Modal Analysis Lecture Notes
Modal Analysis Lecture Notes
Brian J. Olson
Contents
1 The EOM in Physical Coordinates
2 Modal Transformation of the EOM
2.1 The Generalized Eigenvalue Problem .
2.2 Orthonormalization of Normal Modes
2.3 The Expansion Theorem . . . . . . . .
2.4 Modal Transformation . . . . . . . . .
1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
1
1
2
2
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3
3
3
3
4
4
4
5
.
.
.
.
4 Modal Solution
5
5
6
ME 461
Mechanical Vibrations
We begin with the matrix-vector differential equation describing undamped forced vibration of a linear
discrete N degree of freedom (DOF) system, that is,
M
x + Kx = f (t),
(1)
where M and K are the mass and stiffness matrices, respectively. The system solution and forcing vectors
are given by x(t) = (x1 , x2 , . . . , xN ) and f (t) = (f1 , f2 , . . . , fN ) . In general, the system is given some
initial displacements and velocities, which are described by the vectors x(0) = (x1 (0), x2 (0), . . . , xN (0))
and x(0)
= (x 1 (0), x 2 (0), . . . , x N (0)) .
2
2.1
In order to perform a modal analysis, it is first necessary to solve the generalized eigenvalue problem, which
results in the system natural frequencies i and mode shapes u(i) (i = 1, 2, . . . , N ). The eigenvalues, or
square of each natural frequency, can be obtained from the N th order characteristic polynomial in 2
det(K 2 M) = 0.
(2)
Each corresponding eigenvector, or mode shape, can be obtained by employing the equation
i = 1, 2, . . . , N
K i2 M u(i) = 0
2.2
(3)
It is sometimes convenient to work with normalized, or orthonormal mode shapes. In particular, we wish to
scale each mode shape such that
u
(i) M
u(i) = 1,
(4)
where u
(i) is the ith orthonormal mode. To accomplish this, we seek some constant i such that
u
(i) = i u(i) .
(5)
(i)
(i)
Mii
u
Mu
(6)
(7)
where Mii = u(i) Mu(i) , a scalar, is the ith generalized mass. Hence, the ith orthonormal mode is given by
r
1
(i)
u
=
u(i) .
(8)
Mii
ME 461
2.3
Mechanical Vibrations
It is noted that the component equations of motion represented by Equation (1) are, in general, coupled via
the mass matrix M (called dynamic coupling) and/or the stiffness matrix K (called static coupling). This
means that one cannot solve for the j th solution xj (t) without solving simultaneously for all of the other
solutions xi (t) (i = 1, 2, . . . , N 1). It is always possible, however, to find a set of modal coordinates qi (t)
(also called principle or generalized coordinates) so that the equations of motion are uncoupled. This is a
consequence of the Expansion Theorem which can be stated as follows.
2.1 THEOREM. (Expansion Theorem). Let {v(1) , v(2) , . . . , v(N ) } be a set of N linearly
independent vectors that are orthonormal with respect to some square matrix A. If x(t) is an
arbitrary vector in N -dimensional space, it can written as the linear combination
x(t) = c1 v(1) + c2 v(2) + . . . + cN v(N ) =
N
X
ci v(i) = Vc,
(9)
i=1
where c = [c1 , c2 , . . . , cN ] is the N -vector of coordinates of x(t) with respect to the basis
v(1) , v(2) , . . . , v(N ) . The vector c can be obtained by premultiplying Equation (9) by V A and
invoking orthonormality relations. This results in
c = V Ax.
2.4
(10)
Modal Transformation
Since the system mode shapes are linearly independent and orthogonal with respect to the mass matrix, the
Expansion Theorem guarantees that they can be used as a basis in N -dimensional space. To this end, the
solution vector x(t) of Equation (1) is expressed by a linear combination of the system mode shapes, each
multiplied be a time-dependent generalized coordinate, say, qi (t). That is,
x(t) = u(1) q1 (t) + u(2) q2 (t) + . . . + u(N ) qN (t) =
N
X
u(i) qi (t).
(11)
i=1
(12)
Equation (13) is a coordinate transformation from physical coordinates x(t) to modal coordinates q(t).
1 It should be noted that the equality on the RHS of Equation (13) is not a trivial step. Its proof requires invoking the
definitions of a transpose of a matrix and matrix multiplication.
ME 461
Mechanical Vibrations
3.1
Mapping Equation (13) into Equation (1), we arrive at the intermediate relationship
MU
q(t) + KUq(t) = f (t).
(14)
to obtain
U MU
q(t) + U KUq(t) = Q(t),
(15)
Q(t) = U f (t)
(16)
where
is the generalized force vector.
3.2
It can be shown that Equation (15) represents a set of N decoupled ordinary differential equations (ODEs)
in modal space. This important result follows directly from orthogonality of mode shapes, which is discussed
next in terms of both unnormalized and orthonormal (normalized) modes.
3.2.1
Unnormalized Modes
Given two unnormalized normal modes u(i) and u(j) , their orthogonality properties can be described by the
relationships
(
0
i 6= j
(i)
(j)
u
Mu
=
(17)
Mii i = j
(
0
i 6= j
u(i) Ku(j) =
(18)
2
Kii = Mii i i = j
where the constants Mii and Kii are called the generalized mass and stiffness coefficients, respectively.
Equation (17) and Equation (18) indicate that the modal vectors u(i) and u(j) are orthogonal with respect
to both the mass and stiffness matrices. Two important consequences of these orthogonality properties are
(2)
0
M22 . . .
0
Mu(1) u(2) Mu(2) . . . u(2) Mu(N )
u
=
U MU =
.
.
.. = diag(Mii )
..
..
..
.
..
.
.
.
.
.
.
.
.
.
.
.
u(N ) Mu(1)
u(N ) Mu(2)
...
u(N ) Mu(N )
...
MN N
(19)
and
u(1) Ku(1)
(2)
Ku(1)
u
U KU =
..
u(1) Ku(2)
u(2) Ku(2)
..
.
...
...
..
.
u(1) Ku(N )
K11
0
u(2) Ku(N )
=
.
..
..
.
0
K22
..
.
...
...
..
.
0
0
..
.
= diag(Kii ) (20)
0
0
. . . KN N
u(N ) Ku(2) . . . u(N ) Ku(N )
where U = u(1) , u(2) , . . . , u(N ) is the modal matrix and N is the order of the square matrices M and K.
u(N ) Ku(1)
ME 461
3.2.2
Mechanical Vibrations
If the mode shapes are normalized according to Equation (8), then their orthogonality properties can be
described by the relationships
(
0 i 6= j
(i)
(j)
u
M
u
=
(21)
1 i=j
(
0
i 6= j
(i)
(j)
(22)
u
K
u
=
i2 i = j
where i is the ith natural frequency. Equation (21) and Equation (22) indicate that the orthonormal modal
vectors u
(i) and u
(j) are orthogonal with respect to both the mass and stiffness matrices. Two important
consequences of these orthogonality properties are
u(N )
u(2) . . . u
(1) M
u(1) u
(1) M
u
(1) M
1 0 ... 0
(2)
M
u(1) u
(2) M
u(2) . . . u
(2) M
u(N )
u
0 1 . . . 0
MU
=
U
=
I = diag(1)
(23)
.
.
.
..
..
..
..
. . ... .
.
. .
.
.
.
.
u
(N ) M
u(1)
u
(N ) M
u(2)
u
(N ) M
u(N )
0 0
u(N )
u
(1) K
12
(2)
(N )
0
u
K
u
=
.
..
..
.
0
22
..
.
...
...
..
.
0
0
..
.
...
2
N
...
...
and
u(1)
u
(1) K
(2)
K
u(1)
u
KU
=
U
..
u(2)
u
(1) K
(2)
u
K
u(2)
..
.
...
...
..
.
0
u
(N ) K
u(2) . . . u
(N ) K
u(N )
(1) (2)
= u
where U
,u
,...,u
(N ) is the normalized modal matrix.
u
(N ) K
u(1)
3.3
= diag(i2 ) (24)
It is now shown that, due to orthogonality of normal modes, Equation (15) is actually decoupled.
3.3.1
Unnormalized Modes
In light of the structure of Equation (19) and Equation (20) the decoupled equations of motion follow from
Equation (15) and are given by
diag(Mii )
q(t) + diag(Kii )q(t) = Q(t).
(25)
Equation (25) represents a set of N decoupled, second-order, ordinary differential equations in the modal
coordinates qi (t). The ith decoupled equation is given by
Mii qi (t) + Kii qi (t) = Qi (t),
(26)
where Mii (generalized mass) and Kii (generalized stiffness) are the ith diagonal elements of U MU and
U KU, respectively. (See Equation 19 and Equation 20.) The modal forcing, or generalized force Qi (t) is
given by the ith element of Q(t) (see Equation 16), or it may be computed using2
(27)
ME 461
3.3.2
Mechanical Vibrations
in Equation (15) and EquaIf the mode shapes are normalized according to Equation (8) we let U = U
tion (16). Then in light of the structure of Equation (23) and Equation (24) the decoupled equations of
motion are given by
diag(1)
q(t) + diag(i2 )q(t) = Q(t),
(28)
or
I
q(t) + q(t) = Q(t),
(29)
where I is the identity matrix. Equation (29) represents a set of N decoupled, second-order, ordinary
differential equations in the modal coordinates qi (t). The ith decoupled equations is given by
qi (t) + i2 qi (t) = Qi (t),
(30)
Qi (t) = u
(i) f (t).
(31)
Modal Solution
Equation (26) and Equation (30) are recognized to be of the same form as those describing the dynamics
of an undamped single DOF harmonic oscillator. We simply employ the methods of Chapters 2-4 to solve
them. In particular,
qi (t) = qitransient + qiforced ,
(32)
The transient, or free vibration, is described by
qitransient = qi (0) cos i t +
qi (0)
sin i t.
i
(33)
q
Kii
is the ith natural frequency of vibration. The nature of qiforced depends on the nature of
where i = M
ii
the excitation (i.e., harmonic, periodic, arbitrary, etc.).
In order to specify the constants qi (0) and qi (0) in Equation (33) the initial conditions must be transformed
from physical coordinates to modal coordinates.
5.0.3
Unnormalized Modes
q(0) = U1 x(0)
x(0)
= Uq(0)
q(0)
= U1 x(0)
(34)
ME 461
Mechanical Vibrations
where
q(0)
q(0)
x(0)
x(0)
=
=
and
(x1 (0), x2 (0), . . . , xN (0))
(x 1 (0), x 2 (0), . . . , x N (0))
Note that this requires the determination of U1 , which can be computationally expensive.
5.0.4
x(0)
= Uq(0)
q(0)
= U
(35)
(36)
MUq(0)
= U
MU
q(0)
= U
(37)
(38)
But in light of orthogonality with respect to the mass matrix (namely, Equation 23), it follows that
q(0)
q(0)
Mx(0)
= U
Mx(0)
= U
(39)
(40)
Once the N modal solutions qi (t) have been obtained the physical solutions xi (t) are obtained by employing
the modal transformation given by Equation (13).
ME 461
Mechanical Vibrations
1
=
(i)
u
Mu(i)
1
.
Mii
h
i
u(1) , u(2) , . . . u(N )
h
i
= u
(1) , u
(2) , . . . u
(N )
5. (Forced Vibration Problems) Transform the forcing vector f (t) to obtain the modal force vector.
U f (t) (for unnormalized modes)
Q(t) =
f (t) (for orthonormal modes)
U
6. (Free Vibration Problems) Transform the initial conditions.
U1 x(0) (for unnormalized modes)
q(0) =
Mx(0) (for orthonormal modes)
U
q(0)
=
U1 x(0)
U Mx(0)
8. Obtain the solution in modal coordinates using the methods of Chapters 2-4. Each solution qi (t) in
qi (0)
sin i t.
i
Uq(t)