FOUNDMENTALS FOR FINITE ELEMENT

METHOD


CHAPTER1:
The Finite Element Method
A Practical Course

CONTENTS
STRONG AND WEAK FORMS OF GOVERNING EQUATIONS
HAMILTONS PRINCIPLE
FEM PROCEDURE
Domain discretization
Displacement interpolation
Formation of FE equation in local coordinate system
Coordinate transformation
Assembly of FE equations
Imposition of displacement constraints
Solving the FE equations
STATIC ANALYSIS
EIGENVALUE ANALYSIS
TRANSIENT ANALYSIS (reading materials)
REMARKS
STRONG AND WEAK FORMS OF
GOVERNING EQUATIONS
System equations: strong form (PDE), difficult to
solve.
Weak (integral) form: requires weaker continuity
on the dependent variables (e.g., u, v, w).
Weak form is often preferred for obtaining an
approximated solution.
Formulation based on a weak form leads to a set
of algebraic system equations FEM.

HAMILTONS PRINCIPLE
Of all the admissible time histories of
displacement the most accurate solution makes the
Lagrangian functional a minimum.

An admissible displacement must satisfy:
The compatibility conditions
The essential or the kinematic boundary conditions
The conditions at initial (t
1
) and final time (t
2
)
HAMILTONS PRINCIPLE
Mathematically
0
2
1
=
}
dt L
t
t
o
where L=TH+W
f
V U U T
T
V
d

}
=
2
1
V c V
T
V
T
V
d d
2
1

2
1
} }
= =
f s
T
S
b
T
V
f
S f U V f U W
f
d d
} }
+ =
(Kinetic energy)
(Potential energy)
(Work done by
external forces)
Lagrangian functional
FEM PROCEDURE
Step 1: Domain discretization
Step 2: Displacement interpolation
Step 3: Formation of FE equation in local coordinates
Step 4: Coordinate transformation
Step 5: Assembly of FE equations
Step 6: Imposition of displacement constraints
Step 7: Solving the FE equations
Step 1: Domain discretization
The solid body is divided into N
e
elements with proper
connectivity compatibility.
All the elements form the entire domain of the problem
without any gap or overlapping compatibility.
There can be different types of element with different
number of nodes.
The density of the mesh depends upon the accuracy
requirement of the analysis.
The mesh is usually not uniform, and a finer mesh is often
used in the area where the displacement gradient is larger.

Triangular elements
Nodes
Step 2: Displacement interpolation
Bases on local coordinate system,
the displacement within element is
interpolated using nodal displacements.
e i i
n
i
z y x z y x z y x
d
d N d N U ) , , ( ) , , ( ) , , (
1
= =

=
1
2
displacement compenent 1
displacement compenent 2
displacement compenent
f
i
n
f
d
d
d
n




=
`

)
d
1
2
displacements at node 1
displacements at node 2
displacements at node
d
e
n
d
n

=
`

)
d
d
d
d
n
f
: Degree of freedoms at a node

x, u
y, v
1 (x
1
, y
1
)
(u
1
, v
1
)
2 (x
2
, y
2
)
(u
2
, v
2
)
3 (x
3
, y
3
)
(u
3
, v
3
)
A
f
sx

f
sy

n
d
: number of nodes in an element
Step 2: Displacement interpolation
N is a matrix of shape functions
1 2
( , , ) ( , , ) ( , , ) ( , , )

for node 1 for node 2 for node
d
n
d
x y z x y z x y z x y z
n
(
=

+ + +
N N N N
(
(
(
(
(

=
f
in
i
i
i
N
N
N
0 0 0
0 0 0
0 0 0
0 0 0
2
1

N
where
Shape function
for each
displacement
component at a
node
Displacement interpolation
Constructing shape functions
Consider constructing shape function for
a single displacement component
Approximate in the form

1
( ) ( ) ( )
d
n
h
i i
i
T
u p o
=
= =

x x p x
1 2 3
={ , , , ......, }
d
T
n
o o o o
p
T
(x)={1, x, x
2
, x
3
, x
4
,..., x
p
}
(1D)
Basis function
Pascal triangle of monomials: 2D

xy
x
2
x
3
x
4
x
5
y
2
y
3
y
4
y
5
x
2
y
x
3
y
x
4
y
x
3
y
2

xy
2
xy
3
xy
4
x
2
y
3
x
2
y
2
Constant terms: 1
x y
1
Quadratic terms: 3
Cubic terms: 4
Quartic terms: 5
Quintic terms: 6
Linear terms: 2
3 terms
6 terms
10 terms
15 terms
21 terms
{ }
2 2
( ) ( , ) 1, , , , , ,..., ,
T T p p
x y x y xy x y x y = = p x p
Pascal pyramid of monomials : 3D

x
x
2
x
3
x
4
y
y
2
y
3
y
4
xy
z
xz
yz
x
2
y xy
2
x
2
z
zy
2
z
2
xz
2 yz
2
xyz
z
3
x
3
y
x
3
z
x
2
y
2
x
2
z
2
x
2
yz

xy
3
zy
3
z
2
y
2
xy
2
z
xyz
2
xz
3
z
4 z
3
y
1
Constant term: 1
Linear terms: 3
Quadratic terms: 6
Cubic terms: 10
Quartic terms: 15
4 terms
10 terms
20 terms
35 terms
{ }
2 2 2
( ) ( , , ) 1, , , , , , , , , ,..., , ,
T T p p p
x y z x y z xy yz zx x y z x y z = = p x p
Displacement interpolation
Enforce approximation to be equal to the nodal
displacements at the nodes
d
i
= p
T
(x
i
)o i = 1, 2, 3, ,n
d

or
d
e
=P

o
where
1
2
=
d
e
n
d
d
d



`


)
d
T
1
T
2
T
( )
( )
( )
d
n
(
(
(
=
(
(
(

p x
p x
P
p x
,
Moment matrix
Displacement interpolation
The coefficients in o can be found by

e

=
1
P d
Therefore, u
h
(x) = N( x) d
e


| |
1 2
1 1 1 1
1 2
( ) ( ) ( )
1 2
( ) ( ) ( ) ( ) ( )
( ) ( ) ( )
n
T T T T
n
N N N
n
N N N

(
(
= =
(

=
x x x
N x p x P p x P p x P p x P
x x x
Displacement interpolation
Sufficient requirements for FEM shape
functions

( )
1 , 1, 2, ,
0 , , 1, 2, ,
d
i j ij
d
i j j n
N
i j i j n
o
= =

= =

= =

x 1.
(Delta function
property)
1
( ) 1
n
i
i
N
=
=

x
2.
(Partition of unity property
rigid body movement)
1
( )
d
n
i i
i
N x x x
=
=

3.
(Linear field reproduction property)
Step 3: Formation of FE equations in local
coordinates
Since U= Nd
e

Therefore,
c = LU
c = L N d
e
= B d
e
Strain matrix
or where
(Stiffness matrix)
e
T
V
e
T
e e
T T
e
V
e
T
V
e
V c V c V c d d B B d d Bd B d d ) (
2
1
2
1

2
1
} } }
= = =
V c
T
V
e
e
d B B k
}
=
1
2
T
e e e
H = d k d
Step 3: Formation of FE equations in local
coordinates
Since U= Nd
e


e
= U Nd
or
e e
T
e
T d m d

2
1
=
where
(Mass matrix)
1 1 1
d d ( d )
2 2 2
e e e
T T T T T
e e e e
V V V
T V V V = = =
} } }
U U d N Nd d N N d
d
e
T
e
V
V =
}
m N N
Step 3: Formation of FE equations in local
coordinates
e
T
e s
T
e b
T
e f
W F d F d F d = + =
s b e
F F f + =
(Force vector)
d d ( d ) ( d )
e e e e
T T T T T T T T
f e b e s e b e s
V S V S
W V S V S = + = +
} } } }
d N f d N f d N f d N f
d
e
T
b b
V
V =
}
F N f
d
e
T
s s
S
S =
}
F N f
Step 3: Formation of FE equations in local
coordinates
) (
d
d
)
d
d
(
T
e
T
e
T
e
t t
d
d
d o o o = =

t t t
e e
t
t
T
e e e
t
t
T
e
t
t
e e
T
e e e
t
t
T
e
d d d
2
1
2
1
2
1
2
1






d m d d m d d m d d m d

} } }
= = o o o o
0 d ) (
2
1


= +
}
t
e e e e
T
e
t
t
F kd d m d

o
0 d )
2
1
2
1
(
2
1


= +
}
t
e
T
e e e
T
e e e
T
e
t
t
F d d k d d m d

o
e e e e e
f d m d k = +

FE Equation

(Hamiltons principle)
2
1


( )d 0
t
T T T
e e e e e e e e
t
t o o o + =
}
d m d d k d d f
Step 4: Coordinate transformation
e e e e
f d m kd = +

x



y



x'
y'
y'



x'



Local coordinate
systems


Global
coordinate
systems



e e
TD d =
e e e e e
F D M D K = +

T k T K
e
T
e
= T m T M
e
T
e
=
e
T
e
f T F = , ,
where
(Local)
(Global)
Step 5: Assembly of FE equations
Direct assembly method
Adding up contributions made by elements
sharing the node
F D M KD = +

F KD =
(Static)
e e e e e
F D M D K = +

Step 6: Impose displacement constraints
No constraints rigid body movement
(meaningless for static analysis)
Remove rows and columns corresponding
to the degrees of freedom being constrained
K is semi-positive definite
Step 7: Solve the FE equations
Solve the FE equation,

for the displacement at the nodes, D

The strain and stress can be retrieved by
using c = LU and o = c c with the
interpolation, U=Nd
F D M KD = +

STATIC ANALYSIS
Solve KD=F for D

Gauss elimination
LU decomposition
Etc.
EIGENVALUE ANALYSIS
0 = + D M KD
(Homogeneous equation, F = 0)
Assume
) exp( t ie | = D
0 ] [
2
= | M K e
Let
2
e =

0 ] [ = | M K
0 ] det[ = = M K M K
[ K
i
M ] |
i
= 0
(Eigenvector)
(Roots of equation are the
eigenvalues)
EIGENVALUE ANALYSIS
Methods of solving eigenvalue equation
Jacobis method
Givens method and Householders method
The bisection method (Sturm sequences)
Inverse iteration
QR method
Subspace iteration
Lanczos method
TRANSIENT ANALYSIS
Structure systems are very often subjected to
transient excitation.
A transient excitation is a highly dynamic time
dependent force exerted on the structure, such as
earthquake, impact, and shocks.
The discrete governing equation system usually
requires a different solver from that of eigenvalue
analysis.
The widely used method is the so-called direct
integration method.
TRANSIENT ANALYSIS
(reading material)
The direct integration method is basically using
the finite difference method for time stepping.
There are mainly two types of direct integration
method; one is implicit and the other is explicit.
Implicit method (e.g. Newmarks method) is more
efficient for relatively slow phenomena
Explicit method (e.g. central differencing method)
is more efficient for very fast phenomena, such as
impact and explosion.

REMARKS
In FEM, the displacement field U is expressed by
displacements at nodes using shape functions N
defined over elements.
The strain matrix B is the key in developing the
stiffness matrix.
To develop FE equations for different types of
structure components, all that is needed to do is
define the shape function and then establish the
strain matrix B.
The rest of the procedure is very much the same
for all types of elements.
Newmarks method (Implicit)
Assume that
( ) ( )
2 1
2
t t t t t t t
t t | |
+A +A
(
| |
= + A + A +
| (
\ .

D D D D D
( ) ( )
1
t t t t t t
t
+A +A
( = + A +

D D D D
+ + = KD CD MD F
Substitute into
( ) ( )
( ) ( )
{ }
2 1
2
1
t t t t t
t t t t t t t t
t t
t
| |

+A
+A +A +A

(
| |
+ A + A + +
`
|
(
\ .

)
( + A + + =

K D D D D
C D D D MD F
Typically
= 0.5
| = 0.25
Newmarks method (Implicit)
residual
cm t t t t +A +A
= K D F
where
( )
2
cm
t t |
(
= A + A +

K K C M
( ) ( ) ( )( ) { }
2
residual
1
1
2
t t t t t t t t t
t t t |
+A +A

| |
= + A + A + A
`
|
\ .
)
F F K D D D C D D
Therefore,
1
cm
residual
t t t t

+A +A
= D K F
Newmarks method (Implicit)
Start with D
0
and
0
D
Obtain
0
D + + = KD CD MD F
using
1
cm
residual
t t t t

+A +A
= D K F Obtain
t A
D
using
Obtain D
At
and
t A
D
using
( ) ( )
2 1
2
t t t t t t t
t t | |
+A +A
(
| |
= + A + A +
| (
\ .

D D D D D
( ) ( )
1
t t t t t t
t
+A +A
( = + A +

D D D D
March
forward
in time
Central difference method (explicit)
int residual
( = + = =

MD F CD KD F F F
residual
=
1
D M F (Lumped mass no need to solve matrix equation)
( )
2
t t t t t
t
+A A
= A + D D D
( )
2
t t t t t
t
+A A
= A + D D D
( )
( )
2
1
2
t t t t t t
t
+A A
= +
A
D D D D
( )
( )
2
2
t t t t t
t
t
A
A
= A + D D D D
Central
difference
method (explicit)
D,
t
x
x
x
x
x
t
0
At




-At







-At/2










At/2







Find average velocity at time t =
-At/2 using



Find using the average acceleration at
time t = 0.

Find D
At
using the average velocity at time t =At/2

Obtain D
-At
using
D
0
and are
prescribed and
can be obtained from
Use to
obtain assuming .
Obtain using
Time marching in half the time step
0
D
0
D
residual
=
1
D M F
( )
( )
2
2
t t t t t
t
t
A
A
= A + D D D D
/ 2 t A
D
( )
/ 2 / 2 t t t t t
t
+A A
= A + D D D
/ 2 t A
D
( )
/ 2 / 2 t t t t t
t
+A A
= A + D D D
( )
/ 2 / 2 t t t t t
t
+A A
= A + D D D
( )
/ 2 / 2 t t t t t
t
+A A
= A + D D D
t A
D
/ 2 0 t A
= D D
t A
D
residual
=
1
D M F