Finite Element Method: Iso-Parametric Formulation

Iso-Parametric Formulation

Outlines

l It makes formulations for computer program simple

l It allows to create elements with a shape of a straight line or a curved

surface.
Make it possible to choose a variety of factors.

l We will derive the stiffness matrix of simple beam elements and

rectangular elements using an iso-parametric formulation.

l Numerical integration: We will calculate the stiffness matrix of

rectangular elements that is made using an iso-parametric formulation.

l Finally, we will consider several higher-order elements and shape

functions.

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

1 Iso_parametric formulation: Stiffness matrix of a beam element

The term of iso-parametric formulation comes from the usage of shape functions [N] which is
used to determine an element shape for approximation of deformation.

l If a deformation function is u = a1 + a2 s , use a node x = a1 + a2 s on a beam element.

l It is formulated using the natural (or intrinsic) coordinate system, s, defined by
geometry of elements. A transformation mapping is used for the element formulation
between natural coordinate system, s, and global coordinate system, x.
.

Step 1: Determination of element type

(a) (b)

Fig. 1: Linear beam element at node x in (a) natural coordinate system, s , (b) global
coordinate system, x.

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

Relation between s and x coordinate systems: (when s and x coordinate systems are parallel)
L
x = xc + s xc indicates center of element
2
x can be expressed as a function of x1 and x2
1
x = [(1 - s) x1 + (1 + s ) x2 ] = [ N1
R xU
N ]S V
1

2
2
Tx W2
Then shape functions are
1- s 1+ s
N1 = N2 = Note : N1 + N 2 = 1
2 2

Fig. 2: Shape functions in natural coordinate system

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

{u} = [ N1 N 2 ]
RSu UV
1
Step 2: Determination of deformation function
Tu W
2

u and x are called iso-parameter because they are defined by the same shape function at
the same node.

Step 3: Definition of strain-displacement and stress-strain relations

Calculation of element matrix [ B ] :

FG du IJ = [- 1 , 1 ] RSu UV 1

du H ds K 2 2 Tu W
dx F dx I FG L IJ
du du dx = = 2

- By chain rule =
ds dx ds Þ
GH ds JK H 2K
LM 1 1 OP RSu UV
1
\ {e x } = -
N L L Q Tu W
2

LM 1 1 OP
- Therefore, {e} = [ B ]{d } [ B] = -
N L L Q
Mechanics and Design SNU School of Mechanical and Aerospace Engineering
Finite Element Method: Iso-Parametric Formulation

Step 4: Calculation of element stiffness matrix

Eelement stiffness matrix: [ k ] = z

0
L
[ B ] T [ D][ B ] Adx

- In general, matrix [ B ] is a function of s:

where J is Jacobian.
z L

0
f ( x ) dx = z
-1
1
f ( s) J ds

dx L
In case of 1-D, J = J . In case of simple beam element : J= =
ds 2

Ratio of element’s length between global and natural coordinate systems

- Stiffness matrix in a natural coordinate system:

LM OP
[k ] =
L
2 z
-1
1
[ B ]T E [ B ] A ds =
AE 1 -1
L N
E -1 1 Q

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

2 Rectangular plane stress element

Characteristics of rectangular element:

- It is easy to input data, and it is simple to calculate stress.
- Physical boundary conditions are not well approximated at the edge of rectangle.

Step 1: Determination of element type – using natural coordinate ( x , y )

R|u U| 1

||uv ||
1

||v ||
2

{d } = S V
2

||u ||
3

||u ||
v
3

|Tv |W
4

4
(11.2.1)
Four node rectangular element and nodal
displacement

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

Step 2: Determination of deformation function – element deformation functions, u and v , are

linear along the rectangular corner

1
u( x , y ) = [(b - x )(h - y )u1 + (b + x )(h - y ) u2
4bh
+ (b + x )(h + y )u3 + (b - x )(h + y )u4 ]
u( x , y ) = a1 + a2 x + a3 y + a4 xy
Þ 1
v ( x , y ) = a5 + a6 x + a7 y + a8 xy v( x, y) = [(b - x )(h - y ) v1 + (b + x )(h - y )v2
4bh
+ (b + x )(h + y ) v3 + (b - x )(h + y )v4 ]

\ {y } =
RSuUV = [ N ]{d} = LM N
1 0 N2 0 N3 0 N4 0OP{d}
Tv W N0 N1 0 N2 0 N3 0 N Q
4

where shape functions are

(b - x )(h - y ) (b + x )(h - y )
N1 = N2 =
4bh 4bh
(b + x )(h + y ) (b - x )(h + y )
N3 = N4 =
4bh 4bh

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

Step 3: Definition of strain-displacement and stress-strain relationships

Element strain in a 2-D stress state:

R| ¶u U|
R| e x
U| || ¶¶xv ||
{e} º S e V| = S| ¶y V| = [ B]{d}
|Tg y

xy W | ¶u + ¶v |
|T ¶y ¶x |W
where

1 M
L -(h - y ) 0 (h - y ) 0 (h + y ) 0 -(h + y ) 0 OP
4bh M PP
[ B] = 0 - (b - x ) 0 - (b + x ) 0 (b + x ) 0 (b - x )
MN-(b - x) - (h - y ) - (b + x ) (h - y ) (b + x ) (h + y ) (b - x ) - (h + y ) Q

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

Step 4: Calculation of element stiffness matrix and element equation

Element stiffness matrix: [ k ] = zz

h

- h -b
b
[ B ]T [ D][ B ]t dxdy

Element force matrix: { f }= zzz

V
[ N ]T { X } dV + { P} + zz
s
[ N ]T {T}dS

Element equation: l f q = k ld q
Step 5,6, and 7

Step 5, 6, and 7 are constitution of global stiffness matrix, determinant of

unknown deformation, calculation of stress. However, stress in each element
varies in all directions of x and y.

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation
3 Iso-parametric formulation: stiffness matrix of a plane element

A process of iso-parametric formulation is same in all elements

Step 1: Determination of element type

It is possible to numerically integrate the rectangular element defined in natural

coordinate system s - t .

Transformation equation: x = xc + bs y = yc + ht

(a) (b)

Fig. 4: (a) A linear rectangular element in a coordinate system, s - t , (b) A rectangular

element in a coordinate system, x - y , The size and shape of the rectangular
element are defined by coordinates of four nodes.

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation
Transformation equation between a local coordinate system, s - t , and a global coordinate
system, x - y :

1
x = [(1 - s)(1 - t ) x1 + (1 + s)(1 - t ) x2
4
+ (1 + s)(1 + t ) x3 + (1 - s)(1 + t ) x4 ]
x = a1 + a2 s + a3t + a4 st
Þ 1
y = a5 + a6 s + a7t + a8 st y = [(1 - s)(1 - t ) y1 + (1 + s)(1 - t ) y2
4
+ (1 + s)(1 + t ) y3 + (1 - s)(1 + t ) y4 ]

In a matrix form:
R| x U|
1
(1 - s)(1 - t )
|| xy ||
1 N1 =
4

RS x UV = LM N | |
0 O |y |
2
N2 =
(1 + s)(1 - t )

T yW N 0
1 0
N1
N2
0
0
N2
N3
0
0
N3
N4
0
S V
N PQ | x |
2 4
(1 + s)(1 + t )
|| y ||
4 3
N3 =
3 4
|| x ||
4 N4 =
(1 - s)(1 + t )
Ty W
4
4

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

1. Shape function is linear.

2. Any point in rectangular element ( s, t ) can be mapped to the quadrilateral element

point ( x , y ) in Fig. 4(b).

3. Note that for all values of s and t , N1 + N 2 + N 3 + N 4 = 1 .

4. N i (i＝1, 2, 3, 4) is 1 for node i , and 0 for the other nodes.

Two general conditions of shape functions:

n

1. å N =1
i =1
i (i = 1,2,..., n)

2. N i = 1 for node i , N i = 0 for the other nodes.

Additional conditions:

3. Continuity of deformation --- Lagrangian Interpolation

4. Continuity of slope --- Hermitian Interpolation

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

Fig. 5: Change of shape functions in a linear rectangular element

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

Step 2: Determination of deformation

Deformation functions in the element are defined by shape functions that are used to
define element shape.
R|u U|
1

||uv ||
1

RSuUV = LM N | |
0 O |v |
2

Tv W N 0
1 0
N1
N2
0
0
N2
N3
0
0
N3
N4
0
S V
N PQ |u |
2

||v ||
4 3

||u ||
4

Tv W
4

Step 3: Strain-displacement and stress-strain relationships

The derivative of deformation u and v about x and y should be executed using a chain rule of
derivation because the deformation function is expressed with s and t .

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

¶f ¶f ¶x ¶f ¶y
= +
¶s ¶x ¶s ¶y ¶s
Reference: chain rule of f ¶f ¶f ¶x ¶f ¶y
= +
¶t ¶x ¶t ¶y ¶t

Calculating ( ¶f / ¶x ) and ( ¶f / ¶y ) using Cramer’s lure (Appendix. B).

¶f ¶y ¶x ¶f ¶x ¶y
¶f 1 ¶s ¶s , ¶f = 1 ¶s
= ¶s J = ¶s ¶s
¶x J ¶f ¶y ¶y J ¶x ¶f where ¶x ¶y (*)
¶t ¶t ¶t ¶t ¶t ¶t

Element strain:
LM ¶( ) OP
R| e U| M ¶x 0
P
¶( ) P Ru U
V| = MM 0
x

e º Se S V = Bd
|Tg ¶y P T v W
y

xy W MM ¶( ) ¶( ) P
P
MN ¶y ¶x PQ
A formulation to obtain B is required.

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

Using the equation (*) in previous page (use u or v instead of f ):

LM ¶y ¶() - ¶y ¶() OP
R| e U| 1 M ¶t ¶s ¶s ¶t 0
¶x ¶() ¶x ¶() P Ru U
V| = J MM 0 P
x

S| e y SV
¶s ¶t ¶t ¶s P Tv W
-
Tg W MM ¶x ¶() - ¶x ¶() ¶y ¶() ¶y ¶() P
¶t ¶s ¶s ¶t PQ
xy

N ¶s ¶t ¶t ¶s -

LM ¶y ¶() - ¶y ¶() 0
OP
D' = M
1 M ¶t ¶s ¶s ¶t
¶x ¶() ¶x ¶() P
P
J M ¶s ¶t ¶t ¶s P
0 -
e = D' Nd
or where
MM ¶x ¶() - ¶x ¶() ¶y ¶() ¶y ¶() P
N ¶s ¶t ¶t ¶s ¶t ¶s ¶s ¶t PQ
-

Thus,
B = D' N
(3 ´ 8) (3 ´ 2) (2 ´ 8)

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

Step 4: Derivation of element stiffness matrix and equation

Stiffness matrix in a coordinate system, s - t :
[k ] = zz
A
[ B ]T [ D][ B ]tdxdy

Converge the integral region from x - y to s - t :

[k ] = zz
1

-1 -1
1 T
[ B ] [ D][ B ]t J dsdt

LM 0 1- t t-s s -1 OP
J = lX q M PPl q
1 t -1
T 0 s + 1 -s - t
Determinent J is 8 MMs - t
c
-s - 1 0 t +1
Yc
PQ
N1 - s s + t -t - 1 0

R| y U|
1

|
l q Sy Vy |
lX q
2
Y =
|| ||
T c
where c =[ x1 x2 x3 x4 ] , 3

Ty W4

J is a function of s , t in natural coordinate system, and x1 , x2 ,..., y4 in the

known global coordinate system.

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

1
Calculation of B : B ( s, t ) = [ B1 B 2 B 3 B 4 ]
J
where
LMa( N i ,s ) - b ( N i ,t ) 0 OP
B =M 0 c( N i ,t ) - d ( N i , s )P
MNc( N ) PQ
i
i = 1, 2, 3, 4
i ,t ) - d ( N i , s ) a ( N i , s ) - b ( N i ,t

and
1
a = [ y1 ( s - 1) + y2 ( -1 - s) + y3 (1 + s) + y4 (1 - s)]
4
1
b = [ y1 (t - 1) + y2 (1 - t ) + y3 (1 + t ) + y4 ( -1 - t )]
4
1
c = [ x1 (t - 1) + x2 (1 - t ) + x3 (1 + t ) + x4 ( -1 - t )]
4
1
d = [ x1 ( s - 1) + x2 ( -1 - s) + x3 (1 + s) + x4 (1 - s)]
4
1 1
For example, N1,s = (t - 1) N1,t = ( s - 1) (etc.)
4 4

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

Element body force matrix:

{ fb} =
(8 ´ 1)
zz
1

-1 -1
1
[ N ]T { X } t J ds dt
(8 ´ 2) (2 ´ 1)

Element surface force matrix: Length is L , an edge t = 1 (See. Fig. 4(b))

R| f U|
z
s 3s

{ f s} =
1
[ N ]T {T} t
L
ds |S f s 3t |V = z LM N
1 3 0 N4 0 OP RS p UVt L ds
T
s
2
|| f || N 0 Q Tp W 2
-1
, s4 s
-1 N3 0 N4 t
(4 ´ 1) (4 ´ 2) (2 ´ 1)
Tf s4r W
For N 1 = 0 and N 2 = 0 along the edge t = 1 , the nodal force is zero at nodes 1 and 2.

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

4 Gaussian Quadrature (Numerical integration)

One node Gaussian quadrature

I = z
-1
1
ydx » y1 *{(1) - ( -1)}
= 2 y1

If function y is straight line,

it has exact solution.

General equation: I= z
-1
1 n
y dx = å Wi yi
i =1

l Gaussian quadrature using n nodes(Gaussian point) can exactly calculate polynomial

equation which has integral term under 2n - 1 order.

When function f ( x ) is not a polynomial, Gaussian quadrature is inaccurate. However, the more
Gaussian points are used, the more accurate solution is. In general, the ratio of two polynomials is not a
polynomial.

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

l Table 1 Gaussian points for integration from –1 to +1

Number Locations, xi Associated

ofPoints Weights, Wi
1 x1 = 0.000... 2.000
2 x1 , x2 = ±0.57735026918962 1000
.

3 x1 , x3 = ± 0.77459666924148 5 / 9 = 0.555...
x2 = 0.000... 8 / 9 = 0.888...

Fig. 7: Gaussian quadrature with two extraction points

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

2-D problem: Integrate about second coordinate after integrate about first coordinate.

zz
1 1
z 1 LMå W f (s , t )OP
I=
-1 -1
f ( s, t ) ds dt =
-1
Ni
i i
Q
= å Wj
LMå W f (s , t )OP = å å WW f (s t )
j Ni
i i
Q
j
i j
i j i, j

For 2 ´ 2 : I = WW
1 1 f ( s1 , t1 ) + WW
1 2 f ( s1 , t 2 ) + W2W1 f ( s2 , t1 ) + W2W2 f ( s2 , t 2 )

where the sample four points

are located at
si , ti = ± 0.5773....
= ±1/ 3
And the all weight factors
are 1000
. . Thus, the two
summation marks can be
interpreted as one summation
mark for four points of the
rectangle.

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

3-D problem: I= zzz

1 1

-1 -1 -1
1
f ( s, t , z )ds dt dz = å å å WW
i j
i jWk f ( si , t j , zk )
k

NOTE: If the integration limit is z0

1 n
f ( x )dx = åi =1Wi f ( xi ) , the weight factor Wi and the
location xi are different from that of the integration limit which is between -1 and 1 (See
table 2).

Table 2. Gaussian points of the four node gaussian integration (integration from 0 to 1)

Locations, xi Associated Weights, Wi

0.0693185 0.1739274

0.3300095 0.3260725
0.6699905 0.3260725

0.9305682 0.1739274

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

Example 1: Calculate the integration of sin px using numerical integration.

I= z
1
0
sinpx dx

Using table 2, the following can be obtained.

4
I = å Wi sin pxi
i =1

= W1 sin px1 + W2 sin px2 + W3 sin px3 + W4 sin px4

= 01739
. sin p (0.0694) + 0.3261 sin p (0.3300)
0.3261 sin p (0.6700) + 01739
. sin p (0.9306)
= 0.6366

Use four decimal places. The exact value of direct integration is 0.6366. Note that location
xi and weight factor Wi are different from that in table 2 if we use the 3-points Gaussian
integration.

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

5 Calculation of stiffness matrix by Gaussian integration

Element stiffness matrix in 2-D:

k = zz T
B ( x , y ) DB ( x , y )t dx dy

zz
A
1 1 T
= B ( s, t ) DB ( s, t ) J t ds dt
-1 -1

T
The integral term B DB J , which is a function of ( s, t ) , is calculated by the
numerical integration.

T
k = B ( s1 , t1 ) DB ( s1 , t1 ) J ( s1 , t1 ) tWW
1 1
T
+ B ( s2 , t 2 ) DB ( s2 , t 2 ) J ( s2 , t 2 ) tW2W2
T
Using four-points Gaussian integration, + B ( s3 , t 3 ) DB ( s3 , t 3 ) J ( s3 , t 3 ) tW3W3
T
+ B ( s4 , t 4 ) DB ( s4 , t 4 ) J ( s4 , t 4 ) tW4W4

where s1 = t1 = - 0.5773, s2 = - 0.5773, t 2 = 0.5773, s3 = 0.5773, t 3 = -0.5773,

s4 = t 4 = 0.5773 , W1 = W2 = W3 = W 4 = 1000
. .

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

(e)
Fig. 9: Flow chart for obtaining k using Gaussian integration

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

Example 2

Calculate the stiffness matrix of

rectangular element using four-point Gaussian
integration.
E = 30 ´ 106 psi , n = 0.25 .
The unit of length in global coordinate
system is inch, and t =1 in .

Fig. 10: Quadrilateral elements for calculation of stiffness

Using 4-points rule:

( s1 , t1 ) = ( -0.5733, - 0.5773) W1 = 10
.
( s2 , t 2 ) = ( -0.5733, 0.5773) W2 = 10
.
( s3 , t 3 ) = (0.5733, - 0.5773) , W3 = 10
. .
( s4 , t 4 ) = (0.5733, 0.5773) W4 = 10
.

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

Calculation of stiffness matrix:

T
k = B ( -0.5773, - 0.5733) DB ( -0.5773, - 0.5773)
´ J ( -0.5773, - 0.5773) (1)(1000
. )(1000
. )
T
+ B ( -0.5773, 0.5773) DB ( -0.573,0.5773)
´ J ( -0.5773, 0.5773) (1)(1000
. )(1000
. )
T
+ B (0.5773,- 0.5773) DB (0.573,-0.5773)
´ J (0.5773, - 0.5773) (1)(1000
. )(1000
. )
T
+ B (0.5773, 0.5773) DB (0.573,0.5773)
´ J (0.5773, 0.5773) (1)(1000
. )(1000
. )

We need to calculate J and B at Gaussian points

( s1 , t1 ) = ( -0.5733, - 0.5773), ( s2 , t 2 ) = ( -0.5733, 0.5773) ,
( s3 , t 3 ) = (0.5733, - 0.5773), ( s4 , t 4 ) = (0.5733, 0.5773) .

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

Calculation of J :
1
J ( -0.5773, - 0.5773) = [3 5 5 3]
8
LM 0 1 - ( -0.5773) -0.5773 - ( -0.5773) -0.5773 - 1 OP
´ MM -0.5773 - 1 0 -0.5773 + 1 -0.5773 - ( -0.5773)
PP
MN
-0.5773 - ( -0.5773)
1 - ( -0.5773)
-0.5773 - 1
-0.5773 + ( -0.5773)
0
-0.5773 - 1
-0.5773 + 1
0
PQ
R|2U|
|2| .
´ S V = 1000
||4||
T4W
J ( -0.5733, - 0.5733) = 1000
.
J (0.5733, - 0.5733) = 1000
.
Similarly,
J (0.5733, 0.5733) = 1000
.

Generally J ¹ 1 , and it changes within the element.

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

Calculation of B :
1
B ( -0.5733, - 0.5733) = B1 B2 B3 B4
J ( -0.5733, - 0.5733)

LMaN 1, s - bN1, t 0 OP
B1 =M 0 cN1, t - dN1, s PP
Calculation of B1 :
MNcN
1, t - dN1, s aN1, s - bN1, t Q
where
1
a= y1 ( s - 1) + y2 ( -1 - s) + y3 (1 + s) + y4 (1 - s)
4
1
= 2( -0.5773 - 1) + 2( -1 - 0.5773))
4
+4(1 + ( -0.5773)) + 4(1 - ( - - 0.5773))
= 100
.

The same calculation can be used to obtain b, c, d .

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

Also,
1 1
N1, s = (t - 1) =
( -0.5773 - 1) = -0.3943
4 4
1 1
N1,t= ( s - 1) =
( -0.5773 - 1) = -0.3943
4 4
Similarly, B 2 , B 3 , B 4 can be calculated at ( -0.5773, - 0.5773) . And calculate B
repeatedly at other Gaussian points.

Generally a computer program is used to calculate B and k .

Final form of B is.

LM-01057
. 0 01057
. 0 0 -01057
. 0 -0.3943OP
B = M-01057
. -01057
. -0.3743 01057
. 0.3943 0 -0.3943 0 PP
MN 0 0.3943 0 01057
. 0.3943 0.3943 01057
. -0.3943 Q

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

LM1 n 0
OP L32 8 0 OP
MMn 0 P = MM 8
E
0 P ´ 10 psi
6

1- n P
D= 1 32
Matrix D : 1- n2
MN0 0
2 PQ
MN 0 0 12PQ

Finally, the stiffness matrix k:

LM1466 500 -866 -99 -733 -500 133 99OP

MM-500 1466 99 133 -500 -733 -99 -866
PP
MM -86699
99
133
1466
-500
-500
1466
133
99
-99
-866
-733
500
500
-733
PP
MM-733 PP
4
k = 10
-500 133 99 1466 500 -866 -99
MM-500 -733 -99 -866 500 1466 99 133 PP
MM 133 -99 -733 500 -866 99 1466 -500
PP
N 99 -866 500 -733 -99 133 -500 1466 Q

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

6 Higher order shape function

l Higher order shape function can be obtained by adding additional nodes to the each
side of the linear element.
l It has higher order strain distribution in element, and it converges to the exact
solution rapidly with few elements.
l It can more accurately approximate the irregular boundary shape.

Fig. 11: 2nd order iso-parametric element

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

Second order iso-parametric element:

x = a1 + a2 s + a3t + a4 st + a5 s 2 + a6t 2 + a7 s 2 t + a8 st 2
y = a9 + a10 s + a11t + a12 st + a13 s 2 + a14 t 2 + a15 s 2 t + a16 st 2

For the corner node ( i ＝1, 2, 3, 4)

1
N1 = (1 - s)(1 - t )( - s - t - 1)
4
1
N 2 = (1 + s)(1 - t )( s - t - 1)
4
1
N 3 = (1 + s)(1 + t )( s + t - 1)
4
1
N 4 = (1 - s)(1 + t )( - s + t - 1)
4

1
Ni = (1 + ssi )(1 + tti )( ssi + tti - 1)
4
si = -1, 1, 1, - 1 for i = 1, 2, 3, 4
or
ti = -1, - 1, 1, 1 for i = 1, 2, 3, 4

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

For the middle node ( i ＝5, 6, 7, 8),

1
N5 = (1 - t )(1 + s)(1 - s)
2
1
N 6 = (1 + s)(1 + t )(1 - t )
2
1
N 7 = (1 + t )(1 + s)(1 - s)
2
1
N 8 = (1 - s)(1 + t )(1 - t )
2
or
1
Ni = (1 - s 2 )(1 + tti ) ti = -11
, for i = 5, 7
2
1
N i = (1 - ssi )(1 - t 2 ) si = -11
, for i = 5, 7
2
2 2
When edge shape and displacement are function of s (if t is constant) or t (if s is
constant), it satisfies the general shape function conditions.

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

Deformation function:
R|u U|
1

|
0 O|u |
v1 |
RSuUV = LM N 0 N 2 0 N 3 0 N 4 0 N5 0 N 6 0 N 7 0 N8
SV
N PQ|v |
1 2

Tv W N 0 N1 0 N 2 0 N 3 0 N 4 0 N 5 0 N 6 0 N 7 0
|| M ||
8 2

Tv W
8

Strain matrix: e = Bd = D' N d

2nd order iso-parameter with 8 nodes

For the calculation of B and k ,
9-points Gaussian rule is used
(3×3 rule). There is large difference
between 2×2 and 3×3 rule, and 3×3
rule is generally recommended.
(Bathe and Wilson[7])

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

3rd order iso-parametric element:

Shape function of a 3rd order element is based on incomplete 4th order polynomial
(see reference [3]).

x = a1 + a2 s + a3t + a4 st + a5 s 2 + a6t 2 + a7 s 2 t + a8 st 2
+ a9 s 3 + a10t 3 + a11s 3t + a12 st 3

y also has same polynomial equation.

Mechanics and Design SNU School of Mechanical and Aerospace Engineering

Finite Element Method: Iso-Parametric Formulation

1
For the corner nodes ( i ＝1, 2, 3, 4): Ni = (1 + ssi )(1 + tti )[9( s 2 + t 2 ) - 10]
32
si = -1, 1, 1, - 1 for i = 1, 2, 3, 4
where ti = -1, - 1, 1, 1 for i = 1, 2, 3, 4
9
For the nodes ( i ＝7, 8, 11, 12) when s = ±1 : Ni = (1 + ssi )(1 + 9tti )(1 - t 2 )
32
1
where si = ±1 , ti = ±
3
9
For the nodes ( i ＝5, 6, 9, 10) when t = ±1 : Ni = (1 + tti )(1 + 9 ssi )(1 - s 2 )
32
1
where ti = ±1 , si = ±
3
When the shape function of coordinates has lower order than that of deformation, it is called
Subparametric formulation (For example, x is linear, u is 2nd order function). The opposite way is
called Superparametric formulation.

Mechanics and Design SNU School of Mechanical and Aerospace Engineering