You are on page 1of 19

Isoparametric finite elements

• Computation of stiffness matrix is done


on element defined in natural coordi-
nates (η, ξ, ζ = < −1, 1 >): line, ξ
rectangle, cube.
• Element in natural coordinates (η, ξ, ζ)
1
is transformed to real finite in (x, y, z) co- γ
1
ordinates with use of shape functions
Ni
1 1
• Shape functions are used also as ap-
proximation functions of unknown dis-
placements
1
Natural coordinates
t
s
−1 0 +1
1 1

• Lenght of element side is 1 + 1 = 2

• Here we will use natural coordinates s, t (in textbooks are


often called ξ, η)

2
Natural and real coordinates
t
s
−1 0 +1
1 1

y
x1 x2 x3
x

Relation between x and s:


X
x= Ni(s) xi = N1 x1 + N2 x2 + ..., (1)
i

where Ni are shape functions.


3
Shape functions
1 1
1
1 2 3 1 2 3 1 2 3

In order to use equation (2), shape functions Ni have to have


these values:

• in the i point .... 1,

• in any other point ... 0.

4
Deformations by shape functions

1 1
1
1 2 3 1 2 3 1 2 3

X
u= Ni(s) ui, (2)
i
y
x1 x2 x3
u1 u2 x u3

For this particular finite element:

u = N1(s) u1 + N2(s) u2 + N3(s) u3 (3)

5
Truss element (1)
Natural and real coordinates:
t
s
−1 0 +1
1 1

y
x1 x2 x3
x

Unknown displacements:
y
x1 x2 x3
u1 u2 x u3

Theree unknown dispalcement on element: u = u1, u2, u3T


6
Truss element (2)
Shape functions:

1 1
1
1 2 3 1 2 3 1 2 3

s(1 − s)
N1 = −
2
N2 = (1 − s2) (4)
s(1 + s)
N3 =
2
Thus:
s(1 − s) 2 s(1 + s)
x = N1 x1+N2 x2+N3 x3 = − x1+(1−s ) x2+ x3
2 2
(5)
7
Truss element (3)
Shape functions:

s(1 − s) 2 s(1 + s)
x = N1 x1+N2 x2+N3 x3 = − x1+(1−s ) x2+ x3
2 2
(6)

In matrix form:
 
h

i
 x1 


{x} = N1 N2 N3 
x2 
(7)
x3

 

 
 
 x1 

s(1−s)
(1 − s2) s(1+ s)
 
{x} = − x2
2 2  
x3

 

8
Truss element (4)
Approximation of displacements u (the same shape functions
like for x):
X
u = Ni(s) ui, (8)
i
s(1 − s) 2 s(1 + s)
u = N 1 u1 + N 2 u2 + N 3 u3 = − u1 + (1 − s ) u2 + u3
2 2
In matrix form:
 
h

i
 u1 


{u} = N1 N2 N3 
u2 
(9)
u3

 

 
 
 u1 

s(1−s)
(1 − s2) s(1+ s)
 
{u} = − u2
2 2  
u3

 

9
Truss element(5)
Relative deformations:

ε = ∂u (10)

thus:
 
u1
∂u ∂  s(1−s) 
 

2) s(1+s)
 
ε= ={ } − (1 − s u2 (11)
∂x ∂x 2 2  
u3

 

after derivation:
 
"
s(1−s)
#
s(1+s) 
 u1 

∂ (1−s2)

ε= ∂ (− 2 ) ∂ 2 u2 (12)
∂x ∂x ∂x  
u3

 

10
Truss element (6)
Relative deformations:
 
" # u1 
s(1−s) s(1+s) 
 
∂ (1−s2)

ε= ∂ (− 2 ) ∂ 2 u2
∂x ∂x ∂x  
u3

 

in short form ε = B u:
 
 
 u1 

∂N1 ∂N2 ∂N3
 
ε= ∂x ∂x ∂x 
u2 
(13)
u3

 

Problem: Equations for ε includes ∂N


∂x
i , but N is function of s.
i

11
Trruss element (7)
From mathematics:
∂Ni ∂Ni ∂s
= (14)
∂x ∂s ∂x

Equation (2):
X
x= Ni(s) xi
i

After derivation of (2):


dx X ∂Ni
= xi = J (15)
ds i ∂s
The member (auaually matrix) J is Jakobian .
12
Truss element (8)
From eqaution
dx
=J
ds
we can get ds:

1
ds = dx (16)
J

Note: In general problems the member J1 is an inverse of


Jacobian materix: J−1.
13
Truss element (9)
From equations
∂Ni ∂Ni ∂s
=
∂x ∂s ∂x
and
1
ds = dx
J
one can get:
∂Ni ∂Ni ∂Ni ∂Ni 1
=J ⇒ = (17)
∂x ∂s ∂s ∂x J

14
Truss element (10)
From equation (17) it is possible to get B:

∂N1 ∂N2 ∂N3

1 
∂N1 ∂N2 ∂N3

B= ∂x ∂x ∂x
= ∂s ∂s ∂s
(18)
J

Thus:
 
u1
1 
  

∂N1 ∂N2 ∂N3
 
{ε} = ∂s ∂s ∂s
u2 (19)
J 

 u3


15
Truss element (11)
If we know ε:
 
u1
1 
  

∂N1 ∂N2 ∂N3
 
{ε} = ∂s ∂s ∂s
u2 ,
J 

 u3


it is possible to write equation for potential energy of internal


forces:
1Z t
Πi = ε DεdV, (20)
2 V
and equaito for stiffness matrix of finite element:
Z x
3
K=A εtDεdx. (21)
x1

16
Truss element (12)
Equation (22) can be integrated analytically here. It is not possi-
ble for more complicated problems (2D, 3D) thus it is necessary
to utilize numerical integration. In many cases Gauss formula
is used:

m
εtDεwi,
X
K=A (22)
i=1
where wi is weight of integrational point, m is number of inte-
grational points.

17
Isoparametric elements for plane
problems (1)
y η

3 4 3

1.0
4

ξ
1.0

1 2
2

1 1.0 1.0
x

Shape functions:
1
Ni(ξ, η ) = (1 + ξξi)(1 + ηηi), (23)
4
18
Izoparametric element for plame
problems (2)
y η

5 7 6 5

1.0
6 8 4
7
4
ξ
1.0
8
1 2 3
2 3

1 1.0 1.0
x

Shape functions for corners:


1
Ni(ξ, η ) = (1 + ξξi)(1 + ηηi)(ξ ξi + η ηi − 1), (24)
4
and for mid-sides:
ξi2 η 2
Ni(ξ, η ) = (1 + ξξi)(1 − η 2) + (1 − ξ ξi)(1 − η 2). (25)
2 2
19

You might also like