Finite element method for structural dynamic

and stability analyses

Module-2

Finite element analysis of dynamics of planar trusses

and frames

Lecture-4 FEM: motivations. Analysis of axially vibrating rods and

Euler-Bernoulli beams

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

1
 Field equation:  EI  x  v  x, t    m  x  v  x, t   c  x  v  x, t   f  x, t 
ICS: v  x, 0   v0  x  , v  x, 0   v0  x 
BCS: Appropriate geometric and natural BCS

N
v  x, t    a  t    x 
n n
n 1

Recall 
Drive the residue e  x, t  to be small
in some sense (MWR).


Ma  Ca  Ka  P  t  ; a  0 , a  0 

What MWR achieves?

A PDE governing the behavior of a continuous system
has been replaced by an equivalent set of ODE-s (IVP-s)
2
with a view to obtain an approximate solution.
 N
v  x, t    an  t  n  x 
n 1

e  x, t 
L
Least squares 0 an e  x, t  dx  0 for n  1, 2, ,N

L
Collocation    x  x  e  x, t  dx  0 for n  1, 2,
0
n ,N

L
Galerkin    x  e  x, t  dx  0 for n  1, 2,
0
n ,N

Subdomain
collocation
 U  x  x   U  x  x  e  x, t  dx  0
0
n 1 n

for n  1, 2, ,N
L
Petrov-Galerkin   x  e  x, t  dx  0 for n  1, 2,
0
n ,N
3
L

 wn  x  e  x , t  dx  0 for n  1, 2, ,N
0 Weight Residue
Method of weighted resuidues

 Ma  Ca  Ka  P  t  ; a  0  , a  0 

Assumed mode method and Lagrange's equation

Strong (operational) form, Weighted residual form, and

weak (variational) form of governing equations
4
 Limitations

N
v  x, t    an  t  n  x 
n 1

n  x  ,n  1, 2, , N are global in nature, that is, they are

valid for all x   0, L .
Constructing these functions for simple geometries is
relatively easy. Not so when geometries become more complicated.
an  t  , n  1, 2, , N are generalized coordinates and they do not
have direct physical meaning.

5
Exact solutions are not possible.
Approximate solutions using global trial
functions are not easy to develop:
How to select global trial functions?
The generalized coordinates do not have
direct physical meaning. 6
 FEM: motivations and general remarks

• The class of problems which can be solved exactly is limited.

The exact solutions, when possible, serve as benchmarks
against which approximate solutions can be validated.

• Approximate solutions which employ global trial functions are

possible for a limited class of problems.

• We would like to work with methods in which the generalized

coordinates have direct physical meaning. Unknown
coefficients in a series expansion in terms of global trial
functions do not serve this purpose.

• For most practical problems of engineering interest neither

exact solutions are possible nor can we construct global shape
function based approximations.
7
 FEM: motivations and general remarks (continued)

• Some of the complexities that we wish to be able to handle are

– Geometric complexities (curved shapes, cut-outs, large sized

structures).
– Inhomogeneity in structural material and geometric properties.
– Structures assembled out of bars, beams, plates, shells, and 3D
continua.
– Diversity of structural behaviour (vibration, buckling, fracture,
fatigue, creep, parametric instabilities, bifurcations, etc.).
– Geometric and (or) material nonlinearities and time varying
systems
– Multiphysics problems

• Fluid structure interactions

• Thermal stress analysis
• Electro-mechanical systems
• Structures under active control 8
 FEM: motivations and general remarks (continued)

• FEM : a powerful means to tackle these issues

Main idea: approximate the field variables in terms of piecewise
polynomials over subdomains and assemble the solutions suitably

• Drivers
– Advances in computer hardware and software
– Professional software development which
utilize the computer resources for
• Preprocessing
• Solving equations
• Postprocessing of results
– Mathematical foundations being well
understood (Finite Element “Analysis”). 9
 What are “F”, “E”, and “M” in FEM?

• (M) FEM is an approximate numerical method for

analysing mathematical models of physical
phenomena encapsulated through a set of PDE-s
or variational principles.

Note: Finite element analysis can be taken to mean study of

mathematical aspects of the method including questions on how the
various approximations made affect the convergence and accuracy of
the solutions.
10
We generally work in Cartesian coordinate system.

Let u  x, t  be the s 1 vector of dependent field variables

to be determined.  = domain of interest.

FEM treats u  xi , t i 1 as unknowns and develops approximation

N

to u  x, t  in terms of u  xi , t i 1 . The method employs piecewise

N

polynomials to achieve this.

Thus the problem of approximating the infinite number of unknowns,

that is, u  x, t  x   & t  0, is converted into the problem of
determining a finite set of of functions u  xi , t i 1 .
N

11
This is the "F" in FEM.
 xi i 1   xi , yi , zi i 1
N N
are called the nodes.

u  xi , t   degrees of freedom at the node i; these are

generalized coordinates which have a direct physical meaning
since they are the values of the system states at the nodes.

The time variable t is still continuous at this stage. The

discretization done is thus called the semi-discretization.

How u  x, t  is approximated in terms of u  xi , t i 1 ?

N

 Notion of an element. The "E" in FEM.

12
 r
  = i such that i  j  i  j; i : i th element
i 1    
   
Element  
Geometry 
 Number of nodes   
 
Degrees of freedom at each node
Total dof-s for the element    
 
Value of field variables i
within the element   

3 
4 
 
Within an element
i x2
u  x, t    N j  x  u  x j , t 
q

 x1
j 1 2
 x3
1
N j  x  : interpolation functions. Chosen to be polynomials. 13
 Typical steps in FE analysis of structural dynamic problem
• Partition the domain into a set of elements.

• Approximate the behaviour of field variable within each element either

starting from the governing PDE or the relevant variational principle. Trial
functions are piecewise polynomials. This leads to element level structural
matrices (element mass, stiffness and damping matrices and nodal force
vector).

• Assemble element level matrices and force vectors to form the global
matrices and force vector for the structure.

• Impose boundary conditions and obtain governing equilibrium equations

• Obtain solutions to the semi-discretized equilibrium equations using time

stepping methods.

• Compute derived quantities like stresses and strains

14
Several other points to consider
r
  = i such that i  j  i  j; i : i th element
i 1

Also defines division of geometry.

Mesh : Collection of elements which replaces the domain  in
the simulation work.
Inter-element behavior of field variables: communication only
through value of state variables through nodes (and not through
lines or surfaces).
Evaluation of the derived quantities (e.g., strains and stresses)
within an element and their behavior across element boundaries.
Selection of number of elements, order of polynomials,
step size in time integration and the convergence of the solution
 h  p  t  15
 Approach

• Develop structural matrices for different types of

elements

– Axially vibrating bar

– Torque element
– Beams (2D and 3D)
– Plane stress and plane strain elements
– Plate bending element
– Shell element
– 3D solid element

• Study of trusses, planar frames, grids and 3D frames

• 2D and 3D continuum structures 16

 Review of equations of linear elasticity of isotropic material

Independent variables and Coordinate

system
y x2
OR  time t
x x1
z x3
Number of independent variables: 4

x2 x2
x1  cos x1x1 cos x1x2 cos x1x3 
cos x x cos x x cos x2 x3 
x1  x   T t
 x ; T t
  2 1 2 2
x3  cos x3 x1 cos x3 x2 cos x3 x3 
x3 17
Dependent variables
Stress components
 11  12  13 
  x1 , x2 , x3 , t    12  22  23   x1 , x2 , x3 , t  Behavior under
 13  23  33  coordinate transformation
Strain components    T t T
11 12 13     T t T
  x1 , x2 , x3 , t   12  22  23   x1 , x2 , x3 , t  u  T t u
13  23  33 
Displacement components
 u1 
 
u  x1 , x2 , x3 , t   u2   x1 , x2 , x3 , t 
u 
 3
Total number of dependent variables=15 18
 Number of unknowns=15
Numer of equations
Equilibrium: 3
Strain-displacement: 6 Strain-displacement equations
Stress-strain: 6 u1 u2 u3
11  ;  22  ;  33  ;
x1 x2 x3
Equilibrium equations
1  u1 u2 
 11  21  31 12    ;
   X 1   u1 2  x2 x1 
x1 x2 x3
 12  22  32 1  u1 u3 
   X 2   u2 13    ;
x1 x2 x3 2  x3 x1 
 13  23  33 1  u2 u3 
   X 3   u3  23    ;
x1 x2 x3 2  x3 x2 
 ij   ji ; i, j  1, 2,3  ij   ji ; i, j  1, 2,3
19
 Representation of the
Representation of the
constitutive laws-II
constitutive laws-I
I1   11   22   33
e  11   22   33
1
 11   e  2G11 11  1    11  I1 
E
 22   e  2G 22
1
 33   e  2G 33  22  1    22  I1 
E
 12  2G12 1
 33  1    33  I1 
 13  2G13 E
 23  2G 23 1 1 
12   12   12
G  3  2G  2G E
E 1 1 
 G 13   13   13
 2G E
 1 1 
2 G     23   23   23
2G E 20
Equations of elasticity in indicial notations

Equillibrium  ij , j  X i   ui
1  ui u j 
Strain-displacement  ij    

2  x j xi 
Stress-strain  ij  cijkl  kl

21
Alternative representation of equations of elasticity
   11  22  33  12  13  23 
t

  11  22  33 212 213 2 23 

t

1    0 0 0 
  1   0 0 0 

 

 1  0 0 0 

  C
 0 1  2
E 0 
C 
1  1  2  
0 0

0
   C
1  2 
 0

0 0 0

0 

C  C 1
 1  2 
 0 0 0 0 0 
  

1   0 0 0 
  1  0 0 0 

1    1 0 0 0 
C  
E0 0 0 2 1   0 0 
0 0 0 0 2 1   0 
  22
 0 0 0 0 0 2 1   
  
 x 0 0
  D t  X   u
0 
0 Du  
 y 
    c
0 
0   c
 z 
D 
   D t c  X   u
0
 y x   D t cDu  X   u
 
0  
 z y 
 
 0

 z x 
23
 Strain energy in a body under uniaxial state of stress

1 1
V      xx dydz  xx dx     xx xx dv
2 2v
 11   11 
dz    
 xx dydz  xx dydz   22  22 
dy  33    33 
y       

 12  2
 12 
x dx  13   213 
   

 23  
 23 
2
z
1 t
Strain energy in a body   c  V    c dv
under 3D state of stress 2v
1 t t
1 t
V     dv   Du  V   u D cDudv
2v 2v 24
 1 t t
V  t    u D cDudv
2v
1 t
T  t    u  udv
2v
L  T (t )  V (t )
t2
1
A    [u  u u D cDu ]dvdt
t t t

2 t1 v

25
 Axially vibrating rod element
y y

x z
AE, m, l , c
 xx  x, y, z, t    xx  x, t  ; all other stress components are zero.
1
  x, t    xx  x, t  ; u  x, y, z, t   u  x, t 
E

2 2
 u  1  u 
l l
1
V   AE   dx; T   m   dx
0
2  x  0
2  t 
 l 1  u 2
t2 l
1  u  
2

A     m   dx   AE   dx  dt
t1  0 2  t  0
2  x   26
u1  t  u2  t 

P1  t  1 u  x, t  2
P2  t 
Take u  x, t   u1  t  1  x   u2  t  2  x 
u1  t  & u2  t  : Generalized coordinates
1  x  & 2  x  : Trial functions
u  0, t   u1  t   u  0, t   u1  t  1  0   u2  t  2  0 
Take 1  0   1& 2  0   0
u  l , t   u2  t   u  l , t   u1  t  1  l   u2  t  2  l 
Take 1  l   0 & 2  l   1
This ensures that we have satisfied the requirements
on displacements at x  0 & x  l 27
Let us consider the equilibrium of the bar under
the action of support displacements u1  t  & u2  t  .
 1  0   1& 1  l   0
d 21
 AE 2  0; 1  0   1& 1  l   0
dx
x
 1  x   1 
l
d 22
Similarly, we get AE 2  0; 2  0   0 & 2  l   1
dx
x
 2  x  
l
28
 x
1 1  x   1 
l

x
x
1 2  x  
l
x
x x
1  x   1  & 2  x  
l l
 x x
 u  x, t   u1  t  1    u2  t 
 l l
 x x  1 1
u  x, t   u1  t  1    u2  t  & u  x, t   u1  t      u2  t 

 l l  l l
29
 2
 u 
l
1 AE
V   AE   dx   2 1

2
u u
0
2  x  2l
2
1   x x
l
T   m u1  t  1    u2  t   dx 
2   l l 6

ml 2
u1  u u
1 2  u 2
2

L  T V 
ml 2
6
 u1  u1u2  u2  
2 AE
2l
 u2  u1 
2

L ml d  L  ml
  2u1  u2  ;     2u1  u2 
u1 6 dt  u1  6
L ml d  L  ml
  2u2  u1  ;     2u2  u1 
u2 6 dt  u2  6
L AE L AE
  u2  u1  ;    u2  u1 
u1 l u2 l
30
d  L  L ml ml AE
  0 u1  u2   u2  u1   0
dt  u1  u1 3 6 l
d  L  L ml ml AE
  0 u1  u2   u2  u1   0
dt  u2  u2 6 3 l

Element level equation of motion

ml  2 1   u1  AE  1 1  u1   P1 
        
6 1 2  u2  l  1 1  u2   P2 

 Mu  Ku  P

31
Remarks
ml  2 1 
M     element mass matrix
6 1 2
M is called the consistent mass matrix
since the shape function used for computing KE
is same as the one used in computing strain energy.
M  M t
AE  1 1
K    =element stiffness matrix
l  1 1 
K  K t
Two noded bar element with one dof per node
 K and M are non-diagonal
32
 Lumped mass matrix

 
 ml 
 2 0 
M  
0 ml 
 2 
Matrix is diagonal
Excludes inertial coupling between nodal displacements
Lumping is an arbitrary process

33
Remarks  continued 
 x x
We have u  x, t   1   u1  t   u2  t 
 l l
Suppose u1  t   1& u2  t   1. That is, the rod is undergoing
rigid body displacement.
 x x
u  x, t    1     1
 l l
1  x   2  x   1  Partition of unity 
 x x u u2  t   u1  t 
u  x, t   1   u1  t   u2  t   
 l l x l
 Strain is independent of x for any l
 Constant strain condition can be simulated as element
length  0.
These properties have implications on convergence of FE solutio34ns.
Example

P t 
AE, m, l , c
u1  t  u2  t 

P1  t  1 u  x, t  2
P2  t 

u1  t   0
P2  t   P  t 

35
ml  2 1  u1  0  AE  1 1 u1  0   P1  R (t ) 
        
6  1 2  2  l 
u 1 1  2  
u P  
t
Two unknowns: u2  t  & P1  R (t )
ml AE
 u2  u2  R  t 
6 l
ml AE
u2  u2  P  t 
3 l
3P  t  3 AE
u2   u2 
2
; 
ml ml 2
x
u  x, t     u 2  t 
l
u2  t  u2  t  AE
 xx  x, t   ;  xx  x, t   E ;Axial thrust= u2  t 
l l l
36
 5
u  x, t    ai  t  i  x 
i 1

1 2 3 4 5

1  x

 2  x

1
 3  x

1
 4  x

 5  x
1

37
 x
1 1  x   1 
l

1 x
2  x  
l

u e  x, t   u1e  t  1  x   u2e  t  2  x 
1 et 1 et
T  t   u  t  M u  t  ;V  t   u  t  K e u e  t 
e e e e

2 2

4 4
T  t    Ts  t ;V  t   Vs  t 
s 1 s 1 38
 Euler-Bernoulli beam element
y, v y

x, u z
EI , m, l , c
• Bending in the x-y plane
• x-axis coincides with the centroidal axis of the beam
• No coupling exists between bending and torsion

v
u  x, y , t    y ; v  v  x , t 
x
u  2v
 xx   y 2
x x
u v v v
2 xy     0
y x x x 39
 1
V
2v  xx xx dx
L L
1 1
   E xx2 dxdA T     v 2  x, t  dxdA
20A 20A
2
  v
L 2 L
1 1
   E   y 2  dxdA   mv 2  x, t  dx with m=   dA
2 0 A  x  20 A
2
1
L
 v
2
   
2
EI z  2 
dx with I y dA
 x 
z
20 A

L  T t   V t 
2
1
L
1
L
 v 2
  mv  x, t  dx   EI z  2  dx
2

20 20  x  40
 u1  t  , P1  t  u3  t  , P3  t 

 

u2  t  , P2  t  u4  t  , P4  t 

2-noded element with 2dofs per node

v  x, t   u1  t  1  x   u2  t  2  x   u3  t  3  x   u4 t  4  x 

41