CHAPTER 10
FREE VIBRATION OF MDOF SYSTEMS System without Damping
The equation of motion of a two-DOF system in free vibration (no external force) is
mu + ku = 0
The displacements of masses are the solution with an initial condition
( )
0=
u u and u u =
( )
0If a two-DOF system is let to vibrate with an arbitrary initial displacement, the displacement of each mass will be as shown. They are not a simple harmonic as free vibration of SDF system and deflected shape changes with time as the ratio u u1/ 2 varies with time.
12 - 2
An undamped structure would undergo simple harmonic motion without change of deflected shape if free vibration is initiated by appropriate displacement shapes. It will maintain that initial shape and both floors will reach their maximum displacement at the same time.
For the first shape above, both floors move in the same direction.
For the second shape, the two floors move in the opposite directions.
The location where displacement remains zero is called “node.” These shapes are called “natural modes of vibration.” The number of modes increases as number of degrees of freedom increases.
For each mode, all floors move like a simple harmonic motion with the same frequency called “natural circular frequency of vibration” (ωn), or “natural cyclic frequency of vibration” ( fn) and the time to complete one cycle is called “natural period of vibration.”
2
n
n
T π
= ω 1
n n
f =T
The lowest natural frequency is known as the fundamental natural frequency, denoted by ω1. It corresponds to the longest natural period, which is called fundamental natural period (T1) of vibration.
12 - 4
Natural Vibration Frequencies and Modes
Displacement vector u
( )
t can be written as a product of mode shape vector φn and modal coordinate q tn( )
, which is a scalar quantity.( )
t =q tn( )
φnu
Modal coordinate varies with time whereas the mode shape vector does not. In free vibration, the modal coordinate varies like a simple harmonic with frequency equal to the natural frequency.
( )
cos sinn n n n n
q t = A ω t B+ ω t Therefore, u
( )
t =φn[
Ancosωnt B+ nsinωnt]
Substitute this in the equation of motion, we get
( )
2
n n n q tn
ω φ φ
− + =
m k 0
This equation can be satisfied by two ways:
(1) q tn
( )
=0 but this is not very useful so it is called trivial solution (2) −ω φn2m n +kφn =0 or kφn =ω φn2m nThis is called a matrix eigenvalue problem. The matrices k and m are known and we want to determine ωn2 and φn
2
n n n
φ ω φ− =
k m 0 k−ωn2mφn =0
Again, this equation can be satisfied by either φn =0 (trivial) or detk−ωn2m =0
This determinant is a polynomial of order N in terms of ωn2. This equation is called “characteristic equation or frequency equation.” It has N real and positive roots for ωn2 because matrices k and mare symmetric and positive definite.
The N roots for ωn2 are known as eigenvalues, or characteristic values, and they correspond to the natural frequencies of N modes of vibration.
For each value of ωn2, φn can be solved from the equation
2
n n
ω φ
− =
k m 0, but they are not unique because many vectors obtained from scaling a solution of φn will also satisfy the equation. The shape of this vector φn is called natural mode of vibration, or mode shape. It is also known as eigenvector corresponding to the eigenvalue
2
ωn.
12 - 6
Mode and Spectral Matrices
We can put many mode shapes together in a matrix to have them in compact form.
11 12 1
21 22 2
1 2
...
...
...
N N jn
N N NN
φ φ φ
φ φ φ
φ
φ φ φ
= =
Φ # # % #
The matrix Φ is called modal matrix for the eigenvalue problem.
The N eigenvalues ωn2 can be assembled into a diagonal matrix Ω2, which is known as the spectral matrix of the eigenvalue problem.
2 1
2
2 2
2 N
ω ω
ω
=
Ω %
From the eigenvalue problem in vector form
2
n n n
φ =ω φ
k m
It is possible to write them in matrix form
= 2
kΦ mΦΩ
Orthogonality of Modes
The natural modes corresponding to different natural frequencies can be shown to satisfy the following orthogonality conditions.
When ωn ≠ωr,
T 0
n r
φ φk = φnTmφr =0 It can be proved as follows: From
2
n n n
φ =ω φ
k m
Premultiply by φrT
2
T T
r n n r n
φ φk =ω φ mφ
Above is scalar so we can transpose the right hand side
2
T T
n r n n r
φ φk =ω φ mφ Suppose we start from the rth mode,
2
r r r
φ =ω φ
k m
premultiply by φnT, we get
2
T T
n r r n r
φ φk =ω φ mφ Subtract the two equations, we get
(
ωn2 −ω φr2)
nTmφr =0When ωn2 ≠ωr2, φnTmφr =0 and also φ φnTk r =0.
12 - 8
The orthogonality condition implies that the matrices
≡ T
K Φ kΦ M ≡ΦTmΦ are diagonal where the diagonal elements are
T
n n n
K =φ φk Mn =φnTmφn
Because k and m are positive definite so Kn and Mn are positive are related by
2
n n n
K =ω M
by Kn =φ φnTk n =φ ω φnT
(
n2m n)
=ω φn2(
nTmφn)
=ωn2MnInterpretation of Modal Orthogonality
Consider a structure vibrating in nth node with displacements
( ) ( )
n t =q tn φn
u
The corresponding accelerations are un
( )
t =q tn( )
φn and inertia forces are( )
fI n = −mun( )
t = −mφn nq t( )
Next consider displacements of the structures in its rth mode
( ) ( )
r t =q tr φr
u The work done by inertia force is
( )
fI Tn ur = −(
φnTmφr)
q t q tn( ) ( )
rwhich is zero because of modal orthogonality.
Another implication is that the work done by equivalent static forces associated with displacement in nth mode doing through the rth mode displacement is zero.
( )
fS n =kun( )
t =kφn nq t( )
The work done is
( )
fS Tn ur =(
φ φnTk r)
q t q tn( ) ( )
rwhich is zero because of modal orthogonality.
12 - 10
Normalization of Modes
The eigenvector multiplied by a scalar scaling factor is also the eigenvector corresponding to the same eigenvalue.
Therefore, only the shape of mode is essential. A mode can be scaled without changing the shape. This process is called normalization.
Sometimes the mode shape vector is scaled such that the largest element equal to unity.
For a building, mode is often scaled such that the DOF at the roof is unity.
Computer program usually normalizes the modes such that
T 1
n n n
M =φ mφ = or M =ΦTmΦ=I
where I is an identity matrix (diagonal with 1 along the main diagonal).
These modes are called a mass orthonormal set. This results in
2 2
T
n n n n n n
K =φ φk =ω M =ω or K =ΦTkΦ Ω= 2
12 - 12
12 - 14
12 - 16
12 - 18
12 - 20
12 - 22
Modal Expansion of Displacements
Any set of N independent vectors can be used as a basis for representing any other vectors of order N.
1 N
r r r
φ q
=
=
∑
=u Φq
where qr are scalar multipliers called modal coordinates or normal coordinates.
( )
1
T N T
n n r r
r
φ φ φ q
=
=
∑
mu m
Because of modal orthogonality, only the term r n= is nonzero.
( )
T T
n n n qn
φ mu= φ mφ Therefore,
T T
n n
n T
n n n
q M
φ φ
φ φ
= mu = mu
m
This modal expansion is employed to determine the free vibration response and response to excitation of MDF systems.
12 - 24
Free Vibration of Undamped MDF Systems
Given u
( )
0 and u( )
0 determine u( )
tUse modal expansion of initial conditions
( ) ( )
1
0 N n n 0
n
φ q
=
=
∑
u
( ) ( )
1
0 N n n 0
n
φ q
=
=
∑
u
Therefore,
( ) ( )
00
T n n
n
q M
= φ mu
( ) ( )
00
T n n
n
q M
=φ mu
Then, the free vibration response is
( ) ( )
1 N
n n n
t φ q t
=
=
∑
u
( ) ( ) ( )
1
0 cos 0 sin
N n
n n n n
n n
t φ q ω t q ω t
= ω
= +
∑
u
12 - 26
Free Vibration of Damped MDF Systems
If the damping matrix c of the system is such that C=ΦTcΦ is diagonal, we said that the system has classical damping.
The equations of motion can be uncoupled and solved by modal analysis.
If C=ΦTcΦ is not a diagonal matrix, the system is nonclassically damped and must be solved by numerical method and its eigenvalue will be complex numbers.
For classically damped system, the uncoupled equation is
n n n n n n 0
M q +C q +K q = where
T
n n n
C =φ φc The modal damping ratio is
2
n n
n n
C ζ M
= ω
The solution of q tn
( )
will be free vibration of a damped SDF system.2 2 0
n n n n n n
q + ζ ω q +ω q =
( ) ( ) ( )
0( )
00 cos sin
n nt n n n n
n n nD nD
nD
q q
q t e ζ ω q t ζ ω t
ω ω
ω
− +
= +
where
1 2
nD n n
ω =ω −ζ
12 - 28