Finite element method for structural dynamic

and stability analyses

Module-9

Structural stability analysis

Lecture-33 Dynamic analysis of stability and analysis of time

varying systems

Prof C S Manohar
Department of Civil Engineering
IISc, Bangalore 560 012 India

1
 Dynamic analysis of a beam column

f  x, t 
P P

EI , m, l , c 
n

n 2 2 EI 
l2 m

 P
Pncr
2
 Parametrically excited systems

f  x, t 
P t  P t 



u t 
y
ms a, v
d kv cv
m  g  yg  t  d
h  A mu

y

yg  t 
EI , m, c, l

y  x, t 
x

xg  t 

3
 P P
P
P

“Follower”
forces

Line of action of P Line of action of P

remains unaltered as beam remains tangential to the deformed
deforms. Static analysis can beam axis. Static analysis does not lead
be used to find Pcr to correct value of Pcr .
Problem 1
How to characterize resonances in systems governed by equations of the form
M  t  X  C  t  X  K  t  X  0; X  0   X 0 ; X  0   X 0
when the parametric excitations are periodic.
Problem 2
How to arrive at FE models for PDE-s with time varying coefficients?
Problem 3
Are there any situations in statically loaded systems, wherein one needs
to use dynamic analysis to infer stability conditions?

5
Qualitative analysis of parametrically excited systems
u  t   p1  t  u  t   p2  t  u  t   0
u  0   u0 ; u  0   u 0
pi  t  T   pi  t  , i  1, 2
The governing equation is a linear second order ODE with
time varying coefficients. It admits two fundamental solutions.
u  t   p1  t  u  t   p2  t  u  t   0
Let u1  t  and u2  t  be the fundamental solutions of this equation.
 u  t   c1u1  t   c2u2  t 
Consider the governing equation at t  T
u  t  T   p1  t  T  u  t  T   p2  t  T  u  t  T   0
Since pi  t  T   pi  t  , i  1, 2, we get
u  t  T   p1  t  u  t  T   p2  t  u  t  T   0
 If u  t  is a solution  u  t  T  is also a solution.

u1  t  T   a11u1  t   a12u2  t 
u2  t  T   a21u1  t   a22u2  t 
 u  t  T    Au  t 
We are interested in nature of the solution as t  . 7
u  t  T    Au  t 
lim u  t   ?
t 

This is equivalent to asking lim u  t  nT   ?

n 

u  t  T   Au  t 
u  t  2T   Au  t  T   A2u  t 

u  t  nT   Au  t   n  1 T   Anu  t 
 The behavior of lim u  t  nT  is controlled by the
n 

behavior of lim An .
n 

Intutively, one can see that this, in turn, depends upon the
nature of eigenvalues of A.
8
u  t  T   Au  t 
Introduce the transformation u  t   Qv  t 
u  t  T   Au  t   Qv  t  T   AQv  t 
Pre-multiply by Q t
Q t Qv  t  T   Q t AQv  t 
Select Q such that A is diagonalized.
That is, we wish to find Q such that Q t Q & Q t AQ are diagonal.
Consider the eigenvalue problem: A  
Select Q to be the matrix of eigenvectors of A.
Q   1  2 
1 0 
 v t  T     v t 
 0 2 
9
 vi  t  T   i vi  t  , i  1, 2
 vi  t  nT   in vi  t  , i  1, 2
We have
 lim vi  t   lim vi  t  nT   0 if i  1, i  1, 2
t  n 

 lim vi  t   lim vi  t  nT    if i  1, i  1, 2
t  n 

vi  t  is periodic with period T if i  1, i  1, 2

vi  t  is periodic with period 2T if i  1, i  1, 2

10
Reduction to normal form
Consider vi  t  T   i vi  t 
Multiply by exp   i  t  T  
 exp   i  t  T   vi  t  T   i exp   i  t  T   vi  t 
1
We relate i & i as i  exp   iT    i  log e i
T
exp   i  t  T   vi  t  T   exp   i t  vi  t 
  i  t   exp   i t  vi  t  is a periodic function with period T , for i  1, 2.
This leads to
vi  t   exp  i t  i  t  , i  1, 2
vi  t   exp  i t  i  t 
Periodic
function

11
 vi  t   exp  i t   i t 
Could be Periodic
 periodic function
 aperiodic with decay
 aperiodic with explosion

Remarks
1
 i  log e i , i  1, 2 are called the characteristic exponents or
T
the Floquet coefficients.
Any solution u  t  can be expressed as u  t   a1v1  t   a2v2  t 
Behavior of u  t  as t   is governed by the nature of  i , i  1, 2
 i   i  j i ,i=1,2 j 1 
Growth or Oscillatory
Decay behavior

u  t  is periodic if  =0, that is when  i , i  1, 2 are pure imaginary.

i  1  u  t  is periodic with period T
i  1  u  t  is periodic with period 2T 12
Determination of the characteristic exponents
u  t   p1  t  u  t   p2  t  u  t   0
Let u1  t  & u2  t  be two solutions of this equation with
u1  0   1, u1  0   0 and
u2  0   0, u2  0   1
We have
u1  t  T   a11u1  t   a12u2  t 
u2  t  T   a21u1  t   a22u2  t 
u1  0   1  u1 T   a11u1  0   a12u2  0   a11
u1  0   0  u1 T   a11u1  0   a12u2  0   a12
u2  0   0  u2 T   a21u1  0   a22u2  0   a21
u2  0   1  u2 T   a21u1  0   a22u2  0   a22
 u1 T  u1 T  
A 
u
 2  T  u 2  T   13
  u1 T  u1 T  
A 
 2   2  
u T u T
Find eigenvalues of A
Infer nature of solutions by using the criteria
 lim vi  t   lim vi  t  nT   0 if i  1, i  1 and 2
t  n 

 lim vi  t   lim vi  t  nT    if i  1, i  1, or 2
t  n 

vi  t  is periodic with period T if i  1, i  1, 2

vi  t  is periodic with period 2T if i  1, i  1, 2

14
If the condition i  1 occurs, we say that the system has got into
parametric resonance.
Here the motion grows exponentially with time.
Presence of damping does not limit the amplitude of oscillations.
Amplitudes could get limited due to nonlinear effects.
This is contrast with resonance in externally driven systems:
P
2 
x   x  P cos t ; x  0   0; x  0   0  x  t   2
2
cos t  cos t 
 
P Pt
lim x  t   lim 2
     2
cos t  cos t   sin t
  2
 lim lim x  t   
  t 

Resonance response amplitudes are limited by damping.

Nonlinearity would also become important as response grows.
15
Extension to MDOF systems
Consider n-dof system
U  t    P  t   U  t   Q  t   U  t   0
with
 P  t  T     P  t   nn
Q  t  T    Q  t   nn

U  t  T    AU  t 
Recipe
Generate a set of n independent solutions of the governing equation
by using a set of n linearly independent ics.
Form the A matrix
Find eigenvalues of matrix A.
Infer the nature of the solution by examining the nature of the eigenvalues.
16
“Follower” forces

P
P

Line of action of P remains tangential to

the deformed beam axis.
Work done by P is dependent on path of
deformation.
Such forces are called nonconservative forces.
What is the critical value of P?
 Static analysis

P cos 
P EIy  P cos   f  y   P sin   l  x 
f  P
cos   1,sin   
P sin   EIy  P  f  y   P  l  x 
lx  EIy  Py  Pf  P  l  x 
 f  y 
P
 y  k 2 y  k 2  f    l  x   with k 2 
EI
y y  x   A cos kx  B sin kx   f    l  x  

x BCs: y  0   0, y  0   0
y ( L)  f , y   L   
y  x   A cos kx  B sin kx   f    l  x  
y  x    Ak sin kx  Bk cos kx  
BCs: y  0   0, y  0   0
y (l )  f , y  l   
y  0   0  A  f  l  0
y  0   0  Bk    0
y  l   f  f  A cos kl  B sin kl  f
y  l        Ak sin kl  Bk cos kl  
 1 0 1 l   A 
 0 k 0 1   B 
  0
 cos kl sin kl 0 0   f 
  
 k sin kl k cos kl 0 0     19
For non trivial solutions
1 0 1 l
0 k 0 1
0
cos kl sin kl 0 0
k sin kl k cos kl 0 0
1 0 1 l
0 k 0 1
We get    1
cos kl sin kl 0 0
k sin kl k cos kl 0 0
This means that only trivial solution is possible for all values of k .
Structure's state of rest  y  0  is always stable for all values of P.
This defies expectations.
Did we miss something? 20
Idea
The loss of structural stability is accompanied by
oscillations whose amplitude grow in time.
Therefore, include inertial effects in considering
stability of equillibrium state.

Consider the case when P was applied in a conservative

manner.
When dynamic analysis was performed, when P  Pcr ,
the response grew linearly in time with natural frequency=0.
The results from static and dynamic analysis coincided.

21
 Model with distributed mass

Recall
BCs at x  l
P P EIy  0 & EIy  Py  0
A
P P
A A
A

BCs at x  l
EIy  0 & EIy  0
P has zero component
along AA

22
EIy iv  Py  my  0
BCs: y  0, t   0, y  0, t   0, EIy  l , t   0, EIy  l , t   0
y  x, t     x  exp  it 
EI iv  P   m 2  0
 iv  k 2   a 2  0
  x   0 exp  sx 
s 4  k 2 s 2  a 2  0
s 2     2  k 2   a 2  0
k2 k 4
k 2
k 4
1    a 2  & 2    a 2 
2 4 2 4
   x   A cosh 1 x  B sinh 1 x  C cos 2 x  D sin 2 x
  0   0,    0   0,    l   0,    l   0 23
  x   A cosh 1 x  B sinh 1 x  C cos 2 x  D sin 2 x
  0   0,    0   0,    l   0,    l   0
Condition for nontrivial solution
2a 2  k 4  2a 4 cos 1l  k 2 a 2 sinh 1l sin 2l  0
This leads to the relation between P and .
y  x, t     x  exp  it 
Write   a  ib  y  x, t     x  exp  ia  b  t 
EI
Instability when b  0  Pcr  19.739
l2
25

20

15

10

Response for P  Pcr

5
y(t)

-5

-10

-15

-20
24
-25
0 2 4 6 8 10
t
 References
• V V Bolotin, 1963, Nonconservative problems of the
theory of elastic stability, Pergammon, Oxford.

• M A Langthejm and Y Sugiyama, 2000, Dynamic stability

of columns subjected to follower loads: a survey, Journal
of Sound and Vibration, 238(5), 809-851.

25
FE analysis of vehicle-structure interactions

u t 
ms a, v
kv cv mu  unsprung mass
ms  sprung mass
mu

EI , m, c, l

y  x, t 
for 0  t  texit
 
ms u  cv u 

D
Dt 
 
y  x  t  , t    kv u  y  x  t  , t   0

 1 2
EIy  my  cy  f  x, t    x  vt  at 
iv

 2 
 
 
f  x, t    mu  m s  g  kv u  y  x  t  , t   cv u 

D
Dt
y  x  t  , t  

D2
 mu 2 y  x  t  , t 
Dt
f  x, t   wheel force
for t  texit
EIy iv  my  cy  0
with conditions at texit obtained from equations valid for 0  t  texit

Approach: integral and weak formulation

Guide way uneveness

u t 
ms a, v
kv cv mu  unsprung mass
ms  sprung mass
mu

EI , m, c, l

y  x, t 

28
 u1  t  u2  t 
ms1
kv1 cv1 kv 2 cv 2

mu1 mu 2

EI , m, c, l

y  x, t 

Vehicles and the beam interact

29
ms
kv1 cv1 kv 2 cv 2

mu1 mu 2

EI , m, c, l

y  x, t 

Vehicles and the beam interact

30
31
Prelude : Integral and weak formulations for modeling beam vibrations

We consider situations in which the system to be analyzed is described

in terms of a governing differential equation.This is in contrast to our studies
so far wherein we started with Hamilton's principle in formulating the problem.

Consider

 EIy   m  x  y  f  x, t 
y  0, t   0, y  0, t   0, EIy  l , t   M 0  t  ,  EIy    l , t   0
y  x, 0  , y  x, 0   0
We aim to find an approximate solution to this equation in the form
N
y  x, t    an  t  n  x   0  x 
n 1

As we have seen earlier, the substitution of the assumed solution into the
governing equation leads to a residue. 32
 Weighted residual statement
l
 EIy   m x y  f x, t dx  0
0 w  x       


If y  x, t  is the exact solution,  EIy   m  x  y  f  x, t    0.

 
If y  x, t  is replaced by its approximation,

y  x, t    an  t  n  x   0  x    EIy    m  x  y  f  x, t    0.
N

n 1
 
The above statement implies that the error of representation is zero
in a weighted integral sense.
By choosing N independent weight functions, we get N independent
equations for the unknowns an  t  , n  1, 2, ,N
Continuity requirements on w  x  & n  x  are different. The
requirements on n  x  are more stringent.
The weighted integral statement is equivalent to the governing field
equation and does not take into account BCs.
The unknowns an  t  , n  1, 2, , N can be determined by considering
N weight functions wn  x  , n  1, 2, ,N
  
l
  N
  N

0 n  
w  x   EI
n 1
an  t  n
  x   0
  x     m  x   an  t  n  x   f  x, t  dx  0
 n 1 
 
To proceed further with the solution, we need to select the trial functions
n  x  , n  1, 2, , N which possess 4 th order derivatives and satisfy the
prescribed boundary conditions.
There is no such stringent requirements on the weight functions
wn  x  , n  1, 2, ,N
 l
Consider  w  x   EIy   m  x  y  f  x, t  dx and integrate
 
0

the first term by parts:

l l l
  w  x  EIy     w  x  EIy dx   w  x   m  x  y  f  x, t  dx  0
  0
0 0
l l
  w  x  EIy     w  x  EIy   0   w  x  EIy dx
l

  0
0
l
  w  x   m  x  y  f  x, t  dx  0
0

This is known as the weak form.

Notice: differentiability requirement on y  x  [and hence on trial functions
n  x  , n  1, 2. N ] has come down to 2 and the requirement on w( x)
has gone up to 2.
The integration by parts has enabled us to trade the differentiability
35
requirements between trial functions and the weight functions.
 l
Consider the terms  w  x  EIy   &  w  x  EIy   0
l

  0
We can identify two types of BCs: natural and essential.
We call coefficients of the weight function and its derivatives in the above
terms as secondary variables.

Thus,  EIy  &  EIy  are secondary variables.

Specification of secondary variables on the boundaries constitute the
natural (or force) BCs.
The dependent variables expressed in the same form as the weight function
as appearing in the boundary terms are called the primary variables.
Thus y  x, t  & y  x, t  are the primary variables.
Specification of the primary variables on the boundaries constitutes the
essential (or geometric) boundary conditions.

36
Remarks
The number of primary and secondary variables would be equal.
The SVs have direct physical meaning.
 EIy  : bending moment

 EIy  : shear force.

Each PV is associated with a corresponding SV

Secondary variable Primary variable

Bending moment :  EIy  Slope y  x, t 

Shear force: EIy  Displacement y  x, t 

Essential BCs invovle specifying displacement & slope

at the boundaries.
Natural BCs involve specifying BM and SF at boundaries 37
Remarks  Continued 
On the boundary either a pv can be specified or the corresponding
sv can be specified. A given pair of sv and pv cannot be specified
simultaneously at the same boundary.
Thus, at a free end we can specify BM to be zero and slope remains
unspecfied; similarly, SF can be specified to be zero and displacement
remains unspecified.

By denoting:  EIy   V &  EIy   M , we write

l
 w  x  V  x   0   w  x  M  x   0   w  x  EIy dx 
l l

0
l

 w  x  m  x  y  f  x, t dx  0
0
38
 l
 w  x  V  x   0   w  x  M  x   0   w  x  EIy dx 
l l

0
l

 w  x  m  x  y  f  x, t dx  0
0

We now require the weight functions to satisfy the essential BCs

of the problem.
Recall: the BCs we are considering are

y  0, t   0, y  0, t   0, EIy  l , t   M 0  t  ,  EIy   l , t   0
Accordingly, we demand w  0   0, w  0   0
Thus we have
 w  x  V  x   0   w  x  M  x   0  w  l V  l   w  l  M  l   w  l  M 0  l 
l l

The weak form thus reads

l l
 w  l  M 0  t    w  x  EIy dx   w  x   m  x  y  f  x, t  dx  0 39
0 0
 l l
 w  l  M 0  t    w  x  EIy dx   w  x   m  x  y  f  x, t  dx  0
0 0

This is equivalent to the original differential equation and the natural

BCs.
N
Recall that we have y  x, t    an  t  n  x   0  x 
n 1

with an  t  , n  1, 2, , N to be determined. 
l
 N 
 w  l  M 0  t    w  x   EI  an  t  n  x   0  x   dx 
0   n 1 
l
 N  
0 w  x  m  x  
n 1
an  t  n  x    f  x, t  dx  0
 
We can use w  x   n  x  ,1, 2, , N and obtain equations for
an  t  , n  1, 2, , N.
40
Remarks (continued)
The method leads to symmetric coefficient matrices.
The natural boundary conditions are included in the weak form
and the approximate solution needs to satisfy only the essential boundary
conditions.

J N Reddy, 2006, An introduction to the finite element method,

3rd Edition, Tata McGraw-Hill, New Delhi

41
  EIy  m  x  y  f  x, t 
y  0, t   0, y  0, t   0, y  l , t   0, EIy  l , t   0
y  x, 0  , y  x, 0   0

x
1 2 k n

x0 x1 x2 xk 1 xk xn 1 xn
1 3
k

2
xk 1 xk
4

42
Consider the k th element
n 1
Let   x -  li where li  xi 1  xi
i 1

xk 1  x  xk  0    lk
 y 
th

For the k element we have  y  
  
 EIy   m  x  y  f  x, t 
y  0, t   u1  t  , y  0, t   u2  t  , y  lk , t   u 3  t  , y  lk , t   u 4  t 
EIy  0, t   F1  t  , EIy  0, t   F2  t  , EIy  le , t   F3  t  , EIy  le , t   F4  t 
Weighted residual statement
lk
 EIy   m  y  f  , t d   0
0 w         

43
Weak form
lk lk
 w   EIy     w   EIy     w   EIy  d 
0 
lk

  0 
0
lk

  w    m   y  f  , t  d   0
0
lk lk

  w   EIy d    w    m   y  f  , t  d 

0 0

 F1w  0   w  lk  F3  F2  w  0   F4  w  lk   0

F1   EIy   0 
F2   EIy  0 

F3    EIy   lk 
F4    EIy  lk 
44
lk lk

 w   EIyd   w   m   y  f  , t d

0 0

 F w  0   w  l  F  F  w  0   F  w  l   0
1 k 3 2 4 k
4
y  x, t    ui  t  i  
i 1

and by selecting w    i   , i  1, 2,3, 4 we get

 
lk 4 lk 4

0 j
     EI 
i 1
ui  t  i
  d     j    

m    
i 1
ui  t  i     f   , t  d

0

 F1 j  0    j  lk  F3  F2  w j  0   F4  j  lk   0

4 4
  K ij ui  t    M ij ui  t  Pj  t   0, j  1, 2,3, 4
i 1 i 1
lk lk

K ij   EI j   i d ; M ij   m    j   i   d

0 0
lk

Pj  t     j   f  , t d  Fj 45
0
 2 3
1    1  3 2
2 3
;
l l
2 3
2      2  2
;
l l
2 3
3    3 2
2 3
;
l l
2 3
4     
l l2

 156 22l 54 13l   12 6l 12 6l 
 4l 2 3l 2   2
6l 2l 2 
ml  22l 13l EI  6l 4l
M &K  3
420  54 13l 156 22l  l  12 6l 12 6l 
   
 13l 3l 2 22l 4l 2  6l
2
 6l 2l 4l 2 

46
Assembly
Requirements
Inter element continuity of primary variables (deflection and slope in this case)
Inter element equillibrium of secondary variables (BM and SF here).

Imposition of boundary conditions

Primary variables not constrained  the corresponding secondary variables
are zero (unless there are applied external actions)
Free end: deflection and slope are not constrained  BM and SF are zero
unless the free end carries additional forces.
Primary variables are prescribed
to be zero  the secondary variables

to be specified functions of time  determine the reactions

Governing equations of motion 47
for 0  t  texit
 
ms u  cv u 

D
Dt 
 
y  x  t  , t    kv u  y  x  t  , t   0

 1 2
EIy  my  cy  f  x, t    x  vt  at 
iv

 2 
 
 
f  x, t    mu  m s  g  kv u  y  x  t  , t   cv u 

D
Dt
y  x  t  , t  

D2
 mu 2 y  x  t  , t 
Dt
f  x, t   wheel force
for t  texit
EIy iv  my  cy  0
with conditions at texit obtained from equations valid for 0  t  texit

Approach: integral and weak formulation