Chapter 7

FINITE ELEMENT APPROXIMATION

Dr. Pham Minh Tuan

 Interpolation & Shape function
Interpolation is to devise a continuous function expressing a field
quantity for an element, which satisfies the nodal values of the field
quantity at a finite number of nodes of the element.
Interpolation generally provides approximate values.
Interpolating function is almost always a polynomial, can be
represented by shape function.
Shape function can be formulated based on Lagrange polynomials
or the general form of displacement function in an element.
Characteristics of shape function:
1; if i ≡ j n
( )
Ni ξ j =
 ∑ Ni ( ξ j ) =
1
0; if i ≠ j i =1

2nd order shape function: N i ( ξ,η=

) Liξ × Liη
 Shape function of 1-D element
Reference element Real element

1 2 3 1 2 3
ξ x
−1 0 1 x1 x2 x3

( ξ − 0)( ξ − 1) = 1 ( x − x2 ) ( x − x3 )
N1 ( ξ)= ξ ( ξ − 1) N1 ( x ) =
( −1 − 0)( −1 − 1) 2 ( x1 − x2 ) ( x1 − x3 )
 ξ − ( −1)  ( ξ − 1) ( x − x1 ) ( x − x3 )
N 2 ( ξ) = = − ( ξ + 1)( ξ − 1) N2 ( x) =
0 − ( −1)  ( 0 − 1) ( x2 − x1 ) ( x2 − x3 )
 ξ − ( −1)  ( ξ − 0 ) 1 N3 ( x ) =
( x − x1 )( x − x2 )
N 3 ( ξ )= = ξ ( ξ + 1) ( x3 − x1 )( x3 − x2 )
1 − ( −1)  (1 − 1) 2
 Shape function of 2-D element
η
Reference bilinear quadrilateral element:
4 1 3
−1 1 ξ
0
ξ −1 η −1 1
N1 ( ξ,η=
) L1ξ ⋅ L1=
η ⋅ = (1 − ξ)(1 − η) 1 −1 2
−1 − 1 −1 − 1 4
ξ − ( −1) η − 1 1
N 2 ( ξ,η=
) L2ξ ⋅ L2=
η ⋅ = (1 + ξ)(1 − η)
1 − ( −1) −1 − 1 4
ξ − ( −1) η − ( −1) 1
N 3 ( ξ,η=
) L3ξ ⋅ L3=
η ⋅ = (1 + ξ)(1 + η)
1 − ( −1) 1 − ( −1) 4

ξ − 1 η − ( −1) 1
N 4 ( ξ,η=
) L4ξ ⋅ L4=
η ⋅ = (1 − ξ)(1 + η)
−1 − 1 1 − ( −1) 4
 Shape function of 2-D element
y
Real bilinear quadrilateral element:
b 4 3

1 2 x
0 a
x − a y − b ( a − x )( b − y )
N1 ( x, y ) = L1x ⋅ L1 y = ⋅ =
0−a 0−b ab
x − 0 y − b x (b − y )
N 2 ( x, y ) = L2 x ⋅ L2 y = ⋅ =
a−0 0−b ab
x − 0 y − 0 xy
N 3 ( x, y ) = L3 x ⋅ L3 y = ⋅ =
a − 0 b − 0 ab
x − a y − 0 (a − x) y
N 4 ( x, y ) = L4 x ⋅ L4 y = ⋅ =
0−a b−0 ab
 Shape function of 2-Dyelement
Real bilinear quadrilateral element:
b 4 3
General form of displacement function:
u =A + Bx + Cy + Dxy
1 2 x
Substituting coordinates of nodes:  A = u 0 a
1
 u −u
 u1 = A B = 2 1
u = A + Ba  a
 2
 ⇒ u4 − u1
 u3 =A + Ba + Cb + Dab C = b
u4= A + Cb 
 D = u1 − u2 − u3 − u4
 ab
Substituting back into the equation and simplifying to the form:
u = N1 ( x, y ) u1 + N 2 ( x, y ) u2 + N 3 ( x, y ) u3 + N 4 ( x, y ) u4
 Shape function of 2-D element
Reference bilinear triangular element: η
General form of displacement function: 1 3
u= A + Bξ + C η
1 2
ξ
0 1
N1 ( ξ,η) = 1 − ξ − η
N 2 ( ξ,η) = ξ
N 3 ( ξ,η) = η
 Shape function of 2-D element
y
Real bilinear triangular element:
3
1 x1 y1 
A = det 1 x2 y2 
1
2   1
1 x3 y3  2
x
0

( x2 y3 − x3 y2 ) + x ( y2 − y3 ) + y ( x3 − x2 )
N1 ( x, y ) =
2A
( x3 y1 − x1 y3 ) + x ( y3 − y1 ) + y ( x1 − x3 )
N 2 ( x, y ) =
2A

N 3 ( x, y ) =
( x1 y2 − x2 y1 ) + x ( y1 − y2 ) + y ( x2 − x1 )
2A
 Coordinate transformation
Mapping

Real element
Reference element

n n n
x ∑ N i ( ξ,η,ς ) ⋅ xi =
; y ∑ N i ( ξ,η,ς ) ⋅ yi =
; z ∑ N i ( ξ,η,ς ) ⋅ zi
i 1 =i 1 =i 1

where Ni(ξ, η, ς) is shape function.

 Jacobian matrix
Coordinate transformation:
n n n
x ∑ N i ( ξ,η,ς ) ⋅ xi =
; y ∑ N i ( ξ,η,ς ) ⋅ yi =
; z ∑ N i ( ξ,η,ς ) ⋅ zi
i 1 =i 1 =i 1
where Ni(ξ, η, ς) is shape function.
Relationship between reference and real elements:
∂   ∂x ∂y ∂z   ∂ 
 ∂ξ   ∂ξ ∂ξ ∂ξ   
     ∂x 
∂   ∂x ∂y ∂z   ∂ 
   ∂η ∂η ∂η  ⋅  ∂y 
where [J] is Jacobian matrix
 ∂η     
∂   ∂x ∂y ∂z   ∂ 
   ∂ς ∂ς ∂ς   ∂z 
 ∂ς  
[J ]
 η Example 1
1  1 1
1 N1 ( ξ,η
=) ( )( )
1 − ξ 1 − η ⇒ N1  0, =

4 3 4  2 8
1/2 B  1 1
−1 1 ξ N 2 ( ξ ,η
=
1
) (1 + ξ)(1 − η) ⇒ N 2  0, =
0 4  2 8
1  1 3
1 2 N 3 ( ξ ,η
=) (1 + ξ)(1 + η) ⇒ N 3  0, =
−1 4  2 8
y 1  1 3
N 4 ( ξ ,η
=) ( )( )
1 − ξ 1 + η ⇒ N 4  0, =

4 3
4  2 8
3 4
B' xB′
1 1 3 3
= ∑ N i xi = ⋅ 2 + ⋅ 4 + ⋅ 2 + ⋅ 4 = 3
i =1 8 8 8 8
2 1 2 4
x 1 1 3 3
0 2 4 y B′ = ∑ N i yi = ⋅ 2 + ⋅ 2 + ⋅ 3 + ⋅ 3 = 2.75
i =1 8 8 8 8

Find coordinate of B in (x, y) frame. ⇒ B ′ ( 3,2.75)

 Example 2
y
3
5
η
3
3 2
2
1 1 A
3
1
2 4 5 x 0 1 2
0
3 ξ
3 3
=x ∑ N i ( ξ,η) ⋅ xi =
; y ∑ N i ( ξ,η) ⋅ yi
i 1 =i 1
N1 ( ξ,η) = 1 − ξ − η ; N 2 ( ξ,η) = ξ ; N 3 ( ξ,η) = η
1 1
with=
x1 2= , y1 2= , x2 5,=y2 3= , x3 4= , y3 =
5, ξ = ,η
3 3
 11 10 
⇒ A′  , 
3 3
 Example 2 (Cont.)
 ∂x ∂y 
 ∂ξ ∂ξ   J J12  3 1
= [ J ] =  11 ⇒ [J ] =
J 22  2 3
Jacobian matrix:
 ∂x ∂y   J 21 
 ∂η ∂η 

∂  3  ∂
=
J11  ∑ N i ( ξ,=
η) xi   2 (1 − ξ − η) + 5ξ=
+ 4η 3
∂ξ  i =1  ∂ξ
∂  3  ∂
=
J12  ∑ N i ( ξ,=
η) yi   2 (1 − ξ − η) + 3ξ=
+ 5η 1
∂ξ  i =1  ∂ξ
∂  3  ∂
=
J 21  ∑ N i ( ξ,=
η) xi   2 (1 − ξ − η) + 5ξ=
+ 4η 2
∂η  i =1  ∂η
∂  3  ∂
=
J 22  ∑ N i ( ξ,=
η) yi   2 (1 − ξ − η) + 3ξ=
+ 5η 3
∂η  i =1  ∂η
 Interpolation
Assuming ui is a physical quantity at node ith, the value of such
physical quantity at an arbitrary point in the element can be
estimated by interpolation.

In reference element with n nodes:

n
u ( ξ, η
=, ς) ∑ N i ( ξ, η, ς ) ⋅ ui
i =1
where Ni(ξ, η, ς) is shape function in reference frame.

In real element with n nodes:

n
u ( x, y, z )
= ∑ N i ( x, y, z ) ⋅ ui
i =1
where Ni(x, y, z) is shape function in real frame.
 y Example 3
4 3 Temperature at nodes is given as:
3
A T1 = 100°C, T2 = 60°C, T3 = 50°C, T4 = 90°C
2 1 2 Interpolate the temperature at A(2.5, 2.5)
x
0 2 4
Assuming the frame’s center is at node 1, shape functions are:
N1 ( x, y )
) ; N x, y x ( b − y=
( a − x )( b − y= ) ; N x, y xy y (a − x)
2 ( ) 3 ( ) = ; N 4 ( x, y )
ab ab ab ab
where a = 2 and b = 1.
a b
Relative coordinate of A referring to node 1: A  , 
4 2
a b 3 a b 1 a b 1 a b 3
=
Substitution: N1  ,  = ; N2  ,  = ; N3  ,  = ; N4  , 
4 2 8 4 2 8 4 2 8 4 2 8
4
3 1 1 3
⇒ TA = ∑ N i Ti = ⋅ 100 + ⋅ 60 + ⋅ 50 + ⋅ 90 = 85 ( °C )
i =1 8 8 8 8
 Example 4
Node displacement (mm):
u1 = -0.0100, v1 = 0.1000
u2 = -0.0200, v2 = 0.1000
u3 = -0.0200, v3 = 0.2000
where ui, vi represent displacements along
the x and y axes respectively
Displacement at A(300, 250) = ?
( x2 y3 − x3 y2 ) + x ( y2 − y3 ) + y ( x3 − x2 )
N1 ( x, y ) =
x2 y3 − x3 y2 − x1 y3 + x1 y2 + x3 y1 − x2 y1
( x3 y1 − x1 y3 ) + x ( y3 − y1 ) + y ( x1 − x3 )
N 2 ( x, y ) =
x2 y3 − x3 y2 − x1 y3 + x1 y2 + x3 y1 − x2 y1

N 3 ( x, y ) =
( x1 y2 − x2 y1 ) + x ( y1 − y2 ) + y ( x2 − x1 )
x2 y3 − x3 y2 − x1 y3 + x1 y2 + x3 y1 − x2 y1
3 3
uA = ∑ N i ( x A , y A ) ui = −0.0150
= ( mm ) vA N i ( x A , y A ) vi
∑= 0.1250 ( mm )
i =1 i =1
End of Chapter 7