Dynamic structural analysis:

Response history

L. Driemeier
R. Moura
M. Alves
Problem:

• given a known load (in time and space) obtain the structure response (in time and
space)
• or, obtain
– displacement,
– velocity
– acceleration
of a loaded structure

Source: Concepts and applications of finite element

analysis, RD Cook, DS Malkus, ME Plesha, RJ Witt 2
Response history

Modal methods:
It is necessary to
solve an eigen-
problem
Usually no good
for non-linear response
Ritz vectors:
More efficient than
modal methods

Response spectra

Component mode
synthesis
(substructuring) 3
Response history

.. . ..
{D}n 1  f ({D}n ,{D}n 1 ,{D}n 1 ,{D}n ,{D}n )

Implicit direct
integration
Direct integration algorithms
methods
Explicit methods

. ..
{D}n 1  f ({D}n ,{D}n ,{D}n ,{D}n 1 )
[ M ]{D}n  [C ]{D}n  {R int }n  {R ext }n

[ M ]{D}n  [C ]{D}n  [ K ]{D}n  {R ext }n 4

 •Number of multiplications per
time step implicit/explicit:
•2 for 1D
•15-150 2D
•4000 3D
•Because in implicit, matrix
becomes less narrowly banded
•Implicit requires more storage

Explicit methods Implicit methods

•Unconditionally stable
(calculation remains stable
•Conditionally stable (a critical
regardless of time step but
time step must not be exceeded
accuracy may suffer)
to avoid instability)
•Matrices cannot be made
•Matrices can be made diagonal
diagonal (coupling)
(uncoupling)
•High cost per time step but few
•Low cost per time step but
steps
many steps
•Structural dynamics problem:
•Wave propagation problem:
•Response dominated by
•Blast and impact loads
lower modes, eg structure
•High modes are important
vibration, earthquacke
•Response spans over
•Response spans over
small time interval
many fundamental periods
5
Explicit direct integration

t 2
{D}n 1  {D}n  t{D}n  {D}n  1
2
t 2
{D}n 1  {D}n  t{D}n  {D}n  2
2
1
1 - 2 {D}n  {D}n1  {D}n1   {D}n1  {D}n1  2t{D}n
2t
1
2 
1 + 2 { D}  {D}n 1  2{D}n  {D}n 1 
t
n

[ M ]{D}n  [C ]{D}n  {R int }n  {R ext }n

{D}n1 6
Half-step central differences

7
 Use this in the bottom eq. slide 6
to obtain:

THIS IS EXPLICIT: IMPLEMENT IT IN A PROGRAM

8
Remarks on explicit integration

• [M] being diagonal uncouples the system

• In slide 6, the system is uncoupled only if [C] is diagonal
• The critical time step in the previous equation is

t 
2
max  1  2   
  damping ratio It is not
necessary to
max  largest possible "calculated" frequency form [K]

• Internal forces can be calculated from either

Nels

{R }n  [ K ]{D}n
int {Rint }n   {rint }n i {rint }n   [ B]T { }n dV
i 1
Minimize n. of integration
points to 1: be carefull with 9
stress calculation though!
To start the process

t
{D}0 
1
t / 2
{D}0  {D}1/2   {D}1/2  {D}0  {D}0
2


{D}0  [ M ]1 {R ext }0  [ K ]{D}0  [C ]{D}0 

t 2
{D}1  {D}0  t{D}0  {D}0
2

10
Critical time step

Lmesh Lmesh tactual

tcr   Cn 
c E/ tcr

• consistent mass matrix gives higher frequencies than

lumped mass matrix
• lumped mass increases ∆t
• higher order elements have higher frquencies than lower
order ones. Use the latter

11
Exercise I

12
13
Implicit direct integration
(Newmark family methods)

• Most of these methods are unconditionable stable: large time keeps the
solution stable, although may compromise accuracy

Average acceleration method

0    t Integração
t  tn 1  tn com
aceleração
u  u (t ) cte.

d
t  t

d
t 2
dt  dt t 
1  

t t+t
14

0    t
 Linear acceleration method

Integração
com
aceleração
linear

IMPLICIT

At
step
n+1
(=t)

15
Newmark relations

{D}n 1

16
 [ M ]{D}n 1  [C ]{D}n 1  [ K ]{D}n 1  {R ext }n 1

{D}0  [ M ]1 {R ext }0  [ K ]{D}0  [C ]{D}0 

Implicit: to implement

17
Exercise II
1. Atribua dimensões e material realistas à ponte de pedestres
abaixo e calcule suas frequências naturais e modos de vibrar.
Em seguida, adicione o carregamento indicado ao nó 1 e
calcule a amplitude da força para que o deslocamento do nó 2
seja L/10. Use os métodos implícitos e explícitos. Plote a
resposta do nó 2 no intervalo [0..4T ].

Função seno
+
T=metade do primeiro
período natural da estrutura

t
1

+
2
L
18