ASSUMED MODES METHOD FOR DISTRIBUTED PARAMETER SYSTEM

COURSE: Structural Dynamics (2CE611)

Dr S S Mishra ( Dept. of Civil Engineering) , NIT Patna

Assumed mode method is a procedure to discretize of a distributed parameter system. The

discrete model obtained from the assumed mode method is identical to the model obtained from
the Rayleigh-Ritz method.

Rayleigh-Ritz method.- is concerned with discretization of the differential eigen value .It is
mainly concerned with the free fresponse

Assumed mode method – discretisation of the boundary value problem implicitly. It also deals
with the discretised force response problem

Assumed method aims at deriving the equations of motion by first discretizing the kinetic
energy, potential energy, virtual work and then makes use of the Lagrange’s equations.

Let us consider a distributed parameter system and approximate the displacement u(x.t) by the
finite series
n

u ( x, t ) = ∑ψ ( x)q i(t )
i =1
i i (1)

Where ψi(x) are known as trial functions and may be admissible function.

qi(t) are unknown generalized coordinates ( i=1,2,..n)

Assuming that there is no lumped masses at the boundaries, we can discretize the kinetic energy
as follows.
L L n n

∑ψ ( x)q& (t )∑ψ
1 1
T (t ) =
2∫m( x) y& 2 ( x, t )dx =
0
2 ∫
m( x )
0 i =1
i i
i= j
j ( x)q& j (t )dx
 n n L n n

∑∑ ∑∑ m q& (t )q& (t )
1 1
=
2 i =1 i= j
∫
q& i (t )q& j (t ) m( x)ψ i ( x)ψ j ( x)dx =
0
2 i =1 i= j
ij i j (2)

Where mij = m ji =
∫ m( x)ψ ( x)ψ ( x)dx
0
i j i, j= 1, 2,…n (3)

The PE expression varies from system to system.

For example a non-uniform beam fixed at x=0 and supported at x= L on a spring of stiffness k
The PE is given by
L
1 1
V (t ) =
2 ∫
EI ( x)[ y ' ' ( x)]2 dx + ky 2 ( L, t )
0
2
(4)

L n n n n

∑ ∑ ∑ ∑ψ (L)q (t )
1 1
Then, V (t ) =
2
EI ( x)
∫ 0 i =1
ψ i ' ' ( x)qi (t )
j =1
ψ j ' ' ( x)q j (t )dx + k
2 i =1
ψ i ( L)qi (t )
j =1
j j

n n L 
∑∑
1
=
2 i =1 j =1  0 ∫
qi (t )q j (t )  EI ( x)[ψ i ' ' ( x)ψ j ' ' ( x)]dx + kψ i ( L)ψ j ( L)


n n

∑∑ ki q (t )q (t )
1
= j i j (5)
2 i =1 j =1

∫
Where , k ij = k ji = EI ( x)[ψ i ' ' ( x)ψ j ' ' ( x)]dx + kψ i ( L)ψ j ( L)
0
(6)

Finally let f(x,t) = distributed nonconservative force, the virtual work is

L L n

δWnc (t ) =
∫ 0
f ( x, t )δy ( x, t )dx = f ( x, t )
∫
0
∑ψ ( x)δq (t )dx
i =1
i i

= ∑Q i
inc (t )δqi (t ) (7)

Where the generalized nonconservative forces

 L

Qinc (t ) =
∫ f ( x, t )ψ ( x)dx
0
i i=1, 2,..n (8)

The Lagrange’s equation

d  δT  δT δV
  − + = Qk k=1,2,..n (9)
dt  δq& k  δq k δq k

We can write
n n

∑ ∑ k q( j )
δT δV
= mkj q& ( j ), = kj k=1, 2…n (10)
δq& k j =1
δq k j =1

δT
sin ce, = 0 (T does not depend on qk). Replacing the subscript k by I, we obtain the equation of motion
δqk

n n

∑ j =1
mij q ( j ) + ∑ k q( j) = Q (t )
j =1
ij i (11)

These resemble the equation of motion for a n dof system. The form of the mass and
stiffness matrices are same as in Rayleigh-Ritz method. Its another version of the Rayleigh- Ritz
method.

Choice of the Trial function:

Since the rate of convergence depends on the function ψi chosen.

1. ψi must form linearly independent set

2. It must be an admissible function

Admissible function: following are the requirements

(i) A function which satisfies all prescribed geometric boundary condition (BC).
(ii) It should possess derivatives upto the order appearing in the expression for V (PE).

BCs are of two types.

(i) Force or natural BCs.

(ii) Geometric or kinematic BCs
Expressions for PE and KE for an axial bar in terms of assumed modes.

The strain energy in the nununiform axial rod or bar for axial vibration u(x,t) is given as;
v ε L ε L
ε2
∫∫
V = [ (σ dε )]dv = [ Eε dε )]A( x)dx = E
0 0
∫∫ 0
∫
0
2
A( x)dx

L L 2 L
1 Eε 2 A( x)dx 1  ∂u  1
∫ ∫
EA( x)(u ') dx
∫
2
= = EA( x)  dx =
2
0
2  ∂x  2
0 0

L
1
EA( x)(u ') dx
∫
2
Hence V = (12
2
0

Now substituting the Eq(1) for

n

u ( x, t ) = ∑ψ ( x)q i(t )
i =1
i i

We have
n n

∑∑ k
1
V= ij qi q j (13)
2 i =1 j =1

∫
Where , k ij = k ji = EA( x)[ψ i ' ( x)ψ j ' ( x)]dx)
0
(14)

Or in matrix form

1 T
V= q kq
2
  k11 k12 . K1n 
k k 22 k 2 n 
k =  21 (15)
 . . . . 
 
k n1 kn2 . k n1 

Similarly, The KE is obtained as

L L
1 1
∫
ρ ( x)A(u& )2 dx = m(u& ) dx
∫
2
T= (16)
2 2
0 0

Where ρ is mass per unit volume and A = area of cross section.

n n n n

∑∑ ∑∑ m q& q&
1 1
T= ρAq& i (t )q& j (t ) = ij i j (17)
2 i =1 i= j
2 i =1 i= j

Where m i ij = ρAψ iψ j dx
∫ (18)

This mass matrix is called consistent mass matrix.

In matrix form

1
T = q& T mq& (19)
2

If a force is acting the generalized force is obtained through virtual work equation

(i) For a distributed force from Eq(8) repeated here:

L

Qinc (t ) =
∫ f ( x, t )ψ ( x)dx
0
i i=1, 2,..n (20)

(ii) A concentrated force F is acting, say at x =L then

δWnc = Fδu ( L, t ) = F ∑δu(L, t ) = F{ψ (L, )δq (t ) +ψ (L, )δq (t ) + ...ψ (L, )δq (t )}
i =1
1 1 2 2 n n

= F{ψ 1 ( L, )δq1 (t ) + ψ 2 ( L, )δq2 (t ) + ...ψ n ( L, )δq n (t )}

= {P1δq1 (t ) + P2δq2 (t ) + ...Pnδq n (t )}
We have n generalized forces corresponding to n trial function ( assumed modes). These are

P1 = F .ψ 1 ( L)

P2 = F .ψ 2 ( L)

. Pn = F .ψ n ( L)

Or in expression form Pi = F .ψ i ( L) (21)

Similarly if we have distributed damping c(x) it will be obtained same as Eq(20)

Then the damping force, f D ( x, t ) = −c( x)u& ( x, t ) (22)

Qinc (t ) =
∫ f ( x, t )ψ ( x)dx
0
i i=1, 2,..n

Writing FD for generalized damping

L L
 n 
FD (t ) =
∫
0
∫
f D ( x, t )ψ i ( x)dx = − c( x)

0 
∑j
ψ j ( x)q& j (t )ψ i ( x)dx


Or,

n L

FD = − ∑ q& (t )∫ c( x)ψ ( x)ψ ( x)dx

j
j
0
j i (23)

The the generalized damping coefficient is

L

∫
cij = c( x)ψ j ( x)ψ i ( x)dx
0
(24)

However, if we have a concentrated damper with damping coefficient cr at any location

r then ,

Generalized damping coefficient is given as

c= ∑ c ψ (r )ψ (r )
r
r i j (25)
Example:

Use assumed mode method with a polynomial approximation of u(x,t) to obtain a 2DOF
model for axial vibration of a uniform cantilever bar subjected to an end force F(t).

u(x,t) F

F(t)
x
L
L

Solution

Since we have to discretize the bar as a 2-DOF model:

2

Let u ( x, t ) = ∑ψ ( x)q (t )
i =1
i i

Select the shape (Trial) function ψi(x): we have to select two shape or trial function ψ1(x) and
ψ2(x):

Condition for an admissible function :

at least first derivative must be present since the expression for V contains first derivative. It
should also satisfy the BCs eg at x= 0 , ψi(0) =0 and also for ψ2(0):0

x
let us choose ψ 1 ( x) =
L
2
x
and ψ 2 ( x) =  
L

The chosen function satisfy both the above conditions

1
We have ψ '1 ( x) =
L
 2x
And ψ ' 2 ( x) =
L2

∫ AEψ ψ dx
' '
Let us compute the stiffness matrix elements k ij = i j
0

L
1 1 AE
for i=1 and j=1, k11 =
∫
0
AE . dx =
L L L

L
1 2x AE
for i=1 and j =2, k12 =
∫ AE L. L
0
2
dx =
L

AE
for i=2 and j =1, k12 = k 21 =
L
L
2x 2x 4 AE
for i=2 and j =2, k 22 =
∫ AE L . L
0
2 2
dx =
3L

Let us compute the mass matrix elements

∫
mij = ρAψ iψ j dx
0

L
x x ρAL
As before for i=1 and j=1,
∫
m11 = ρA . dx =
L L
0
3

L
x x2 ρAL
∫
for i=1 and j =2, m12 = ρA . 2 dx =
L L
0
4

ρAL
for i=2 and j =1, m12 = m21 =
4
L
x2 x2 ρAL
∫
for i=2 and j =2, m22 = ρA 2 . 2 dx =
L L
0
5

Let us determine the generalized forces :

L
We have generalized forces: P1 = F (t )ψ 1 ( L) = F (t ) = F (t )
L
 L2
P2 = F (t )ψ 2 ( L) = F (t ) = F (t )
L2

The Equation of Motion:

 m11 m12   q&&1   k11 k12   q1   P1 (t ) 

m  +  = 
 21 m22  q&&2  k 21 k 22  q 2   P2 (t )

Substituting the values we get the required 2-DOF model:

1 1
 1 1 
ρAL  3 4   q&&1  + AE  4   q1  =  F (t )
1   
1 q&&2  L 1    
   3   q 2   F (t )
4 5

Using the admissible function, We land in to generalized coordinate system q(t).

However using the Rayleigh –Ritz approach we directly form the eigen value problem.