CE 595 Section 1 - Purdue University

Download Report

Transcript CE 595 Section 1 - Purdue University

Section 4: Implementation of Finite Element
Analysis – Other Elements
1. Quadrilateral Elements
2. Higher Order Triangular Elements
3. Isoparametric Elements
Implementation of FEA:
Other Elements
-1-
Section 4.1: Quadrilateral Elements
 Refers in general
to any four-sided,
2D element.
 We will start by
considering rectangular
elements with sides
parallel to coordinate
axes. (Thickness = h)
Implementation of FEA:
Other Elements
-2-
4.1: Quadrilateral Elements (cont.)
Normalized Element Geometry –
 “Standard” setting
for calculations:
 Mapping between real and normalized coordinates:
x  xc
y  yc
Implementation of FEA:

; 
 x  a  xc ; y  b  yc . -3Other Elements
a
b
4.1: Quadrilateral Elements (cont.)
First Order Rectangular Element (Bilinear Quad):
 4 nodes; 2 translational
d.o.f. per node.
 Displacements interpolated as follows:
u   x  ,  y    a1  a2  a3  a4
v   x  ,  y    a5  a6  a7  a8
Implementation of FEA:
Other Elements
“Bilinear terms” – implies that all
shape functions are products of
linear functions of x and y.
-4-
4.1: Quadrilateral Elements (cont.)
Shape Functions:
N1  ,  
1   1    ; N 2  ,   14 1   1    ;
N 3  ,   14 1   1    ; N 4  ,   14 1   1    .
Implementation of FEA:
Other Elements
 N k  ,   14 1   k  1   k  .
1
4
-5-
4.1: Quadrilateral Elements (cont.)
 Displacement interpolation becomes:
d 
 u  x, y    N1  , 
0
N 2  , 
0
N 3  , 
0
N 4  , 
0
 1 


  .
v
x
,
y
0
N

,

0
N

,

0
N

,

0
N

,











1
2
3
4

 
d 
 8
" N x   "
 Need to compute [B] matrix:
B  x        N  x  
 x

 0
 y

0
 N1  , 
0
N 2  , 
0
N 3  , 
0
N 4  , 
0

 
   ? .
y  
0
N

,

0
N

,

0
N

,

0
N

,





1
2
3
4
 
x 
Implementation of FEA:
Other Elements
-6-
4.1: Quadrilateral Elements (cont.)
 Chain rule:
Nk
x
Nk
y
 ,   N
N

,


  
k

x
 Nk
k

y
 Nk
0
1  k  ;
a

1
1


1



*

k  b
y
4 k 
4 ab 1   k   .

x
 14  k 1  k  * a1 
bk
4 ab
k
0
 Resulting [B(x)] matrix:
0
b 1   
0
b 1   
0
b 1   
0
 b 1   



0
a 1   
0
a 1   
0
a 1   
0
a 1    
B  x    41ab 
 a 1    b 1    a 1    b 1    a 1    b 1    a 1    b 1    
 Recall general expression for [k]:
k   h * 
Implementation of FEA:
Other Elements
area
B  x  C B  x  dA
T
Express in terms of  and !
-7-
4.1: Quadrilateral Elements (cont.)

Can show that
 1  1
 dx
 dy

dA  dxdy  
d 
d   ab * d d   dA    ab * d  d .
 d
 d

area
 1  1
k   h * ab *
 1  1
  B CB d d.
T
Everything in terms of  and !
 1  1

Can also show that
 f    N  x   b  x   dV  
T
Ve
 h * ab *
 
 1  1
 1

 1
 1
h * b
Implementation of FEA:
Other Elements
T
Ae ,
 1  1
h * b
 N  x    t  x   dA

 1
 N  ,    b  ,   d d  h * a
T
 N   1,    t   1,   d  h * a
T
 1

 1
 1

 1
 N  ,  1   t  ,  1  d
T
 N  ,  1   t  ,  1  d
T
 N   1,    t   1,   d .
T
-8-
4.1: Quadrilateral Elements (cont.)
Gauss Quadrature:
 Let’s take a closer look at one of the integrals for the
element stiffness matrix (assume plane stress):
 1  1


E
h
2
2
2
1 2
k22 
*
a
1



b
1


1


  2    d d.
2


1  16ab  1  1
 Can be solved exactly, but for various reasons FEA
prefers to evaluate integrals like this approximately:

Historically, considered more efficient and reduced coding errors.
Only possible approach for isoparametric elements.

Can actually improve performance in certain cases!

Implementation of FEA:
Other Elements
-9-
4.1: Quadrilateral Elements (cont.)
Gauss Quadrature:
 Idea: approximate integral by a sum of function values at
predetermined points with optimal weights –
1
n
1
i 1
weights = known constants, depend on n
1D case:     d  Wi i .
Gauss points = known locations, depend on n
 n = order of quadrature; determines accuracy of integral.
(Note: any polynomial of order 2n-1 can be integrated exactly using nth order Gauss quadrature.)
Implementation of FEA:
Other Elements
-10-
4.1: Quadrilateral Elements (cont.)
Gauss Quadrature:
 Have tables for weights and Gauss points:
 2D case handled as two 1D cases:
1 1
    ,  d d  WW   , .
1 1
Implementation of FEA:
Other Elements
n
n
j 1 i 1
i
j
i
j
-11-
4.1: Quadrilateral Elements (cont.)
Higher Order Rectangular Elements
 More nodes; still 2 translational d.o.f. per node.
 “Higher order”  higher degree of complete polynomial
contained in displacement approximations.
 Two general “families” of such elements:
Implementation of FEA:
Other Elements
Serendipity
Lagrangian
-12-
4.1: Quadrilateral Elements (cont.)
Lagrangian Elements:
 Order n element has (n+1)2 nodes arranged in square-
symmetric pattern – requires internal nodes.
 Shape functions are products of nth order polynomials
in each direction. (“biquadratic”, “bicubic”, …)
 Bilinear quad is a Lagrangian element of order n = 1.
Implementation of FEA:
Other Elements
-13-
4.1: Quadrilateral Elements (cont.)
Lagrangian Shape Functions:
 Uses a procedure that automatically satisfies the
Kronecker delta property for shape functions.

Consider 1D example of 6 points; want function = 1 at 3  0.3
and function = 0 at other designated points:
0  1;
1  .75;
 2  .2;
3  .3;
 4  .6;
5  1.
Implementation of FEA:
Other Elements
L(5)
3   
  0   1    2    4   5  .
3  0 3  1 3   2 3   4 3  5 
-14-
4.1: Quadrilateral Elements (cont.)
Lagrangian Shape Functions:
 Can perform this for any number of points at any
designated locations.
( m)
k
L
m
  0    1     k 1    k 1     m 
  i 



.
  
k  0 k  1  k  k 1 k  k 1  k  m  i0 k  i 
ik
No -k term!
Implementation of FEA:
Other Elements
Lagrange
polynomial
of order m
at node k
-15-
4.1: Quadrilateral Elements (cont.)
Lagrangian Shape Functions:

Use this procedure in two directions at each node:
(3)
6
V
H
(3)
6
  5   7   8 

  
6  5 6  7 6  8 
Implementation of FEA:
Other Elements
  2    10    14 

  
6  2 6  10 6  14 
N 6  ,   H 6(3)  V6(3)   .
-16-
4.1: Quadrilateral Elements (cont.)
Notes on Lagrangian Elements:
 Once shape functions have been identified, there are no
procedural differences in the formulation of higher order
quadrilateral elements and the bilinear quad.
 Pascal’s triangle for the Lagrangian quadrilateral elements:
Implementation of FEA:
Other Elements
3x3
nxn
-17-
4.1: Quadrilateral Elements (cont.)
Serendipity Elements:
 In general, only boundary nodes – avoids internal ones.
 Not as accurate as Lagrangian elements.
 However, more efficient than Lagrangian elements and
avoids certain types of instabilities.
Implementation of FEA:
Other Elements
-18-
4.1: Quadrilateral Elements (cont.)
Serendipity Shape Functions:
 Shape functions for mid-side nodes are products of an
nth order polynomial parallel to side and a linear function
perpendicular to the side.

E.g., quadratic serendipity element:
Implementation of FEA:
Other Elements
N6  12 1    1   2  ; N7  12 1   2  1    .
-19-
4.1: Quadrilateral Elements (cont.)
 Shape functions for corner nodes are modifications of
the shape functions of the bilinear quad.



Step #1: start with appropriate bilinear quad shape function, Nˆ 1 .
Step #2: subtract out mid-side shape function N5 with appropriate
weight Nˆ 1  node #5  12
Step #3: repeat Step #2 using mid-side shape function N8 and weight Nˆ 1  node #8  12
Implementation of FEA:
Other Elements
Nk 
1
4
1   k 1  k  k  k  1 ; k  1, 2,3, 4.
-20-
4.1: Quadrilateral Elements (cont.)
Notes on Serendipity Elements:
 Once shape functions have been identified, there are no
procedural differences in the formulation of higher order
quadrilateral elements and the bilinear quad.
 Pascal’s triangle for the serendipity quadrilateral elements:
Implementation of FEA:
Other Elements
3x3
mxm
-21-
4.1: Quadrilateral Elements (cont.)
Zero-Energy Modes (Mechanisms; Kinematic Modes) –
 Instabilities for an element (or group of elements) that
produce deformation without any strain energy.
 Typically caused by using an inappropriately low order of
Gauss quadrature.
 If present, will dominate the deformation pattern.
 Can occur for all 2D elements except the CST.
Implementation of FEA:
Other Elements
-22-
4.1: Quadrilateral Elements (cont.)
Zero-Energy Modes –
 Deformation modes for a bilinear quad:



#1, #2, #3 = rigid body modes; can be eliminated by proper constraints.
#4, #5, #6 = constant strain modes; always have nonzero strain energy.
#7, #8 = bending modes; produce zero strain at origin.
Implementation of FEA:
Other Elements
-23-
4.1: Quadrilateral Elements (cont.)
Zero-Energy Modes –
 Mesh instability for bilinear quads using order 1 quadrature:
“Hourglass modes”
Implementation of FEA:
Other Elements
-24-
4.1: Quadrilateral Elements (cont.)
Zero-Energy Modes –
 Element instability for quadratic quadrilaterals using 2x2
Gauss quadrature:
“Hourglass modes”
Implementation of FEA:
Other Elements
-25-
4.1: Quadrilateral Elements (cont.)
Zero-Energy Modes –
 How can you prevent this?



Use higher order Gauss quadrature in formulation.
Can artificially “stiffen” zero-energy modes via penalty functions.
Avoid elements with known instabilities!
Implementation of FEA:
Other Elements
-26-
Section 4: Implementation of Finite Element
Analysis – Other Elements
1. Quadrilateral Elements
2. Isoparametric Elements
3. Higher Order Triangular Elements
Note: any type of geometry can be used for isoparametric elements;
we will only look at quadrilateral elements.
Implementation of FEA:
Other Elements
-27-
Section 4.2: Isoparametric Elements
 For various reasons, need elements that do not “fit” the
standard geometry.
Curved boundaries
Implementation of FEA:
Other Elements
Transition regions
-28-
4.2: Isoparametric Elements (cont.)
 Problem: How do you map a general quadrilateral onto
the normalized geometry?
 x, y   F  ,    ,   F1  x, y  ; F  ?
Implementation of FEA:
Other Elements
-29-
4.2: Isoparametric Elements (cont.)
 Idea: Approximate the mapping using “shape functions”.
x  N1*  ,  x1  N*2  ,  x2  N*3  ,  x3 
 N*n  ,  xn ;
y  N1*  ,  y1  N*2  ,  y2  N*3  ,  y3 
 N*n  ,  yn .
*
 Require N k  ,  to have Kronecker delta property.
*
N
 k  ,  not required to be the actual shape functions of
the element; n can be as large or as small as you want.
Implementation of FEA:
Other Elements
-30-
4.2: Isoparametric Elements (cont.)
 Approximate “serendipity element” shown using bilinear
quad shape functions and approximation points at corners
1   1    x1  14 1   1    x2
 14 1   1    x3  14 1   1    x4 .
x
Implementation of FEA:
Other Elements
1
4
-31-
4.2: Isoparametric Elements (cont.)
 For an isoparametric element, the number of approximation
points equals the actual number of nodes for the element;
also, the approximation functions are the actual shape
functions for the element:
x  N1  ,  x1  N 2  ,  x2  N 3  ,  x3   N n  ,  xn ;
y  N1  ,  y1  N 2  ,  y2  N 3  ,  y3 
 N n  ,  yn ;
u  x, y   N1  ,  u1  N 2  ,  u2  N 3  ,  u3 
 N n  ,  un ;
v  x, y   N1  ,  v1  N 2  ,  v2  N 3  ,  v3 
 N n  ,  vn ;

If # of approx. pts. > # of nodes, element is called
superparametric; if # of approx. pts. < # of nodes,
element is called subparametric.
Implementation of FEA:
Other Elements
-32-
4.2: Isoparametric Elements (cont.)
Formulating an Isoparametric Element:
 Recall the formulation of “standard” bilinear quad:
 1  1
 dx
 dy

dA  dxdy  
d 
d   ab * d d   dA    ab * d  d .
 d
 d

area
 1  1
k   h * ab *
 1  1




1 1
Implementation of FEA:
Other Elements
B CB d d.
T
How does this work for an
isoparametric element?
-33-
4.2: Isoparametric Elements (cont.)
Formulating an Isoparametric Element:
 Calculating the [B] matrix (assume isoparametric bilinear
quad element):
d 
 u  x, y    N1  , 
0
N 2  , 
0
N 3  , 
0
N 4  , 
0
 1 


  .
v
x
,
y
0
N

,

0
N

,

0
N

,

0
N

,

 



  
1
2
3
4
 
 d8 
" N x   "
 x

B  x     0
 y

0
 N1  , 
0
N 2  , 
0
N3  , 
0
N 4  , 
0

 
   ? .
y  
0
N

,

0
N

,

0
N

,

0
N

,





1
2
3
4
 
x 
Need to apply the chain rule!
Implementation of FEA:
Other Elements
-34-
4.2: Isoparametric Elements (cont.)
Formulating an Isoparametric Element:
 Chain rule: compute inverse rule first –
N k N k x N k y N k N k x N k y


;


.

x  y  
x  y 
 Using the approximate mapping:
n
n
N i
N i
x
x
x   N i  ,  xi 

 ,  xi ;    ,  xi .
 i 1 
 i 1 
i 1
n
n
n
N i
N i
y
y
Similarly,

 ,  yi ;    ,  yi .
 i 1 
 i 1 
Implementation of FEA:
Other Elements
-35-
4.2: Isoparametric Elements (cont.)
Formulating an Isoparametric Element:
 Put all of this together –
 N k
 

 N k
 

  x
  

  x
  
 
y   N k
   x

y   N k
   y
 n N i
    ,  xi
i 1
 n
 N i
    ,  xi
 i 1
Implementation of FEA:
Other Elements






N i
  N k

,

y
  i 

i 1 
  x
n
N i
  N k
 ,  yi   y

i 1 

n
The Jacobian matrix [J] of the mapping.


.



-36-
4.2: Isoparametric Elements (cont.)
Formulating an Isoparametric Element:
 Can now compute the regular chain rule –
 N k
 

 N k
 


 N k

 x
  J 

 N k
 y



  N k
  x

  N k
  y
 
 n N i
 ,  yi


1  i 1 
1
J  
J  n N i
    ,  xi
 i 1
Implementation of FEA:
Other Elements
 N k

 

1
  J 
 N k


 




.



N i


 ,  yi 
i 1 
 ; J  det  J .
n
N i

 ,  xi  “Jacobian” of

i 1 

n
the mapping
-37-
4.2: Isoparametric Elements (cont.)
Formulating an Isoparametric Element:
 J is a (nonconstant) scaling factor that relates area in
original geometry to area in normalized geometry; can
show that dxdy  J * d d.
 For a well-defined mapping, J must have same sign at
all points in normalized geometry.
 Large variations in J imply highly distorted mappings –
leads to badly formed elements.
Implementation of FEA:
Other Elements
-38-
4.2: Isoparametric Elements (cont.)
Formulating an Isoparametric Element:
 x 
 
 u   N  x   d    ε     y   B  x   d  ;
 
 xy 
 Calculating the [B] matrix:
 ux   ux 
  x  1 0 0 0   u   u 
1

J


  
  y   y 
  y   0 0 0 1   v  ;  v   
   0 1 1 0   x   x    0
  v   v 
 xy  
 y   y 
1 0 0 0 
1
 B   0 0 0 1 

J
0 1 1 0 
 n N
iy
 
 i 1  i

 n Ni
xi
 
 i 1 


0



0


Implementation of FEA:
Other Elements
n N
iy

i
i 1 
n N
ix

i
i 1 
0
0
 u   u   N1
0   u   u   N1

;

1  v   v 
 J          0
 v   v 
       0
0
0
n N
iy

i


i 1
n N
ix

i


i 1





0


n N 
iy 

i


i 1

n N

i
xi 


i 1 
0
 N1
 N1
 
0

 0
0
0
N 2

N 2

0
0
N3

N3

0
0
N 4

N 4

N1

N1

0
0
N 2

N 2

0
0
N3

N3

0
0
0
0
N 2

N 2

0
0
N3

N3

0
0
N 4

N 4

N1

N1

0
0
N 2

N 2

0
0
N3

N3

0
0
  d1 
 
  d2 .
N 4
 


N 4  
d
 8 
 
0
0
0

0
N 4 .


N 4 
 

-39-
4.2: Isoparametric Elements (cont.)
Calculating the element stiffness matrix:
 1  1
k   h *   B CB * J  ,  d d.
T
 1  1
area scaling factor –
polynomial function of (,)
 Note: [B] is proportional to J-1:
polynomials functions of  ,  
.
B 
J  , 
new polynomials functions of  ,  
k   h *  
d d .
J  , 
 1  1
 1  1
In general, you are integrating ratios of polynomial functions, which typically
don’t have exact integrals  use Gauss quadrature to evaluate!
Implementation of FEA:
Other Elements
-40-
4.2: Isoparametric Elements (cont.)
Calculating the element nodal forces:
 f    N  x  b  x   dV  
T
Ve
N  x   t  x   dA  ?
T
Ae ,
 Body force contribution:
 N  x  b  x  dV  h *
T
Ve
What do you do with this?
 1  1
 
 1  1
 N  ,   b  x, y   * J  ,  d d.
T
 Surface traction contribution:

 N  x  
T
 t  x   dA  
Ae ,
all edges
h*

edge # k
 N  , 

edge
#
k


T
 t  x, y 
edge # k
d
edge # k
What do you do with these?
Implementation of FEA:
Other Elements
-41-
4.2: Isoparametric Elements (cont.)
Converting body force and surface tractions:
 Idea #0: If body force = constant and/or surface traction on
edge #k = constant, do nothing!
 Idea #1: Use the isoparametric mapping to modify force functions:
n
  n

x   N i  ,  xi , y   N i  ,  yi   b  x, y     b   N i  ,  xi ,  N i  ,  yi   .
i 1
i 1
i 1

  i 1
" bˆ  ,   "
 1  1
T
T
   N  x    b  x   dV  h *    N  ,   bˆ  ,  * J  ,  d d .
n
n

 1  1
Ve

 Idea #2: Make an isoparametric approximation for the forces:
n
 b  x, y     N  ,  *  b  x , y  
 1  1
i 1
i
i
i
T
 n

   N  x    b  x   dV  h *    N i  ,  *  N  ,   b  xi , yi    * J  ,  d d .

Ve
 1  1  i 1
T
Implementation of FEA:
Other Elements
-42-
4.2: Isoparametric Elements (cont.)
Converting dℓ on edge #k:
 In general:
d 
 dx    dy 
2
2
; dx 
x
x
y
y
d 
d and dy 
d 
d .




 On the given edge #k, d  0 :
2
d
edge # k
2
2
2
 x 
 y 
 n Ni
  n Ni

 d  
 
 d  
 ,  1 xi      ,  1 yi  .
  edge # k   edge # k
 i 1 
  i 1 

Implementation of FEA:
Other Elements
L  
-43-
4.2: Isoparametric Elements (cont.)
 Thus, the contribution from surface tractions on edge #k is:
h*

edge # k
 N  , 

edge # k 

T
 t  x, y 
 1
edge # k
d
edge # k
Idea #1!
n
  n

 h *   N  ,  1   t   N i  ,  1 xi ,  N i  ,  1 yi   * L   d .
i 1

  i 1
 1
T
 Note: Ni  ,  1  0 unless i = k or i = k+1 !
Implementation of FEA:
Other Elements
-44-
4.2: Isoparametric Elements (cont.)
Example: Formulating an Isoparametric Bilinear Quad –
 0.4*  8  y  
t  x, y   
 ksi
0


 Given: 4-node plane stress element has E = 30,000 ksi,  = 0.25, h =
0.50 in, no body force, and surface traction shown.
 Required: Find [k] and (f). Use 2 x 2 Gauss quadrature for [k].
Implementation of FEA:
Other Elements
-45-
4.2: Isoparametric Elements (cont.)
Solution:
 Isoparametric mapping:
x  14 1   1    x1  14 1   1    x2  14 1   1    x3  14 1   1    x4
 14 1   1    * 4  14 1   1    *8  14 1   1    *11  14 1   1    * 2
= 254  134   14   54  ;
y  14 1   1    *3  14 1   1    * 4  14 1   1    *10  14 1   1    *8
= 254  34   114   14  ;
 Jacobian matrix and Jacobian:
 x
 
 J    x

 

Implementation of FEA:
Other Elements
y 
   134  54 

y   14  54 
 
 14  
35
27
1
;
J

det
J






4
8
8 .

11
1
4  4
3
4
-46-
4.2: Isoparametric Elements (cont.)
Solution:
 [B] matrix:
1 0 0 0 
1
B   0 0 0 1 
J
0 1 1 0 
 n N
iy
 
 i 1  i

 n Ni
xi
 
 i 1 


0



0


n N
iy

i


i 1
n N
ix

i


i 1
0
0
n N
iy

i
i 1 
n N
ix

i
i 1 
0
0
 11 1
 
1 0 0 0   4 4
 1  5
8


=
0 0 0 1  4 4

 0
70    27
0 1 1 0   0

 3  1
4 4
13  5
4 4
0
0
11 1 
4 4
 1  5
4 4

 4 6  2
1

=
0


70    27  63 9
Implementation of FEA:
Other Elements
0
63 9
4 6  2





0


n N 
iy 

i
i 1  
n N

ix 

i
i 1 

0
7 5  2
0
7  2 9
0
0
0
7  2 9
7 5  2
 N1
 N1
 
0

 0
N 2

N 2

0
0
N1

N1

 1 1
   4  4

0   1  1
 4 4
 3  1   0
4 4 
13  5   0
4 4  
45 
0
6 2  4
N 2

N 2

0
0
0
0
0
6 2  4
45 
0
0
1  1
4 4
 1  1
4 4
0
0
1  1
4 4
 1  1
4 4
 1  1
4 4
1
  1
4 4
0
N3

N3

0
0
7 6 
0
7 3  4
0
0
0
0
N3

N3

N 4

N 4

0
0
1  1
4 4
1  1
4 4
0
0
1  1
4 4
1  1
4 4
0
0
0

0
N 4


N 4 
 

 1  1
4 4
1  1
4 4
0
0


7 3 4 
76  
0
-47-



0 

 1  1 
4 4 
1  1 
4 4 
0
4.2: Isoparametric Elements (cont.)
Solution:
 [k] matrix:
0 
32000 8000
C   8000 32000 0  ksi;
 0
0
12000 
1 1
k    0.5 in    B  CB  * Jd d
T
1 1


31.25* 236276 196 315 2 354  275 2

8
*
70  27 

sym
1 1

1 1



31.25* 70+231 203 90 2  231  43 2

31.25* 539588 490 180 228 131
2

2







d  d
k  ,  
Implementation of FEA:
Other Elements
-48-
4.2: Isoparametric Elements (cont.)
Solution:
 2 x 2 Gauss quadrature:
Wi  W j  1; i, j  1, 2.
1 
1 
  1 1    1
  1 1 
 
   1
k   WW
i
j * k     i ,   j    k   3 ,  3    k   3 , 3    k  3 ,  3    k  3 , 3  
2
2
i 1 j 1
Implementation of FEA:
Other Elements
7028.9
k   

1260.6
7136.6
Note: k exact  

1263.9
1260.6 
 kips/in.

8489.9 
1263.9 
 kips/in.

8499.0 
-49-
4.2: Isoparametric Elements (cont.)
Solution:
 Element nodal forces:
Implementation of FEA:
Other Elements
-50-