Sie sind auf Seite 1von 26

Introduction to the Finite Element Method 4:

Energy Principles, Interpolation and


Numerical Integration

Goangseup Zi

Introduction to the Finite Element Method

Field Equations of One-Dimensional Elastostatic


Problems

d
A bA 0
dx
E
u, x
subjected to
u ( x) u on u

n j ti on t
u t and u t

Introduction to the Finite Element Method

Strong Form to Weak Form in One Dimension


Multiply a test function u to the
governing equation.

d
u A u bA 0
Here the test function vanishes at
dx
u (displacement)
x) U , U u ( x) u C 0 ( x), u u on u
the essential (or

boundary.

u ( x) U 0 , U 0 u ( x) u C 0 ( x), u 0 on u
Integrate the governing equation
d

u
bA
dx 0
over the total length of the one

L
dx

dimensional structure.

Rearrange the first term using the


trick of partial integration.

d u
d

L dx u A dx A u bA dx 0
Introduction to the Finite Element Method

Strong Form to Weak Form in One Dimension,


Contd Each of the terms in the weak form

represents a virtual work due to the


virtual displacement. The test
function is often called a virtual
displacement to indicate that it is not
the actual displacement; according
to Websters dictionary, virtual
means being in essence or effect,
not in fact.

d u
d

L dx u A dx A u bA dx 0

d u
(Belytschko et al. 2001)
u A 0
Adx u bAdx 0
L dx
L
L

ext

int

Introduction to the Finite Element Method

Weak Form to Strong Form in One Dimension


d u
Adx
L dx

u t A u bAdx
L

Rearrange the internal work


L
d u
d

Adx

u
dx weak
0 L form and
Athe
So the strong
L dx
dx
form are equivalent each other.

Then

u t A u bAdx utA 0

d
u A dx
L
dx

L u dx A bA dx 0
Introduction to the Finite Element Method

Field Equations of Multi-Dimensional Elastostatic


Problems

ji , j bi 0 in
ji ij
ij Cijkl kl

ij

1
ui , j u j ,i

subjected to
ui (x) ui on u

ij n j ti on t

u t and u t

Introduction to the Finite Element Method

Strong Form to Weak Form in Multi-Dimensions


Multiply a test function to the
governing equation.

u(x) U, U u(x) u C0 (x), u u on u

u(x) U0 , U0 u(x) u C0 ( x), u 0 on u

Integrate the governing equation


over the domain of the structure.

ui ji , j ui bi d 0

Rearrange the first term using the


trick of partial integration.

ui ji , j d ui ji , j d ui , j ji d

ui ji n j d ui , j ji d

u t d

Work
u t dby
internal
ustress
d

u b d u d

ui ti dforce

Workby external
t

i i

i i

i, j

i, j

ji

ji

Introduction to the Finite Element Method

Weak Form to Strong Form in Multi-Dimensions

uiti d ui bi d ui , j ji d

ui ti d ui bi d ui ji , j d ui ji , j d

ui ji n j d ui ji , j d
t

ui ji , j bi d 0

By the fundamental lemma of variational principle,

ji , j bi 0
Introduction to the Finite Element Method

The Method of Total Potential Energy

PVW is a very
powerful tool which
does not depend on the
type of the constitutive
law of the system.

uiti d ui bi d ij ij d

strain energy density


ij

U ij d ij
0

ij

We can develop the


method of the total
potential energy for
elastic media.

The extremization of
the total potential
energy w.r.t. admissible
deformation field is
necessary for
equilibrium to exist.

U
ij

ui ti d ui bi d ij

U
d
ij

variation
U Lofp the
0load potential
The first
The first variation of the internal energy

U Ud

Lp ui ti d ui bi d
t

Introduction to the Finite Element Method

A Comment about the Choice of


Interpolation Functions
Lets assume that we are using Cn function to interpolate the displacements.

Strong form

ji , j bi 0
The strain is the symmetric part of the
first derivative of the displacement.
Therefore it is in Cn-1 and so is the
stress.
To define the above equilibrium
equation mathematically, we should
be able to differentiate Cn-1 functions
since the stress is proportional to the
strain.
I.e. the set of the interpolation should
have Cn-1 continuity.

Weak form

ui ti d ui bi d

ui , j ji d

Here we do not have to differentiate the


stress but the virtual displacement.

Note that the strain comes from the


derivative of the displacement.

Therefore no problemo as long as the


displacement and the virtual displacement
are differentiable, and the differentiations
are integrable.

The displacement field for the total


potential energy method is subjected to the
same requirement.

Introduction to the Finite Element Method

10

The General Procedure for the


Finite Element Method
1. Describe the equilibrium equation using the principle of the
total potential energy of an elastic body or the principle of
virtual work.

W W
int

ext

U Lp 0
2. Choose a set of C0 functions to approximate the displacement
field and the displacement variation.

3. Find the best function which satisfies the equilibrium


condition which is defined by the principle of the total
potential energy of an elastic body or the principle of virtual
work in the finite element method.
Introduction to the Finite Element Method

11

Applications of Interpolation

N 2 ( x) sign ( ( x))

N3 ( x) sign ( ( x))

N 3 ( x) sign( ( x)) sign( ( x3 ))

N 2 ( x) sign( ( x)) sign( ( x2 ))

Introduction to the Finite Element Method

12

Lets See How to Construct a Function

Introduction to the Finite Element Method

13

1D Linear Lagrange Interpolation


Linear interpolation
U ( x) a1 x a2
Lagrange Interpolants

U ( x) U1 N1 ( x) U 2 N 2 ( x)
where N1 and N 2 are linear functions of x
U1 U ( x1 ) and U 2 U ( x2 )

The above equations can be written collectively as


U ( x1 ) U1 N1 ( x1 ) U 2 N 2 ( x1 )
U ( x2 ) U1 N1 ( x2 ) U 2 N 2 ( x2 )
N1 ( x1 ) 1, N 2 ( x1 ) 0, N1 ( x2 ) 0 and N 2 ( x2 ) 1

N I ( xJ ) IJ
1 I J
where IJ
0 I J

Kronecker Delta.

Introduction to the Finite Element Method

14

Construction of Linear Lagrange Interpolation


A line between two points with the Kronecker Delta property
N1

x x2
x1 x2

N2

x x1
x2 x1

Therefore
U ( x) U1 N1 ( x) U 2 N 2 ( x)
2

N I ( x)U I
I 1

Note that the function is C0.

Introduction to the Finite Element Method

15

An Example of Linear Lagrange Interpolation


x

1.0

0.20

2.0

0.40

3.0

0.45

4.0

0.43

5.0

0.40

for 1 x 2

Points (2,3)

for 2 x 3 or element 2

Points (3, 4)
Points (4,5)
0.5
Points
(5, 6)
0.4

6.0

0.37

x2
x 1
0.20
0.40
1 2
2 1
x 3
x2
U ( x)
0.40
0.45
23
3 2
x4
x 3
U ( x)
0.45
0.43
3 4
43
x 5
x4
U ( x)
0.43
0.40
45
54
x6
x 5
U ( x)
0.40
0.37
56
65

Points (1, 2) U ( x)

or element 1

for 3 x 4 or element 3
for 4 x 5 or element 4
for 5 x 6 or element 5

0.3
0.2
0.1
0
0

Introduction to the Finite Element Method

16

Approximation of Displacement Field

u( x) N I ( x) uI N u uT NT
I

u ( x) N I ( x)uI Nu uT NT
I

N ( x) 1
I

Introduction to the Finite Element Method

17

The Weak Form with the Approximated


Displacement

int

int

ext

ext

du d u
AE
dx
0
dx dx
T
L
dN dN
T
u AE
dx u
0
dx dx
L

u FL u
T

bAN dx

Introduction to the Finite Element Method

18

The Discretized Equilibrium Equation Using PVW

int

ext

T
L

dN
T
u AE
0
dx

T
L
dN
int
f AE
0
dx

ext

dN
T
T
dx u FL bAN dx 0
0
dx

dN
dx u Ku
dx

FL bANT dx
T

Introduction to the Finite Element Method

19

Element-wise Operation
Note that we can have the test function del u vanish outside an
element. Therefore one can integrate the internal and the
external forces element-by-element. Then assemble them to
the corresponding positions.

int

ext

dN dN
AE
dx u
dx dx

FL bAN dx
T

Introduction to the Finite Element Method

20

Stiffness Matrix

(1)

x2

x1

dNT dN
AE
dx
dx dx

x x2
N1
x1 x2

x x1
N2
x2 x1

1
x2 x x x1
L
dN 1
1 1
dx L

1
K
1 1 dx

x1
1
for constant AE
(1)

(1)

x2

1
AE 2
L

AE 1 1

1
1
L

SAME AS BEFORE!

Introduction to the Finite Element Method

21

External Force

ext, (1)

x2

bAN dx
T

x1

ext, (1)

x2

x1

1 x2 x
bA
dx

L x x1
x2

x x2
N1
x1 x2
1
N x2 x
L

x x1
N2
x2 x1
x x1

1 2

x2 x x

bA
bA 2 1
2


L 1 2
2 L 1

x x1 x
2
x
1
1
1
bAL
2
1

Introduction to the Finite Element Method

22

Numerical Integration: Trapezoidal Rule


x2

K (1) AE
x1

x2

dN dN
dx
dx dx

f ( x)dx

H. W. Read
Simpsons 1/3 Rule and
Simpsons 3/8 Rule.

x1

N 1

h
f ( x)dx f (a) 2 f (a jh) f (b)
2
j 1

Introduction to the Finite Element Method

23

Numerical Integration: Gauss Quadratures


K

(1)

AE
1

dN dx dN dx dx
d

d d d d d
T

f x( ) J d
1

I f ( )d
1

wk f ( xk )
k 1

N
1
2

i
wi
0.000000000 2.000000000
0.577350269 1.000000000

0.000000000 0.888888889
0.774596669 0.555555556

H. W. Get Gauss points and


weights from reference
books.

Introduction to the Finite Element Method

24

Derivation of Gauss Quadratures

Introduction to the Finite Element Method

25

Homework
Integrate the following function using the Trapezoidal rule and
Gauss quadrature in the range of [-0.5 1.5].

f ( x) x

Increase your sampling points until the error gets less than 1e3. The error is defined as

E f ( x)dx wI f I
b

Introduction to the Finite Element Method

26

Das könnte Ihnen auch gefallen