Finite element method for structural dynamic

and stability analyses

Module-1 Approximate methods and FEM

Lecture-1 Equations of motion using Hamilton’s principle

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

1
 What is this course about?

• FEM in the context of structural mechanics problems.

• What are the issues to be dealt with, when the FEM, as
developed for analysis of static problems, is extended to
deal with problems of structural dynamic and elastic
stability analyses
• New questions
• New phenomena
• New numerical tools
• Application areas

2
• This is not a first course in
• FEM
• Structural dynamics
• Structural Stability analysis

Pre-requisites
• Matrix methods of static structural analysis
• A first course in theory of vibrations
• Elements of elastic stability analysis
• Matrix algebra and ode-s

The course will contain brief overviews of the main

ideas of these subjects
3
 Different Facets of FEM
• A numerical method to obtain approximate solutions to
boundary value problems (ODE-s, PDE-s).
• A modeling and simulation tool
• A tool in computer aided engineering that exploits the
power of computers
• Modeling (pre-processing)
• Numerical solutions
• Scrutiny of results (post processing)

• Combined with sensing and actuation, a tool for

• Active control of structures
• System identification and health monitoring
• An important component in hybrid testing methods

4
 Topics proposed to be covered in this course

• Approximate methods and FEM

• Dynamics of truss and planar frame structures
• Damping models and analysis of equilibrium equations
• Dynamics of Grids and 3D frames
• A few computational aspects (solution of equilibrium equations,
eigenvalue problems, model reduction, and substructuring)
• Dynamic stiffness matrix and transfer matrix methods
• Dynamics of plane stress/strain, plate bending, shell and 3d elements
• Applications (earthquake engineering and vehicle structure
interactions)
• FEA of elastic stability problems
• Treatment of nonlinearity
• FE model updating
• FEM in hybrid simulations

5
Three intertwining themes

• Modeling
• Numerical solutions
• Applications

Target audience

• Graduate level students

• Research students
• Engineers who use FEM as a tool in their work

6
Books
1. M Petyt, 1990, Introduction to finite element vibration analysis, CUP,
Cambridge.
2. W Weaver and P R Johnston, 1987, Structural dynamics by finite
elements, Prentice-Hall, Englewood Cliffs.
3. M Geradin and D Rixen, 1997, Mechanical vibrations, 2nd Edition, Wiley,
Chichester.
4. R W Clough and J Penzien, 1993, Dynamics of Structures, 2nd Edition,
McGraw-Hill, New York.
5. K J Bathe, 1996, Finite element procedures, Prentice Hall of India, New
Delhi.
6. I H Shames and C L Dym, 1991, Energy and finite element methods in
structural mechanics, Wiley Eastern Limited, New Delhi.
7. R R Craig, 1981, Structural dynamics: an introduction to computer
methods, Wiley, NY
8. C H Yoo and S C Lee, 2011, Stability of structures, Butterworth-Heinmann,
Burlington.
9. G J Simitses and D H Hodges, 2006, Fundamentals of structural stability,
Elsevier, Amsterdam.
7
Additional references
1. L Meirovitch, 1997, Principles and techniques of vibrations, Prentice-Hall,
New Jersey
2. O C Zienkiewicz and R L Taylor, 1989, The finite element method, Vols-I
and II, 4th Edition, McGraw-Hill, London.
3. R D Cook, D S Malkus, and M E Plesha, 1989, Concepts and applications
of finite element analysis, 3rd Edition, John Wiley, New York
4. J N Reddy, 2006, An introduction to the finite element method, 3rd Edition,
Tata-McGraw-Hill, New Delhi.
5. S S Rao, 1999, The finite element method in engineering, 3rd Edition,
Butterworth-Heinemann, Boston.
6. T J R Hughes, 2000, The finite element method, Dover, Mineola.
7. M Paz, 1985, Structural dynamics, 2nd Edition, CBS Publishers, New Delhi
8. W McGuire, R H Gallagher, and R D Ziemian, 2000, Matrix structural
analysis, 2nd Edition, John Wiley, New York.
9. P Seshu, Textbook on finite element analysis, 2003, Prentice Hall India,
New Delhi.
10. G Strang and G J Fix, 2008, 2nd Edition, An analysis of the finite element
method, Wellesley-Cambridge Press, Wellesley.
8
Dynamic loads

The magnitude, direction, and (or) point of application of the load

change with time.

Under the action of dynamic loads the structure vibrates, that is,

(a) the structure develops significant level of inertia forces

(b) significant level of mechanical energy is stored as kinetic energy

Note: Not all time varying loads need to be dynamic in nature;

for example,
(a) Load on a dam due to filling of a reservoir
(b) Load on a spectator gallery as a stadium gets filled up.

9
Examples:
1. Earthquake loads on buildings, bridges, dams, power plants etc.
2. Wind loads on long span bridges, tall chimneys etc.
3. Loads due to blast and impact.
4. Running machineries in buildings.
5. Vehicle moving on a bridge.

10
Degree of freedom (DOF)

Number of independent coordinates required to specify the

displacements at all points in the system and at all times.

 
u t 
g x 2  y 2  L2
k
L 
m f t 
y
m

x

11
  

Two-dof systems z
 Particle in space
DOF=3

y
z
x

y
Rigid body in space
12
x DOF=6
 This chunk has mass, stiffness and damping
DOF  

N
w  x, t    an  t  n  x   DOF  N
n 1
Unknown Known
function function
of time of x

Two-dof system 13
Remarks:

•The DOF is not an intrinsic property of a given structure.

•It is a choice exercised by the modeler.
•SDOF systems: systems with one DOF.
•MDOF systems: systems with DOF>1
•MDOF systems could be discrete (finite DOFs) or
continuous (infinite DOFs)
•The coordinates need not have direct physical meaning
(generalized coordinates)
•DOF=a number or the coordinate
FEM: replaces a system with infinite dofs by a system with
finite dofs.

14
Entities in a simple mathematical model for a vibrating system

Whenever a structure vibrates, it not only displaces, but also

accelerates. Besides, the mechanical energy is converted to heat
and/or sound.

Stiffness: resists displacement

Inertia: resists acceleration
Damping: dissipates energy

Inputs: external forcing and/or initial conditions; supply mechanical

power

•Mass: offers resistance to acceleration, stores kinetic energy

•Damper: dissipates mechanical energy
•Spring: offers resistance to displacement, stores potential energy
•Force and/or initial displacement, velocity: supplies mechanical power
15
Position of the mass
when the spring is not u t 
stretched k
m f t 

•Mass: offers resistance to acceleration, stores kinetic energy

•Damper: dissipates mechanical energy
•Spring: offers resistance to displacement, stores potential energy
•Force and/or initial displacement, velocity: supplies mechanical power

1
Mass (kg) KE= mu 2 Force=mu
2
1 2
Spring (N) PE= ku Force=ku
2
1 2
Damper (Ns/m) DE= cu Force=cu 16
2
 u t 

Force=mu  t 
Mass  kg  m mu  t  1
KE=T  mu 2  t 
2

Force=k u2  t   u1  t  
Spring  N/m  u1  t    u2  t  1
k u2  t   u1  t  
2
PE=V 
k 2

c
Force=c u2  t   u1  t  
Damper  Ns/m  u1  t    u2  t 
1
DE  c u2  t   u1  t  
2

17
 How does the motion take place?

Hamilton's principle
q  t   n 1 vector of system dof-s
Lagrangian : L  q, q   T  q, q   V  q, q 
T  q, q   total kinetic energy of the system
V  q, q   total strain energy of the system
Among all the dynamic paths which satisfy the boundary condtions
(on prescribed displacements) at all times, and
with the actual values at two arbitrary instants of time t1 and t2 at
every point of the body, the actual dynamic path minimizes the functional
t2

   T  q, q   V  q, q   dt
t1
18
 u t 

u  t2  

u t 


u  t1 

t1 t2
t

19
Remarks
Functional: function of functions
Domain: set of admissible functions
A functional can also be viewed as a function of infinite set of variables
t2

   T  q  t  , q  t    V  q  t  , q  t    dt
t1

Optimization of functionals is studied in the subject of calculus

of variations.
Presently we have not included external forces and damping forces.

Immediate objective:
Gain familiarity with the application of the principle
by considering equations of motion of simple oscillators and
structural elements 20
Example-1
u t 
k

Hamilton's principle: Minimize the functional

t2 t2

   L  t  dt   T  t   V  t   dt
t1 t1
t2
1
   mu 2  t   ku 2  t   dt
t1
2
 u  t   u (t )    t 
u (t )
u (t )

u  t2 


u  t1 

 t 
t
t1 t2

u t   u t     t 
Admissible functions Unknown optimal solution Variation
u  t1 u  t1 &u  t2 u  t2   arbitrary real number
  t  arbitrary fucntion   t1    t2   0

  t   sufficiently smooth, reasonably small, &  t1     t2   0

22
but otherwise arbitrary
 u t   u t     t 
Admissible functions Unknown optimal solution Variation
 x  t1   x t1 & x  t2   x t2   arbitrary real number
  t   arbitrary function   t1    t2   0
t2 2 t
1
       mu 2  t   ku 2  t   dt   L u  t  , u  t   dt
t1
2 t1

d     L u L u 
t2

 0    dt  0 for   0
d  0 t1 
u  u  

 L L
t2

    t     t   dt  0
t1 
u u 
 L  d  L  L 
t t

2 2

    t          t  dt  0
 u  t1 t1  dt  u  u 
d  L  L
    0  because   t  is arbitrary 
dt  u  u
 d  L  L
   0
dt  u  u
1 L d  L  L
L   mu  t   ku  t   
 2 2
  mu;    mu;  ku
2 u dt  u  u
 mu  ku  0

d  L  L
The equation     0 is called Lagrange's equation.
dt  u  u
For discrete systems, with n dofs, Lagrange's equation can
be generalized as
d  L  L
    0; i  1, 2, ,n
dt  ui  ui
where L  L  qi  t  , qi  t  ; i  1, 2, , n  .
Remarks
t2

   T  u  t  , u  t    V  u  t  , u  t    dt is a functional
t1
t2
1
       mu 2  t   ku 2  t   dt : one parameter family of functions
t1
2
d   
The condition  0 is not a sufficient condtion
d  0

for minimum of     . The discussion on whether this condition

implies minimum, maximum, or or stationary value of     is not
considered here. We will refer to (0) as being the extremum and
the optimizing function as the extermizing function.
Even when (0) is a minimum, it need not be the absolute minimum.
Refer, for a further discussion on these issues,
R Weinstock, 1974, Calculus of variations, Dover, NY. 25
Remarks (continued)

The condition for stationarity

t2
The condition for stationarity
of the functional    L u  t  , u  t   dt of the function F  u , v  is
F F
t1

d  L  L  0& 0
is    0 u v
dt  u  u

26
 Example-2 u1  t  u2  t 

n  2;
m1 m2
k1 k2 DOFs: u1  t  & u2  t 

1
T  t    m1u12  m2u22 
2
1 2
V  t   k1u1  k2  u2  u1  
2

2 
1 1 2
L u1 , u2 , u1 , u2    m1u1  m2u2   k1u1  k2  u2  u1  
2 2 2

2 2 
d  L  L
    0; i  1, 2 
dt  ui  ui
m1u1  k1u1  k2  u1  u2   0   m1 0   u1   k1  k2 k2   u1 
        0
m2u2  k2  u2  u1   0   0 m2  u2   k2 k2  u227
Example-3 n  2;
DOFs: u  t  &   t 
k
M


g  L



Y 
m u
 

 X 

T (t ) 
1 1 1
Mu 2  mX 2  mY 2 n  2; q   u, 
2 2 2 X  u  L sin  ; Y  L 1  cos   ;
1 2
V  Ku  mgY X  u  L cos  ; Y  L sin 
2 28
 1 1 1
T (t )  Mu  mX  mY 2
2 2

2 2 2
1 2
V  Ku  mgY
2
n  2; q   u ,  
X  u  L sin  ; Y  L 1  cos   ;
X  u  L cos  ; Y  L sin 
1 1 2 1 2
   
T  Mu  m u  L cos    m  L sin  
2

2 2 2
1 2
V  Ku  mgl 1  cos  
2

L  L u, , u, 
1 1 1 1
L  Mu  m u  L cos    m  L sin    Ku 2  mgL 1  cos  
2 2
2

2 2 2 2
 1 1 1 1 2
L  Mu  m u  L cos    m  L sin    Ku  mgL 1  cos  
2 2
2

2 2 2 2
d  L  L
  
dt  u  u
 0   M  m  u mL  cos    
2
sin   ku  0

d  L  L
    0  mL 2
  mLu cos   mLu sin   mgL sin   0
dt    

Remark

The approach can handle nonlinear systems

30
Example-4
Mr 2
J
m M 2
k1 k2
 r  x2

x1

x1
 
  No slip  r =  x2  x1 
x2
1 1 1  
 x2  x1 
T  mx12 + Mx22 + J 2
2 2 2 r
2
1 2 1 Mr
= mx1 + Mx22  2
2 2 4
Mr 2  x2  x1 
2
1 2 1
= mx1 + Mx2  2

2 2 4 r2
1 1
V  k1 x12  k2  x2  x1 
2
31
2 2
 Mr 2  x2  x1  1 2 1
2
1 2 1
L  mx1 + Mx2      
2 2
k x
1 1 k 2 x2 x1
2 2 4 r2 2 2
d  L  L M
   0  mx1   x1  x2   k1 x1  k2  x1  x2   0
dt  x1  x1 2
d  L  L M
    0  Mx   x2  x1   k2  x2  x1   0
dt  x2  x2
2
2

 M M
 m  
2 2   x1    k1  k2 k2   x1 
     0
 M 3M   x2   k2 k2   x2 
 2 2 
Note
1 t
T  t   x  t  Mx  t 
2
1 t
V  t   x  t  Kx  t 
2 32
 Distributed parameter systems : Axially vibrating rod

u  x, t 
x

L, AE  x  , m  x 

L
1
Kinetic energy: T  t    m  x  u 2  x, t  dx
20
L
1
Potential energy: V  t    AE  x  u 2  x, t  dx
20
L L
1 1
Lagrangian L  T  V   m  x  u 2  x, t  dx   AE  x  u 2  x, t  dx
20 20
L
 L   F u  x, t  , u   x, t   dx
0
33
Let at t  t1 , u  x, t   u1  x  & t  t2 , u  x, t   u2  x 
Admissible functions: satisfy the above conditions
u  x, t   u  x, t     x, t 
Admissible Unknown optimal Variation
function solution

  x, t  is such that   x, t2     x, t1   0.
We will shortly come to the question of what condtions
  x, t  must satisfy at x  0 and x  L for t  t1 &t  t2 .
1 L
t2
1
L

         m  x  u  x, t  dx   AE  x  u   x, t  dx dt
2 2

t1  
20 20

34
d
 0 at   0
d
t2 L

   [m  x  u  x, t   x, t   AE  x  u   x, t    x, t ]dxdt  0
t1 0
t2

Consider  m  x  u  x, t   x, t  dt 
t1


t2

 m  x  u  x, t   x, t   t    m  x  u  x, t     x, t  dt
t2

1
t1
t
L
Similarly  AE  x  u   x, t    x, t  dx 
0


L
 AE  x  u   x, t   x, t   0    AE  x  u   x, t   x, t  dx
L

0
x
35
d
 0 at   0 
d
L t2

   m  x  u  x, t   x, t   t dx    AE  x  u   x, t   x, t   0 dt
t2 L

1
0 t1

   2u 
t2 L

     AE  x  u   x, t   m  x  2   x, t  dxdt  0
t1 0 
x t 
Since   x, t  is chosen arbitrarily, we take that each of these terms are
separately equal to 0. That is
   2u 
t2 L

t 0  x  AE  x  u  x, t   m  x  t 2   x, t  dxdt  0
1

 m  x  u  x, t   x, t 
t2
t1
dx  0
0
t2

  AE  x  u  x, t   x, t  0 dt  0
L

36
t1
Consider
   2u 
t2 L

t 0  x  AE  x  u  x, t   m  x  t 2   x, t  dxdt  0
1

   2u 
  x, t  is arbitrary    AE  x  u   x, t   m  x  2   0
 x t 
Valid over the length of the bar 

Consider
 m  x  u  x, t   x, t   t  m  x  u  x, t 2   x, t 2   m  x  u  x, t1   x, t1 
t2

  x, t2     x, t1   0   m  x  u  x, t   x, t   t  0
t2

37
Boundary condtions
Consider the term
 AE  x  u   x, t   x, t   0  AE  L  u   L, t   L, t   AE  0  u   0, t   0, t 
L

The two terms on the RHS need to be zero individually

(again, because the variation is arbitrary)

Two situations arise:

 The unknown function, that is, u  x, t  , is not specified at the boundaries

(as in the case of a free end of the rod). Here it is "natural" to expect that the term
AE  L  u   L, t   or AE  0  u   0, t   is zero.

When the unknown function, that is, u  x, t  , is specified to be zero (as in the
case of a clamped end), the variation, in order that it confirms with the stipulated
geometric constraints, needs to be zero.
38
   u   2u
 AE  x    m  x  2 ; IC - s: u  x, 0  & u  x, 0  specified
x  x  t

Boundary AE  0 u  0, t   0 & AE  L  u  L, t   0
conditions

AE  0  u  0, t   0 & u  L, t   0

u  0, t   0 & AE  L  u  L, t   0

u  0, t   0 & u  L, t   0

Geometric, forced, or kinematic boundary condition: u  x, t   0 on the boundary

Free or natural boundary condition: AE  x  u  x, t   0 on the boundary
39
Remarks

 AE  x  u   x, t  at x  0 & L represents the axial thrust at x  0 & L

respectively. From a physical stand point, it is clear that if the
displacement is specified to be zero, that is the end is clamped, there
would be an axial thrust, representing the reaction, at the boundaries.

For a free end, displacement at the end is not specified.

The boundary conditions that we obtain at ends where the unknown

field variable is not specified are called the natural boundary conditions
(also called additional or dynamical boundary conditions).

40
Remarks (continued)

The boundary conditions that we obtain at ends where the unknown

field variable is specified are called the geometric boundary conditions
(also called the essential or imposed boundary condtions).

The variational method identifies the required boundary condtions

consistent with the physics of the problem along with the governing
field equation.

41