EMAE 415 Lectures

Finite Element Analysis -

Basic Concepts

Based in part on slides posted by

Dr. Ara Arabyan, University of Arizona;
Dr. Kenneth Youssefi, San Jose State University
 Basics of Finite Element Analysis

FEA in Biomechanics
FEA models have proven to be powerful tools
in analyzing biomechanical structures and to
evaluate designs for implants, prostheses and
musculoskeletal constructs.
The advantages are the ability to account for
complex geometries and material behaviors.
Always remember, FEA is a numerical
technique. Its answers are only as good as the
formulation of the problem.
 Basics of Finite Element Analysis

FEA is a mathematical representation of a

physical system and the solution of that
mathematical representation.

Always remember, FEA is a numerical

technique. Its answers are only as good as
the formulation of the problem.

GIGO (Garbage in-garbage out)

 FEA Stress Analysis
After approximating the object by finite elements,
each node is associated with the unknowns to be
solved.
For a two dimensional solid (e.g. a beam) the
displacements in x and y would be the unknowns.
This implies that every node has two degrees of
freedom and the solution process has to solve 2n
degrees of freedom.
Once the displacements have been computed, the
strains are derived by partial derivatives of the
displacement function and then the stresses are
computed from the strains.
 Finite Element Analysis
Consider the cantilever beam
shown.

Finite element analysis starts with an approximation of the region of

interest into a number of meshes (triangular elements). Each mesh is
connected to associated nodes (black dots) and thus becomes a finite
element.
 Displacement Fields
The displacements of different points in
the structure set up a displacement field

The displacement field expresses the

displacement of any point in the
structure as a function of its position
measured in a reference frame
Strains versus Displacement Fields
(2-dimensional case)

Knowing the displacement fields allows you

to determine the strain fields.
Stiffness of a Uniform Rod
Recall from elementary mechanics of solids that a
uniform rod of length L, cross sectional area A, and
elasticity modulus E can modeled as a linear spring
of stiffness keq

P A, E P P P

L keq

with AE
keq
L
 Nodal Displacements, Forces
Consider a linear spring of stiffness k. Let the
displacements of its two ends, called nodes, be
denoted by ui and uj, known as nodal displacements.
Let the forces acting at its two ends, called nodal
forces, be denoted by fi and fj.
Nodal
ui uj displaceme
y nt
fi fj
x i j Nodal
k force
Referenc
e frame
Nodes
Force-Displacement Relations

The relationships between the nodal

forces and displacements (as shown
below) are given by:
f k u u
i i j

f k u u
j j i

ui uj

fi fj

k
Element Stiffness Matrix
These relations can be written in matrix
form as
k k u i fi
k k u f
j j

or more briefly as

ku f
Element Stiffness Matrix (contd)

In this relation
k k
k
k k
is known as the element stiffness matrix (always
symmetric);
u i
u
u j

is known as the element nodal displacement vector; and

Element Stiffness Matrix (contd)

f i
f
f j

is known as the element nodal force vector.

The element nodal displacements are also known as

element nodal degrees of freedom (DOF)
Singularity of Element Stiffness Matrix

The equation ku = f cannot be solved for the nodal

displacements for arbitrary f because the matrix
k is singular.
Physically this means that, in static equlibrium, the
displacements of the endpoints of a spring
cannot be determined uniquely for an arbitrary
pair of forces acting at its two ends.
One of the ends must be fixed or given a specified
displacement; the displacement of the other end
can then be determined uniquely.
 Solution for Single Rod Element
If node i is fixed (i.e. its displacement is set to 0) then
ku = f reduces to
kuj f j

and the displacement of node j is easily determined as

fj
uj
k
which is the expected solution
ui=0 uj

fj

k
 Matrix Reduction
Note that when a displacement or DOF is set
to zero, rows and columns of k associated
with that displacement are eliminated and
only the remaining set is solved
Row(s)
k k u i fi associated
k k u f with ui
j j

Column(s)
associated
with ui
 Multiple Elements
Now consider two springs of different stiffness linked
to each other
Globally numbered
elements

u1 ui(1) u2 u(1)
j ui
(2)
u3 u(2)
j
1 2
F1
1 2 F2 3 F3
k1 k2

Globally
numbered
nodes
 Continuity Relations
When two elements are joined together the
joined nodes become one and must have the
same displacement
u2 u(1)
j u(2)
i

where the subscript denotes the global node

number, the superscript denotes the global
element number, and i and j denote local
element numbers
Force Balance Relations
The external nodal forces acting at each node must
equal the sum of the element nodal forces at all
nodes
F1 fi
(1)

F2 f j
(1)
i
(2)

F3 f j(2)

where F1, F2, F3 are external nodal forces numbered

globally
 Assembly of Equations
When these continuity and force balance relations
are imposed the resulting global equilibrium
equations are
k1 k1 0 u1 F1
k k k k u
1 1 2 2 2 F2
0 k2 k2 u 3
F3

More briefly this can be written as

KU F
 Global Stiffness Matrix
In this relation
k1 k1 0
K k1 k1 k2 k2

0 k2 k2

is known as the global stiffness matrix (always

symmetric);
u 1

U u 2
u
3

is known as the global nodal displacement vector; and

Global Stiffness Matrix (contd)
F 1

F F 2
F
3
is known as the global nodal force vector or the global
load vector

The global nodal displacements are also known as

global degrees of freedom (DOF)
Assembly of Global Stiffness Matrix

Note that the global stiffness matrix is assembled

from element matrices as follows
Stiffness matrix from
element 1

k1 k1 0
K k1 k1 k2 k2

0 k2 k2

Stiffness terms from two

Stiffness matrix from
matrices add at
element 2
coinciding DOF
 Singularity of Global Stiffness Matrix
As in the case of individual element matrices the
global stiffness matrix K is singular. (You can
check this out for this small example by calculating
the determinant of the matrix; the result will be
zero.)
Some nodes of the structure need to be constrained
(i.e. fixed or given known displacements) to make it
statically determinate or overconstrained. Then the
remaining DOF can be determined.
Constraining some nodes in the structure
corresponds to applying boundary conditions.
Solution for Global Structure
If node 1 is fixed (i.e. its displacement is set to 0)
then the equilibrium equations reduce to
k1 k2 k2 u2 F2
k u
2 k2 3 F3

u1 0 u2 u3
1 2

1 2 F2 3 F3
k1 k2
Solution for Global Structure (contd)

The displacements of nodes 2 and 3 can now be

found from
1
u 2 k1 k2 k 2 F2
F
u
3 k2 k
2 3

It can be shown that constrained global stiffness

matrix is not singular
 Matrix Reduction
Note that when a DOF is set to zero rows and
columns of K associated with that DOF are
eliminated and only the remaining set is solved

k1 k1 0 u1 F1
k k k k u Row(s)
1 1 2 2 2 F2
associated
0 k2 k2 u 3
F3
with u1

Column(s)
associated
with u1
 Rod Elements
The spring models introduced thus far
constitute one class of finite elements and are
known as rod, spar (ANSYS), or truss
(ALGOR) elements
In the form shown these elements can be
used to model only unidimensional (one
dimensional) problems
The more general form of these elements can
be used to model two or three dimensional
problems
 Example
An aluminum (E = 10.4 106 psi) rod of variable
cross section is subjected to a point load of 1000 lb at
its narrower end. Determine the deflection of the rod
at its loaded end.

1000 lb
Al = 0.250 in2

Ar = 0.125
12 in in2
 Example, contd.
The variable cross section rod can be approximated
as a number of rods of constant cross section. For
this example let us choose three rods of equal length
that span the length of the original rod.

1000 lb
A1 A2 A3

4 4 4
in in in
 Example, contd.
The cross-sectional area of each rod
can be assumed to be the average of
the segment each rod spans:
1 A Ar 4 0.229in2
A1 A l A l l
2 12
1 Al Ar Al A r
A 2 Al 4 Al 8 0.188in2
2 12 12
1 A Ar 8 A 0.146in2
A 3 Al l r
2 12
 Example, contd.
The equivalent stiffness of each rod can now be
computed from
Ai Ei
ki
Li
which results in

k1 5.95 105 lb/ in

k2 4.89 105 lb/ in
k3 3.80 105 lb/ in
 Example, contd.
The spring (finite element) model of the problem is
now

u1 0 u2 u3 u4
1 2 3

1 2 3 4 1000 lb
k1 k2 k3
 Example, contd.
The equilibrium equations can now be written as:

k1 k1 0 0 u1 R1
k k k k2 0 u 2 0
1 1 2

0 k2 k2 k3 k3 u 3 0
0 0 k3 k3 u 4 1000

Row and
column
associated
with u1
 Example, contd.
Solving the remaining equations we obtain

1
u 2 10.84 4.89 0 0 1.68

u 3 105 4.89 8.69 3.80 0
3
3.73 10 in


u 4 0
3.80 3.80 1000 6.36

Since the deflection of the right end is required the

answer to the problem is

ur u4 6.36 103 in
Comparison with Exact Results

Using exact analytical methods the displacement of

any point (u(x)) on the rod is given by

PL Al
u x ln
E Al Ar A Al Ar
x
l
L
u x

1000 lb

x
Comparison with Exact Results (contd)

Using this expression the displacement of the three

nodes in the finite element model can be computed
as
u 4 1.68 103 in
u 8 3.74 103 in
u 12 6.40 103 in

Evidently the results from the finite element method

are very close to those produced by exact methods
Comparison with Exact Results (contd)
The plot below charts the variation of displacements
across the bar for the two solutions
-3
x 10
7
Exact solution u3
FE solution
6

5
Displacement (in)

4 u2

2 u1

0
0 2 4 6 8 10 12
x (in)
 Formulation of the Finite Element Method

f B Body forces (forces distributed over the volume of the body;

gravitational, inertia, or magnetic forces)
f S surface forces (pressure of one body on another, or hydrostatic
pressure)
f i Concentrated external forces
Formulation of the Finite Element Method
Denote the displacements of any point (X, Y, Z) of the object from
the unloaded configuration as UT

The displacement U causes the strains

and the corresponding stresses

The goal is to calculate displacement, strains, and stresses from

the given external forces.
 Example a plate under load
Derive and solve the system equations of a plate loaded as
shown.Plate thickness is 1 cm and the applied load Py is
constant.

using two triangular elements,

 Example, contd
Displacement within the triangular element with three
nodes can be assumed to be linear.
 Example, contd
Displacement for each node,
 Example, contd
Solve the equations simultaneously for and ,
 Example, contd
Substitute x1= 0, y1= 0, x2=10, y2= 0, x3= 0, y3=4 to
obtain displacements u and v for element 1,

Element 1
 Example, contd
Rewriting the equations in the matrix form,
 Example, contd

Similarly the displacements within element 2 can be

expresses as
 Example, contd

The next step is to determine the strains using 2D

strain-displacement relations,
 Example, contd

Differentiate the displacement equation

to obtain the strain,
 Example, contd
Element 2
 Example, contd
Using the stress-strain relations for homogeneous,
isotropic
plane-stress, we have
 Formulation of the Finite Element Method
Equilibrium condition and principle of virtual displacements

The left side represents the internal virtual work done and the
right side represents the external work done by the actual
forces as they go through the virtual displacement.

The above equation is used to generate finite element

equations. And by approximating the object as an assemblage
of discrete finite elements, these elements are interconnected
at nodal points.
 Formulation of the Finite Element Method
The equilibrium equation can be expressed using matrix
notations for m elements.

where
B(m) represents the rows of the strain displacement matrix
C(m) is the elasticity matrix of element m
H(m) is the displacement interpolation matrix
U is a vector of the three global displacement components at
all nodes
F is a vector of the external concentrated forces applied to
the nodes
 Formulation of the Finite Element Method
The previous equation can be rewritten as follows,

The above equation describes the static equilibrium problem. K is the stiffness matrix.
Stiffness matrix for element 1
 Example, contd
Calculating the stiffness matrix for element 2.
 Example, contd
The stiffness of the structure as a whole is obtained by combing
the two matrices.
 Example, contd
The load vector R, equals Rc because only concentrated
loads act on the nodes

where Py is the known external force and F1x, F1y, F3x, and F3y are
the unknown reaction forces at the supports.
 Example, contd
The following matrix equation can be solved for nodal point
displacements
 Example, contd
The solution can be obtained by applying the boundary conditions
 Example, contd
The equation can be divided into two parts,

The first equation can be solved for the unknown nodal

displacements, U3, U4, U7, and U8. And substituting these
values into the second equation to obtain unknown reaction
forces, F1x, F1y, F3x, and F3y .
Once the nodal displacements have been obtained, the
strains and stresses can be calculated.
 Finite Element Analysis

FEA involves three major steps

Pre-Processing
Solving Matrix (solver)
Post-Processing
 Summary of Pre-Processing
Build the geometry
Make the finite-element mesh
Add boundary conditions; loads and
constraints
Provide properties of material
Specify analysis type (static or dynamic,
linear or non-linear, plane stress, etc.)
 FEA Pre-Processing
Mesh Development
The FEA mesh is your way of communicating
geometry to the solver, the accuracy of the solution
is primarily dependent on the quality of the mesh.
The better the mesh looks, the better it is.
A good-looking mesh should have well-shaped
elements, and the transition between densities
should be smooth and gradual without skinny,
distorted elements.
 FEA Pre-Processing - Example

Coarse mesh Refined mesh - better?probably

depends on the loading and
boundary conditions
 FEA Pre-Processing
Finite elements supported by most finite-element codes:
 FEA Pre-Processing
Material Properties
Material properties can be specified for element regions,
elements or even within elements in most large-scale FEA
codes.
The material properties required for an isotropic, linear
static FEA are: Youngs modulus (E), Poissons ratio (v),
and shear modulus (G).
G = E / 2(1+v)
Only two of the three properties can be specified
independently.
 FEA Pre-Processing
Nonlinear Material Properties

A multi-linear model requires the input of stress-strain

data pairs to essentially communicate the stress-strain
curve from testing to the FE model

Highly deformable, low stiffness, incompressible materials,

such as rubber and other synthetic elastomers require
distortional and volumetric constants or a more complete set
of tensile, compressive, and shear force versus stretch curve.

A creep analysis requires time and temperature

dependent creep properties. Plastic parts are extremely
sensitive to this phenomenon
 FEA Pre-Processing
Boundary Conditions

In FEA, the name of the game is boundary

condition, that is calculating the load and
constraints that each component experiences in its
working environment.
The results of FEA should include a complete
discussion of the boundary conditions
 Boundary Conditions
Loads
Loads are used to represent inputs to the system.
They can be in the forms of forces, moments,
pressures, temperature, or accelerations.
Constraints
Constraints are used as reactions to the applied
loads. Constraints can resist translational or
rotational deformation induced by applied loads.
 Boundary Conditions
Degrees of Freedom

Spatial DOFs refer to the three translational and three rotational

modes of displacement that are possible for any part in 3D
space. A constraint scheme must remove all six DOFs for the
analysis to run.

Elemental DOFs refer to the ability of each element to transmit

or react to a load. The boundary condition cannot load or
constrain a DOF that is not supported by the element to which
it is applied.
 View Animated
Displacements
Post-Processing

Does the shape of deformations make sense? No Review Boundary

Yes Conditions
View Displacement
Fringe Plot

Are magnitudes in line with your expectations? No Review Load Magnitudes

and Units
Yes
View Stress
Fringe Plot
Review Mesh Density
Is the quality and mag. Of stresses acceptable? No and Quality of Elements
Yes
View Results Specific
To the Analysis