Lecture 12

FEM in 2D: Bilinear and Cubic Shape Functions

(Lecture notes taken by Sriram Pakkam and Kishore Krishna B. S.)

·  Bilinear Shape Functions:

uex,y=α1+α2x+α3y+α4xy

Let the element be rectangular. e rectangular.

η 3

y 4 2 ξ

b 1 a

x

ueξ,η= ∝1+∝2ξ+∝3η+∝4ξη

For point 1, ξ=0 and η=0

For point 2, ξ=a and η=0

For point 3, ξ=a and η=b

For point 4, ξ=0 and η=b

u1eu2eu3eu4e=101a00001a10babb0∝1∝2∝3∝4

A matrix

ue=A∝

A-1=B=10-1a1a0000-1b01ab-1ab01b1ab-1ab

ue=N1ξ,ηu1e+N2ξ,ηu2e+N3ξ,ηu3e+N4ξ,ηu4e

Where for instance,

1=N10,0

0=N1a,0

0=N1a,b

0=N10,b

N1ξ,η=1+-1a ξ+-1b η+1abξ η

=1-ξa1-ηb

=L1(ξ) L1(η)

L2(η) L2(η)

Hence it is bilinear. The functions here are analogous to the 1-D shape functions.

·  Element matrices:

ke=a(Nj,Ni)

4 X 4

As in the quadratic shape function case, one can represent this element matrix as

ke= BTGB.

G is derived from-uxx-uyy=f

G=000ab000ab22000ab22aba2b2a2b2a3b+ab33

The following formula is used in the computation of the above components

hm,n=ξmηn

=am+1m+1*bn+1n+1

·  Cubic Shape Functions:

Let the elements be triangles and use ten terms in 10 terms

ueξ,η=∝iξmiηni

mi+ni≤3

mi=0,1,2,3

η

3 2 ξ

uc 10x1 10x10 10x1

1 centroid u1u1ξu1ηu2u2ξu2ηu3u3ξu3ηuc=ue=A∝

where for point 1 we haveu1u1ξu1η for point 2, u2u2ξu2η and point 3,u3u3ξu3η

·  Element matrix ke is of dimension 10X10:

Let B=A-1 and G is obtained from the differential equation

ke=RTBTGBR

R is a result of the partial derivatives and the chain rule

R=CZZZCZZrZrCr ccc1

C=0000cosθsinθ0-sinθcosθ c=000

Z=000000000 r=000

·  Gauss Quadrature Rules:

1-D case on [-1,1]:

-11fxdx≅inωif(xi)

Choose the weights ωi and the points xi so that the error is zero, that is,

-11P2n-1(x)=ωif(xi)

Example 1: Use n = 2 and require the errors to be zero for


1,x,x2,x3

-111= ω1*1+ω2*1

2= ω1+ω2

-11x= ω1x1+ω2x2

0= ω1x1+ω2x2

-11x2= ω1x12+ω2x22


23= ω1x12+ω2x22

-11x3= ω1x13+ω2x23

0=ω1x13+ω2x23

Solving the above equations gives

ω1=ω2=1 and x1=-13 , x2=13.

Example 2: Use n = 3

xi / ωi
-0.77454 / 0.5555
0 / 0.88885
0.77454 / 0.5555

Gauss Quadrature of Triangle:

P=1,s,t,s2,st,t2

By requiring the errors to be zeros for the above six polynomials and solving weights and (s,t)

s1,t1=16,16


s2,t2=23,16

s3,t3=16,23

ω1=ω2=ω3=16

See chapter seven in Gockenbach’s book for additional examples of quadrature rules.