Numerical

Evaluation of
Dynamic
Response
Chapter 5

September 26, 2021 CC Chang / HKUST

 Time-Domain Analysis
e n
iv
g
mx (t)  cx (t)  kx(t)  p(t) pi

Loading p
p1 p pi+1
given 2

mx i  cx i  kx i  pi
time

Displacement x
Time stepping k
e
mx i 1  cx i 1  kx i 1  pi 1 xi se
x1
seek xi+1
x2
t1 t 2 t i tj time
 Time-Domain Analysis
• Interpolation of Excitation Method

• Central Difference Method

• Newmark’s Method

• Wilson- Method

• Commercial Software
 Interpolation of Excitation
pi
linear pi  
mx i  cx i  kx i  pi t i pi+1

pi p i
t i  t  t i 1
t i
Δpi
mx t  cx t  kx t  pi  τ
Δt i

mx i 1  cx i 1  kx i 1  pi 1 ti  ti 1
Interpolation of Excitation
pi
mx   cx   kx   pi  
t i
constant linear ramp

Solution of x(t) consists of 3 parts

- Step response due to pi
Pi

- Ramp response due to t i
- Free vibration response
 Interpolation of Excitation
• Displacement
x i 1  A  x i  B  x i  C  p i  D  p i 1
where

A  e  t ( sin D t  cos D t )
2
1 
1
B  e  t ( sin D t )
D
1  2  t 1  2
2
 2 

C  e [(  ) sin D t  (1  ) cos D t ]
k  t D t 1  2 t 

2
1 2  t 2  1 2 
D  1  e ( sin D t  cos D t )
k  t D t t 
 Interpolation of Excitation
• Velocity
x i 1  A   x i  B  x i  C  p i  D  p i 1
where

A   e  t ( sin D t )
2
1 

B  e  t (cos D t  sin D t )
2
1 
1  1  t   1 

C    e [(  ) sin D t  cos D t ]
k  t 1   2
t 1   2 t 


D 
1
1 e t (  sin D t  cos D t )
kt 1  2
Interpolation of Excitation
 Interpolation of Excitation
• Note:

– Linear system

– SDOF system

– Accuracy depends on the linear

interpolation of loading p
 Central Difference Method
• Finite difference x i 1
xi
x
mx i  cx i  kx i  p i x i 1 x i x i 1  x i 1
x  x i 1  x i 1
2t
  i 2 t
 x  2x i  x i 1
x i  i 1
 t 2
t t t
x i 1  2x i  x i 1 x i 1  x i 1
m c  kx i  pi
2 2t
( t )
m c m c 2m
[  ]x i 1  pi  [  ]x i 1  [ k  ]x i
( t ) 2 2t ( t ) 2 2t ( t ) 2
k̂ unknown P̂i
 Central Difference Method
m c m c 2m
[  ]x i 1  pi  [  ]x i 1  [ k  ]x i
( t ) 2 2t ( t ) 2 2 t ( t ) 2

find need
At t=0, given x 0 and x 0 To find x1 we need x0 and x-1
mx i  cx i  kx i  p i
p  cx 0  kx 0 x  x i 1  x i 1
x 0  0
m   i 2t
x  x1  x 1  x  2x i  x i 1
 0

2 t x i  i 1
 t 2
 x  2x 0  x 1
x 0  1
 t 2
( t ) 2
x 1  x 0  tx 0  x 0
2
 Central Difference Method
• Summary
Given m, c, k, p, x 0 and x 0
p0  cx 0  kx 0 ( t ) 2
x 0  x 1  x 0  tx 0  x 0
m 2
m c m c 2m
k̂   a  bk
( t ) 2 2 t ( t ) 2 2 t ( t ) 2
For step i
p̂i  pi  a  x i 1  b  x i
p̂
x i 1  i
k̂
x  x i 1 x  2 x i  x i 1
x i  i 1 x i  i 1
2 t t 2
 Time-Domain Analysis
e n
iv
g
mx (t)  cx (t)  kx(t)  p(t) pi

Loading p
p1 p pi+1
given 2

mx i  cx i  kx i  pi
time

Displacement x
Time stepping k
e
mx i 1  cx i 1  kx i 1  pi 1 xi se
x1
seek xi+1
x2
t1 t 2 t i tj time
 Newmark’s Method
Given mx i  cx i  kx i  pi
Seek mx i1  cx i1  kx i1  pi1
Too many unknowns ! Need assumptions.
x i 1 x i 1

( x i  x i 1 ) ( x i 1  x i )

x i x   x i x   x i 
2 t
x i 1  t
x i 1  x i  ( x i  x i 1 )
t 2
x i x i  ( x i  x i 1 ) x i
2

x i 1  x i  x i t x i 1  x i  x i t

t 2 t 2 t 2
 ( x i  x i 1 )  x i  x i 1
xi xi 3 6
4

 
 Newmark’s Method
Given mx i  cx i  kx i  pi
Seek mx i1  cx i1  kx i1  pi1

x i 1  x i  [(1   ) t ]x i  ( t )x i 1


 1 2 2
x
 i 1  x i   tx i  [(   ) t ]x
 i  (   t )x i 1
2
 ,  are Newmark’s parameters
 Newmark’s Method
1 1
• Average acceleration  , 
2 4
x i 1
( x i  x i 1 )
x  
2 ( x i  x i 1 )
x i x  
2
 x i 1  x i 
t x i 1 
( x i  x i 1 )
 2 x i 
t
( x i  x i 1 )
 x i 2

 t 2
 x i 1  x i  x i t  ( x i  x i 1 ) x i 1  x i  x i t
4
t 2
xi  ( x i  x i 1 )
4
 Newmark’s Method
1 1
• Linear acceleration  , 
2 6
x i 1
( x i 1  x i )
x   x i 
t ( x i 1  x i )
x i x   x i 
t
t
x
 i 1
  x
 i  ( x i  x i 1 ) t
 2 x i 1  x i  ( x i  x i 1 )
 2
 x i
 t 2 t 2
 x i 1  x i  x i t  x i  x i 1 x i 1  x i  x i t
3 6
t 2 t 2
 x i  x i 1
xi 3 6
 Newmark’s Method
Incremental form
(1)
 mx i  cx i  kx i  pi
 mx  cx
 i 1 i 1  kx i 1  pi 1 (2)
 x  x  [(1   )t ]x  ( t )x
 i 1 i i i 1 (3)

 i 1 i
x  x   tx i  [(
1
  )  t 2 
]x i  (   t 2 
) x i 1 (4)
2
(2) - (1) gives mx i  cx i  kx i  pi (5)
(3) gives x i  x i 1  x i  tx i  ( t )x i (6)
1 2 2 
(4) gives x i  x i 1  x i  t x
 i   t x
 i  ( t )x i (7)
2
 Newmark’s Method mx i  cx i  kx i  pi (5)
x i  tx i  ( t ) x i (6)
From (6) and (7) x i  tx i 
1 2
t x i  (t 2 ) x i (7)
2
1 1 1
x i  x i  x i  x i (8)
 t 2 t 2
  
x i  x i  x i  t (1  ) x i (9)
 t  2
(8) & (9) into (5)
k̂x i  p̂i
 1
k̂  k  c m
 t 2
 t
m c m 
p̂i  pi  (  ) x i  [  t (  1)c]x i

 t  2 2
 Newmark’s Method
- Summary
Given m, c, k, p, x 0 and x 0
1 1 1
Assume  and  (   2 , 6    4 ) t

1) Initialize 2) For each time step i ( i = 0,1,2,…)

p  cx 0  kx 0
x 0  0 p̂i   pi 1  pi   ax i  bx i
m
 1
k̂  k  c m x i  p̂i / k̂
t 2
 t
m c 1 1 1
a  x i  x i  x i  x i
t   t 2  t 2
m    
b  t (  1)c x i  x i  x i  t (1  ) x i
2 2  t  2
 Newmark’s Method
- Example
• Case study:
– Given:
• Physical properties of an SDF system:
2
m  0.253 kip  sec / in, k  10 kips / in,
T  1 sec (   6.28 rad / sec),   0.05
• Half-cycle sine pulse force (shown right)
p, kips 10 sin(t / 0.6)
10

t, sec
0.6
 Newmark’s Method
- Example
1 1
• Average acceleration using t  0.1 sec  , 
2 4
1.0 Initial calculations:
p  cx 0  kx 0
m  0.253 k  10 c  0.159 x 0  0
m
u 0  0 u 0  0 p0  0  1
k̂  k  c m
t 2
p  cx 0  kx 0 t
1.1 x 0  0 0 m c
m a 
t 
1.2 t  0.1
m 
2 4 b  t (  1)c
1.3 k̂  k  c m  114 .5 2 2
t 2
( t )
4
1.4 a m  2c  10.45; and b  2m  0.507.
t
 Newmark’s Method
- Example
• Average acceleration using t  0.1 sec
2.0 calculations for each time step
2.1 p̂i  pi  ax i  bx i  pi  10.45x i  0.507x i

p̂ i p̂ i
2.2 x i  
k̂ 114 .5
2
2.3 x i  x i  2 x i  20x i  2 x i
t
4
2.4 x i  ( x i  tx i )  2x i  400( x i  0.1x i )  2x i
2
( t )

2.5 x i 1  x i  x i , x i 1  x i  x i , x i 1  x i  x i

 Newmark’s Method
- Example
• Average acceleration using t  0.1 sec
Results
 Newmark’s Method
- Example
1 1
• Linear acceleration using t  0.1 sec  , 
2 6
1.0 Initial calculations:
p  cx 0  kx 0
m  0.253 k  10 c  0.159 x 0  0
m
x 0  0 x 0  0 p0  0  1
k̂  k  c m
p  cx 0  kx 0 t t 2
1.1 x 0  0 0
m m c
a 
t  0.1 t 
1.2
m 
b  t (  1)c
3 6 2 2
1.3 k̂  k  c m  166.8
t 2
( t )
6 t
1.4 a m  3c  15.68; and b  3m  c  0.768.
t 2
 Newmark’s Method
- Example
• Linear acceleration using t  0.1 sec
2.0 calculations for each time step

2.1 p̂i  pi  ax i  bx i  pi  15.68x i  0.7679x i

p̂i p̂i
2.2 x i  
k̂ 166.8
3 t
2.3 x i  x i  3x i  x i  30x i  3x i  0.05x i .
t 2
6
2.4 x i  ( x i  tx i )  3x i  600( x i  0.1x i )  3x i
2
( t )
2.5 x i 1  x i  x i , x i 1  x i  x i , x i 1  x i  x i
 Newmark’s Method
- Example
• Linear acceleration using t  0.1 sec
Results
Newmark’s Method - Stability
To keep the algorithm stable
t 1 1

T  2   2
1 1
Average acceleration  , 
2 4
Unconditional stable
1 1
Linear acceleration  , 
2 6
t
Conditional stable  0.55
T
Usually for accuracy t  0.1T or less!
 Newmark’s Method
for nonlinear systems
EOM of nonlinear system
f Linear
mx ( t )  cx ( t )  f ( x )  p( t )
( k i )T
mx i ( t )  cx i ( t )  f ( x i )  p i f i 1
f i
fi
mx i 1 ( t )  cx i 1 ( t )  f ( x i 1 )  p i 1 x i
Nonlinear
mx i  cx i  f i  pi
f i  f ( x i 1 )  f ( x i ) xi x i 1 x
f i
  x i  ( k i ) T x i
x i
Tangential stiffness
function of displacement
 Newmark’s Method
for nonlinear systems
mx i ( t )  cx i ( t )  (k i ) T x i  p i
Newmark’s Scheme
x i x i 1, x i 1, x i 1
Correction (minimize the unbalanced force)
Original mx i ( t )  cx i  f i  pi
Approximate mx i ( t )  cx i  ( k i )T x i  pi
Unbalanced f i  ( k i )T x i  R
( j) R ( j)
x i 
( k i )T Newton-Raphson Iteration
x i t  x i  x i (1)  x i ( 2)  
 Newmark’s Method
for nonlinear systems
Given m, c, k, p, x 0 and x 0
1 1 1
Assume  and  (   2 , 6    4 ) t

1) Initialize
p  cx 0  (fs )0
x 0  0
m
m c
a 
 t 
m 
b  t (  1)c
2 2
 Newmark’s Method
for nonlinear systems R (i )  p̂i

2) For each time step i ( i = 0,1,2,…) x i 1(0)  x i

 1
p̂i  pi  ax i  bx i k̂ i  k i  c m
 t 2
 t
R ( j) j=1, 2, ...
x ( j)
 x i 1( j)  x i 1( j1)  x ( j)
k̂ T

f ( j)  fs( j)  fs( j1)  ( k̂ T  k T )x ( j) R ( j1)  R ( j)  f ( j)

 1
k̂ T  k T  c m (Newton-Raphson iteration)
 t 2
 t
1 1 1
x i  x i  x i  x i
2  t 2
 t
  
x i  x i  x i  t (1  ) x i
 t  2
 Newmark’s Method
for nonlinear systems
• Case study
– Given:
• Physical properties of an SDOF system:
m  0.253 kip  sec 2 / in,   0.05
• Force-deformation relation is elastoplastic with x y  0.75 in

• Half-cycle sine pulse force

f, kips b
a
7.5 p, kips 10 sin( t / 0.6)
e 10
x, in.
o
0.75
10 10 t, sec
1 1 0.6
d c
 Newmark’s Method
for nonlinear systems
• Average acceleration without iteration using t  0.1 sec
1.0 Initial calculations:
m  0.253 k  10 c  0.159
x 0  0 x 0  0 p0  0

p0  cx 0  kx 0
1.1 x 0  0
m
1.2 t  0.1
4
1.3 a m  2c  10.45; and b  2 m  0.507.
t
 Newmark’s Method
for nonlinear systems
• Average acceleration without iteration using t  0.1 sec
2.0 calculations for each time step
2.1 p̂i  pi  ax i  bx i  pi  10.45x i  0.507x i

2.2 ki  k for branches oa, bc and de; and ki  0 for ab and cd

2 4
2.3 k̂ i  k i  c  m  k i  104.5
t 2
( t )
p̂i
2.4 x i 
k̂
2
2.5 x i  x i  2 x i  20x i  2x i
t
2.6 x i 1  x i  x i , x i 1  x i  x i

2.7 (fs )i 1  (fs )i  k i x i .

p  cx i 1  (fs )i 1
2.8 x i 1  i 1 .
m
 Newmark’s Method
for nonlinear systems
• Average acceleration without iteration using t  0.1 sec
Results
 Newmark’s Method
for nonlinear systems
• Newton-Raphson iteration using t  0.1 sec
1.0 Initial calculations:
m  0.253 k  10 c  0.159
x 0  0 x 0  0 p0  0

p  cx 0  kx 0
1.1 x 0  0 0
m
1.2 t  0.1
4
1.3 a m  2c  10.45; and b  2m  0.507.
t
 Newmark’s Method
for nonlinear systems
• Newton-Raphson iteration using t  0.1 sec
2.0 calculations for each time step
2.1 p̂i  pi  ax i  bx i  pi  10.45x i  0.507x i

2.2 ki  k for branches oa, bc and de; and k i  0 for ab and cd

2 4
2.3 k̂ i  k i  c
2
m  k i  104.5
t ( t )
( j)
R
2.4 x ( j)  x i 1( j)  x i 1( j1)  x ( j)
k̂ T
f ( j)  fs( j)  fs( j1)  ( k̂ T  k T ) x ( j) R ( j1)  R ( j)  f ( j)
 1
k̂ T  k T  c m
t 2
2 t
2.5 x i  x i  2 x i  20x i  2 x i
t
4 4
2.6 x i  x i  x i  2x i  400x i  40 x i  2x i
2 t
( t )
2.7 x i 1  x i  x i ,
 Newmark’s Method
for nonlinear systems

No iteration
• Linear acceleration using t  0.1 sec
Results
 Wilson- Method
Modification of the Newmark’s linear acceleration method
1 1
Incremental form  , 
2 6
mx i  cx i  kx i  pi
x i  tx i  ( t ) x i x
1 2
x i  tx i  t x i  (t 2 ) x i
2 x i 1
x i
Linear acceleration and force x i
between ti and ti+ x i
mx i  cx i  kx i  pi t
1
x i  tx i tx i ti t i 1 ti time
2
1 1 t  t
x i  tx i  t 2 x i  t 2 x i
2 6
 Wilson- Method
mx i  cx i  kx i  pi
p̂i
x i 
k̂
where
3 6
k̂  k  c m
 t   t  2
6 t 
p̂i   pi 1  pi    m  3c  x i   3m  c  x i
 t   2 
then
3 t
x i  x i  3x i  x i
t 2
6 6 1
x i 
 x i  x i  3x i
   x
 i  x i
t 2 t 
 Wilson- Method
- Summary
Given m, c, k, p, x 0 and x 0 x i  p̂i / k̂
Assume t and     1 normally   1.4
1) Initialize 2) For each time step i ( i = 0,1,2,…)
p  cx 0  kx 0 p̂i   pi 1  pi   ax i  bx i
x 0  0
m p̂i
x i 
3 6 k̂
k̂  k  c m
 t   t  2 3 t
x i  x i  3x i  x i
6 t 2
a m  3c 6 6 1
 t x i  x i  x i  3x i x i  x i
1 t 2 t 
b  3m  tc
2 1
x i  tx i  tx i
2
1 1
x i  tx i  t 2 x i  t 2 x i
2 6
 Commercial Software - MATLAB
Demonstration - SDOF structure under earthquake excitation
1940 El Centro EQ
4
2

2
ground acc. m/s

0 m
-2

-4

k
0 5 10 15 20
time s

c
75-story building
m  4.61  107 kg
c  1.04  106 N - s/m (1%)
x g
k  5.83  107 N/m
 Commercial Software - MATLAB
Response of a tall building under earthquake
4
clear all;

2
2

ground acc. m/s

load elcen; 0

-2

m=4.61*10^7; c=1.04*10^6; k=5.83*10^7; -4

0 5 10 15 20

A=[0 1; -k/m -c/m]; B=[0;-1]; C=[1 0]; D=[0]; 0.4

0.2

[x temp]=lsim(A,B,C,D,acc,time);

disp. m
0

-0.2
subplot(2,1,1); -0.4
plot(time,acc); 0 5 10
time s
15 20

ylabel('ground acc. m/s^2');

title('Response of a tall building under earthquake');
subplot(2,1,2);
plot(time,x);
xlabel('time');
ylabel('disp. m');
 Commercial Software - MATLAB
The lsim command operates in the STATE-SPACE domain

Original 2nd order mx  cx  kx  p

State-Space 1st order mx  cx  kx  p

 x  x
x   0 1  x   0 
     k  c      1 p
x   m m  x
  
m 

x 
output y  1 0     0 p
x 
Commercial Software - MATLAB
  AX  BU
X x 
X 
Y  CX  DU x 
 0 1   0 
A k c B 1 
   m 
 m m 
C and D depend on the output desired.

[x temp]=lsim(A,B,C,D,-acc,time); Demonstration !
Summary mx (t)  cx (t)  kx(t)  p(t)
• 5 methods are introduced
– Interpolation of Excitation Method
– Central Difference Method
– Newmark’s Method
– Wilson- Method
– Commercial Software
• Accuracy
• Extension to multiple degrees and/or nonlinear
structures
– Newmark’s Method
– Wilson- Method
– Commercial Software