Modeling in Dynamic Analysis

Escolha dos modelos matemáticos em Mecânica das

Estruturas
Modelos estruturais Tipo da análise
M1 - Treliça A1 – Estática
M2 - Estado plano de tensões A2 - Dinâmica

M3 - Estado plano de deformações Cinemática

M4 - Axissimétrico K1 - pequenos deslocamentos e
deformações
M5 - Viga de Timoshenko
K2 - grandes deslocamentos e
M6 - Viga de Bernoulli-Euler pequenas deformações
M7 - Reissner-Mindlin K3 - grandes deslocamentos e
deformações
M8 - Elasticidade tridimensional
Comportamento constitutivo
Carregamentos
C1 - Elástico, linear e isotrópico
Condições de contorno C2 - Elástico, linear e ortotrópico
•
•
•
PEF3302 – Mecânica das Estruturas I – Prof. Miguel Luiz Bucalem
 Modeling in Dynamic Analysis
Loads varing with time • “slowly” - quasi-static analysis
(resulting acceleration are “small”)
• “rapidly” - dynamic analysis
(acceleration should be accounted for
and significantly influence the
t s response)
f

f − t ρ u(t ) d´Alembert´s principle

B

t B
f
 RB = ∑ ∫
V m
H
( m )T
[f B(m)
− ρ (m) H
( m)
]
U dV ( m )
m

M = ∑∫
( m )T
ρ (m)
H H
(m)
dV ( m )
Vm
m
M U + K U = R
Linear Analysis
M U (t ) + K U (t ) = R(t )
If damping is taken into account
f − ρ u − k u
B
Velocity dependent damping forces

M U (t ) + C U (t ) + K U (t ) = R(t )
And initial conditions
U (0) = U 0
U (0) = U 0

• Direct Integration Methods

• Mode Superposition
 Apoio a PEF-2401 – slides selecionados de
apresentação preparada pelo Prof. João Cyro

5
Sistemas com vários graus de
liberdade
U(0 ) = U 0
 + KU = 0
MU com : 
U (0 ) = U
0

Solução : U = Û cos (ωt − θ )

Substituindo a solução na equação do movimento:

(K − ω Μ )Uˆ = 0
2

6
 Sistemas com vários graus de
liberdade - sistema não-amortecido
(K − ω Μ )Uˆ = 0
2

Problema de autovalores associado :

K − ω 2M = 0 ou

A − ω 2I = 0 com A = M −1K

7
 VGL: Exemplo 8

U2 U3

L
U1
EI = 80000 Nm 2
L L
L=2m
m = 50 kgm -1

8
 VGL - E 8 : Matriz de rigidez
6EI 6EI 8EI 2EI 2EI 8EI
L2 L2 L L L L
24EI 1 6EI 6EI
1 1 L3 L2 1 L2

1 1

a) b) c)

 24 6 L 6L 
EI  2
K = 3 6 L 8L 2
2L
L  
6 L 2 L2 8 L2 
9
 VGL - E 8 : Matriz de massa
3
22mL2 22mL2 8mL3 3
- 3mL - 3mL 8mL3
420 420 420 420 420 420
1
1 1 22mL2 1 22mL2
2 ×156 1 mL 420
+ 420
420
1 1

a) b) c)

 732 22 L 22 L 
mL  2
M= 22 L 8 L2
− 3L
420  
22 L − 3L2 8 L2 
10
 VGL - E 8 : Modo 1
− 3.633
 
ˆ 1 =  1.000 
ω1 = 32.1 rd/s ⇒ U
 1.000 
 

3,633

1
1

11
 VGL - E 8 : Modo 2
 0.000 
ˆ  
ω 2 = 151.4 rd/s ⇒ U 2 =  1.000 
− 1.000
 
1 1

1 1

12
 VGL - E 8 : Modo 3
− 0.109
 
ˆ 3 =  1.000 
ω3 = 326.8 rd/s ⇒ U
 1.000 
 
0,109

1
1

13
 Avaliação no ADINA de
frequências e modos de vibração

14
 VGL - E 8 : Matriz modal
− 3.633 0 − 0.109
 
Φ= 1 1 1 
 1 − 
Propriedades :  1 1 
2145 0 0
M * = Φ T MΦ =  0 21 0 ⇒ M ij* = φiT Mφ j = 0
 
 0 0 7 

2210650 0 0 
K * = Φ T KΦ =  0 480535 0  ⇒ K ij* = φiT Kφ j = 0
 
 0 0 751012
15
 MSM: Exemplo 9
U2 U3

U1 R 1(t)
L

L L

EI = 80000 Nm 2
L=2m
m = 50 kgm -1
R1 (t ) = 1000 sen(ω t ) com ω = ω1

16
 Direct Integration Methods

0 T t

• no transformation of coordinates
• equation is satisfied at discrete times positions
• solution is obtained at these time positions

march in time

T t
initial conditions

How we obtain the solution of the “next” time position is dependent of the
particular intergration scheme used
Central Difference Method
t - ∆t t t + ∆t

 1 t − ∆t t + ∆t
U = 2 ( U −2 U+ U)
t t
(a)
∆t
 1 t + ∆t t − ∆t
t
U= ( U− U) (b)
2∆t

Substituting (a) and (b) in the “equilibrium” equation at time t

M U + C U + K U = R
 tt t t

Then
 1 1  t + ∆t  2 t  1 1  t − ∆t
 2M+ U = R − K − 2 M  U −  2 M − C
t
C U
 ∆t 2∆t   ∆t   ∆t 2∆t 
 It is an Explicit Method
• Equilibrium is imposed at time t
• No factorization of the stiffness matrix is required

Since explicit methods requires small time steps they are efficient when one
can consider:
•lumped mass matrices
•neglet damping
Then  1  t + ∆t t
 M U = Rˆ
 ∆t 
2

t  2 t  1  t − ∆t
Rˆ = R −  K − 2 M  U −  2 M 
t
U
 ∆t   ∆t 
Resulting  ∆t2

t + ∆t
U i = Ri 
ˆ t

 mii 

Since triangularization of K is not required, it is not necessary to assemble the

stiffness matrix
K U =∑ K U =∑ F
t (i ) t t (i )

i i
 Implicit Method
• Newmark - Widely use scheme
t + ∆t
U = U + [(1 − δ ) U +δ U ]∆t
  
t + ∆t t t

  1 t  t + ∆t
 
U = U + U ∆t +  − α  U + α U  ∆t
t + ∆t t t

 2  
For example consider δ 1 1
= ,α = then
2 4
1/2
1/2
t + ∆t
U = U + [(1 − δ ) U +δ U ]∆t
  
t + ∆t t t
Constant-average
accelerations
v v0 α∆t

  1 t  t + ∆t
  2
U = U + U ∆t +  − α  U + α U  ∆t
t + ∆t t t

s s0 v0 ∆ t  2  
1/2α (∆ t)2
Integração direta no ADINA

21
 Sistemas com vários graus de
liberdade - sistema amortecido
U(0 ) = U 0
 + CU
MU  + KU = 0 com : 
U (0 ) = U
0

problema de autovalores
C qualquer ⇒ 
 no campo complexo

problema de autovalores
Ctipo Rayleigh = a0M + a1K ⇒ 
 do sistema amortecido

22
Método da Superposição Modal
U(0 ) = U 0
MU  + KU = R (t )
 + CU com : 
U (0 ) = U
0

n
Mudança de variáveis: U(t ) = Φ Y = ∑ φi Yi (t )
i =1

MΦ Y  + CΦ Y  + KΦ Y = R (t )
Φ T MΦ Y  + Φ T KΦ Y = Φ T R (t )
 + Φ T CΦ Y
 + K *Y = R * (t )
 + C*Y
M *Y
23
Método da Superposição Modal
Amortecimento do tipo Rayleigh:

C = a0M + a1K
*
C = Φ CΦ = Φ
T T
(a0M + a1K )Φ = a0M* + a1K *
C*i = a0M *i + a1K *i

C*i
ξi = *
2M i ω i

24
Método da Superposição Modal
para amortecimento do tipo Rayleigh
Problema de n graus de liberdade
=
n problemas com 1 grau de liberdade

 i + K *i Yi = R *i (t )
i + 2ξi M *i ωi Y
M *i Y
φiT MU 0  i0 = φiT MU 0
com : Yi 0 = e Y
M *i M *i

visto que : φiT MU 0 = φiT MΦ Y0 = φiT Mφi Yi 0

25
 MSM: Exemplo 9
U2 U3

U1 R 1(t)
L

L L

EI = 80000 Nm 2
L=2m
m = 50 kgm -1
R1 (t ) = 100 sen(ω t ) com ω = ω1

26
 MSM: Exemplo 9
- 3.61813247757214d0 1.8558971585539d-7 - 0.10877823357968d0
1.0d0 1.0d0 1.0d0
Φ= 1.00000258496357d0 - 1.00000012829729d0 0.99999997976364d0

2139.46 0 0 2205108 0 0
*
M* = 0 21.0 0 K = 0 480000 0
0 0 7.02 0 0 750626

ξ1 = 0 ,050
* 
C = 6879.94
0
0
1497.6
0
0 ⇒ ξ 2 = 0 , 235
0 0 2341.95 ξ = 0 ,511
 3
- 3618.13247757214d0
*
R = 0
1.8558971585539d-4
- 108.778233579682d0 27
 MSM: Exemplo 9
parcela permanente da resposta

Y1 (t ) = - 0.01638302582894d0 sin(32.1 t - 1.56812127993292d0)

Y2 (t ) = 4.02662053664706d-10 sin(32.1 t - 0.10429109406732d0)

Y3 (t ) = - 1.45584235043844d-4 sin(32.1 t - 0.10081750827976d0 )

U 1 (t ) Y1 (t )
   
U = U 2 (t ) = Φ Y2 (t )
U (t ) Y (t )
 3   3 
28
Stability and Accuracy
M U (t ) + C U (t ) + K U (t ) = R(t )
wi frequencie s
Kφ =w Mφ 
2

 φ 1 , φ 2 , ... , φ n node shapes

n
U (t ) = ∑ xi (t )φ i
i =1
and defining ri = φ i R
T

xi + 2ξ iω i xi + (ω i ) 2 xi = ri
mi = φ i M φ i
T
ki
ci = φ i C φ i
T
i
r i(t)
ki = φ i K φ i
T

ci
ki ci
wi = ; ξi =
mi 2mi wi
If all equation (i=1,2, ..., n) are integrated numerically with the same
∆t then the solution of the mode superposition approach is exactly
the same as integrating directly

M U (t ) + C U (t ) + K U (t ) = R (t )
with the same numerical scheme and with this time step ∆t.

Therefore the considerations of accurancy and stability can be

made in the uncoupled system.
 Accuracy
• How many modes we want to represent accurately?
• Depends on the frequency content of the load

Finite element discretization

n degrees of fredom
p is the number of modes to be represented accurately
In general: p << n
2π
The period of the mode T p =
pth
wp
T π
Accuracy ∆t = p =
20 10 w p
However for mode n
∆t ∆t T p ∆t T p
= = = 50
Tn Tn T p T p Tn can be
≈
1 can be ≈ 1000
20
No accurancy for modes → n. But the response on these higher modes need to be
bounded ⇒ Stability.
Analytically
ki u (0) = 0
i
r i(t) u (0) = 0
ci

In the computer
ki
i
u (0) ≠ 0
r i(t) (finite digit computations)
u (0) ≠ 0
ci

Unconditionally stable - the solution remains bounded for any ∆t (ri=0)

Conditionally stable - the solution remains bounded as long as ∆t ≤ ∆t cr (ri=0)

∆t cr - critical time step

Central Difference Method
Tn
∆tcr =
π

Newmark Method

Unconditionally stable as long as δ ≥ 0,5 α ≥ 0,25(δ + 0,5) 2

Note that for CDM ∆tcr 1

≤
Tn π
∆tcr T p ∆tcr Tn 1
= ≤
Tn T p Tp Tp π
≈ 1000

∆tcr 1 Tp
1 Tp (much smaller
1000 ≤ ∆tcr ≤ ∆tcr ≤
Tp π π 1000 3141,6 than Tp /20)
 xi + 2ξ iω i xi + (ω i ) 2 xi = ri
ω
β=
ω

ω
xi + 2ξ iω i xi + (ω i ) xi = ri
2 •
ω
grande, resposta bem pequena
ri = φ i R ω
T

• pequeno, resposta praticamente estática

ω
Qual é o conteúdo em freqüência de ri ?
Selecionar ∆t para representar p modos com precisão.
Modos de vibrar:

Frequências naturais:
 Frequência natural BE (4 elementos) Chapa (20 elementos) Valor analítico
1 3,9482 3,9560 3,9478
2 15,8433 15,9240 15,7914
3 36,1262 36,2169 35,5306
4 69,9223 65,3914 63,1655
5 109,5290 104,2780 98,6960
6 110,9720 108,1190 142,1223
7 175,1270 153,9630 193,4442
8 261,6290 215,5530 252,6619
9 317,7380 288,9910 319,7752
10 345,5230 324,3660 394,7842
11 627,6540 367,2850 477,6889
12 907,6420 411,3270 568,4892
13 - 540,7190 667,1853
... - ... ...
207 - 54 878,5000 16 9161,1000
Obs.: Valores de frequências em rad/s.