Revision I
By Tom Irvine
Email: tom@vibrationdata.com
Introduction
The effective modal mass provides a method for judging the significance of a vibration
mode.
Modes with relatively high effective masses can be readily excited by base excitation.
On the other hand, modes with low effective masses cannot be readily excited in this
manner.
Consider a modal transient or frequency response function analysis via the finite element
method. Also consider that the system is a multi-degree-of-freedom system. For brevity,
only a limited number of modes should be included in the analysis.
How many modes should be included in the analysis? Perhaps the number should be
enough so that the total effective modal mass of the model is at least 90% of the actual
mass.
Definitions
The equation definitions in this section are taken from Reference 1.
Consider a discrete dynamic system governed by the following equation
M x K x F (1)
where
1
function
A solution to the homogeneous form of equation (1) can be found in terms of eigenvalues
and eigenvectors. The eigenvectors represent vibration modes.
is given by
The systems generalized mass matrix m
T M
m (2)
Let r be the influence vector which represents the displacements of the masses resulting
from static application of a unit ground displacement. The influence vector induces a
rigid body motion in all modes.
L T M r (3)
The modal participation factor matrix i for mode i is
Li
i
ii
m
(4)
m eff , i
The effective modal mass for mode i is
Li 2
m eff , i
ii
m
(5)
m ii
Note that = 1 for each index if the eigenvectors have been normalized with respect
to the mass matrix.
m ,i j
Furthermore, the off-diagonal modal mass ( i j ) terms are zero regardless of the
normalization and even if the physical mass matrix M has distributed mass. This is due to
the orthogonality of the eigenvectors. The off-diagonal modal mass terms do not appear
in equation (5), however. An example for a system with distributed mass is shown in
Appendix F.
2
3
Example
x2
m2
k3
x1
k2 m1
y
k1
Figure 1.
Table 1. Parameters
Variable Value
m1 2.0 kg The homogeneous equation of motion is
m2 1.0 kg
m1 0 x1 k1 k 3 k3 x1 0
k1 1000 N/m 0 m x k x 0
2 2 3 k 2 k3 2
k2 2000 N/m (6)
k3 3000 N/m
The mass matrix is
2 0
M kg
0 1 (7)
The stiffness matrix is
4000 3000
K N/m
3000 5000 (8)
4
The eigenvalues and eigenvectors can be found using the method in Reference 2.
det K 2M 0
(9)
f 1 4.78 Hz (12)
f 2 12.4 Hz
(15)
0.6280 0.3251
0.4597 0.8881 (16)
The eigenvectors were previously normalized so that the generalized mass is the identity
matrix.
T M
m (17)
5
0.6280 0.4597 1.2560 0.6502
m 0.4597 0.8881
0.3251 0.8881 (19)
1 0
m
0 1 (20)
Again, r is the influence vector which represents the displacements of the masses
resulting from static application of a unit ground displacement. For this example, each
mass simply has the same static displacement as the ground displacement.
1
r
1 (21)
L T M r (22)
0.6280 0.4597 2 0 1
L 0 1 1
0.3251 0.8881 (23)
0.6280 0.4597 2
L
0.3251 0.8881 1 (24)
1.7157
L kg
0.2379 (25)
i
The modal participation factor for mode i is
6
Li
i
ii
m
(26)
1.7157
0.2379 (27)
The coefficient vector L and the modal participation vector are identical in this
example because the generalized mass matrix is the identity matrix.
m eff , i
The effective modal mass for mode i is
Li 2
m eff , i
ii
m
(28)
For mode 1,
m eff , 1
1.7157 kg 2
1 kg (29)
m eff , 1 2.944 kg
(30)
For mode 2,
m eff , 2
0.2379 kg 2
1 kg (31)
m eff , 2 0.056 kg
(32)
Note that
7
m eff , 1 m eff , 2 3 kg
(34)
Thus, the sum of the effective masses equals the total system mass.
Also, note that the first mode has a much higher effective mass than the second mode.
Thus, the first mode can be readily excited by base excitation. On the other hand, the
second mode is negligible in this sense.
From another viewpoint, the center of gravity of the first mode experiences a significant
translation when the first mode is excited.
On the other hand, the center of gravity of the second mode remains nearly stationary
when the second mode is excited.
Each degree-of-freedom in the previous example was a translation in the X-axis. This
characteristic simplified the effective modal mass calculation.
In general, a system will have at least one translation degree-of-freedom in each of three
orthogonal axes. Likewise, it will have at least one rotational degree-of-freedom about
each of three orthogonal axes. The effective modal mass calculation for a general system
is shown by the example in Appendix A. The example is from a real-world problem.
Aside
References
8
6. T. Irvine, Longitudinal Vibration of a Rod via the Finite Element Method, Revision B,
Vibrationdata, 2008.
9
APPENDIX A
x
m, J
z
0
kz1
kz2
kx1
kx2
ky2
ky1
kz3
kx3 ky3
The mass and inertia are represented at a point with the circle symbol. Each isolator is
modeled by three orthogonal DOF springs. The springs are mounted at each corner. The
10
y
springs are shown with an offset from the corners for clarity. The triangles indicate fixed
constraints. 0 indicates the origin.
x
a1 a2
C. G.
b 0
c1
c2
All dimensions are positive as long as the C.G. is inside the box. At least one
dimension will be negative otherwise.
11
12
The mass and stiffness matrices are shown in upper triangular form due to symmetry.
m 0 0 0 0 0
m 0 0 0 0
m 0 0 0
M
Jx 0 0
Jy 0
J z
(A-1)
4k x 0 0 0 2k x c1 c 2 4k x b
4k y 0 2k y c1 c 2 0 2k y a 1 a 2
4k z 4k z b 2k z a 1 a 2 0
4k z b 2 2k y c12 c 2 2 2k z a 1 a 2 b k y a1 a 2 c1 c 2
2k x c12 c 2 2 2k z a12 a 2 2 2k x c1 c 2 b
K =
4k x b 2 2k y a12 a 2 2
(A-2)
The equation of motion is
x x 0
y y 0
z z 0
M K
0
0
0
(A-3)
The variables , and represent rotations about the X, Y, and Z axes, respectively.
Example
A mass is mounted to a surface with four isolators. The system has the following properties.
M = 4.28 lbm
Jx = 44.9 lbm in^2
Jy = 39.9 lbm in^2
Jz = 18.8 lbm in^2
kx = 80 lbf/in
ky = 80 lbf/in
kz = 80 lbf/in
a1 = 6.18 in
a2 = -2.68 in
b = 3.85 in
c1 = 3. in
c2 = 3. in
14
Let r be the influence matrix which represents the displacements of the masses resulting from
static application of unit ground displacements and rotations. The influence matrix for this
example is the identity matrix provided that the C.G is the reference point.
1 0 0 0 0 0
0 1 0 0 0 0
0 0 1 0 0 0
r
0 0 0 1 0 0
0 0 0 0 1 0
0 0 0 0 0 1
(A-4)
L T M r (A-5)
The modal participation factor matrix i for mode i at dof j is
Lij
ij
ii
m
(A-6)
ii
m
Each coefficient is 1 if the eigenvectors have been normalized with respect to the mass
matrix.
m eff , i
The effective modal mass vector for mode i and dof j is
m eff , i j
Li j 2
ii
m
(A-7)
The natural frequency results for the sample problem are calculated using the program:
six_dof_iso.m.
15
16
six_dof_iso.m ver 1.2 March 31, 2005
Enter m (lbm)
4.28
Enter Jx (lbm in^2)
44.9
Enter Jy (lbm in^2)
39.9
Enter Jz (lbm in^2)
18.8
Enter kx (lbf/in)
80
Enter ky (lbf/in)
80
Enter kz (lbf/in)
80
Enter a1 (in)
6.18
Enter a2 (in)
-2.68
Enter b (in)
3.85
Enter c1 (in)
3
Enter c2 (in)
3
m =
0.0111 0 0 0 0 0
0 0.0111 0 0 0 0
0 0 0.0111 0 0 0
0 0 0 0.1163 0 0
0 0 0 0 0.1034 0
0 0 0 0 0 0.0487
k =
1.0e+004 *
0.0320 0 0 0 0 0.1232
0 0.0320 0 0 0 -0.1418
0 0 0.0320 -0.1232 0.1418 0
0 0 -0.1232 0.7623 -0.5458 0
0 0 0.1418 -0.5458 1.0140 0
0.1232 -0.1418 0 0 0 1.2003
Eigenvalues
lambda =
1.0e+005 *
The following definition is taken from Reference 3. Note that the mode shape functions are
unscaled. Hence, the participation factor is unscaled.
P
p ( x , t ) o p( x ) f ( t )
L (B-1)
1 L
L 0
i p( x ) i ( x ) dx
(B-2)
where
Let
Yn ( x ) = mass-normalized eigenvectors
m(x) = mass per length
L
n m( x ) Yn ( x )dx
0 (C-1)
2
L
0 m ( x ) Yn ( x ) dx
m eff , n
L
0 m(x ) Yn (x ) dx
2
(C-2)
L
0 m( x ) Yn ( x ) 2 dx 1
(C-3)
Thus,
2
L
m eff , n n 2 m( x ) Yn ( x )dx
0 (C-4)
APPENDIX D
2 EI 2 8
L
1 2 L
L2 2
2 EI
2 4 0 0
L2
2 EI 2 8
L
3 9 2 L
L2 3 9 2
2 EI
4 16 0 0
L2
2 EI 2 8
L
5 25 2 L
L2 5 25 2
2 EI
6 36 0 0
L2
2 EI 2 8
L
7 49 2 L
L2 7 49 2
95% of the total mass is accounted for using the first seven modes.
Table D-2. Bending Vibration, Fixed-Free Beam
Natural Participation Effective Modal
Mode
Frequency n
Factor Mass
2
1.87510 EI
L 0.6131 L
1 0.7830
L
2
4.69409 EI
L 0.1883 L
2 0.4339
L
2
5 EI
L 0.06474 L
3 2L 0.2544
2
7 EI
L 0.03306 L
4 2L 0.1818
90% of the total mass is accounted for using the first four modes.
APPENDIX E
E
c
(E-1)
93% of the total mass is accounted for by using the first three modes.
APPENDIX F
Consider an aluminum rod with 1 inch diameter and 48 inch length. The rod has fixed-free
boundary conditions.
A finite element model of the rod is shown in Figure F-1. It consists of four elements and five
nodes. Each element has an equal length.
E1 E2 E3 E4
N1 N2 N3 N4 N5
Figure F-1.
dU
0
dx x L
(Free end) (F-2)
The natural frequencies and modes are determined using the finite element method in
Reference 6.
The resulting eigenvalue problem for the constrained system has the following mass and
stiffness matrices as calculated via Matlab script: rod_FEA.m.
Mass =
0.0016 0.0004 0 0
0.0004 0.0016 0.0004 0
0 0.0004 0.0016 0.0004
0 0 0.0004 0.0008
Stiffness =
1.0e+006 *
1.3090 -0.6545 0 0
-0.6545 1.3090 -0.6545 0
0 -0.6545 1.3090 -0.6545
0 0 -0.6545 0.6545
n fn(Hz)
1 1029.9
2 3248.8
3 5901.6
4 8534.3
r =
1
1
1
1
L T M r (F-3)
where
L =
0.0867
0.0233
0.0086
-0.0021
The modal participation factor matrix i for mode i is
Li
i
ii
m
(F-4)
ii
m
Note that = 1 for each index since the eigenvectors have been previously normalized
with respect to the mass matrix.
Thus, for the sample problem,
iLi (F-5)
m eff , i
The effective modal mass for mode i is
Li 2
m eff , i
ii
m
(F-6)
Thus
m eff , i L i 2
(F-7)
m eff
=
0.0075
0.0005
0.0001
0.0000
The models total modal mass is 0.0081 lbf sec^2/in. This is equivalent to 3.14 lbm.
Thus, the four-element model accounts for 83% of the true mass. This percentage can be
increased by using a larger number of elements with corresponding shorter lengths.
APPENDIX G
L1 L2
k1 k2
x y
Figure G-1.
k 1 ( y - x - L1 ) k 2 ( y - x + L2 )
) )
Figure G-2.
F m x (G-1)
m x k1 ( y x L1) k 2 ( y x L 2 ) (G-2)
m x k1 ( y x L1) k 2 ( y x L 2 ) 0 (G-3)
M J (G-6)
J k1L1 k 2 L 2 x k1L12 k 2 L 2 2 k 1L1 k 2 L 2 y (G-10)
m 0 x k 1 k 2 k 1L1 k 2 L 2 x k 1 k 2
0 J k L k L k L k L y
1 1 2 2 k 1L12 k 2 L 2 2 1 1 2 2
(G-11)
k 1 k 2 k 1L1 k 2 L 2 x k 1 k 2
k L k L y
k 1L1 k 2 L 2 k 1L12 k 2 L 2 2 1 1 2 2 (G-12)
k 1 k 2 k 1L1 k 2 L 2 r1 k 1 k 2
r k L k L
k 1L1 k 2 L 2 k 1L12 k 2 L 2 2 2 1 1 2 2 (G-13)
r1 1
r 0
2 (G-14)
z=xy (G-15)
x=z+y (G-16)
m 0 z m y
0 J 0
k 1 k 2 k 1L1 k 2 L 2 z k 1 k 2 k 1L1 k 2 L 2 y
k L k L 0
k 1L1 k 2 L 2 k 1L12 k 2 L 2 2 1 1 2 2 k 1L12 k 2 L 2 2
k 1 k 2
y
k 1L1 k 2 L 2
(G-17)
m 0 z k 1 k 2 k 1L1 k 2 L 2 z m y
0 J k L k L k 1L12 k 2 L 2 2
0
1 1 2 2
(G-18)
m 0 z k 1 k 2 k 1L1 k 2 L 2 z m 0 r1
0 J k L k L 0 J r y
1 1 2 2 k 1L12 k 2 L 2 2 2
(G-19)
Solve for the eigenvalues and mass-normalized eigenvectors matrix using the
homogeneous problem form of equation (G-9).
z 1
2 (G-20)
1 k 1 k 2 k 1L1 k 2 L 2 1
m 0
0 J
m 0 r1
y
2 k 1L1 k 2 L 2 k 1L12 k 2 L 2 2 2 0 J r2
(G-21)
T
Then premultiply by the transpose of the eigenvector matrix .
T m0 0
J
1
2
T
k 1 k 2
k 1L1 k 2 L 2
k 1L1 k 2 L 2
k 1L12 k 2 L 2 2
1
2
T
m 0 r1
y
0 J r2
(G-22)
1 0 1 12
0 1
2 0
0 1
2
2 2
T m 0 r 1 y
0 J r
2
(G-23)
m 0 r 1
T
0 J r 2 (G-24)
m 0 1 m
T T
0 J 0 0 (G-25)
Example
Consider the system in Figure G-1. Assign the following values. The values are based on a
slender rod, aluminum, diameter =1 inch, total length=24 inch.
Natural Frequencies =
133.8 Hz
267.9 Hz
-4.4 1.029
0.1486 0.6352
Participation Factors =
0.2156
0.0504
The total modal mass is 0.0490 lbf sec^2/in, equivalent to18.9 lbm.
x
APPENDIX H
Repeat the example in Appendix G, but use the left end as the coordinate reference point.
L1 L2
x1 k1 k2
y
Figure H-1.
k 1 ( y - x1 ) k 2 ( y x1 + L)
Figure H-2.
The free-body diagram is given in Figure H-2. Again, the displacement and rotation are
referenced to the left end.
Sign Convention:
F m x (H-1)
m x k1 ( y x1 ) k 2 ( y x1 L) (H-2)
m x k1 ( y x1 ) k 2 ( y x1 L) 0 (H-3)
m x k1 k 2 x1 k 2 L (k1 k 2 ) y (H-
5)
x x1 L1 (H-6)
m x 1 L1 k1 k 2 x1 k 2 L ( k1 k 2 ) y (H-7)
m x 1 m L1 k1 k 2 x1 k 2 L (k1 k 2 ) y (H-8)
M1 J1 (H-9)
J 1 k 2 L ( y - x1 + L ) mL 1 x - x 1 (H-10)
J 1 mL 1 x - x 1 k 2 L ( y - x1 + L ) 0 (H-11)
J 1 mL 1 x - x 1 k 2 Ly - k 2 Lx 1 k 2 L2 0 (H-12)
J 1 mL 1 x - x 1 - k 2 Lx 1 k 2 L2 k 2 Ly (H-13)
x x1 L1 (H-14)
J 1 mL 1 x 1 L1 - x 1 - k 2 Lx 1 k 2 L2 k 2 Ly (H-15)
J 1 mL 1 x 1 - k 2 Lx 1 k 2 L2 k 2 Ly (H-16)
m mL 1 x 1 k 1 k 2 k 2 L x1 k 1 k 2
mL J1 k L
k 2 L2 k 2 L
y
1 2 (H-17)
Note that
2
J1 J mL 1
(H-18)
m mL 1 x 1 k 1 k 2 k 2 L x1 k 1 k 2
2 y
mL 1 J mL 1 k 2 L k 2 L2 k 2 L
(H-19)
k 1 k 2 k 2 L x1 k 1 k 2
y
k 2L k 2 L2 k 2 L
(H-20)
r 1 1
r 0
2 (H-22)
The natural frequencies are obtained via a Matlab script. The results are:
Natural Frequencies
No. f(Hz)
1. 133.79
2. 267.93
ModeShapes =
5.5889 4.0527
0.1486 0.6352
Participation Factors =
0.2155
-0.05039
0.04642
0.002539
The total modal mass is 0.0490 lbf sec^2/in, equivalent to18.9 lbm.