the inverse of A and is
denoted as A-1.

The inverse of a matrix is unique

 Uniqueness
Inverse of a
matrix

The inverse of a matrix is unique

Assume that B and C both are inverses of A

AB = BA = I
AC = CA = I
(BA)C = IC = C
B(AC) = BI = B
B=C
Hence a matrix cannot have two or more
inverses.
 Some properties
Inverse of a
matrix

Property 1: If A is any invertible square

matrix the inverse of its inverse is the matrix A
itself
(A )
-1 − 1
=A
Property 2: If A is any invertible square
matrix and k is any scalar then

(k A ) −1 1 -1
= A
k
Inverse of a Properties
matrix
Property 3: If A and B are invertible square
matrices then
−1
(A B ) −1
=B A
-1

(AB) (AB )
−1
=I
Premultiplying both sides by A-1
A (AB) (AB ) = A −1
-1 −1

(A A)B(AB )
-1 −1
= A −1
B(AB ) = A −1
−1

Premultiplying both sides by B-1

(AB )−1 = B −1A −1
 What is a determinant?
The determinant of a square matrix is a number
obtained in a specific manner from the matrix.

For a 1x1 matrix:

A = a 1 1  ; det( A ) = a 1 1
For a 2x2 matrix:

 a 11 a 12 
A=  ; det( A ) = a 11a 22 − a 12 a 21
 a 21 a 22 

Product along red arrow minus product along blue arrow

Example 1
1 3 
Consider the matrix A= 
 5 7 

Notice (1) A matrix is an array of numbers

(2) A matrix is enclosed by square brackets

1 3
det( A ) = = 1  7 − 3  5 = −8
5 7

Notice (1) The determinant of a matrix is a number

(2) The symbol for the determinant of a matrix is
a pair of parallel lines
Computation of larger matrices is more difficult
Duplicate column method for 3x3 matrix
For ONLY a 3x3 matrix write down the first two
columns after the third column

 a 11 a 12 a 13   a11 a12 a13  a11 a12

A = a 21 a 22 a 23  a 
 21 a 22 a 23  a 21 a 22
a 31 a 32 a 33  a 31 a 32 a 33  a 31 a 32
Sum of products along red arrow
minus sum of products along blue arrow

det( A ) = a 11a 22a 33 + a 12a 23a 31 + a 13a 21a 32

− a 13a 22a 31 − a 11a 23a 32 − a 12a 21a 33
This technique works only for 3x3 matrices
Example

2 4 - 3  2 4 − 3 2 4
A = 1 0 4  1 
  0 4 1 0
 2 - 1 2   
 2 − 1 2  2 −1

0 -8 8 0 32 3

Sum of red terms = 0 + 32 + 3 = 35

Sum of blue terms = 0 – 8 + 8 = 0
Determinant of matrix A= det(A) = 35 – 0 = 35
 Finding determinant using inspection
Special case. If two rows or two columns are proportional
(i.e. multiples of each other), then the determinant of the
matrix is zero

2 7 8
3 2 4 =0
−2 −7 −8

because rows 1 and 3 are proportional to each other

If the determinant of a matrix is zero, it is called a

singular matrix
 What is a cofactor?
Cofactor method
If A is a square matrix
 a 11 a 12 a 13 
A = a 21 a 22 a 23 
a 31 a 32 a 33 
The minor, Mij, of entry aij is the determinant of the submatrix
that remains after the ith row and jth column are deleted from A.
The cofactor of entry aij is Cij=(-1)(i+j) Mij

a 21 a 23 a 21 a 23
M 12 = = a 21a 33 − a 23a 31 C12 = − M 12 = −
a 31 a 33 a 31 a 33
What is a cofactor?

Sign of cofactor + - +
- + - 
 
+
 - +
Find the minor and cofactor of a33
2 4 - 3 
A = 1 0 4 
  Minor 2 4
 2 - 1 2  M 33 = = 2  0 − 4  1 = −4
1 0

Cofactor C = ( − 1 ) ( 3 + 3 ) M = M = − 4
33 33 33
 Cofactor method of obtaining the
determinant of a matrix

The determinant of a n x n matrix A can be computed by

multiplying ALL the entries in ANY row (or column) by
their cofactors and adding the resulting products. That is,
for each 1  i  n and 1  j  n

Cofactor expansion along the jth column

d e t( A ) = a 1 j C 1 j + a 2 j C 2 j +  + a n j C n j

Cofactor expansion along the ith row

det( A ) = a i 1C i 1 + a i 2 C i 2 +  + a i n C i n
 Example: evaluate det(A) for:
1 0 2 -3
A= 3 4 0 1
det(A) = a11C11 +a12C12 + a13C13 +a14C14
-1 5 2 -2
0 1 1 3

4 0 1 3 0 1 3 4 1
det(A)=(1) 5 2 -2 - (0) -1 2 -2 +2 -1 5 -2
1 1 3 0 1 3 0 1 3
3 4 0
- (-3) -1 5 2 = (1)(35)-0+(2)(62)-(-3)(13)=198
0 1 1
 Example : evaluate
1 5 -3
det(A)= 1 0 2
3 -1 2
By a cofactor along the third column

det(A)=a13C13 +a23C23+a33C33

-3* (-1) 4 1 0 1 5 1 5
det(A)= +2*(-1)5 +2*(-1)6
3 -1 3 -1 1 0

= det(A)= -3(-1-0)+2(-1)5(-1-15)+2(0-5)=25
Quadratic form

The scalar U =d k d
T d = vector
k = square matrix
Is known as a quadratic form

If U>0: Matrix k is known as positive definite

If U≥0: Matrix k is known as positive semidefinite
Quadratic form

 d1   k 11 k 12 
Let d = k = 
Symmetric
d 2   k 21 k 22  matrix

Then
 k11 k12   d 1 
U = d k d = d1 d 2 
T
  
k12 k 22  d 2 
 k11d1 + k12 d 2 
= d1 d 2  
k12 d 1 + k 22 d 2 
= d1 (k11d1 + k12 d 2 ) + d 2 (k12 d 1 + k 22 d 2 )
= k11d1 + 2k12 d1 d 2 + k 22 d 2
2 2
Differentiation of quadratic form

Differentiate U wrt d1
U
= 2 k 11 d 1 + 2 k 12 d 2
d 1
Differentiate U wrt d2
U
= 2 k 12 d 1 + 2 k 22 d 2
d 2
Differentiation of quadratic form

Hence

 U 
U  d 1   k11 k12   d 1 
  = 2   
 d  U   k12 k 22  d 2 
 d 2 
=2k d
Drag Force Analysis
of Aircraft

• Question
What is the drag force distribution on the aircraft?
• Solve
– Navier-Stokes Partial Differential Equations.
• Recent Developments
– Multigrid Methods for Unstructured Grids
 San Francisco Oakland Bay Bridge

Before the 1989 Loma Prieta earthquake

San Francisco Oakland Bay Bridge

After the earthquake

San Francisco Oakland Bay Bridge

A finite element model to analyze the

bridge under seismic loads
Courtesy: ADINA R&D
Engine Thermal
Analysis

Picture from
http://www.adina.com

• Question
– What is the temperature distribution in the engine block?
• Solve
– Poisson Partial Differential Equation.
• Recent Developments
– Fast Integral Equation Solvers, Monte-Carlo Methods
Electromagnetic
Analysis of Packages

Thanks to
Coventor
http://www.cov
entor.com

• Solve
– Maxwell’s Partial Differential Equations
• Recent Developments
– Fast Solvers for Integral Formulations
 Example
E 2E
a A b A P

L L

Determine displacements of materials a and b if

the load P is applied to the end of the bar given
the above information.