Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Methods
1
Formulation of FEM Model
Direct Method
Formulation of FEM Model Variational Method
Weighted Residuals
• Several approaches can be used to transform the physical
formulation of a problem to its finite element discrete analogue.
3
Physical problems
Axially loaded elastic bar
x
x
x=0 x=L
Boundary conditions (examples)
u0 at x 0 Dirichlet/ displacement bc
u 1 at x L
u 0 at x 0
du
EA F at x L Neumann/ force bc
dx
Differential equation + Boundary conditions = Strong form of the
“boundary value problem”
5
Physical problems
Flexible string
d 2u
S 2 p 0; 0 xL
dx
Second order differential equations
Requires 2 boundary conditions for solution
6
Physical problems
Heat conduction in a fin
y
A(x) = cross section at x
Q(x) = heat input per unit length per
x
unit time [J/sm]
x k(x) = thermal conductivity [J/oC ms]
x=0 x =L T(x) = temperature of the fin at x
Q(x)
d dT
Ak Q 0; 0 xL
dx dx
Second order differential equations
Requires 2 boundary conditions for solution
7
Physical problems
Heat conduction in a fin
y
x
x
x=0 x=L
Q(x)
8
Physical problems
Fluid flow through a porous medium (e.g., flow of water through a dam)
y
10
Physical problems
11
Formulation of FEM Model
Observe:
1. All the cases we considered lead to very similar differential
equations and boundary conditions.
2. In 1D it is easy to analytically solve these equations
3. Not so in 2 and 3D especially when the geometry of the domain is
complex: need to solve approximately
4. We’ll learn how to solve these equations in 1D. The approximation
techniques easily translate to 2 and 3D, no matter how complex the
geometry
12
Formulation of FEM Model
A generic problem in 1D
d 2u
2
x 0; 0 x 1
dx
u 0 at x 0
u 1 at x 1
Analytical solution
1 7
u ( x) x 3 x
6 6
Assume that we do not know this solution.
13
Formulation of FEM Model
A generic problem in 1D
A general algorithm for approximate solution:
Guess u( x) a0j o ( x) a1j1 ( x) a2j 2 ( x) ...
where jo(x), j1(x),… are “known” functions and ao, a1, etc are constants
chosen such that the approximate solution
Satisfies the differential equation
Satisfies the boundary conditions
i.e.,
d 2jo ( x) d 2j1 ( x) d 2j 2 ( x)
a0 2
a1 2
a2 2
... x 0; 0 x 1
dx dx dx
a0jo (0) a1j1 (0) a2j 2 (0) ... 0
a0jo (1) a1j1 (1) a2j 2 (1) ... 1
Solve for unknowns ao, a1, etc and plug them back into
This is your approximate
u ( x) a0j o ( x) a1j1 ( x) a2j 2 ( x) ... solution to the strong form
14
Formulation of FEM Model
Solution of Continuous Systems – Fundamental Concepts
Exact solutions
limited to simple geometries and boundary & loading conditions
Approximate Solutions
Reduce the continuous-system mathematical model to a discrete idealization
Lu 0
L is a differential operator
d du
eg. For Axial element EA P( x) 0
dx dx
d d
L EA P( x)
dx dx
Exact Approximate
ERROR L u R
Objective:
Define u so that weighted average of Error vanishes
w L(u) dV 0
V
or w R dV 0
V
17
Weighted Residual Methods
Weighted Residual Formulations
w
w L(u ) dV 0
V
w 1
ERROR
18
Weighted Residual Methods
Weighted Residual Formulations
w w L(u ) dV 0
V
ERROR
19
Appendix
20
Weighted Residual Methods
x j Power series
sin jx, cos jx Fourier series
w j ( x) L j ( x) Lagrange
H j ( x)
Hermite
T ( x ) Chebychev
j
21
Weighted Residual Methods
Assume certain profile (trial or shape function) between nodes
J
u ( x) a j j ( x)
j 1
22
Weighted Residual Methods
In general, we deal with the numerical integration of trial or
interpolation functions
Trial functions:
constant, linear, quadratic, sinusoidal, Chebychev
polynomial, ….
Weighting functions:
subdomain, collocation, least square, Galerkin, ….
w( x, y, z) R( x, y, z)dxdydz wL(u)dv 0
V V
23
General Formulation
24
Weighted Residual Methods
wm ( x, y, z ) R( x, y, z )dxdydz 0, m 1, 2, ,M
25
Weighted Residual Methods
Weak form – integral formulation
R 0, but “weighted R” = 0
Choices of shape or interpolation functions?
Choices of weighting functions?
26
Subdomain Method
1, in Dm
wm
0, outside Dm
27
Collocation Method
R xm R xm , ym , zm
No control on the residuals between nodes
28
Least Square Method
am V V
am
R Square error
wm x R2 0
am
1
wm x R x dx 2 am dx 0 R2 0
2
R
29
Galerkin Method
wm x m x
w x R x dx x R x dx
m m
30
Numerical Accuracy
31
Application to an ODE
y (0) 1
Use global method with only one element
Select a trial function of the form of
N
y 1 ajx j
j 1
32
Application to an ODE
33
Residual
dx j 1 j 1 j 1
34
Subdomain Method
Zero average error in each subdomain
D1 D2 D3
Uniform spacing
x0 x1 x2 x3
xm
1 xm a a a
0
wm Rdx
xm1
R( x)dx 0 (a1 1) x (a2 1 ) x 2 (a3 2 ) x 3 3 x 4
2 3 4 xm1
1/ 3 5 8 11 1
m 1 : 0
Rdx 0
18
a1 a2
81 324
a3
3
1.0156
ai 0.4219
2/3 3 20 69 1
m 2 : Rdx 0 a1 a2
18 81 324
a3
3
1/ 3
0.2813
1 1 26 163 1
m 3 :
2/3
Rdx 0 a1
18 81
a2
324
a3
3
36
Subdomain Method
Nonuniform subdomains?
D1 D2 D3
x0 x1 x2 x3
m 1 : x1
x0
Rdx 0
x2
Different coefficients
m 2 : Rdx 0 for different choices of
x1
subdomains
m 3 : x3
x2
Rdx 0
37
Least Square Method
38
Least Square Method
R
2x x 2 Rdx 0
1
m 2, w2 2x x 2
a2 0
R
3x x 3x 2 x 3 Rdx 0
1
m 3, w3
2 3
a3 0
39
Least Square Method
40
Least Square Method
41
Least Square Method
Weighted average error =
Net area under curve = 0
42
Galerkin Method
m ( x ) x 0
, x 1
, x 2
, x 3
, , x N 1
m
w ( x ) m ( x ) x m1
N
R( x) 1 am (mx m1 x m )
m 1
jx x m j 1 dx
N
dx a j
1 1 1 1
wm Rdx x R( x)dx 0 x
m 1 m 1 m j 2
0 0 0 0
j 1
N
j 1 1
j 1
j
a
j m 1 j m m
N
S
j 1
mj a j dm SA D
43
Galerkin Method
R a1 1 2a2 a1 x 3a3 a2 x 2 a3 x 3
m 1, W x0 1 1
Rdx 0
1 0
1
m 2, W2 x x
1
xRdx 0
0
m 3, W3 x 2
1
x 2 Rdx 0
0
44
Galerkin Method
Cubic trial function
3
j 1 1
a j
j 1 m j 1 m j m
1 2 3
m 1, 2
a1 a2 a3 1
3 4
1.0141
a1 a2 a3 ai 0.4225
1 5 11 1
m 2,
6 12 20 2
0.2817
1 3 13 1
m 3, a1 a2 a3
12 10 30 3
y( x) 1 1.0141x 0.4225x 0.2817 x
2 3
R ( x ) 0.0141 0.1691x 0.4226 x 2
0.2817 x 3
W1 = 1
Weighting functions
W2 = x
W3 = x2
Weighted residuals xR
x2R
46
Galerkin Method
Order of
approximation:
Linear, Quadratic, and Cubic
Trial functions
47
Galerkin Method
48
Galerkin Method
Alternative choice of weighting functions
R (a1 1) (2a2 a1 ) x (3a3 a2 ) x 2 a3 x 3
3
1 (1 x)a1 (2x x 2 )a2 (3x 2 x 3 )a3 1 a j j
j 1
m 1, W 1 x 1
(1 x ) Rdx 0
1 0
1
m 2, W2 2x x 2
(2x x 2 ) Rdx 0
0
m 3, W3 3x 2 x 3
1
(3x 2 x 3 ) Rdx 0
0
But y yexact at
collocation points
y(1) = 2.7142857 e
51
Taylor-series Expansion
Truncated Taylor-series
2 3
1 1
ai 1 / 2 0.5
x x
e 1 x
x
2 ! 3!
1 / 6 0.1667
52
Interpolation Functions
53
Numerical Accuracy
54
Interpolation Functions
Comparison of
numerical errors
for weighted
residual methods
55