LINEAR FINITE ELEMENT
ANALYSIS
WEIGHTED RESIDUAL METHOD
FOR 1D BAR PROBLEMS
INTRODUCTION
Direct stiffness method is limited for simple 1D problems
FEM can be applied to many engineering problems that are
governed by a differential equation
Need systematic approaches to generate FE equations
Weighted residual method
Energy method
Ordinary differential equation can be solved using the weighted
residual method, in particular using Galerkin method
Principle of minimum potential energy can be used to derive
finite element equations
1
1D BAR PROBLEM
d du
( EA ) p 0, 0 x L
dx dx
For uniform cross-section:
d 2u
EA 2 p ( x ) 0, 0 x L
dx
Boundary conditions:
u (0) 0 (EBC)
du
EA ( L) P (NBC)
dx
1D BAR PROBLEM
Uniform elastic bar loaded by axial tip load P and distributed axial
load p(x)
Boundary value problem: differential equation + boundary conditions
Displacements in a uniaxial bar subject to a distributed force p(x)
d 2u
EA p 0, 0 x L
dx 2
u(0) 0
du Boundary conditions
EA ( L) P
dx
Essential BC: The displacement at a point is prescribed (displacement BC or
kinematic BC)
Natural BC: The axial force is given at a point (traction BC or stress BC)
Exact solution u(x): twice differentiable function (C2 function)
2
Weighted Residual Method: Galerkin Method
One-dimensional problem:
d 2u
EA p 0, 0 x L
dx 2
du
BC: u(0) 0, EA ( L) P
dx
Approximate solution:
n
u ( x) cii ( x)
i 1
where i(x), i = 1, 2, , n are the basis (weight) functions which satisfy
the essential BC, but not necessarily for natural BC
d 2u
Residual in domain: R( x ) EA p
dx 2
Galerkin Method
Galerkin method:
L
L d 2u
0 i ( x ) R ( x ) dx 0
0
i ( x )
EA
dx 2
p dx 0
Using integration by parts to reduce the order of differentiation in
integrals (weaker form):
d i EA du p dx EA du 0
L L
i
dx 0
i
0 dx dx
d i EA du p dx ( L) P 0, i 1,2,
L
i i , n solve ci
0 dx dx
3
Galerkin Method
L d d j
n
0 EA dxi dx dx c j
L
j 1
0 i ( x ) pdx i ( L) P, i 1,2, ,n
n
Kij c j Fi , i 1, ,n
j 1
where
L
d i d j
Kij
EA dx
0 dx dx
L
Fi 0 i ( x ) pdx i ( L) P
EXACT SOLUTION OF EXAMPLE 1
Example 1:
d 2u
EA p0 0, 0 xL ( p ( x ) p0 )
dx 2
u(0) 0
du Boundary conditions
EA ( L) P
dx
Exact solution:
p0 x 2 p0 Lx Px
u( x )
2 EA EA EA
4
Example Problem 1
Approximate solution
2
u ( x) cii ( x) c1x c2 x 2
i 1
Coefficient matrix and RHS vector
L
K11 EA 1 2 dx EAL p0 L2 PL
0 F1
L 2 EA EA
K12 K 21 EA 12 dx EAL2
0 p L3 PL2
L 4 F2 0
K 22 EA 2 2 dx EAL3 3EA EA
0 3
Example Problem 1
K11c1 K12 c2 F1
K12 c1 K 22 c2 F2
Matrix equation
p0 L2
L L2 2 PL
[K ] EA
4 3 {F}
L2 L p0 L3 2
3 PL
3
Approximate solution
P pL p
u( x ) x 0 x 0 x2
EA EA 2 EA
10
5
Example Problem 2
Uniform elastic bar loaded by axial tip load P and linearly-
distributed axial load p(x) = ax
d 2u
EA ax 0, 0 x L
dx 2
u(0) 0
du Boundary conditions
EA ( L) P
dx
u( x ) c1 x c2 x 2
1 x, 2 x 2
P 7aL2 aL 2
u( x ) x x x
EA 12 EA 4 EA
P aL2 a 3
Exact solution: u( x ) x x x
EA 2 EA 6 EA
11
FORMAL PROCEDURE
Galerkin method is still not general enough for computer code
Apply Galerkin method to one element (e) at a time
Introduce a local coordinate
x xi xx d 1
x xi (1 ) x j (e) i
x j xi L dx L(e)
Approximate solution within the element
N 1( ) N 2 ( )
u( x ) N1 ( x )ui N 2 ( x )u j [ N]{d}
N1 ( ) 1
N 2 ( ) Element e
xx xi xj
N1 ( x ) 1 ( e ) i (e )
L L
x xi
N 2 ( x) (e)
L
12
6
FORMAL PROCEDURE cont.
Interpolation property
N1 ( xi ) 1, N1 ( x j ) 0 u ( xi ) ui
N 2 ( xi ) 0, N 2 ( x j ) 1 u( x j ) u j
Derivative of approx. solution
du dN1 dN
ui 2 u j
dx dx dx
du dN1 dN 2 u1 1 dN1 dN 2 u1
(e)
dx dx
dx u2 L d d u2
1 u1
( e ) 1 1 [ N]{d}
L u2
xj
dN k du xj du du
EA dx p( x ) N k ( x )dx EA ( x j ) N k ( x j ) EA ( xi ) N k ( xi ), k 1,2
xi dx dx xi dx dx
13
FORMAL PROCEDURE cont.
Change variable from x to
EA 1 dN k dN1 dN 2 u1 1
( e ) 0 d d
d L( e ) p( x ) N k ( )d
L d
2
u 0
du du
EA ( x j ) N k (1) EA ( xi ) N k (0), k 1,2
dx dx
Element-level matrix equation
EA du ( x )
dx i 1 N1 ( )
[k ( e ) ]{u( e ) } {f p( e ) } {f p( e ) } L( e ) p( x ) d
EA du ( x j )
0 N 2 ( )
dx
dN1 2 dN1 dN 2
EA 1 d d d
d EA
1 1
k ( e ) ( e )
22 L 0 dN 2 dN1 dN 2
2 L 1 1
(e)
d d d
14
7
FORMAL PROCEDURE cont.
Need to derive the element-level equation for all elements
Consider Elements 1 and 2 (connected at Node 2)
(1) (1)
EA du ( x )
k11 k12 u1 f p1 dx 1
k21 k22 u2 f p 2 EA du ( x2 )
dx
(2) (2)
EA ( x )
du
k11 k12 u2 f p 2 dx 2
k21 k22 u3 f p 3 EA du ( x3 )
dx
Assembly
EA ( x1 )
du
k11
(1) (1)
0 u1 f p(1)
k12 1 dx
(1) (2) (1)
(2)
k21
(1)
k22 k11
(2)
k12 u2 f p2 f p2 0
(2) u
3
(2) (2) du
0 k21 k22 f p3 EA ( x3 )
dx
15
FORMAL PROCEDURE cont.
Assembly of NE elements (ND = NE + 1)
(1) (1)
0 u f p(1) EA du ( x1 )
k11 k12 0
1
1
dx
k (1) (1)
k22 k11
(2) (2)
k12 0 u2 f p 2 f p 2
(1) (2)
0
21
0 (2)
k21 (2)
k22 k11
(2)
0 u3 f p(2) (3)
3 f p3
0
0 0 0 NE
k21 ( NE )
k22 uN ( NE ) du
N D 1
f EA ( x )
pN
N D 1 dx N
ND ND N D 1
[K s ]{Ds } {Fs }
Coefficient matrix [Ks] is singular; it will become non-singular after
applying boundary conditions
16
8
EXAMPLE
Use three equal-length elements (EA = 1, L(e) = 1/3)
d 2u
x 0, 0 x 1 u (0) 0, u (1) 0
dx 2
All elements have the same coefficient matrix
1 1 1 3 3
k (e) (e) , (e 1, 2,3)
22 L 1 1 3 3
Change variable of p(x) = x to p(): p( ) xi (1 ) x j
N1 ( ) 1
d L( e ) xi 1 x j
1 1
{f p( e ) } L( e ) p( x ) d
0 N 2 ( ) 0
xi x j
3 6
L( e ) , ( e 1,2,3)
xi x j
6 3
17
EXAMPLE cont.
RHS cont. f p(1)
1 1 1 f p 2 1 4
(2)
f p 3 1 7
(3)
(1) , (2) , (3)
f p 2 54 2 f p 3 54 5 f p 4 54 8
Assembly
1 EA du (0)
54 dx
3 3 0 0 u1 2 4 Element 1
3 3 3 Element 2
3 0 u2 54 54
Element 3
0 3 33 3 u3 7 5
0
0 3 3 u 54 54
4 8 du
EA (1)
54 dx
Apply boundary conditions
Deleting 1st and 4th rows and columns
6 3 u2 1 1 u2 4
81
3 6 u 9 2 u3
3 5
81
18
9
EXAMPLE cont.
Approximate solution
4 x, 0 x
1
27 3 0.08
1 1 1 2
u-approx.
u-exact
u ( x ) x, x
27 27 3 3 0.06
5 5 x, 2
x 1
27 27
u(x)
3 0.04
Exact solution 0.02
x 1 x2
1
u ( x) 0
6 0 0.2 0.4
x
0.6 0.8 1
Three element solutions are poor
Need more elements
19
PROBLEMS: 3.1, 3.3, 3.7
Solution of problem 3.1:
(a) One-term solution
(b) Two-term solution
Solution of problem 3.3:
Solution of problem 3.7:
20
10
