MCE 466 Introduction to

Finite Element Methods

Instructor: David G. Taggart

Dept. of Mechanical Engineering
University of Rhode Island
 Introduction
Finite Element Method a numerical method for
obtaining approximate solutions to field
problems

Field Problems A problem in which the

behavior in a well defined region, R, is described
by a governing set of differential equations with
certain quantities prescribed on the boundary, C.
(Also called a boundary value problem BVP)
 Introduction (cont.)
Boundary Value Problem:

R
C

Examples:
Stress analysis (includes system of springs, bars under axial loads, trusses,
beams, frames, 2-D & 3-D continua, plate bending)
Stability analysis (buckling), structural dynamics, plasticity, viscoelasticity
Heat transfer
Fluid flow
General PDE problems
 Finite Element Approach

Divide the region into may discrete sub-regions

(discretization) of finite size (elements). Select certain
points (nodes) at which the solution is desired. Assume
a functional form for the solution between the nodes
(interpolation)

element

nodes
 Important points to remember
Finite element analyses provide approximate solutions in
nearly all cases
Results can be very accurate or totally meaningless
(easy to make mistakes)
Method is very susceptible to user error
Stress averaged results may conceal inaccuracies in the
results
Modern commercial codes make it very easy to get
answers
Responsibility of user:
Anticipate results (perform approximate calculations by hand)
Explain discrepancies
Be aware of limitations
 MCE 466 Course Structure

Appropriate balance of FEA theory and use of

commercial codes (primarily Abaqus)
Text problems hand calculations with support
of Matlab as needed
Several Abaqus - based computer assignments
Project individual or teams of 2
 History of Finite Elements*
1943 R. Courant (mathematician) proposed a torsion solution using
triangular subregions (not implemented since no computers were
available)
Early 50s digital computers which solved large systems of equations
became available. Aerospace companies bega solving structures
problems (mostly unpublished)
1960 term Finite Element Method first coined
Early 60s new elements formulated, method became accepted
Mid 60s Applied to heat transfer and fluid flow problems
Late 60s, early 70s large general purpose programs became
commercially available (NASTRAN, ANSYS, SAP)
Early 80s Interactive graphics made using FEA easier
Late 80s Inexpensive PC versions became available
90s and beyond Widespread application in numerous industries,
Integration with CAD packages.
Most recent emphasis multiphysics event simulation

* "Concepts and Applications of Finite Element Analysis," by R. D. Cook, D. Malkus and M. Plesha
 Review Matrix Algebra
(see Appendix A, B)
We will see that the finite element method requires solving large systems
of linear equations using matrix algebra tools.

First, review matrix multiplication

1 2 5 6 7 1 5 2 8 1 6 2 9 1 7 2 10 21 24 9
3 4 8 9 10 2 5 4 8 2 6 4 9 2 7 4 10 47 54 25

2x2 2x3 2x3

In general

[A] [B] = [C]

mxn nxp mxp
 Review Matrix Algebra
(see Appendix A, B)

To review some basic matrix algebra, consider the following system of equations

2x y 6 10

x 4 y 17 9

or 7

6
(x,y)= (1,4)

y 2 x 6 5

1 17
y x
3

4 4 2

0
0 1 2 3 4 5 6
 Matrix Algebra (cont.)

Matrix form:
2 1 x 6
1 4 y 17 or a x b

where

a
2 1
x
x 6
, , b
1 4 y 17

Matlab code Matlab output

a=[2 1;1 4] a =
b=[6;17] 2 1
determinant=det(a) 1 4
solution=a\b b =
6
17 Note |a| 0
determinant =
7
solution =
1.0000
4.0000
 Matrix Algebra (cont.)

Now consider a different system of equations

4 x 2 y 16 10

2x y 6 8

7
or 6

y 2 x 8
4

y 2 x 6
2

0
0 1 2 3 4 5 6
 Matrix Algebra (cont.)

Matrix form:
4 2 x 16
2 1 y 6 or a x b

where

a
4 2
x
x 16
, , b
2 1 y 6

Matlab code Matlab output

a=[4 2;2 1] a =
b=[16;6] 4 2
determinant=det(a) 2 1
solution=a\b b =
16
6 Note |a| = 0
determinant =
0
Warning: Matrix is singular to working precision.

solution =
Inf
-Inf
 Matrix Algebra (cont.)

System of three equations (Matlab format)

Code:

a=[-1 1 2;3 -1 1;3 3 1] 3

b=[2;6;2]
solution=a\b 2.5

2
Output:

a = 1.5

-1 1 2
3 -1 1 1
3 3 1 -0.9

b = -0.95 1.15
1.1
2 -1 1.05
6 -1.05 1
0.95
2 -1.1 0.9
solution =
1
-1
2
 Matrix algebra applied to FEA

For displacement based stress analysis

K11 K12 ... K1n D1 P1

K K 22 D2 P2
12


K1n ... K nn Dn Pn

Nodal Forces
Stiffness matrix -
Symmetric since Kij = Kji Nodal Displacements
 Other matrix terminology
Banded matrix
If all non-zero terms are contained within a band
along the diagonal, the matrix is said to be banded
Sparse matrix
If a matrix has relatively few non-zero terms (as is
common in FEA), the matrix is said to be sparse
Singular matrix
If the determinant of the matrix equals zero, the matrix
is said to be singular. As we saw, if [A] is singular,
then the system of equations [A]{x}={b} has no unique
solution.
 Other matrix terminology (cont.)
Transpose
The transpose of a matrix is found by interchanging
rows and columns, e.g.

1 2
1 3 5
A 3 4 , A
T

5 6 2 4 6

Transpose of a product

AB T BT AT
 Steps in the Finite Element Method

1. Discretize the region and select element type

2. Select a displacement function
3. Define the strain/displacement and stress/strain
relations
4. Derive the element equations
Direct Stiffness Method
Energy Methods
Method of Weighted Residuals (Galerkins method)
5. Assemble global equations and impose boundary
conditions
6. Solve for unknown nodal displacements
7. Solve for element strains and stresses
8. Interpret results
 Demo of a Commercial FEA Package
Determine maximum stress for the following problem:

r/h = 0.10

Source: Advanced Mechanics of Materials, Boresi et al, 1993.