The Finite Element Method For The Analysis of Linear Systems y y
The Finite Element Method For The Analysis of Linear Systems y y
The Finite Element Method For The Analysis of Linear Systems y y
Page 1
Page 2
C t t off T
Contents
Today's
d ' Lecture
L t
Natural coordinates
Page 3
Th Method
The
M th d off Finite
Fi it Elements
El
t
Th principle
The
i i l off virtual
i t l displacements:
di l
t
We may now express the principle of virtual displacements in a more
general form:
dV = U f dV +
T
SfT
f dS + U iT R iC
Sf
Sf
Page 4
Th Method
The
M th d off Finite
Fi it Elements
El
t
Fi it Element
Finite
El
t Equations:
E
ti
We now consider the volume modeled as an assemblage of N elements
connected in the nodal points on the element boundaries
The displacements within the individual elements are measured in a
convenient local coordinate system x,y,z
For element m we now write the displacements within the element as a
function of the total set of global nodal displacements U
u ( m ) ( x, y , z ) = H ( m ) ( x, y , z ) U
T = [U , V , W , U , V ,W ,... U , V , W ] = [U , U ,...U ]
U
1 1
1
2
2
2
1
2
N
N
N
n
Method of Finite Elements 1
Page 5
Th Method
The
M th d off Finite
Fi it Elements
El
t
Fi it Element
Finite
El
t Equations:
E
ti
For element m we now write the strains within the element as a function
of the total set of global nodal displacements U
( m ) ( x, y , z ) = B ( m ) ( x, y , z ) U
The stresses are then:
( m ) ( x, y, z ) = C ( m ) ( x, y, z ) + i ( m ) ( x, y, z )
Page 6
Th Method
The
M th d off Finite
Fi it Elements
El
t
Fi it Element
Finite
El
t Equations:
E
ti
We can now write the equlibrium equations for the total volume by
summing up over the N elements
N
( m )T
(m)
dV
(m)
m =1 V ( m )
N
m =1
m =1 V ( m )
S f ( m )T
S f (m)
S (f m1 ) , S (f m2 ) ,...
( m )T i ( m ) dV ( m ) +
m =1 V ( m )
iT
i
U
R
C
i
dS ( m )
U ( m )T f B ( m ) dV ( m ) +
Page 7
Th Method
The
M th d off Finite
Fi it Elements
El
t
Fi it Element
Finite
El
t Equations:
E
ti
As a next step we represent both the real unknown displacement fields
as well as the virtual displacement fields through the interpolation
functions (provides symmetrical stiffness matrixes )
T
U
T
U
( m )T ( m ) ( m )
(m)
B C B dV U =
m =1 V ( m )
N
N
( m )T B ( m )
(m)
( m )T S f ( m )
(m)
dV +
H f
dS
H f
m =1 S ( m ) , S ( m ) ,..
f1
f2
m =1 V ( m )
B ( m )T i ( m ) dV ( m ) + R C
m =1 V ( m )
Page 8
Th Method
The
M th d off Finite
Fi it Elements
El
t
Fi it Element
Finite
El
t Equations:
E
ti
Now we may finally simplify as
KU = R
N
K=
R = R B + R S R I + RC
( m )T
(m)
(m)
C B dV
(m)
m =1 V ( m )
RB =
m =1 V ( m )
N
RS =
m =1 S
N
RI =
H ( m )T f
S f (m)
(m)
, S f 2( m ) ,..
f1
m =1 V ( m )
RC = RC
Method of Finite Elements 1
H ( m )T f B ( m ) dV ( m )
B ( m )T i ( m ) dV
d (m)
dS ( m )
Page 9
G
Generalized
li d coordinate
di t models
d l
G
Generalized
li d coordinate
di t models:
d l
The principle behind these models is:
Formulate displacement field in terms of polynomials
, , : Genenralized
G
li d coordinates
di
one-dimensional
u ( x) = 1 + 2 x + 3 x 2 + 4 x3 +"
two-dimensional
u ( x, y ) = 1 + 2 x + 3 xy + 4 x 2 +"
v( x, y ) = 1 + 2 x + 3 xy + 4 x +"
2
plate bending
w( x, y ) = 1 + 2 x + 3 xy + 4 x 2 +"
Method of Finite Elements 1
u =
Page 10
G
Generalized
li d coordinate
di t models
d l
G
Generalized
li d coordinate
di t models:
d l
Next step is to relate the generalized coordinates to the nodal
displacements:
We insert the nodal coordinates into
u =
and get:
u = A = A 1u
= E
= C
Page 11
O the
On
th choice
h i off shape
h
functions
f
ti
R
Requirements
i
t to
t shape
h
functions:
f
ti
First of all the functions which we want to represent (displacements
and their derivatives) need to be able to represent the physics of the
type of problem we are aiming to model within the individual elements.
Secondly we also need to be concerned with continuity over the
borders of the elements.
We here introduce continuity requirements:
C0 continuity
C1 continuity
Cm continuity
Page 12
O the
On
th choice
h i off shape
h
functions
f
ti
Sh
Shape
functions:
f
ti
The displacement e.g. u or v at any location within the element (x,y)
can be represented as a function of the nodal displacements (u or v):
u ( x, y ) = f1 (u )
v( x, y ) = f 2 ( v )
4 u4 , v4
u1 , v1 1
3 u3 , v3
2 u2 , v2
Page 13
O the
On
th choice
h i off shape
h
functions
f
ti
Sh
Shape
functions:
f
ti
In general we may write the approximate relation between the field
representation and the nodal displacements as:
u ( x, y ) = H ( x, y )u
T
Displacements
in the u-direction
scalar
Shape/interpolation
functions
vector with dimension n
Nodal
N
d l point
i t displacements
di l
t
in the u-direction
vector with dimension n
O the
On
th choice
h i off shape
h
functions
f
ti
Sh
Shape
functions:
f
ti
Polynomials are usually applied for the development of shape
functions (polynomials are easily differentiated analytically)
- Langrange polynomials
complete
l t polynomial
l
i l expansions
i
- Serendipity polynomials
incomplete polynomial expansions
- Hermitian polynomials
polynomials including derivatives
Page 14
Page 15
O the
On
th choice
h i off shape
h
functions
f
ti
Sh
Shape
functions:
f
ti
Lagrange polynomials (one dimensional element):
n
H i ( x) = Li ( y ) u ( x) = Li ( x)ui
i =1
n +1
Li ( x) =
j =1,
j
x xj
xi x j
u ( x) = L1 ( x)u1 + L2 ( x)u2
( x2 x)
( x x1 )
=
u1 +
u2
x2 x1
x2 x1
Method of Finite Elements 1
O the
On
th choice
h i off shape
h
functions
f
ti
Sh
Shape
functions:
f
ti
Lagrange polynomials (general):
n
u ( x, y, z ) = Li ( x, y, z )ui
i =1
n
u ( x, y, z ) = ui
i =1
Li ( x j , y j , z j ) = 1, i = j
Li ( x j , y j , z j ) = 0, i j
n
L ( x, y , z ) = 1
=1
Page 16
Page 17
O the
On
th choice
h i off shape
h
functions
f
ti
Sh
Shape
functions:
f
ti
Lagrange polynomials (four node rectangular element):
Products of two one dimensional first order Lagrange polynomials
result in the bi-linear four-node element
u ( x, y ) = a1 + a2 x + a3 y + a4 xy
= (b1 + b2 x)(b3 + b4 y )
u4 , v4
u1 , v1 1
3 u3 , v3
2 u2 , v2
Page 18
O the
On
th choice
h i off shape
h
functions
f
ti
Sh
Shape
functions:
f
ti
Lagrange polynomials (four node rectangular element):
4
u4 , v4
u1 , v1 1
3 u3 , v3
2 u2 , v2
u ( x, y ) =
x x3 y y2
x2 x y3 y
x x1 y4 y
x4 x y y1
u1 +
u2 +
u3 +
u4
x2 x1 y3 y1
x2 x1 y4 y2
x4 x3 y3 y1
x4 x3 y4 y2
Method of Finite Elements 1
Page 19
O the
On
th choice
h i off shape
h
functions
f
ti
Sh
Shape
functions:
f
ti
From Pascals triangle we can see how many nodes are required for
the representation of displacement fields of an
any order and
completeness:
x
x2
x3
x4
First order
y2
xy
x2 y
x3 y
Zero order
xy 2
x2 y2
S
Second
d order
d
y3
xy 3
Third order
y4
Fourth order
Page 20
O the
On
th choice
h i off shape
h
functions
f
ti
Sh
Shape
functions:
f
ti
Products of Lagrange polynomials (bi-linear four node rectangular)
1
x
x2
x3
x4
First order
y2
xy
x2 y
x3 y
Zero order
xy 2
x2 y2
S
Second
d order
d
y3
xy 3
Third order
y4
Fourth order
Page 21
O the
On
th choice
h i off shape
h
functions
f
ti
Sh
Shape
functions:
f
ti
Products of Lagrange polynomials (quadratic nine-node rectangular)
This requires
q
an inner node
a difficulty !
x
x2
x3
x4
First order
y2
xy
x2 y
x3 y
Zero order
xy 2
x2 y2
S
Second
d order
d
y3
xy 3
Third order
y4
Fourth order
Page 22
O the
On
th choice
h i off shape
h
functions
f
ti
Sh
Shape
functions:
f
ti
Serendipity shape functions are constructed by incomplete
pol nomials avoiding
polynomials
a oiding inner nodes
1
x
x2
x3
x4
First order
y2
xy
x2 y
x3 y
Zero order
xy 2
x2 y2
S
Second
d order
d
y3
xy 3
Third order
y4
Fourth order
O the
On
th choice
h i off shape
h
functions
f
ti
Sh
Shape
functions:
f
ti
Whereas difficulties may arise (inner nodes) when aiming to develop
quadratic shape functions for rectangular elements using Lagrange
polynomials the shape functions developed by incomplete
polynomials (serendipity shape functions) less terms necessitates
less nodes !
An bi-quadratic eight node rectangular element can be constructed !
Page 23
Page 24
O the
On
th choice
h i off shape
h
functions
f
ti
Sh
Shape
functions:
f
ti
Hermitian shape functions relate not only the displacements at nodes
to displacements within
ithin the elements but
b t also the first order
derivatives
u1
u2
u1 ,
u2 ,
ui
u ( x) = H 0i ( x)ui + H1i ( x)
i =1
H 01
H 02
Slope=1
H11
H12
Slope=1
Page 25
N t
Natural
l coordinates
di t
Sh
Shape
functions
f
ti
Natural
N t
l coordinates:
di t
As we have seen we are able to establish shape functions in global or
local coordinate systems
s stems as we
e please.
please Ho
However,
e er for the p
purpose
rpose of
standardizing the process of developing the element matrixes it is
convenient to introduce the so-called natural coordinate system.
s
-1, 1
1, 1
r
x
Method of Finite Elements 1
-1, -1
1, -1
Page 26
N t
Natural
l coordinates
di t
Sh
Shape
functions
f
ti
Natural
N t
l coordinates:
di t
Let us consider the simple bar element
The relation between the x-coordinate
and the r-coordinate is given as:
x2
x1
1
1
x = (1 r ) x1 + (1 + r ) x2
2
2
2
= hi xi
r = 1
i =1
u=
r =0
r =1
1
1
(1 r )u1 + (1 + r )u2
2
2
= hi ui
i=1
u2
u1
Page 27
N t
Natural
l coordinates
di t
Sh
Shape
functions
f
ti
Natural
N t
l coordinates:
di t
Let us consider the simple bar element
We need to be able to establish the strains meaning we need to be
able to take the derivatives of the displacement filed in regard to the
x coordinate
x-coordinate
x
du du dr
=
=
dx dr dx
du d 1
1
1
= ( (1 r )u1 + (1 + r )u2 ) = (u2 u1 )
2
2
dr dr 2
1
1
dx d 1
= ( ((1 r ) x1 + ((1 + r ) x2 ) = ( x2 x1 )
dr dr 2
2
2
du (u2 u1 ) (u2 u1 )
=
=
dx ( x2 x1 )
L
x1
u2
u1
r = 1
r =0
r =1
Page 28
N t
Natural
l coordinates
di t
Sh
Shape
functions
f
ti
Natural
N t
l coordinates:
di t
Let us consider the simple bar element
The strain-displacement matrix then becomes:
B=
1
[ 1 1]
L
AE 1
K = 2 [ 1 1] Jdr ,
L 1 1
1
AE 1 1
K=
L 1 1
Method of Finite Elements 1
J=
dx L
=
dr 2