MANE 4240 & CIVL 4240

Introduction to Finite Elements

Mapped element geometries
and shape functions: the
isoparametric formulation
Prof. Suvranu De
Reading assignment:

Chapter 10.1-10.3, 10.6 + Lecture notes
Summary:

Concept of isoparametric mapping
1D isoparametric mapping
Element matrices and vectors in 1D
2D isoparametric mapping : rectangular parent elements
2D isoparametric mapping : triangular parent elements
Element matrices and vectors in 2D
For complex geometries
2002 Brooks/Cole Publishing / Thomson Learning
General quadrilateral elements
Elements with curved sides
s
t
1
1
1 2
3 4
1
1
Consider a special 4-noded rectangle in its local coordinate
system (s,t)
4 4 3 3 2 2 1 1
4 4 3 3 2 2 1 1
v N v N v N v N v
u N u N u N u N u
+ + + ~
+ + + ~
Displacement interpolation
) 1 )( 1 (
4
1
) , (
) 1 )( 1 (
4
1
) , (
) 1 )( 1 (
4
1
) , (
) 1 )( 1 (
4
1
) , (
4
3
2
1
t s t s N
t s t s N
t s t s N
t s t s N
+ =
=
+ =
+ + =
Shape functions in local coord system
Recall that
t t N t N t N t N
s s N s N s N s N
N N N N
= + + +
= + + +
= + + +
4 4 3 3 2 2 1 1
4 4 3 3 2 2 1 1
4 3 2 1
1
Rigid body modes
Constant
strain states
Goal is to map this element from local coords to a general
quadrilateral element in global coord system
s
t
1
1
1 2
3 4
1
1
Local coordinate system
s
y
1
2
3
4
x
s
t
Global coordinate system
) , (
) , (
t s y y
t s x x
=
=
) , (
) , (
y x t t
y x s s
=
=
) , (
1
t s N ) , (
1
y x N
) , ( )) , ( ), , ( ( ) , (
1 1 1
y x N y x t y x s N t s N =
In the mapped coordinates, the shape functions need to satisfy
1. Kronecker delta property

=
nodes other all at
i node at
N
i
0
1
y y N
x x N
N
i
i
i
i
i
i
i
i
=
=
=

1
Then
2. Polynomial completeness

The relationship
i
i
i
i
i
i
y t s N y
x t s N x
) , (
) , (

=
=
Provides the required mapping from the local coordinate system
To the global coordinate system and is known as isoparametric
mapping
(s,t): isoparametric coordinates
(x,y): global coordinates
s
t
1
1
1
1
s
t
y
x
Examples
s
t
1
1
y
x
s
t
1D isoparametric mapping
1 2
x
2
x
1
x

1 2
s

1 1
Local (isoparametric)
coordinates
3 noded (quadratic) element
3
3
x
3
( )
( )
2
3
2
1
1 ) (
2
1
) (
2
1
) (
s s N
s s
s N
s s
s N
=
+
=

=
Isoparametric mapping

=
=
3
1
) (
i
i i
x s N x
( ) ( )
( )
3
2
2 1
1
2
1
2
1
x s x
s s
x
s s
x +
+
+

=
( ) ( )
( )
3
2
2 1
1
2
1
2
1
x s x
s s
x
s s
x +
+
+

=
NOTES
1. Given a point in the isoparametric coordinates, I can obtain the
corresponding mapped point in the global coordinates using the
isoparametric mapping equation
2
3
1
; 1
; 0
; 1
x x s At
x x s At
x x s At
= =
= =
= =
Question
x=? at s=0.5?
2. The shape functions themselves get mapped
In the isoparametric coordinates (s) they are polynomials.
In the global coordinates (x) they are in general nonpolynomials
Lets consider the following numerical example
4 ; 6 ; 0
3 2 1
= = = x x x
1 2
x

3
4 2
Isoparametric mapping x(s)
( ) ( )
( )
( ) ( )
( )
2
1 2 3
2
2
1 1
1
2 2
1 1
0 6 1 4
2 2
4 3
s s s s
x x x s x
s s s s
s
s s
+
= + +
+
= + +
= +
Inverse mapping s(x)
2
4 25 3 x
s

=
Simple polynomial
Complicated function
Now lets compute the shape functions in the global coordinates
( )
( )
) (
4 25 2 10
2
1
2
4 25 3
1
2
4 25 3
2
1
2
1
) (
2
2
x N
x x
x x
s s
s N
=
=
|
|
.
|

\
|

+
|
|
.
|

\
|

=
+
=
Now lets compute the shape functions in the global coordinates
N
2
(x) is a complicated function
However, thanks to isoparametric mapping, we always ensure
1. Knonecker delta property
2. Rigid body and constant strain states
1 2
s

1 1
3
N
2
(s)
1 2
x

3
4 2
1
1
N
2
(x)
N
2
(s) is a simple polynomial
( )
2
1
) (
2
s s
s N
+
=
( ) x x x N 4 25 2 10
2
1
) (
2
=
Element matrices and vectors for a mapped 1D bar element
1 2
x

1 2
s

1 1
3
3
Displacement interpolation
d N u N u N u N u = + + ~
3 3 2 2 1 1
Strain-displacement relation d B u
dx
dN
u
dx
dN
u
dx
dN
dx
du
= + + ~ =
3
3
2
2
1
1
c
Stress
d B E E = = c o
The strain-displacement matrix
(

=
dx
dN
dx
dN
dx
dN
B
3 2 1
The only difference from before is that the shape functions are in the
isoparametric coordinates
( )
( )
2
3
2
1
1 ) (
2
1
) (
2
1
) (
s s N
s s
s N
s s
s N
=
+
=

=
And we will not try to obtain explicitly the inverse map.
How to compute the B matrix?
We know the isoparametric mapping

=
=
3
1
) (
i
i i
x s N x
Do I know
?
) (
ds
s dN
i
Do I know
?
dx
ds
I know

=
=
3
1
) (
i
i i
x s N x
Hence
) (
) (
3
1
mapping of Jacobian J x
ds
s dN
ds
dx
i
i
i
=

=
ds
s dN
J dx
s dN
i i
) ( 1 ) (
=
From (*)
dx
ds
ds
s dN
dx
s dN
i i
) ( ) (
=
Using chain rule
(*)
What does the Jacobian do?
ds J dx =
Maps a differential element from the isoparametric coordinates to
the global coordinates
The strain-displacement matrix
(

=
(

=
ds
dN
ds
dN
ds
dN
J
dx
dN
dx
dN
dx
dN
B
3 2 1
3 2 1
1
For the 3-noded element
3 2 1
3
1
2
2
1 2
2
1 2 ) (
sx x
s
x
s
x
ds
s dN
J
i
i
i

+
+

= =

=
(

+
= s
s s
J
B 2
2
1 2
2
1 2 1
The element stiffness matrix
}
}

= =
=
1
1
2
1
Jds dx Jds B B EA
dx B B EA k
T
x
x
T

NOTES
1. The integral on ANY element in the global coordinates in now an
integral from -1 to 1 in the local coodinates
2. The jacobian is a function of s in general and enters the integral.
The specific form of J is determined by the values of x
1
, x
2
and x
3
.
Gaussian quadrature is used to evaluate the stiffness matrix
3. In general B is a vector of rational functions in s
Isoparametric mapping in 2D: Rectangular parent elements
2002 Brooks/Cole Publishing / Thomson Learning
Parent element
Isoparametric mapping
Mapped element in
global coordinates
i
i
i
i
i
i
y t s N y
x t s N x
) , (
) , (

=
=
1
2
3
4
1
( , ) (1 )(1 )
4
1
( , ) (1 )(1 )
4
1
( , ) (1 )(1 )
4
1
( , ) (1 )(1 )
4
N s t s t
N s t s t
N s t s t
N s t s t
=
= +
= + +
= +
Isoparametric mapping
i
i
i
i
i
i
y t s N y
x t s N x
) , (
) , (

=
=
Shape functions of parent element in isoparametric
coordinates
NOTES:
1. The isoparametric mapping provides the map (s,t) to (x,y) , i.e.,
if you are given a point (s,t) in isoparametric coordinates, then
you can compute the coordinates of the point in the (x,y)
coordinate system using the equations





2. The inverse map will never be explicitly computed.

i
i
i
i
i
i
y t s N y
x t s N x
) , (
) , (

=
=
s
t
1
1
3 4
1 2
1
1
7
8
5
6
8-noded Serendipity element
2002 Brooks/Cole Publishing / Thomson Learning
8-noded Serendipity element: element shape functions in
isoparametric coordinates
1 5
2 6
3 7
4 8
1 1
( , ) (1 )(1 )( 1) ( , ) (1 )(1 )(1 )
4 2
1 1
( , ) (1 )(1 )( 1) ( , ) (1 )(1 )(1 )
4 2
1 1
( , ) (1 )(1 )( 1) ( , ) (1 )(1 )(1 )
4 2
1 1
( , ) (1 )(1 )( 1) ( , ) (1 )
4 2
N s t s t s t N s t t s s
N s t s t s t N s t t t s
N s t s t s t N s t t s s
N s t s t s t N s t s
= = +
= + = + +
= + + + = + +
= + + = (1 )(1 ) t t +
NOTES
1. N
i
(s,t) is a simple polynomial in s and t. But N
i
(x,y) is a complex
function of x and y.
2. The element edges can be curved in the mapped coordinates
3. A midside node in the parent element may not remain as a
midside node in the mapped element. An extreme example
s
t
1
1
1 2
3 4
1
1
5
6
7
8
x
y
1
8
5
7
2,6,3
4. Care must be taken to ensure UNIQUENESS of mapping
s
t
1
1
1 2
3 4
1
1
x
y
1
2
3
4
Isoparametric mapping in 2D: Triangular parent elements
s
t
2
3
1
1
1
P(s,t)
s
t
) 1 (
2
1
2
2
2
1
12
23
31
123
t s
s
t
P
P
P
= A
= A
= A
= A
t s L
t L
s L
P
P
P
=
A
A
=
=
A
A
=
=
A
A
=
1
123
12
3
123
31
2
123
23
1
Parent element: a right angled triangles
with arms of unit length

Key is to link the isoparametric
coordinates with the area coordinates
Now replace L
1
, L
2
, L
3
in the formulas for the shape functions of
triangular elements to obtain the shape functions in terms of (s,t)

Example: 3-noded triangle
y
x
s
t
2
3
1
1
1 (x
1
,y
1
)
s
t
Parent shape functions
t s N
t N
s N
=
=
=
1
3
2
1
Isoparametric mapping
3 3 2 2 1 1
3 3 2 2 1 1
) , ( ) , ( ) , (
) , ( ) , ( ) , (
y t s N y t s N y t s N y
x t s N x t s N x t s N x
+ + =
+ + =
2 (x
2
,y
2
)
3 (x
3
,y
3
)
1
Element matrices and vectors for a mapped 2D element
Stress approximation
Recall: For each element
Element stiffness matrix
Element nodal load vector
d N u =
d B D = o
d B =
}
=
e
V
k dV B D B
T


S
e
T
b
e
f
S
S
T
f
V
T
dS T dV X f
} }
+ = N N
Displacement approximation
Strain approximation
S
T
e
e
In isoparametric formulation
1. Shape functions first expressed in (s,t) coordinate system
i.e., N
i
(s,t)
2. The isoparamtric mapping relates the (s,t) coordinates with the
global coordinates (x,y)





3. It is laborious to find the inverse map s(x,y) and t(x,y) and we do
not do that. Instead we compute the integrals in the domain of the
parent element.
i
i
i
i
i
i
y t s N y
x t s N x
) , (
) , (

=
=
NOTE
1. N
i
(s,t) s are already available as simple polynomial functions
2. The first task is to find and
x
N
i
c
c
y
N
i
c
c
Use chain rule
t
y
y
N
t
x
x
N
t
y x N
s
y
y
N
s
x
x
N
s
y x N
i i i
i i i
c
c
c
c
+
c
c
c
c
=
c
c
c
c
c
c
+
c
c
c
c
=
c
c
) , (
) , (
In matrix form

c
c
c
c
(
(
(

c
c
c
c
c
c
c
c
=

c
c
c
c
y
N
x
N
t
y
t
x
s
y
s
x
t
N
s
N
i
i
i
i
Can be
computed
We want to
compute these
for the B matrix
This is known as the
Jacobian matrix (J)
for the mapping
(s,t) (x,y)

c
c
c
c
=

c
c
c
c
y
N
x
N
J
t
N
s
N
i
i
i
i

c
c
c
c
=

c
c
c
c

t
N
s
N
J
y
N
x
N
i
i
i
i
1
How to compute the Jacobian matrix?
Start from
i
i
i
i
i
i
y t s N y
x t s N x
) , (
) , (

=
=


c
c
=
c
c
c
c
=
c
c
c
c
=
c
c
c
c
=
c
c
i
i
i
i
i
i
i
i
i
i
i
i
y
t
t s N
t
y
y
s
t s N
s
y
x
t
t s N
t
x
x
s
t s N
s
x
) , (
;
) , (
) , (
;
) , (
(
(
(
(

c
c
c
c
c
c
c
c
=
(
(
(

c
c
c
c
c
c
c
c
=


i
i
i
i
i
i
i
i
i
i
i
i
y
t
t s N
x
t
t s N
y
s
t s N
x
s
t s N
t
y
t
x
s
y
s
x
J
) , ( ) , (
) , ( ) , (
Need to ensure that det(J) > 0 for one-to-one mapping
3. Now we need to transform the integrals from (x,y) to (s,t)
Case 1. Volume integrals
} } } }

= =
1
1
1
1
) det( ) , ( ) , ( ) , ( dsdt J h t s f dA h t s f dV t s f
e e
A V
h=thickness of element
This depends on the key result
dsdt J dxdy dA ) det( = =
s
t
1
1
1 2
3 4
1
1
Problem: Consider the following isoparamteric map
y
1
2
3
4
(3,1) (6,1)
(6,6)
(3,6)
x ISOPARAMETRIC COORDINATES
y
GLOBAL COORDINATES
4 4 3 3 2 2 1 1
4 4 3 3 2 2 1 1
v N v N v N v N v
u N u N u N u N u
+ + + ~
+ + + ~
Displacement interpolation
) 1 )( 1 (
4
1
) , (
) 1 )( 1 (
4
1
) , (
) 1 )( 1 (
4
1
) , (
) 1 )( 1 (
4
1
) , (
4
3
2
1
t s t s N
t s t s N
t s t s N
t s t s N
+ =
=
+ =
+ + =
Shape functions in isoparametric coord system
1 1 2 2 3 3 4 4
1 1 2 2 3 3 4 4
( , ) ( , ) ( , ) ( , )
( , ) ( , ) ( , ) ( , )
x N s t x N s t x N s t x N s t x
y N s t y N s t y N s t y N s t y
= + + +
= + + +
The isoparamtric map
(1 )(1 ) (1 )(1 ) (1 )(1 ) (1 )(1 )
6 3 3 6
4 4 4 4
3(1 )
2
(1 )(1 ) (1 )(1 ) (1 )(1 ) (1 )(1 )
6 6 1 1
4 4 4 4
7 5
2
s t s t s t s t
x
s
s t s t s t s t
y
t
+ + + +
= + + +
+
=
+ + + +
= + + +
+
=
In this case, we may compute the inverse map, but we will NOT do that!
The Jacobian matrix
NOTE: The diagonal terms are due to stretching of the sides along the x-and y-
directions. The off-diagonal terms are zero because the element does not shear.
3
0
2
5
0
2
x y
s s
J
x y
t t
c c (
(
c c
= (
c c
(
(
c c
(
(
= (
(
(

3(1 )
2
7 5
2
s
x
t
y
+
=
+
=
since
1
2/ 3 0
15
and det( )
0 2/ 5 4
J J

(
= =
(

Hence, if I were to compute the first column of the B matrix along the
positive x-direction
I would use
1
1
1
0
N
x
N
y

c

c


=
`

c

c

)
B
1
1
1
1
1
1 1
2/ 3 0
6 4
0 2/ 5 1 1
4 10
N
N t t
x
s
J
N
N s s
y
t

c c + +


( c c
= = =
` ` ` `
(
c
c + +


c
c ) ) )
)
1
1
1
1
6
0 0
1
10
N t
x
N s
y

c +


c



= =
` `

c +

c )
)
B
Hence
The element stiffness matrix
1 1
T T
1 1
B D B dV B D B det( ) dsdt
e
V
k J h

= =
} } }