Finite Element Method 1 – lecture notes 8_node quadrilateral element Page 1 of 10

_________________________________________________________________________________________________________________
8-node quadrilateral element. Numerical integration
The technique used for the formulation of the linear triangle can be formally extended to construct quadrilateral elements as well as higher order
triangles. But it is connected with some difficulties:
1. The construction of shape functions satisfying consistency requirements for higher order elements with curved boundaries becomes
increasingly difficult.
2. Computations of shape function derivatives to evaluate the strain-displacement matrix.
3. Integrals that appear in the expressions of the element stiffness matrix and consistent nodal force vector can no longer be carried out in
closed form.
The 8-node element is defined by eight nodes having two degrees of freedom at each node: translations in the nodal x (u) and y directions (v).
It provides more accurate results and can tolerate irregular shapes without much loss of accuracy. The 8-node are well suited to model curved
boundaries.

η
y 3 ( x3 , y 3 )
7
4 1 7 3 4
6 ( x6 , y 6 )

8 6 8

-1 1 ξ
1
( x1 , y1 ) 5
( x2 , y 2 ) y
( x5 , y 5 ) 2
1 -1 5 2

( ξ ,η ) → ( x , y )
( −1, −1) → ( x1 , y1 ) (1, −1) → ( x2 , y2 ) (1,1) → ( x3 , y3 ) ( −1,1) → ( x4 , y4 )
( 0, −1) → ( x5 , y5 ) (1, 0 ) → ( x6 , y6 ) ( 0,1) → ( x7 , y7 ) ( −1, 0 ) → ( x8 , y8 )
Finite Element Method 1 – lecture notes 8_node quadrilateral element Page 2 of 10
_________________________________________________________________________________________________________________

8
x (ξ ,η ) = ∑ N i (ξ ,η ) xi
i =1
8
N2(ξ, η) N6(ξ, η) y (ξ ,η ) = ∑ N i (ξ ,η ) yi
i =1
η η

N1 (ξ ,η ) = −
1
4
(1 − ξ )(1 − η )(1 + ξ + η ) N 5 (ξ ,η ) =
1
2
(1 − ξ 2 ) (1 − η )
ξ ξ
N 6 (ξ ,η ) = (1 + ξ ) (1 − η 2 )
1 1
N 2 (ξ ,η ) = − (1 + ξ )(1 − η )(1 − ξ + η )
4 2
N 7 (ξ ,η ) = (1 − ξ 2 ) (1 + η )
1 1
N 3 (ξ ,η ) = − (1 + ξ )(1 + η )(1 − ξ − η )
4 2
N8 (ξ ,η ) = (1 − ξ ) (1 − η 2 )
1 1
N 4 (ξ ,η ) = − (1 − ξ )(1 + η )(1 + ξ − η )
4 2

Shape functions N2 and N6

 x1   x1 
y  y 
 1  1
 x2   x2 
   
y2  y2 
 x   N1 0 N2 0 N3 0 N4 0 N5 0 … N8 0     
x
 = x    = [ N ]  x3  = [ N ]{ xy}
y  0 N1 0 N2 0 N3 0 N4 0 N5 … 0 N 8   3  y y 
y
 3  3
⋮  ⋮ 
x  x 
 8  8
 y8 e  y8 
Finite Element Method 1 – lecture notes 8_node quadrilateral element Page 3 of 10
_________________________________________________________________________________________________________________
8
u (ξ ,η ) = ∑ N i (ξ ,η ) ui
i =1

8
v (ξ ,η ) = ∑ N i (ξ ,η ) vi
i =1

 u1 
v 
 1
u2 
 
v2
u   N1 0 N2 0 N3 0 N4 0 N5 0 … N8 0    u 
 = u    = [ N ]{q}e
v   0 N1 0 N2 0 N3 0 N4 0 N5 … 0 N8   3  v 
v
 3
⋮
u 
 8
 v8 e

∂ 
 0  ∂N ∂N 2 ∂N 3 ∂N 8 
 ∂x   1 0 0 0 0 
 ∂x ∂x ∂x ∂x
…
 
∂
{ε } = [ R ]{u} =  0   N (ξ ,η )  {q} = [ B ]{q} 
[ B] =  0
∂N1 ∂N 2 ∂N 3 ∂N 8 
3×1 3×2 2×1  ∂y   2×16  16×1e 3×16 16×1 ∂y
0
∂y
0
∂y
… 0
∂y 


∂ ∂  ∂N1 ∂N1 ∂N 2 ∂N 2 ∂N 3 ∂N 3 ∂N 8 ∂N 8 
   
 ∂y ∂x   ∂y ∂x ∂y ∂x ∂y ∂x ∂y ∂x 
…
3×2
Finite Element Method 1 – lecture notes 8_node quadrilateral element Page 4 of 10
_________________________________________________________________________________________________________________
Partial derivatives of shape functions with respect to the Cartesian coordinates x and y are required for the strain and stress calculations. Since
the shape functions are not directly functions of x and y but of the natural (local) coordinates ξ and η, the determination of Cartesian partial
derivatives is not trivial.
We need the Jacobian of two-dimensional transformations that connect the differentials of {x,y} to those of {ξ,η} and vice-versa

 ∂x ∂y   8 ∂N i 8
∂N i 
 ∂ξ ∑ ⋅ xi
∂ξ   i =1 ∂ξ
∑ δξ ⋅ yi 
[J ] =  = i =1
 =  J (ξ ,η ) 
 ∂x ∂y   8 ∂N i 8
∂N i 
⋅ xi ⋅ yi 

 ∂η ∂η   ∑
i =1 ∂η
∑
i =1 η 
Matrix J is called the Jacobian matrix of (x,y) with respect to (ξ,η), whereas J-1 is the Jacobian matrix of (ξ,η) with respect to (x,y).
J and J−1 are often called the Jacobian and inverse Jacobian, respectively. The scalar symbol J means the determinant of J: J =|J|=det J.
Jacobians play a crucial role in differential geometry.

∂N i ∂N i ∂ξ ∂N i ∂η  ∂Ni   ∂ξ ∂η   ∂N i   ∂N i 
= ⋅ +  ∂x   ∂x ∂x   ∂ξ  
−1  ∂ξ 

∂x ∂ξ ∂x ∂η ∂x  ∂N  =  ∂ξ 
∂η   ∂N i 
= [ J ]  
∂N i ∂N i ∂ξ ∂N i ∂η  i   ∂N i 
= ⋅ +  ∂y   ∂y ∂y   ∂η   ∂η 
∂y ∂ξ ∂y ∂η ∂y

 ∂N i   ∂x ∂y   ∂N i   ∂N i 
 ∂ξ   ∂ξ 
∂ξ  ∂x   ∂x 
 =   = [ J ]  ∂N 
 ∂N i   ∂x ∂y   ∂N i   i
 ∂η   ∂η 
∂η   ∂y   ∂y 
Finite Element Method 1 – lecture notes 8_node quadrilateral element Page 5 of 10
_________________________________________________________________________________________________________________

{ε } =  B (ξ ,η )  {q}e

1 1
Ue = ∫ ε  {σ } dxdy = ∫  q  e [ B ] [ D ][ B ]{q}e dxdy
T

Ωe ( x , y ) )
2 2 Ωe ( x , y )

∫
A( x , y )
f ( x, y ) dxdy = ∫ξ η f (ξ ,η ) det [ J ] dξ dη
A( ,
dxdy = det [ J ] dξ dη
)

 B (ξ ,η )  [ D ]  B (ξ ,η )  det  J (ξ ,η )  d ξ dη {q}e
T
U e = 1/ 2  q  e ∫
Ωe (ξ ,η ) 16×3 3×3 3×16

1
Ue =  q  e [ k ]e {q}e
2

1 1

[k ]e = ∫ [ B ] [ D ][ B ] dxdy = ∫ ∫  B (ξ ,η )  [ D ]  B (ξ ,η )  det  J (ξ ,η )  dξ dη ∫  B (ξ ,η ) 
T T

Ωe ( x , y ) −1 −1 16 x 3 3×3 3×16 16×3

Finite Element Method 1 – lecture notes 8_node quadrilateral element Page 6 of 10
_________________________________________________________________________________________________________________

Nodal forces of the Ωe element equivalent to the body load:

Wzx =
Ωe
∫  X  {u} d Ω e = ∫  X  [ N ]{q} d Ω
Ωe
e e =  F x  {q}e ,
e

 F x  =
e
Ωe
∫  X  [ N ] d Ω .e

X

y F/3

F = Ae X


F/12
F/3
F/3
F/3
F/3 F/3 F/3

F/12
F/12 F/12
x

Work-equivalent nodal forces for uniform constant body load in the case of CST element and 8-node quadrilateral element
Finite Element Method 1 – lecture notes 8_node quadrilateral element Page 7 of 10
_________________________________________________________________________________________________________________

Finite element method results: continuous displacement field and discontinous stress field

Displacement component (e.g. u(x,y)) interpolation over two finite elements

∂u ∂u
=
∂x′ 1 ∂x′ 2 (ε )1 = (ε x′ )2
x′

∂u ∂u
⇒ (ε )1 ≠ (ε y′ )2
y′
⇒ (σ ) ≠ (σ )
ij 1 ij 2

2 ≠
∂y ′ 1 ∂y′ 2
1 y′
x′
Finite Element Method 1 – lecture notes 8_node quadrilateral element Page 8 of 10
_________________________________________________________________________________________________________________
Numerical Gauss integration in FE algorihms
The use of numerical integration is essential for evaluating element integrals of isoparametric elements. The standard practice has been to use
Gauss integration1 because such rules use a minimal number of sample points to achieve a desired level of accuracy. This property is important
for efficient element calculations because we shall see that at each sample point we must evaluate a matrix product.
[k ]e = ∫ [ B ] [ D ][ B ] dxdy = ∫  B (ξ ,η )  [ D ]  B (ξ ,η )  det  J (ξ ,η )  d ξ dη
T T

Ωe (ξ ,η ) Ωe ( x , y ) 16×3 3×3 3×16

The numerical intergration have to be also performed for finding the equivalent nodal forces.

One dimensional integration

In general: F(x)
b n

∫ F ( x ) dx = ∑ αi Fi ( xi ) + Rn
a i =1

Introducing the new variable -1 ≤η≤ 1

x=
( a + b ) + b − a ⋅η dx =
b−a
η
2 2 2
x1 x2 x3 x4 x5 xn
b−a b−a
b 1 1

∫ F ( x ) dx = ∫ f (η )
a −1
2
dη =
2 −∫1
f (η ) dη
η
x

The Gauss integration

x=b
 d 2n f 
1 n x=a
∫ f (η ) dη = ∑ wi f (η i ) + Rn Rn = 0  dη 2 n  η=-1
x1 η=1
−1 i =1  

Here n ≥ 1 is the number of especially defined Gauss integration points, wi are the integration weights, and ηi are sample-
point abcissae in the interval [−1,1]. The use of the interval [−1,1] is no restriction, because an integral over another range,
from a to b can be transformed to the standard interval via a simple linear transformation of the independent variable, as
shown above. The values ηi and wi are defined in such a way to aim for best accuracy. Indeed, if we assume a polynomial
expression, it is easy to check that for n sampling a polynomial of degree 2n – 1 can be exactly integrated .
Finite Element Method 1 – lecture notes 8_node quadrilateral element Page 9 of 10
_________________________________________________________________________________________________________________

Table below shows the positions and weighting coefficients for gaussian integration.
f(η )
Abscissae and weight coefficients of the gaussian Quadrature
n çi (i=1,n) Wi (i=1,n)
1 0 2
2 −1/ 3 1

+1/ 3 1 η1 η2 η3 η4
3 − 0.6 5/9
0 8/9 1 η
-1
5/9
+ 0.6
4 -0.861136311594953 0.347854845137454
-0.339981043584856 0.652145154862546
+0.339981043584856 0.652145154862546
+0.861136311594953 0.347854845137454
5 η
Remarks: The sum of weighing coefficients is always 2 1
The integration gives the exact solution for polynomials of 2n-1 degree.
8 5 40 5 5 25
5 5 25 w6* = × = w9* = × =
w3* = × = 9 9 81 9 9 81
9 9 81
Numerical integration – rectangular region:
1 1 1
 n 
j ∑ wi f (ξi ,η j ) =
n n
1
∫∫ f ( ξ ,η ) d ξ dη ≈ ∫  ∑ i i 
f ( ξ ,η ) w dη ≈ ∑ w -1
−1 −1 −1  i =1 j =1 i =1 5 8 40 8 8 64 5 8 40
w2* = × = w5* = × = w8* = × = ξ
= ∑∑ wi w j f (ξi ,η j ) = ∑ w f (ξk ,ηk )
n n m 9 9 81 9 9 81
* 9 9 81
k
i =1 i =1 k =1
8 5 40 5 5 25
5 5 25 w4* = × = w7* = × =
w1* = × = 9 9 81 9 9 81
9 9 81
Finite Element Method 1 – lecture notes 8_node quadrilateral element Page 10 of 10
_________________________________________________________________________________________________________________

RESULTS OBTAINED USING 8-NODE ELEMENTS - AVERAGING

Example –2D FE model of the cantilever beam

Y Y

Z X Z X

-3045 -1691 -336.56 1018 2372

-2368 -1014 340.577 1695 3049

Bending stress σx distribution (element solution)

ANSYS 11.0
MAY 18 2010 ANSYS 11.0
PRES-NORM 10:19:31 MAY 18 2010
PLOTPRES-NORM
NO. 5 10:20:11
PLOT NO. 6

MAY 18 2010

PLOT NO. 3

Y Y
Z X Z X

Y Y
-743.468 -413.194 -82.921 247.352 577.626 -743.468 -413.194 -82.921 247.352 577.626
Z X Z X -578.331 -248.058 82.216 412.489 742.762 -578.331 -248.058 82.216 412.489 742.762

σx element and nodal solution σy element and nodal solution