(VIII).

MDOF Systems

(A). Direct Formulation of Motion Equation (Free Body Diagram)

A shear building

m1
p1 u1

k1
m2
p2 u2
k2
m3
p3 u3
k3

 m1 0 0  u1   k1 k1 0  u1   p1 

       
 0 m2 0  u2    k1 k1  k2 k2  u2    p2  mu  ku  p ------(1)
   
0 0
 m3  u3   0  k2 k2  k3  u3   p3 

8-1
(B). FEM Formulation

(i) Influence coefficient

Apply a unit displacement along DOF j, holding the displacement of all other DOFs to zero. The

stiffness influence coefficient kij is the force required along DOF i due to unit displacement at DOF

j. (造成只有 jth DOF 有一單位位移，其他 DOF 位移為 0，在 ith DOF 需要的力)

8-2
Dynamic equilibrium at any time instant

fI  fD  fS  p
   
Elastic force matrix

 k11 k12 k13  k1N 

 f S1  k11u1  k12u2    k1N uN  k  u1 
  ...
fS 2  
 ... 
  21  u 
fS      f S   k31 ...   2  -----(2)
     ...     
    ...  u 
 f SN  N 1  ... 
 k N 1 ...   N 

f S1 第一個自由度的 spring force

Denote kij as the elastic force developed at i due to a unit displacement at j. Then,

f S 1  k11u1  k12u2    k1N uN


f S  ku ------(2)
 

Similarly, denote cij as the damping force developed at i due to a unit velocity at j. Then,

f D 2  c21u1  c22u2    c2 N u N

f D  cu ------(3)
 

Finally, let mij be the inertia force developed at i due to a unit acceleration at j and have

f I 2  m21u1  m22u2    m2 N uN


f I  mu ------(4)
 

8-3
Example 1：Use concept of influence coefficient to determine kij of the shear building, where

12 EI
k1  k2  k3  2 
L3
【Solution】

 k11 k12 k13 

k   k21 k 22 k 23 
  
 k31 k32 k33 

u1=1,u2=0,u3=0
k1

k1

u1=0,u2=1,u3=0
k1
k1
k2

k2

u1=0,u2=0,u3=1

k2

k2
k3

k3

 k1  k1 0   1 1 0 
  12 EI  
  k1 k1  k 2 k2   2  3   1 2 1 
 0 L
  k2 k2  k3   0 1 2 
 

8-4
(ii) Stiffness matrix
u2 u5 u6
u1 u4

u3
Assumption: loads are located at nodes only.
First, consider the shape functions related with the axial direction.
y

x
u1 u4

Assume u ( x )  ax  b (i.e. assume EAu( x )  C )

 x  0, u1  b
B.C.: 
 x  L, u4  aL  b
u u x x
 u ( x)  4 1 x  u1  (1  )u1  ( )u4
L L L
x x
Define 1  1  4 
L L
u  1u1  4u4

Find 2 (x)

C B

2 (x)
u2=1

EIy "  M ( x)  Bx  C
B C
EIy  x 3  x 2  Dx  E ------(a)
6 2
 x  0, y (0)  1, y '(0)  0
B.C.:   (a)
 x  L, y ( L)  0, y '( L)  0
2 EI 3EI
 EIy  3 x3  2 x 2  EI
L L
x x
2  y  1  3( ) 2  2( )3 ------(2)
L L

8-5
 2
x x x x x
Similarly, change B.C., we can get 3  x(1  )2 5  3( ) 2  2( )3 6  (  1)
L L L L L
Finally, use shape function to determine kij .

Use the determination of k14 as an example.

k14=? 4 (x)
u4=1

d u1

Actual displacement field:

 u  u11  u44  0  1  1 4  4

u 4
 = = =4 '
x x

Apply a virtual displacement d u1 , then the virtual displacement filed:

 d u  d u11  d u44  d u11

d u d u11
d = = =d u11'
x x

4 
d WI    d dAdx =   E    d dAdx =   E  u ' d u ' dAdx =   E (d u1 1 )dAdx
x x
=  EAd u14 '1 ' dx

d WE  k14  d u1

d WE  d WI

 k14   EA4 '1 ' dx  EA 1 '4 ' dx

L
 kij  EA i ' j ' dx for i, j  1, 4 ------(3)
0

L 1 1  EA
Check k14  EA   dx 
0 L L L

8-6
Determination of k23

The force developed at 2 due to a unit rotation at 3, u3 =1 .

d u2
3 (x)
u3=1
k23=?

Actual displacement field: u  3u3  3 1  3

Apply a virtual displacement d u2

 23   22  2
d WI    d dAdx    E    d dAdx    E   y  d  y  dAdx    E d
 2 2  y dAdx
u
x 2  x 

  EId u23 2dx ------(d)

d WE  k23  d u2 ------(c)

6EI
(c)=(d) k23   EI3 2dx 
L2

kij   EIi j dx for i, j  2, 3, 5, 6

 AL2 AL2 
 0 0  0 0 
 2I 2I 
 6 3L 0 6 3L 
 2 L2 0 3L L2  2 EI
k 
  AL2  L3
 0 0 
 2I 
 sym. 6 3L 
 2 L2 


8-7
(iii) Mass matrix

Consistent mass matrix

Determine m14

 2u
u  4  x   u4  t   2  u  4  u4  4
t

f I  m  x  u x   m  x  4  x  u4  m  x  4  x 

Assume uniform m(x)

f I  m4  x 

Apply a virtual displacement d u1

d WE  m14d u1
 d WI   f I d udx   m4  x  d u11  x   dx

mL
m14   m14 dx 
6

 mij   m  x  i j dx

Therefore, in general,
L
 mij   m  x  i  j dx i, j  1 ~ 6
0

140 0 0 70 0 0 
 156 22 L 0 54 13L 

 4 L2 0 13L 3L2  mL
m  , if uniform m(x)
  140 0 0  420
 sym. 156 22 L 
 
 4 L2 

8-8
Example:
mL
Determine mass matrix m. Each element has a unit length, therefore  1 for the left element
 420
mL
and  2 for the right element.
420

m  420 m  840

2 5 8

4
1 7
3 6 9

1 2 3 4 5 6 4 5 6 7 8 9
140 0 0 70 0 0  280 0 0 140 0  0
 156 22 0 54 13   312 44 0 108 26 
   
 4 0 13 3   8 0 26 6 
m1    m2   
  140 0 0    280 0 0 
 156 22   312 44 
   
 4   8 

By superposition,

8-9
Lumped mass matrix

Assumption: the structure is lumped at the nodal coordinates where the translational displacement is
defined.
L Lx
mi   ( )m( x)dx
0 L
L x
mj   m( x)dx
0 L
m m
If m( x)  m , a constant, then mi  L , mj  L
2 2

mL/2 mL/2
mL/2

mL/2

 mL 
 2 
 
 mL 
 2 
 0 
m   
  mL 
 
 2 
 mL 
 2 
 
 0

8-10
Example:
Determine mass matrix m by the lumped-mass approach.


 210   420 
 210   420 
   
 0   0 
m1    m2   
  210    420 
 210   420 
   
 0  0

1 2 3 4 5 6 7 8 9
 210 
 210 
 
 0 
 
 630 
m1  m2   630 
   
 0 
 420 
 
 420 
 0 


4 5 6 7 8 9
 630 
 630 
 
 0 
m 
  420 
 420 
 
 0

Example:
Determine mass matrix m by the lumped-mass approach and neglect the axial deformation.

5 6 8 9
630 
 0 
m  
  420 
 
 0
8-11
Example:
Determine mass matrix m by the lumped-mass approach and consider the axial deformation. The

mass density for all members is the same, m .

 2 1 
 
 2 
 2 1 
 
2
m  mL  
  0 
 
 1 
 1 
 
 0

Example: A mass of M is placed at the node.

0 
 0 
 
 0 
m 
  M 
 M 
 
 0

Example: Assume the elements to be axially rigid.

3m 0 
m
  0 m  22

8-12
(iv) Direct Stiffness Method

The stiffness of the complete structure can be obtained by merely adding the element stiffness
coefficients properly.

5
y* 4
x
y u 2* 6
2
u3* q
1 u1*
3 x*

The relationship between the local coordinates and the global coordinates (denoted with an asterisk)
are as follows.

u1  u1* cos q  u2* sin q u2  u1* sin q  u2* cos q u3  u3*

Let   cos q   sin q , we have

 u1     0   u1* 
u      0 0   *
 2    u 2 
u3   0 0 1  u3* 
    *    u    B  u 
*

u 4     0  u 4 
u5   0   0 u5* 
     
u6   0 0 1  u6* 

i.e. u  Bu*
 
1
U  u T ku
2  
1
 u*T BT kBu *  u  Bu*, uT  u *T BT 
2          
1
 u*T k *u*
2  
 k *  BT kB
  
similarly
m*  BT mB
  

8-13
Example:

For element 12, q  270

   cos q  0

   sin q  1
B is obtained by

  0   0 1 0 
   0 0   
  1 0 0 0 
 0 0 1  0 0 1 
B  
    0  0 1 0 
 0   0  0 1 0 0
   
 0 0 1  0 0 1

6 0 3L 6 0 3L 
 
 AL2 AL2
0 0  0 
 2I 2I 
 2 L2 3L 0 L2 
2 EI
k *  BT kB  3  
   L  6 0 3 L 
 
 AL2
0 
 2I 
 2 
2L 


8-14
Example:
Determine the stiffness matrix of the shear building by the DSM.
2

3 1

1
5
6 4

2
8
9 7
3

6 0 3L 6 0 3L 
 
0 AL2 AL 2

0 0  0
 2I 2I 
3 L 0 2 L2 3L 0 L2 
 
 6 0 3L 66 0 3L  3L 6 0 3L 
 
2 EI AL2 2
AL AL 2
AL2
ka *  kb *  kc *  3  0  0 0  0 0  0 
   L  2I 2I 2I 2I 
3 L 0 L2 3L  3 L 0 2 L2  2 L2 3 L 0 L 2 
 
 6 0 3L 66 0 3L  3L 
 
 AL2 AL2 AL2 
0  0 0  0
 2I 2I 2I 
 3L 0 L2 3L  3L 0 2 L2  2 L2 


u2  u5  u8  0
If shear building, then
u3  u6  u9  0

 6 6 0   1 1 0 
2 EI  12 EI
k  3  6 12 6   3  1 2 1
 
 L L
 0 6 12   0 1 2 

8-15
Example:
Determine k and m of the following frame, if the axial deformation of all elements is neglected.
 

Element stiffness matrices for elements 21&34 and element 23 are:

6 0 3L 6 0 3L 
 
0 AL2 AL2
0 0  0 
 2I 2I 
3 L 0 2 L2 3L 0 2 
L 2 EI
For elements 21, q  270 ,  
 6 0 3L 6 0 3L  L3
 2 
 0  AL AL2
0 0 0 
 2I 2I 
3 L 0 L2 3L 0 2 
2L 

6 0 3L 6 0 3L 
 
0 AL2 AL2
0 0  0 
 2I 2I 
3 L 0 2 L2 3L 0 2 
L 2 EI
For elements 34, q  270 ,  
 6 0 3L 6 0 3L  L3
 2 
 0  AL AL2
0 0 0 
 2I 2I 
3 L 0 L2 3L 0 2 
2L 


8-16
  AL2 1.5 AL2 1.5 
  0 0   0 0 
 2 I 2 2I 2 
 1 1
 6 3L 0 6  3L 
2 2
 
2 L2  2 0 3L L2  2  2 EI
For elements 23, q  0 , 
 2
AL 1.5  L3
  0 0 
 2I 2 
 1 
 6 3L 
 2 
 2 L2  2 

from element 21, allocate local DOF of 1, 2, 3 to global DOF of 1, 2, 3

from element 34, allocate local DOF of 1, 2, 3 to global DOF of 4, 5, 6
from element 23, allocate local DOF of 1, 2, 3, 4, 5, 6 to global DOF of 1, 2, 3, 4, 5, 6

 3 AL2 3 AL2 
 6 0 3L  0 0 
 8I 8I 
 AL2 
 3  3L 0 3 3L 
2I
 
2 EI  4 L2  2 L2 0 3 L 2 L2 
k 3  
 L 3 AL2
 6 0 3L 
 8I 
 AL2 
 sym. 3 3L 
 2I 
 6L2 

If axial deformation is neglected, then

 3 AL2 3 AL2 
 f S1   8 I  6 3L 
8I
0 
u 
f    1
2 
  2 EI  6 L2 0 2 L   u3 
 u 2  u5  0  S 3
 3   u 
 fS 4  L 3 AL2
  6 3L   
4
 
 fS 6   8I   6 
 u
 sym. 2
6L 


 f S1  f S 4  6  6 3L  0 0  3L   u1 
  2 EI  3L
 u1  u4   f S 3   3  6 L2 2 L2  u3 
 f  L
 S6   3L 2 L2 6 L2  u6 

Hint:

8-17
 2 EI  3 AL2   3 AL2  
f S1  3 
 6  1 
u   u4  3Lu3  0u6 
L  8 I   8I  

2 EI  3 AL2   3 AL2  
fS 4  3    u1    6  u4  0u3  3Lu6 
L  8 I   8I  

f S1  f S 4  ...

2 3

 156 0 22 L 
 0 140 0 
 
mL  22 L 0 4 L2

ma *  BT ma B    m *c  m *a
    420     
 
 
 

140  3 0 0 70  3 0 0 
 0 156  3 22 L  6 0 54  3 13L  6 

mL  0 132 L 4 L2 12 0 13L  6 3L2  12 
mb *   
 420  210 0 0 140  3 0 0 
 0 162 78 L 0 156  3 22 L  6 
 
 0 78 L 36 L2 0 132 L 4 L2  12 

8-18
  576 0 22 L 210 0 0 
 0 608 132 L 0 162 78 L 

mL  22 L 132 L 52 L 36 L2 
2
0 78L
m   
 420  210 0 0 576 0 22 L 
 0 162 78L 0 608 132 L 
 
 0 78 L 36 L2 22 L 132 L 52 L2 

786 11L 11L 

mL 
 u2  u5  0
 m4x 4 
 m3x3 u1  u4
  11L 26 L2 18L2 
210
11L 18L2 26 L2 

Note: if lumped mass, then

3mL 3mL mL
mL 2 2 2 2mL
2mL
2
3mL 3mL 2mL
mL 2 2 mL 2mL
2 2

mL mL mL mL
2 2 2 2

If axially rigid, then

u2  u5  0
u1  u4

 4mL 0 0 
  0 0 0 
 0 0 0 

8-19
(v) Static condensation
Express the dependent (secondary) DOF in terms of the remaining independent (primary) DOF.
The relation between the dependent and the independent DOF is found by establishing the static
relation between them.
The motion equation in the last example is rearranged as
3 6
p
1

0 0 0   u3   6 L2 2 L2 3 L   u3   M 3  0 
mL  2 EI  2 
 0 0 0  u6   3 2L 6L 2
3L  u6    M 6  0 
210 L
0 0 840  u1   3L 3L 12   u1   P1  p 

We want to condense it to only the primary DOF “1” left.

0 0   us   kss ksp   us   ps  ...eq (1)

i.e. 0 m  u    k    
k pp  u p   p p  ...eq(2)
   pp    p    ps    

from eq(1), kss  us  ksp  u p  ps ------(a)
    

us  k ss 1 ( ps  ksp  u p )  kss 1  ps  kss 1  k sp  u p

         

Define T   kss 1  ksp 

 u s  k ss 1  ps  T  u p ------(1)
       

from eq(2), m pp  up  k ps  us  k pp  u p  p p ------(b)

      
(1)  (b)

 m pp  up  k ps  ( k ss 1  ps  Tu p )  k pp  u p  p p ------(2)
       

 m pp  up   k ps  T  k pp   u p  p p  k ps  k ss 1  ps
        

 m pp  up  kˆ pp  u p  pˆ p


8-20
Condensation of the motion equation
1
2 EI  6 L2 2 L2  L3  3 L  1  3 
T  3  2       
 L 2L 6 L2  2 EI  3 L  8 L  3 

2 EI 1  3 2 EI 19.5EI
kˆpp  3 3L 3L      3 12 
L 8L  3 L L3

pˆ p  p1  0  p1

m pp  4mL

19.5EI
Motion Eq.  4mL  u1  u1  p1
L3

p
1

8-21
Alternate Approach

Eq. (1) and (2) are rewritten in a matrix form.

0 0   us   I T   u s   k ss 1 ps 
0 m  u  
              1 
ˆ
   pp    p   0 k pp  u p   p p  k ps  kss ps 
     
Compare to the original equation

0 0   us   k ss ksp   us   ps 
0 m  u    k    
k pp  u p   p p 
   pp    p    ps    

 kss ksp   I T 
We use Gauss-Jordan elimination to convert    into a form of   ˆ   . Then T , kˆ pp ,
 k ps k pp  0 k pp   
  

and pˆ p are obtained automatically. An inverse in kˆ pp is not necessary!!

 

6 L2 2 L2 3L 
 6 L2 2 L2 3L   
  2 EI 16 L2 2 EI
k   2 L2 6 L2 3L  3   0 2L  3
L  3  L
 3L 3L 12   

 0 2 L 10.5

 2   1 1   3  2 EI 3 2 EI 
6 L 2 L2 3L  1 3 2L   1 0
8L   L3 0
8 L L3 
       
16 L2 2 EI 3  2 EI 3  2 EI 2 EI 3 2 EI 
 0 2L   0 1 
 0 1  0
 3  L3  8L  L3  8L  L3  L3 8 L L3 
       
39 0 39  0 0 39   0 39 2 EI 
 0 0  0 0
 4  4   4   4 L3 
 3 
1 0 8L 
 
 3 
 0 1
 8L 
 
0 0 39 2 EI 
 4 L3 

 3
 8L 
T   
Then we get  3
 8 L 
2 EI 39 19.5
kˆpp  3   3 EI
 L 4 L

8-22
Example:
Use static condensation to reduce the first coordinate.

u1 , f1

m1 m2 u2 , f 2
k k

kˆ
m̂ p̂
Condense to 

 m1 0   u1   2k  k   u1   f1 
 0 m  u    k 
k  u2   f 2 
 2 2 
Method 1:
k
kˆpp  k pp  k ps  kss 1k sp  k  (k )(2k ) -1 ( k ) 
2
f
pˆ p  p p  k ps  kss 1 ps  f 2  (k )(2k )-1 f1  1  f 2
2

Method 2:

 1  f 
 2k k   f1  1    1 
 2k k   f1   2 , 2k
,  k  ,  f1    
 k
 k   f 2   0   f2   k   f1
 2 2  0  f2 
 2   2 

1 k f
 T  , kˆpp  , pˆ p  1  f 2
 2 2 2

8-23
A note on reduced stiffness matrix kˆ pp

Theorem : kˆ pp  t kt
T

 
T 
in which t    
 I 

Actually whether ps is zero or not does not alter kˆ pp . Thus, the relationship of u s and u p ,
  
u s  Tu p  k ss 1 ps , can be simplified as us  Tu p , i.e. set ps  0 . Accordingly,
       

 us  T   kss 1  ps   kss 1  ps 
u      u p       tu p       tu p
  p   I    0    0  
 
 0 0   us   kss k sp   us   ps  0 
 0 m  u    k     
k u

       p
       p 
pp p ps pp p

0 0   k ss k sp   0 
tT    tup  t T  k   tu p  t T   
 0 m pp      ps k pp     p p 
   
 m pp up  t T kt  u p  p p
    
 m pp  up  kˆpp  u p  p p
    

8-24
In previous two examples

T  kss 1  ksp
  

 3 / 8L   6 L2 2 L2 3 L   3 / 8 L 
T  
t       3 / 8L   3
t kt  
T 3  2 EI  2
1 2L 6L 2  EI
3L   3 / 8L   19.5 3  kˆpp
 I     8 L 8 L  L3  L
  1   3L
 3L 12   1 

u1 , f1

m1 m2 u2 , f 2
k k

1  1 1
 1   2k k     k  k
t  2 kˆ pp   1  2  0 2  , as before
     2    k k     2    2
1  1 1

8-25
Express PE, KE, and the virtual work of the damping forces in terms of the independent coordinates

up .

T 
t      kˆpp  t T kt u  t up
 I       

1 T 1 T T 1 T
PE : Potential Energy = u ku  u p t  k  tu p  u p kˆ pp u p ------(i)
2    2     2  
1 T 1 T T 1 T
KE : Kinetic Energy = u mu  u p t mtu p  u p m
ˆ pp u p ------(ii)
2    2     2  
dD  d uT cu  d u pT t T ctu p  d u p T cˆ pp u p ------(iii)
       
From (i), (ii), (iii)

mˆ pp  up  cˆ pp  u p  kˆpp  u p  pˆ p
     
Hence, kˆ, mˆ , and cˆ, may be regarded, respectively, as the stiffness, mass, and damping matrices

of the structure corresponding to the independent coordinates. In other word, the same
transformation based on static condensation for the reduction of stiffness matrix is also used in
reducing the mass and damping matrices.
However, this process may produce error. We will discuss it later.

8-26
(vi) Loading Matrix

If a load is not placed on the nodes, use a consistent load or a lumped load.

Consistent load

p ( x, t )
6
2 5

1 4
3 L

For example
L L
p2 (t )d u2   p( x, t )  d udx   p( x, t )  d u22 (t )dx
0 0

Therefore
L
pi   p( x, t )  i (t )dx
0

Note: Pay attention to the direction of loading and the direction of DOFs.

Lumped load

Assume constant load, L=1

p (t ) p(t )
2 2

8-27
Example:
The floor of a frame is subjected to a load of p( x, t )  A  sin(ωt ) kN/m
Determine the consistent load matrix.
p ( x, t )  A  sin(t ) kM/ m
2 6
5
1 4
3
L 1

Consistent:
L
p1     A  sin(t )   1dx  0  p4
0

L  x x   A  sin(t )
p2     A  sin(t )   1  3( )2  2( )3  dx   p5
0
 L L  2
A
p3    sin t   p6
12

Lumped:
A
p2  sin t  p5
2

p1 = p4 = p3 = p6 =0

A A
s in  t s in  t
2 2

8-28
Example: (E9.2)

Solution:
Loading Matrix

Similarly
Stiffness Matrix

Mass Matrix

MOE:
8-29
Example: (E9.3)

Solution:
Stiffness Matrix

Mass Matrix

MOE:

Application of transformation matrix:

Results: the same as the previous example.

8-30