Professional Documents
Culture Documents
Finite Element Procedures For Solids and Structures Linear Analysis
Finite Element Procedures For Solids and Structures Linear Analysis
Finite Element
Procedures for Solids
and Structures
Linear Analysis
Klaus-JOrgen Bathe
Professor of Mechanical Engineering, MIT
PREFACE
The analysis of complex static and dynamic problems in
volves in essence three stages: selection of a mathematical
model, analysis of the model, and interpretation of the results.
During recent years the finite element method implemented on
the digital computer has been used successfully in modeling
very complex problems in various areas of engineering and
has significantly increased the possibilities for safe and cost
effective design. However, the efficient use of the method is
only possible if the basic assumptions of the procedures
employed are known, and the method can be exercised
confidently on the computer.
The objective in this course is to summarize modern and
effective finite element procedures for the linear analyses of
static and dynamic problems. The material discussed in the
lectures includes the basic finite element formulations em
ployed, the effective implementation of these formulations in
computer programs, and recommendations on the actual use
of the methods in engineering practice. The course is intended
for practicing engineers and scientists who want to solve prob
lems using modem and efficient finite element methods.
Finite element procedures for the nonlinear analysis of
structures are presented in the follow-up course, Finite Element
Procedures for Solids and Structures - Nonlinear Analysis.
In this study guide short descriptions of the lectures and
the viewgraphs used in the lecture presentations are given.
Below the short description of each lecture, reference is made
to the accompanying textbook for the course: Finite Element
Procedures in Engineering Analysis, by K.J. Bathe, Prentice
Hall, Inc., 1982.
The textbook sections and examples, listed below the
short description of each lecture, provide important reading
and study material to the course.
Contents
Lectures
l.
1-1
2.
2-1
3.
3-1
4.
4-1
5.
5-1
6.
6-1
7.
7-1
8.
8-1
9.
9-1
10-1
1l.
11-1
12.
12-1
10.
SOME BASIC
CONCEPTS OF
ENGINEERING
ANALYSIS
LECTURE 1
46 MINUTES
I-I
1-2
INTRODUCTION TO LINEAR
ANALYSIS OF SOLIDS AND STRUCTURES
use
lectures is:
to introduce to you finite
element methods for the
linear analysis of solids
and structures.
["Iinear" meaning infinitesi
mally small displacements and
linear elastic material proeer
ties (Hooke's law applies)j
to consider
13
=Automatic Dynamic
Incremental Nonlinear Analysis
Physical problem
... - -
Revise (refine)
the model?
I
I,-__S_ol_v_e_th_e_m_o_d_el_ _
~-
1-4
- - iL-_I_n_te_r.;..p_re_t_t_h_e_re_s_u_lt_S_....J
10 ft
15 ft
I,.
12 at 15
K~~~~~-~,-Fault
\\(no restraint assumed)
Altered' grit
E= toEc
.,
Analysis of dam.
15
t
W
....
E~~;;C=-------_
16
l,W
p
-0.2
PINCHED CYLINDRICAL
SHELL
OD;:...,...----.---~~~~~~C
16x 16 MESH
Mil
EtW -50
P- 100
""= 0.1
-150
16x 16 MESH
1-7
18
SYSTEMS
DISCRETE
CONTINUOUS
response is
described by
variables at a
finite number
of points
response is
described by
variables at
an infinite
number of
points
set of alge
braic - equations
set of differ
ential
equations
ANALYSIS OF DISCRETE
SYSTEMS
Steps involved:
reduction of continuous
system to discrete form
- system idealization
into elements
powerful mechanism:
- evaluation of element
equilibrium requirements
- element assemblage
- solution of response
19
Physical layout
ELEMENTS
U3
U1
~ : ~\l)
1
F(4)
1
k, u1 -- F(' )
u2
-,
'2 [
4' ]
.
[F1
'4 [-11-1]["1]
1 U
F(4)
3
1-1]["I]fF}]
1 u
F( 2 )
-1
u,
u2
k3
'3 [ ]
-1
-r ]f
1
F(S)
2
F(S)
3
1
P
u
F(3)l]
2
2
[F(5l]
'5 [-11-t]
1 u = F1S )
F(3)
-- 2
F(3)
110
F(4)
F(2 )
--- 2
F(2)
---,
..
..
Element interconnection
requirements :
F(4) + F(S) = R
3
3
3
These equations can be
written in the form
KU
= E.
Equilibrium equations
KU= R
(a)
T
U = [u
1
T = [R
-R
+k4
k + k + k ~ -k - k
... 1 2 3.'" ....2.......3 .......: ..-k....4 ..... ...
..
-k 2 - k3
k2 + k3 + k S~ -k S
.
... .........................................
...
..
111
~=
i =1
~(i)
where
: :]
o
etc ...
112
u,
................. :
:.............. .
...
.
.....
..
u,
...
K1
K=
...
:.............. .
...
:
U1
............................... :
~ ~~.~
~~
:...............
~1---.JI.l\fl--r/A~,1\1\1\~~r/A
113
u,
...
K=
.:
..
..
..
'O
u,
+ K4 ;
K 1 +K 2 + K 3 ;-K 2 -K 3
'O'O'O
.
..
..
..
'O'O'O'O'O'O'O'O'O'O'O'O'O'O:'O'O'O'O'O'O'O'O'O'O'O'O'O'O:'O'O'O'O'O'O'O'O'O'O'O'O'O'O
K=
'O'O
.
+
u,
K4 ;
K 1 +K 2 + K 3 ~-K2 -K 3
-K
'O'O
K=
-K 2 -K 3
~K2+ K3 + K5
'O'O :
114
-K
W =
an -
~-
(b)
U=~UT!!!
T
W = U R
Invoking (b) we obtain
KU = R
115
PROPAGATION PROBLEMS
main characteristic: the response
changes with time ~ need to
include the d'Alembert forces:
m,
a
a
M=
a a
m2 a
a m3
EIGENVALUE PROBLEMS
we are concerned with the
general ized eigenvalue problem
(EVP)
Av
ABv
116
Let
U
Then we obtain
_w
~~ sin W(t-T)
+ -K
<p
sin W(t-T)= -0
w,
,~,
(l)2 ' ~2
eigenpairs
w3 ' ~3
117
ANALYSIS OF
CONTINUOUS SYSTEMS;
DIFFEBENTIAL AND
VABIATIONAL
FOBMULATIONS
LECTURE 2
59 MINUTES
2-1
LECTURE 2
TEXTBOOK:
22
BASIC CONCEPTS
OF FINITE
ELEMENT ANALYSIS
CONTINUOUS SYSTEMS
Some additional
basic concepts are
used in analysis of
continuous systems
We discussed some
basic concepts of
analysis of discrete
systems
CONTINUOUS SYSTEMS
differential
formulation
variational
formulation
Ritz Method
Weighted residual
methods
Galerkin _.._-----41~_
least squares
....
finite element method
23
~Lt: )
Young's modulus, E
mass density,
cross-sectional area, A
R.. ~-------
~ .-;+~~ dx
I~
aA Ix + A ~a
dx - aA Ix
oX X
= p
au
= E ax
24
au
~
u(O,t}
EA ~~ (L,t) = RO
u(x,O}
~
at (x ' O) =
In general, we have
highest order of (spatial) deriva
tives in problem-governing dif
ferential equation is 2m.
highest order of (spatial) deriva
tives in essential b.c. is (m-1)
highest order of spatial deriva
tives in natural b.c. is (2m-1)
Definition:
We call this problem a Cm-1
variational problem.
25
fL
II = J } EA
o
au 2
(--)
ax
dx -
L
u f
dx - u
and
o=0
and we have 0 II
=0
L
au
au
rL.B
JO(EA ax)(6 ax) dx -)0 6u t- dx
- 6u
R = 0
or
26
LR
a8u
ax
8 au
for
, re-
ax
dx + [EA
- EA ~\
dX
~I
ax x=L
x=o
and
au I x=L=R
EAaxAlso
a2u
= -A p -
at 2
and
hence we have
27
Total Potential IT
~
Principle of Virtual
Displacements
solve
problem
Integration by parts
Differential Equation
of Equilibrium
and natural b.c.
28
_solve
problem
(3.6)
with the B.C.
B.[</>]
=
1
q.,
1
= 1 ,2,
at boundary (3.7)
i
(3.10)
where the f i are linearly indepen
dent trial functions and the ai
are multipliers that are deter
mined in the analysis.
(3.11 )
29
fD f.
R dD=O
;=1,2, ,n
(3.12)
a
aa.1
;=1,2, ,n
an
aa.
210
;=1,2, ,n
(3.14)
Properties
The trial functions used in the
Ritz analysis need only satisfy the
essential b.c.
Since the application of oIl = 0
generates the principle of virtual
displacements, we in effect use
this principle in the Ritz analysis.
By invoking 0 II = 0 we minimize
the violation of the internal equilibrium
requirements and the violation of
the natural b.c.
A symmetric coefficient matrix
is generated, of form
KU
=R
Example
Area = 1 em
(
........_x,u
.F-
--- --.;;B;",.,.
--
-- -
R = 100 N
~---r;;;-==-e-
I-...--~~----I..- --------I
100 em
80 em
Fig. 3.19. Bar subjected to
concentrated end force.
211
180
12 EA(~)2
ax
IT
dx -
100 u Ix = 180
Case 1
= a1x
+ a
Case 2
~
= I1JO
oIT =
oIT = 0
~~) o(~~)
dx - 100 OU Ix=180
=0
180
ET
212
(~~u)( EA ~~)
T
-
dV = IT. F.
1
dx = 100 OU x=180
Exact Solution
Using integration by parts we
obtain
~
ax
(EA
EA
~
ax
~)
ax
= 0
= 100
x=180
The solution is
a = 100;
a
100
; 100 < x < 180
(l+x-l00)2
40
213
dx+ I
180
100
I nvoking that
orr = 0
E [0.4467
116
we obtain
34076
116
and
a1
128.6
---=E=---
a 2 -- - 0.341
E
214
12C.6
E
x -
0.341
E
128.6 - 0.682 x
(1+ x-l00)2
40
Case 2
Here we have
100
E
n=2
(1+ x -l00)2
40
100
Invoking again
240
180
1
2
(100 u B) dx+ I2
on = 0
[15.4 -13]
-13
we obtain
[~:] [~oo]
=
13
= 10000
E
11846.2
and
o = 100
o = 1846.2 = 23.08
80
x> 100
2-15
EXACT
15000
-- --- ~
-::.:--~
10000
~~.-.-.
"
Sol ution 2
5000
---..I~
~X
180
100
CALCULATED DISPLACEMENTS
(J
100 -I=:::==-==_==_:=os:=_=_=,==_=_==
""
50
"I~
~
-< ,J
L._._.
-+
EXACT
SOLUTION 1
~._._
---,~--------r-------~X
100
CALCULATED STRESSES
218
SOLUTION 2
180
B.
1
for a em- variational problem
we only need continuity in the
(m-1)st derivatives of the func
tions; in this problem m = 1 .
edomains A - Band B - e are
finite elements and
WE PERFORMED A
FINITE ELEMENT
ANALYSIS.
217
FORMULATION OF THE
DISPLACEMENT-BASED
FINITE ELEMENT
METHOD
LECTURE 3
58 MINUTES
31
32
FORMULATION OF
THE DISPLACEMENT
BASED FINITE
ELEMENT METHOD
- A very general
formu lation
-Provides the basis of
almost all finite ele
ment analyses per
formed in practice
-The formulation is
really a modern appli
cation of the Ritz/
Gelerkin procedures
discussed in lecture 2
-Consider static and
dynamic conditions, but
linear analysis
33
fB
fB
= fBy
fS
fB
f~
i
FX
= fSy
Fi = Fyi
fS
i
FZ
(4.1)
uT = [u
w]
(4.2)
34
(4.3)
where
ITT = [IT If w]
(4.6 )
35
,,
x,u
""
Finite element
!!(m) (x, y, z)
"T
!!
= [U, V, W,
z)
(4.8)
U2V2W2 UNVNW N]
... Un]
.(m) (x, y, z)
=~(m) (x, y,
!.(m)
3&
=!:!.(m) (x, y,
=f(m)~(m)
+ -rI(m)
(4.9)
z)
!!
(4.'0)
(4.'1)
(4.12)
____---..ll.c=~~-----I
j-
'iTl
If
~
~
B(m) Tc(m)B(m)dv(m)j U=
v(m) l-
j [I
T
I 1
_"
-(m) T
--
L l(m)
T (m)
!!(m) 1.B
dV(m)
m
V
I
, .
( )T
B(m)TTI(m) dv(m)j
m
-___.r__.........1
"<I:::
--
_~m
El;rm) -
= B(m) l..u)
(- )(m)
y:(m)
=!!(m)
(m) T
-US
-(m) T
------.
... ~
(4.13)
37
KU
=R
(4.14)
where
R=.Ba + Rs - R1 + ~
~f
K=
( 4. 15)
B(m)Tc(m)B(m)dV(m)
(4.16)
m V(m)- -
R =
"'1.
~ lm) -
H(m)TfB(m)dV(m)
(4.17)
"'1
= "'1
-1
R
-S
HS (m)Tfs(m)dS(m) (4.18)
~ ~m) B(m)TT1(m)dV(m)
~ V(m) -
(4.19)
=-F
(4.20)
~B = ~
(m)T -B(m)
V(m) .!:!.
[1.
_ p(m).!:!.(m)~]dV(m)
MD+KU= R
38
B(m)
1.
(4.21 )
(4.22)
-B(m)
1.
(m)
- p!!
~a ~b
~a ~b
+
~a
t!t>b
-~
(4.38)
..
..
~a~+~a~=~-~b~-~b~
(4.39)
~=~a~+~b~+~a~+~b~
(4.40)
I
!
Global degrees
of freedom
;~
(restrained\
COS
T=
[
'/..
Transformed
C,-:~:e)
rl'f
a -sin
a]
sin a
U= T
\eedom
degrees of
f
cos a
IT
39
.th
1
column
For the transformation on the
total degrees of freedom we use
1.
i th row
(4.41 )
so that
1
cos a.
T=
..
Mu+Ku=R
(4.42)
}h
sin a.
where
L
Fig. 4.11. Skew boundary condition
imposed using spring element.
k U. = k b
where
k k ..
"
___ 310
(4.44)
.th
-s ina.
1
cos a.
1
Example analysis
area
=1
100
area
=9
100
80
element
Finite elements
J~
I"
-I
100
80
~I
311
1.0
I ...
--I
H(l} = [(l-L)
100
!:!.(2} = [
!!(l)=[
!!(2) = [
312
100
(1- L)
a]
v(m} = H(m}U
80
:0]
1
100
1
100
a]
1
80
1
80]
av = B(m}U
ay -
stiffness matrix
- 1
100
5.= (1 HEllO
a
o}Y
a
U
- 80
1
80
Hence
=240
[ 2.4
-2.4
-2.4
15.4
-13
313
GENERALIZED
COORDINATE FINITE
ELEMENT MODELS
LECTURE 4
57 MINUTES
41
4-2
DERIVATION OF SPECIFIC
FINITE ELEMENTS
Generalized coordinate
finite element models
~(m) =
In essence, we need
aW) =
V(m)
,-
'-
V(m)
R(m)
!!S
f
S
Convergence of
analysis results
(m) -
etc.
43
\
I
1ZI
I
\
-\
\
u(x,y), v(x,y)
are non-zero
w= 0 , E zz = 0
Fig. 4.14. (e) Plane strain condition:
long dam subjected to water pressure.
44
j(
I
I
I
I,
I
\-I
(before deformation)
(after deformation)
SHELL
Fig. 4.14. (e) Plate and shell structures.
45
Displacement
Components
Problem
u
w
Bar
Beam
Plane stress
Plane strain
Axisymmetric
Three-dimensional
Plate Bending
u, v
u, v
u,v
u,v, w
w
Strain Vector ~T
Problem
(E"...,)
Bar
[IC...,]
Beam
Plane stress
(E"..., El'l' )' "7)
Plane strain
(E..., EJ"7 )'..7)
Axisymmetric
[E..., E"77 )'''7 Eu )
Three-dimensional [E..., E"77 Eu )'''7 )'76
Plate Bending
(IC..., 1(77 1("7)
)'...,)
au
au + au
a/ )'''7 = ay
ax'
aw
aw
aoy
w
, IC..., = -dx ' IC = - OyZ,IC.., = 2
Nolallon:
E..
au
= ax' 7 =
77
0x
4&
Problem
Bar
Beam
Plane stress
Plane strain
Axisymmetric
Three-dimensional
Plate Bending
[T;u,]
[Mn ]
[Tn TJIJI T"'JI]
[Tn TJIJI T"'JI]
[Tn TJIJI T"'JI Tn]
[Tn TYJI Tn T"'JI TJI' Tu ]
[Mn
MJIJI M"'JI]
Problem
Material Matrix.
Bar
Beam
Plane Stress
E
1-1':&
E
El
1 v
v 1
[o 0 1
~.]
47
(4.47)
w(x,y)
= Y,
u (x,y,z)
= a,
48
Hence, in general
(4.50)
ex
(4.51/52)
(4.53/54)
(4.55)
Example
lp
Nodal point 6
9
Element
0
5
Y.V
Y.V
CD
V7
7
X.V
V7
X.V
(bl Finite element idealization
ZZ
Zy
ZX
=0
49
LJ2.= US
2
--II--.......- - - - - - - - -....~
= structure
element
U{X,y)] (2)
[ v{x,y)
where
uT = [U 1
410
= H(2) u
--
or
U(X,y)]
[ v(x,y)
=_~l!.
where
=[~
~}!= [1
x y xy]
and
Defining
we have
Q = Aa.
Hence
H=iPA- 1
411
Hence
fll4
(1+x ) ( Hy) :
I I
: (1 +x )( 1+y) :
and
z
t':
U3 U4 Us
U6
UJ
H'ZJ
[0
U2
I
0 : H IJ
VJ
: H ZI
(a)
U7 Us
HI.
H 16
HIs: 0
H zs : 0
Element layout
0
0
Hu :
Ha :
(b)
OJ offreedom
O
2x18
412
v.
U9 U1a
0 0: HI.
o : H ZJ H 21 : H:: H: 6 : 0 0: H..
VI -element degrees of freedom
H 17
:H II
u.
where
- au . E _ av.
_ au + av
xx - ax' yy - ay , Yxy - ay ax
Hence
where
[~
1 0 y'O
I
0
0 1
I 0
10
I
413
1
MECHANICAL IDEALIZATION
KINEMATICS, e.g. truss
plane stress
three-dimensional
Kirchhoff plate
etc.
MATERIAL, e.g. isotropic linear
elastic
Mooney-Rivlin rubber
etc.
LOADING, e.g. concentrated
centrifugal
etc.
BOUNDARY CONDITIONS, e.g. prescribed
displacements
etc.
YIELDS:
GOVERNING DIFFERENTIAL
EQUATIONS OF MOTION
e.g.
.!!!)
..!..
ax (EA ax
= - p(x)
1
FINITE ELEMENT SOLUTION
CHOICE OF ELEMENTS AND
SOLUTION PROCEDURES
YIELDS:
APPROXIMATE RESPONSE
SOLUTION OF MECHANICAL
IDEALIZATION
414
SECTION
discussing
error
ERROR
ERROR OCCURRENCE IN
DISCRETIZATION
4.2.5
NUMERICAL
INTEGRATION
IN SPACE
evaluation of finite
element matrices using
numerical integration
5.8. 1
6.5.3
EVALUATION OF
CONSTITUTIVE
RELATIONS
6.4.2
SOLUTION OF
DYNAMIC EQUILI-.
BRIUM EQUATIONS
9.2
9.4
SOLUTION OF
FINITE ELEr1ENT
EQUATIONS BY
ITERATION
8.4
8.6
9.5
10.4
ROUND-OFF
8.5
415
CONVERGENCE
Assume a compatible
element layout is used,
then we have monotonic
convergence to the
solution of the problem
governing differential
equation, provided the
elements contain:
1) all required rigid
body modes
2) all required constant
strain states
~ compatible
LW
CD
incompatible
layout
t:=
no. of elements
If an incompatible element
layout is used, then in addition
every patch of elements must
be able to represent the constant
strain states. Then we have
convergence but non-monotonic
convergence.
416
layout
"
(
1-- -
"
--
" 'r,;
"
>
......~_Q
I
I
I
Rigid body
translation
and rotation;
element must
be stress
free.
(b) Analysis to illustrate the rigid
body mode condition
Fig. 4.24. Use of plane stress element
in analysis of cantilever
417
1.
-l
I
I
I
I
\
\
-- -
_-I
---
('
\
...--,.".-~\
-------
_1
Poisson's
ratio" 0.30
--
I
I
I
01
r------,
Young's
modulus = 1.0
10
--
.J
..... .....
.....
'J
=0
I
I
f
\
\,.
I
I
Flexural mode A4
=0.57692
~-
......
'-
\...~
\
\
.----"'"="""- \
-,
--~
Flexural mode As
= 0.57692
,-----,
I
I
=0.76923
I
I
I
I
I
I
IL
=0.76923
I
.JI
418
.... .... \~
r-------- 1
Stretching mode A7
I
I
I
I
I
I
I
I
Shear mode A.
"
= 1.92308
(0
11
G)
~
17
IT
'c.
,I>
-.
/
@
.f:
20
IS
@)
b)
B
C
2
Oyy(N/m )
1066
716
359
1303
1303
419
IMPLEMENTATION or
METHODS IN
COMPUTER PROGRAMS;
EXAMPLES SAP, ADINA
LECTURE 5
56 MINUTES
51
TEXTBOOK: Appendix
52
IMPLEMENTATION OF
THE FINITE ELEMENT
METHOD
K(m) = 1.
B(m) C(m)B(m) dV(m)
V(m)-R(m) = 1.
H(m)T fB(m) dV (m)
-B
v(m) -
H(m)
B(m)
kxN
.hN
N = no. of d.o.f.
of total structure
where
K = ~ K(m) ; R = ~ R ( m)
m-B m!..!B
~ , ~B'
nxn nxl
tl
kxn
n = no. of
element d.o.f.
~
R,xn
53
Sz :: 6
t W :: 3
Z
r----y
V::2 Sy:: 5
x
/U:: 1
Fig. A.1. Possible degrees of
freedom at a nodal point.
/Sx:: 4
ID(I,J) =
_.
Degree of
freedom
54
nodal point
...
....i
l00"C
aOem
a t6
@
E lOS N/em
2 .,-0.15
t~l1
Element
E. 2 x l(Jl1 N lem2
number
II'" 0.20
10
4
8 -9
5 _ _3
<D
E = l(Jl1 N/cm 2
., - 0.15
2 E = 2 x l(Jl1 N Icm2
=0.20
t"
4 _1
8
-7
Node
Temperature at
bottom face = 70'C
Degree of
freedom
number
10
0 0
55
and then
0 0
0 0
10=
0 0
7 9 11
2 4
6 8 10 12
0 0
0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0
Also
XT = [ 0.0
0.0
YT = [ 0.0 40.0
ZT = [0.0
TT
56
= [70.0
0.0
0 . 0 60.0 60.0
80.0
0.0
0.0 40.0
0.0
0.0
0.0
40.0
80.0]
0.0
0.0
0.0
0.0]
For compacted
matrix
For !1
LMT
= [3
1 2 3 4 5 6 7 8
..,
oJ
0 0 0 0 1 2
4 0 0 0 0 1 2]
57
L MT = [5
4]
7 8]
for element 3,
T
L M = [9 10
LM
,.
mK =3
k l l k 12
" 1
k 23
k 33
k 34
kn
k 14
k 44
K=
= [11
12
9 10]
SkYline
.~
o"
o
J"
0
k 36
k 45
k 46
k ss
k S6
------m
=3
'0
0
6
'0" 0
k 66
Symmetric
A(l)
A(3)
A(91
A(2)
A(S) A(8)
A(4) A(7)
A=
A(15t
A(lO) A(13)
10
A(12) A(17)
12
A(16)
16
58
18
22
I
X 0
o 0
xix
XIX
x = NONZERO
ELEMENT
0= ZERO ELEMENT
XIX
I
SYMMETRIC
I
1
0 0 10 0 0
0 0:0 0:0
x 010 0 x
0 010 0 0
0 0 X 0 0
0 X 10 0 0
xxlxXIO
xix
XiX
IX
X X
XIX
IX
ELEMENTS IN ORIGINAL STIFFNESS MATRIX
Fig. 10. Typical element pattern in
a stiffness matrix using block storage.
X0
X0
XIX
XiX
XiX
BLOCK 1
BLOCK 2
~---~
~_BLOCK
59
,~
14
~
21
24
23
25
10
26
27
11
18
17
16
15
22
28
29
12
20
19
30
13
31
32
33
26
29
31
11
16
19
17
12
14
21
24
22
27
32
~,
, 3
10
13
15
18
20
23
5 283033
2
510
START
READ NEXT DATA CASE
END
5-11
-'-------4
ONE - DIMENSIONAL
ELEMENT
RING ELEMENT
; - - . - - - - - - - - - - - - -... y
Fig. 12. Truss element
p. A.42.
y
Fig. 13. Two-dimensional plane
stress, plane strain and axisymmetric
elements.
p .. A.43.
5-12
---5
y
Fig. 15. Three-dimensional beam
element
p A.45.
513
-- --. __e_
--
---L~---
3-16 NODES
TRANSITION
ELEMENT
514
FOBMULATION AND
CALCULATION OF
ISOPABAMETBIC
MODELS
LECTURE 6
57 MINUTES
61
62
FORMULATION AND
CALCULATION OF ISO
PARAMETRIC FINITE
ELEMENTS
interpolation matrices
and element matrices
isoparametric
elements
lsoparametric Elements
Basic Concept: (Continuum Elements)
Interpolate Geometry
N
x=L
i=l
h. x. ; y=
I
L
i
=1
h. y. ;
I
z=L
i=l
h. z.
I
Interpolate Displacements
N
u=
1:
=1
h. u.
I
v=
L
i == 1
h.v.
I
w=
L
i
=1
h.w.
I
N = number of nodes
&3
1/0 Element
Truss
2/0 Elements
Plane stress
Continuum
Plane strain
Elements
Axisymmetric Analysis
3/0 Elements
Three-dimensional
Thick Shell
64
~~==-=l======~I=-==r======r ~1
==
65
S
Ill(
ll(
r
1
1 - D Element
2 Nodes:
-11.0
~~
-+-r
-r
.. _
1
h1 = %(1 + r)
-- - -
1.0
-...
e_----......:::::...::::=-2
2 - 0 Element
4 Nodes:
/-r-r----+~~-r
h1
=~(1 + r)(1 + 5)
Similarly
h =%(1- r)( 1 + 5)
2
h3 =%(1- r)(1- 5)
h4 = %(1 + r)(1-s)
6-7
S:
.....
....--+-+--+~_
-+--+-I--I--I---I--------I--.._r
2 :
tfF--.-~-_-_-~-r~~_-
-----..1L...4.
!1.0
= %(1 + r)(1 + s)
-%h S
Sim.
h = %(1- r)(1 + s)
2
-%h S
68
\
\
---;q-----
s=o
---..
8
r
\
\
3
\
\
r =-1
=0
r = +1
i =5
i =6
i = 7
i = 8
h,
~(l+r)(l+s)
-~hs
h2
~(l-r)(l+s)
-~hs
h3
~(l-r) (1 -s)
h.
-~hq
~(1 - r 2 ) (1 +s)
-ihq
hs =
1 .- ~hs
I::
-;h
-~ h<j
... -~h6
-~ her
-~h7
-ih q
'h 6
~ (1 - S2) (1 - r)
-1 h<j
h7
~(1 - r 2 ) (1 -s)
-th"
hs
i (1 -
-th<f
h~::
s2) (1 + r)
69
x==r
y == s
z == t
610
ah
ah
ar
E
E
SS
[Y
rs
ah
as
u
1
ar
ah
ah
as
ah
3h
as
ah
as ar
at'
\..
v1
u
v4
vB
We note again
r==x
s=y
GENERAL ELEMENTS
r = +1
s = +1
Y,v
r - - - t - - - 4 _r
611
[ : ] = [::
as
as
:]l~]
as
ay
Aside:
cannot use
a a ar
ax ar ax + ...
or
---
-aar = -J
a
ax
(in general)
a _ 1 a
a-x- J- ar
(5.25)
.!!.
612
= det J
dr ds dt
Element 1
z.._----+----......- " r
3,
'4"""1--1-----------t..~1
6 em.
+-
...
Element 2
2.
+--_1<
1+
.....------....
.;
cDG-W\
'0'
I=>
J
1
2
=
213
&-13
Element 3
\c.1V\
2.
1c.W'I
-+---~~- -,c
: '3,.~
't'
.....,..
.I.
...L... -"'------1431
I
l"l1
-,
(1
+5)]
(3+r)
r=-I
Natural space
,-.
-I'
L/4
614
x=L:
h . x. 9
1 1
L
x =-4(1
+r ) 2
i =1
hence
-J
[!:..2 + !'-2 LJ
and
or
Since
r =
2.Jf- 1
We note
/x
singularity at X = 0 !
615
Numerical Integration
Gauss Integration
Newton-Cotes Formulas
' " ak
F k
-K = !:J
IJ -IJ
I,J,k
x
x
616
FORMULATION OF
STRUCTURAL
ELEMENTS
LECTURE 7
52 MINUTES
71
72
FORMULATION OF
STRUCTURAL
ELEMENTS
Strength of Materials
Approach
straight beam
elements
use beam theory
including shear
effects
isoparametric
approach for
interpolations
plate elements
use plate theory
including shear
effects
(ReissnerIMindlin)
Continuum
Approach
Use the general
principle of virtlial
displacements, but
e.g.
shell
73
Neutral
axis
Beam
section
..
Deformation of cross-section
wi
x-0
wi
x+0
dw
dx
_ dw
-0 - dx
+0
./
Neutral
axis
WI
- Wi
x- O
x+ O
Beam
section
Deformation of cross-section
74
dw
S=--y
dx
(5.48)
(5.49)
_ (L
pw dx
-Lo
m S dx
(5.50)
+ GAkJ
-i
(~~ - S) o(~~ - S)
-io
dx
oW dx
m oS dx = 0
(5.51)
75
A = ab
I = ab
k = ..
6 '
,
12
76
q
W
= ~ h.w.
L..J
i
=,
1 1
~
L..J
i
w=1
H- /U'
-'
h.e.
=,
(5.52)
1 1
B = .:...:.sH U
(5.53)
dW
dX
BU'
1-/ - '
~ =B U
~-
dX
Where
Q. = [w,
~ = [h,
hq 0
~ = [0
0 h,
and
!!w = J-
8,
8qJ
OJ
hqJ
(5.54)
_ _, f,LO... a dr'
dh,
dh q ]
... ar (5.55)
~- J
77
So that
K = E1
1 T
~ ~
det J dr
-1
+ GAk
(~-tla) (~-~)det J
-1
dr
(5.56)
and
R=
det J dr
-1
+/
~ m det J dr
(5.57)
-1
ex. =
GAk
IT
Hence
- use parabolic (or higher-order)
elements
. discrete Kirchhoff theory
- reduced numerical integration
7-8
(5.60)
Here we use
Q,y(r,s,t)
+~ ' b h Q,V k
2 L.- k k sx
k=l
q
k=l
k=l
(5.61)
k=l
q
Q,z(r,s,t) =
L
k=l
hk Q,Zk +
iL
a k hk
Q,V~Z
k=l
q
+ ~2 'LJ
"
k
bkhk V sz
k=l
79
So that
u (r,s,t) =
0
x- x
v (r,s,t) = ly _ 0y
w (r,s,t) =
(5.62)
1z- 0z
and
u(r,s, t) =
k=l
t .E
bkh k
V~x
k=l
q
t q
v(r,s,t)=L: hkv k +2
k=l
k=l
L
q
+tL:
k=l
q
w(r,s,t)=L:
k=l
(5.63)
710
vk = ~
e
...:..s
ak
(5.65)
'is
where
exk
eyk
e =
(5.66)
ezk
nn
Yni; =
~!4~
(5.67)
k=l
Ynl;;
where
T=
[Uk vk wk exk eyk ezk ]
(5.68)
T
nn
Tn~
TnI';;
nn
Gk a
Yn~
Gk
Ynl;;
(5.77)
7-11
.... -- ----
and w=w(x,y)
(5.78)
Hence
dl\
dX
E
XX
E
yy
dS
_-.1.
dS X
Yxy
dy
dS y
dX
dy - Sy
(5.80)
712
dW
Yyz
Yzx
(5.79)
dy
dW+ S
dX
x
1 v
LXX
yy
z_E_ v
2
a
a
l-v
a a
xy
l-v
2
(5.81)
aw
ay - By
yz
E
= 2(1+v)
(5.82)
aw + B
ax x
ZX
1
2
dz dA
II=L
xy
+~
2 A -h/2
dx dA
-fw P dA
A
(5.83)
713
IT =
iT
.<q,
.<dA +
t // f,;
y dA
-I:
P dA
(5.84)
where
as
_ .-J.ay
; y
aw + s
ax x
as x _ ~
ax
ay
C
~
=.
(5.86)
Eh 3
12(l-v 2 )
1-v
f.s
714
Ehk
= 2{1+v)
[ 0
(5.87)
-fw
p dA
=0
(5.88)
w=~h.w.
LJ
1 1
i=l
S
y
=~
LJ h.1 exi
(5.89)
i=l
and
q
x
= LJ
~h.x.
1 1
i=l
Y=~h.y.
LJ 1 1
;=1
715
s
Mid-surface
\....-~-----t~
(5.90)
716
(5.91)
where
(5.92)
To express
Y~
in terms of
k
V
-1
(5.93a)
then
V
..:...n
= - Vk
~
k
O',k + V
-1
Sk
(5.94)
717
(5.100)
l = ~h ~
~h=~h
Jl
1-v
-2-
1-v
-2-
1_~2
!2sh
1-v
2
symmetric
(5.101)
16 node parent element with cubic interpolation
I-
-I
64>-[>
000
o \'.' .\
Variable - number - nodes shell element
718
a) Shell intersections
719
NUMERICAL
INTEGRATIONS,
MODELING
CONSIDERATIONS
LECTURE 8
47 MINUTES
81
82
NUMERICAL INTEGRATION.
SOME MODELING CONSIDERATIONS
We had
K = f BT C B dV
V - --
(4.29)
M = J p HT H dV
V --
(4.30)
R=
f HT fB dV
V- -
-s = Sf
R
(4.31 )
HS fS dS
-
Rr= f ~T !.r
V
dV
(4.32)
(4.. 33)
83
= det.4 dr ds dt
dV
+1 +1
!$ =
f f~
~ ~
det A dr ds
-1 -1
+1 +1
ff
M=
tl T tt det J dr ds
-1 -1
etc...
84
-K=L~a..F
. 4J lJ -lJ..
1 J
where
i, j
a. .. = weight coefficients
IJ
F
-IJ
= -IJ
B T -C -IJ
B detJ
~J
r = O.577
= O.577
r = O.775
r= 0
5 = 0.775
5=0
85
L.--
---.~
I
I
.-8
a+b
-2-
actual function F
2 nd order interpolating
~~~~polynomial in x .
a+b
2
etc....
J{a F(r)dr=(b-a)~C.nF.+R
LJ
n
;=0
1
(5.123)
n = number of intervals
Ci n = Newton - Cotes constants
interpolating polynomial is of
order n.
87
Number of
Intervals n
q
1
"2
2
3
4
5
6
Cn2
C5
Cn6
4
6"
3
"8
"8
32
90
75
288
216
840
90
10-I(b-a}lF"(r)
1
6"
1
19
288
41
840
cn3
Upper Bound on
Error R. as
a Function of
the Derivative of F
1
6"
3
10-3(b-a)5PV(r)
1
"8
"8
12
90
50
32
US
27
840
90
50
288
272
840
1O-3(b-a)5F'V(r)
7
90
75
ill
27
840
10-6(b-a)7FVI(r)
19
288
216
840
10-6(b-a)7Fv'(r)
41
840
lO-'(b-a)'FVIU(r)
(5.124)
88
rj
1
2
3
O. (I5 zeros)
0.57735 02691 89626
0.77459 66692 41483
0.0oo
4
5
0.86113
0.33998
0.90617
0.53846
0.0oo
0oo
/X,
0oo
63115 94053
10435 84856
98459 38664
93101 05683
0oo
0oo
2.
(I5 zeros)
1.0oo
0oo
0oo
0.55555
0.88888
0.34785
0.65214
0.23692
0.47862
0.56888
0.17132
0.36076
0.46791
55555
88888
48451
51548
68850
86704
88888
44923
15730
39345
55556
88889
37454
62546
56189
99366
88889
79170
48139
72691
Now let,
ri be a sampling point and
eli be the corresponding weight
for the interval -1 to +1.
Then the actual sampling
point and weight for the
interval a to bare
a + b + b - a r. and b - a el.
-22
1
2 I
and the ri and eli can be
tabulated as in Table 5.2.
89
we use
+1
-1
+1
I:
F(r,s) dr ds =
-1
"1
+1
F(ri's) ds
-1
(5.131)
or
f f
+1
-1
+1
F(r,s)drds=
-1
I:
,,;,,/(ri'sj)
i ,j
(5.132 )
ff1
+1
-1
+1
-1
+1
F(r,s,t}drdsdt
-1
= LJ
~a.a.a.kF(r.,s.,tk)
1 J
1 J
i,j,k
(5.133 )
and ak = a. Q. Q .
IJ
I J k
810
811
Demonstrative example
Stress calculations
(5.136)
stresses can be calculated at
any point of the element.
stresses are, in general, discon
tinuous across element
boundaries.
8-12
thickness = 1 cm
A
1~[
I
:...
e.
<3>
-p
CD
..,-
1>
3 Coft'1.
2
= 3xl0 7 N/cm
\) = 0.3
3c.m.
= 300 N
-of
'100 N!Crrt'l.
/
8 ...
a.
813
'?
,~
C.
s_
a~
4l
0.3
"?
lOON
"
"
&
Co
<D
174+
/lA/e-t'-
<D
Co
<D
Co
~I ".00 "'Ie-."
8-14
TYPE OF PROBLEM
ELEMENT
TRUSS OR CABLE
2-node
TWO-DIMENSIONAL
PLANE STRESS
PLANE STRAIN
AXISYMMETRIC
8-node or
9-node
THREE-DIMENSIONAL
20-node
3-D BEAM
3-node or
4-node
PLATE
9-node
SHELL
9-node or
16-node
D
D
-=
.....
-/
L7
~~
815
I
1
S node
4/'1ode
elEJmerrt
e 1er1l(1'It.
.
i
g I'\oole
~
~kl'7ll'"t
119
U.s
8
4- I\oc:(e
VA
ele,"~"t.
4 node
e Iem tnt""
Constraint
equations:
vA = (vC + vB)/2
/.
!
816
U,
uA = (uC + uB)/2
c)
'Ve-
4- node
el .... ~I\"t
llA
SOLUTION OF
FINITE ELEMENT
EQUILIBRIUM
EQUATIONS
IN STATIC ANALYSIS
LECTURE 9
60 MINUTES
91
LECTURE 9
92
SOLUTION OF
EQUILIBRIUM
EQUATIONS IN
STATIC ANALYSIS
- static condensation
- substructuring
- frontal solution
- .L Q. .bT factorization
- Cholesky decomposition
- Crout
- column reduction
(skyline) solver
Iterative methods,
e.g. Gauss-8eidel
Direet methods
these are basically
variations of
Gauss elimination
-4
-4
,
0
-4
,
-4
U,
U2
,
(8.2)
-4
U3
-4
U4
93
-4
olI l!
5
-5
16
29
-4
-4
r-----------I
I
o I_~
:
(8.3)
o:
-4
5-5
14
16
r-------~ _20
0:
0: _ 20
I
7
65
14
I
I
94
o
=
(8.4)
STEP 3:
-4
14
U1
-s
U2
15
20
U4
"6
16
7
.- 8 (8.5)
U3
"7
-T
r--I
I
I
I
"6
u4 ,
12
=5
1 - (- 156) U3 - (1) U4
U =--------:;-;;----2
14
_ 13
-S
(8.6)
5
19
36
7
(-4) 35 - (1)15 - ( 0 )"5 _ 8
U1 =----~---5
- "5
o-
95
~a
~-ac]
[!!a] [ Ba]
[
.!Sea ~-e c !!c = Be
(8.28)
Hence
and
-1)
---------
aa
~a - ~ilC
.!Sec.!Sea
!!a = Ba - ~c .!Sec-1 ~
Example
tee
I
I
I
r:~a
-4
U1
-4
-4
U2
-4
-4
U3
-4
U4
---+------------
'---y----'
~c
1
so that
14
5
16
-5
16
-5
29
5
-4
-4
~a
-4
-a. a
,-
-4
[1/5]
[-4
0]
~
Kaa
= -4
-4
-4
'--
98
1
0
1....-
-4
-4
14
-!
5
6 -4
-4
-5
-4
U3
U4
:1
:1
U2
-!
29
VI
U2
6 -4
"5
-4
-4
U3
V4
Fig. 8.1
Physical systems
considered in the Gauss elimination
solution of the simply supported beam.
9-7
......--.-?-?
-~-o--o
-6
l
e--c>---n
50x50
32x32
Example
~~
6L
98
17
-20
-20
48
-28
EA, [ '7
3
6"L
-20
EA,
6L
Ir
3] -[-20] [lJ[-20
7
3
25
-28
48
or
ll.
9
EA, [ 1
L
-1
and
99
Multi-level Substructuring
I' 'I'
L
2A
~I ,
L
4A,
,I.
.1
-\&-o=2:E~f'
'n-~
u.
\
Ur;,
U2
U6
Us
U7
Ug
Rs
U3
I 16A,
'~
SA,
-I
U,
u2
---I
U,
1-
u3
.-u
---I
U,
1-
u3
_I
U,
Us
I
1-
Us
_I
U,
-.
1-
1-
Us.Rr;,
U,
'-10
ug
ug
Elementq
Element q + 1
Elementq + 2
Elementq + 3
-------
~I
m+3
Element 1
N:"
Element 4
4
Wave front
for node 1
Wave front
for node 2
911
L- 1
K
--1 -K = -1
Example:
-4
-4
a
-4
-5
-4
a a a
-4
-4
-4
~4
a _16
5
a
16
5
29
-4
5
-4
We note
L-1-1
1
5
912
-5
5
~1
a
a
S-
1.2-1 1.1-1
S :=
x x x x
x x x
x .......
x
x
K:= S
x
x
x
x
upper
triangular
matrix
Hence
or
Also, because
is symmetric
where
0:=
d .. := s ..
11
11
913
where
L D~
SOLUTION OF EQUATIONS
Using
T
K= L 0 L
(8.16)
we have
L V= R
o LT
(8.17)
V
(8.18)
where
V := L-I
-n-l
(8.19)
and
(8.20)
914
-4
-4
-4
5
4
5
14
5
14
-4
6
-5
-4
5
-4
-4
5
5
-5
14
1
5
8
7
15
T
-4
5
4
-5
14
1
5
8
-7
15
T
-4
5
915
COLUMN HEIGHTS
o 0
o 0
'-----,
SYMMETRIC
000
000
X 000 X
o 0 000
o 0 x 0 0
o X 000
X X X X 0
X X X X
X XX
X
X
ELEMENTS IN ORIGINAL STIFFNESS MATRIX
Typical element pattern in
a stiffness matrix
SKYLINE
o
o
L...-_
X
X
X
X
0
0
0
0
0
X
000
000
0 0 X
0 0 X
X 0 X
X 0 X
X X X X X
X X X X
X X X
X X
X
ELEMENTS IN DECOMPOSED STIFFNESS MATRIX
Typical element pattern in
a stiffness matrix
9-16
ELEMENT
0= ZERO ELEMENT
COLUMN HEIGHTS
-x
o
0 0
0 0
xix x
XlX 0
xIx
0
x 0
SYMMETRIC
I
0 10
0:0
010
010
0 x
X\O
0:0
010
0 x
0 0
0 0
0 0
x xix XIO
xix xix
Ix XlX
xIx
Ix
917
SOLUTION OF
FINITE ELEMENT
EQUILIBRIUM
EQUATIONS
IN DYNAMIC ANALYSIS
LECTURE 10
56 MINUTES
101
102
DIRECT INTEGRATION
SOLUTION OF EQUILIBRIUM
EQUATIONS IN DYNAMIC
ANALYSIS
MU+CU+KU=R
-- -- --
explicit, implicit
integration
selection of solution
time step (b. t)
computational
considerations
some modeling
considerations
MU + C U + K U = R
(9.1)
or
103
time
--
time
to = _l_(_t-tlt u + t+tlt U)
2tlt
-
104
(9.4)
~t
-(-'-2 -M_-'c)t-~tu2~t ~t
(9.6)
where we note
! t!! =(~!(mT!!
=
~ (l5-(m) t lL)
~t(m)
Computational considerations
to start the solution. use
(9.7)
in practice. mostly used with
lumped mass matrix and low-order
elements.
105
l'It er =
Tn
TI ; Tn
smallest natural
period in the system
- e
e=~
106
(9.28)
{9.29J
an implicit integration scheme solution
is obtained using
a.
= la
=~
which is the
constant-average-acceleration
method
(Newmark's method)
~n m2 + 2 n mt
107
Accuracy considerations
time step !'1 t is chosen based
on accuracy considerations only
Consider the equations
~1U+KU=R
and
where
2
K . :: w1
--1
~1 <p.
--1
Using
"1
0. 2
where
..
108
= ~.
~1-
i=l, ... ,n
..
i=l, ... ,n
x1 + w.1 x1
with
n
U=
~<I>.x.
~-l
i =1
..
x+w x=r
..
x+
a
ox=
x=
-w
109
19.0
Houbolt
method
15.0
..
15.0
11.0
11.0
le
....
5..
E!:.
C
.g
'"CC/I
le
7.0
5.0
"0
~
>
"iii
"8
.;:
'"Co
'"
:!
c
l:!
'"
'"
'"
7.0
5.0
'"
"0
~
.~
C/I
Q.
E 3.0
3.0
'"8.
tf
'.01~4t
~
1.0
:!
c
'"
l:!
'"
1.~
Q"
1.0
PE
0.06
0.10
0.14
0.18
0.06
0.10
0.14
0.18
4t-----r----:--r--r----r-----,...-----,-----,
equation
31----+--f-+-+----t-----t-----.,t----'-1
...
o
'0
static
response
"t:J
CtI
CJ
'ECtI
c::
>
1010
7T
D:.r 1.05
nYNAMIC RESPONSE
_ .. - STATIC RESPONSE
~ =0.05
gi
r.it
z~~.:.::::7'--
!2C 1\
'
'"- .j.,
fs!
,;1
1
!
T
74- ._--+-- -t-" - -- .... __..--t-
81
'c.oe
o.."~
I I ,.,~.
fl. 'JO
n. 7~
I. 00
DLF .. 0.50
DYNAMIC RESPONSE
--- STATIc.: RESPONSE
.f... = 3.0
w
.... ,
---- ---~
/"7--_____ . .
/'
./
-----=~---':....;-,,---=
__
==_7'~--....
_.~~.:.--==---/-
g,
::i- +-~--+---+---
c.':;:
C.25
"
~.I)C
.. -------t-----+---+I- - t - _ _---+--+1~--+--+I::-:---+----,+1:::---+----,+1::-:---+------::+-'::-:---+-----:<'
:."L
:.00
: . .?~
I.SO
1.75
2.00
2.25
2.50
2.75
3.00
TIllE
1011
(9.100)
(9.101 )
c /':,t
1012
(9. 102)
a = , / (aAt 2 )
O
a = 0/ ex.- ,
4
as = -a 3
a,=O/(aAt)
a3 = , / (2ex. )- ,
as = I1t(O/ex.- 2)/2
a 7 = -a 2
a g = I1t(' - 0)
a, = '/2I1t
...
0
O 0
2. Inltlahze !!., !!., !!. ;
For central difference method
only, calculate I1t u from
initial conditions: -
M = a~ +
a,f.
1013
increments;
in implicit time integration:
1014
Newmark Method:
1015
MODE SUPERPOSITION
ANALYSIS; TIME
BISTORY
LECTURE 11
48 MINUTES
111
112
!L = 1:. !(t)
nxl nxn nxl
P
transformation matrix
Using
!L(t)
1:. !(t)
(9.30)
on
MU +
c 0 + KU= R
(9.1)
we obtain
~ R(t) +
i(t) +
R!(t)
~(t)
(9.31)
where
C fT ~ f ;
R = PT R
(9.32)
113
(9.34)
=0
Using
(9.36)
M'" "
-<P 1" _.:t:..
J 1== 0'
i =j
i ., j
(9.37)
< W
114
(9.38)
Defining
(9.39)
we can write
(9.40)
and have
T M
(9.41)
Now using
!L(t)
(9.42)
~Jt)
!(t) +
!T ~!
!(t) +
r;i ~(t)
!T !S.(t)
(9.43)
The initial conditions on ~(t) are
obtained using (9.42) and the
M - orthonormality of ; i.e.,
at time 0 we have
(9.44)
115
(9.45)
i.e., n individual equations of
the form
.x 1.(t)
+ w.2 x. (t) = r. (t )
1 1
1
i
where
= ',2, ... ,n
(9.46)
with
X'I1 t=O =
X'I1
T a
lj). -M
-1
(9.47)
.T
t=O
-U
=-'-- -1
cp.M
-
O'
-U
(9.48)
+ a..1
sin1
w.t 1
+ 8. cos wt
1
where a.i and
8i
are determined
(9.49)
11-&
x + 2E;,wx. + W2X = S 1. n P t
-+-_ _--.,
....
0
.....
u
static
response
.....
CtI
-0
CtI
CtI
r:::::
>-
~= \-0
Hence we use
uP
--
~ . x1 (t)
~--l
i =1
where
uP - U
(9.50)
117
~_=LriUl~)
i =1
Hence
r.1
T R
-1-
Then
and
K flU
fiR
!(t) + !T f!i(t) +
fi !(t)
!T ~(t)
(9.43)
T
. C (po = 2w. E;,. cS. .
1 1 1J
(9.51)
-1 ---J
and we have
w~1 x.(t)
1
= r
(t)
i=l, ... ,n
(9.52)
118
(9.56)
~1
--
+BK
(9.55)
example:
Assume
~, =
0.02
w, = 2
calculate a and
We use
2w.1
~.
or
'/
a + Bw:-
2w.1
~.
1
119
a + 4ii = 0.08
a + 913 = 0.60
-0.336 M + 0.104 K
2
a + 13 w. = 2w. [, .
1
2
a + SW.
1
E,. =
2w.
a + -13 w
2w.
2 i
= -
1110
x.1
+ 2w. E,. x +
w~
x. = r.
111111
with
r1
TO
xi I t = 0 "--. !i!i .!:L
ITO'
xi t = 0 = !i f1 .!:L
and then
uP
P
~. x. (t)
LJ-1
i =1
Practical considerations
mode superposition analysis
is effective
- when the response lies in a
few modes only, P n
- when the response is to be
obtained over many time in
tervals (or the modal response
can be obtained in closed form).
e.g. earthquake engineering
vibration excitation
- it may be important to
calculate E p (t) or the
static correction.
1111
SOLUTION METHODS
FOR CALCULATIONS
OF FREQUENCIES
AND MODE SBAPES
LECTURE 12
58 MINUTES
121
122
r! = \ !
nxn
Generalized EVP:
!sP.-=\!i!
(\=w
Quadratic EVP:
n> 500
m> 60
r sP.- = w2 !i!
If zero freq. are present we can
use the following procedure
sP.-
+ )1
or
(r+)1
!i)sP.- = \ !i sP.
or
\ = w2 + )1
=\-)1
123
p(A)
p(A) = det(K - A ~)
In buckling analysis
.!$.!=A~!
where
p(A) = det
p(A)
124
(~ -
A ~)
..
K:
.. --
(~ - ~ ~)! = n
.K 2.
.Ki=;\!ii
M=I::I::T
i=hTjJ
hence
:t =
;\
i ; K = 1::- 1
K [- T
or
T
M= W02 W
etc ...
125
1.3
eigenpairs ( Ai'
are required
. ..
AM!
1n
1. i)
i=l,
or i=r,
~P_=A~~
! ~+l = M ~
~+l
Forward Iteration
Rayleigh Quotient Iteration
can be employed to cal
culate one eigenvalue
and vector, deflate then
to calculate additional
eigenpair
Convergence to " an eigenpair",
which one is not guaranteed
(convergence may also be slow)
126
,p
,s
~ = A ~ ~ ~ (K - A M)
Hence
p(A)
,,
Newton Iteration
p(A)
(A-An)
= det L D LT = II d ..
-- -
II
Pi+l = Pi
12-7
]J.1- 1
p (A) / (A-A,)
I
I
III
Convergence guaranteed to A1 ' then
A2 , etc. but can be slow when we
calculate multiple roots.
Care need be taken in L D LT factor
ization.
128
3rd associated
constraint problem
2nd associated
constraint problem
1st associated
constraint problem
<p =
A!11
:::}
.
9
~
..
3 4
. . ..
; ;
.
.. -..... .
... .
129
Calculate
~ - ].lS. ~ = h
Qh
,,
"
,/
].ls
].lS2
Tf
..
Need to take care in L D L aetonzatlon
---
4)
TRANSFORMATION METHODS
<PTK<P=A
~!=A~!--T--
<P
Construct
<P
iteratively:
<P = [~,... ~ J;
1210
-M<P
_H
n
= [Al ...
'n]
T
T T
~ ... ~ ~l f f1 ~ ... ~-~
T
... ~ f 1 !i f 1
~ ... ~-l
1211
p(A)
/
A
= det L D LT = n d ..
-;
11
11;+1 = ].1; -
n P(l1;) - P(11;_1)
11;-11;_1
n is normally = 1.0
1212
....
~.
....
....
....
lJi+ 1
k =1,2, ...
~+l
= - ~+l
T ~
(~+l !i ~+l)
- T
(~+l)
~+l !i ~k
=- T
~+l !i ~+l
1213
Advantage:
Calculates only eigenpairs actually
required; no prior transformation
of eigenproblem
Disadvantage:
Many triangular factorizations
Effective only for small banded systems
We need an algorithm with less
factorizations and more vector iterations
when the bandwidth of the system is large.
k=1,2, ...
-T
~+1 = 4+1
-T
~+1 = 4+1
~1
-~+1
4+1
1214
~+1 ~+1
CONDITION:
eigenvalue
p eigenvalues
!5. - flS t1
flS
T
~ Q~
<
tal
1215
Starting Vectors
Two choices
1)
x.
~l
e.,
random vector
2) Lanczos method
Here we need to use q much
larger than p.
Checks on eigenpairs
2.
E:.=
[I
~9,+l) II
-1
1216
MIT OpenCourseWare
http://ocw.mit.edu
The following may not correspond to a particular course on MIT OpenCourseWare, but has been
provided by the author as an individual learning resource.
For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.