Professional Documents
Culture Documents
MODAL VECTORS
Lars Aigner
Volvo Car Corporation
Complete Vehicle
Dept: 98255, PV4A2
S-405 08 Gothenburg
Sweden
Very often the selected points for a set of modal vectors are 1. INTRODUCTION
evenly spread over the structure and they represent nearly
the same amount of mass. Hence the diagonal elements of The modal assurance criteria (MAC) was first presented in
the mass matrix are nearly equal and the MAC value will be 1980 [1] by Allemang and Brown. In general it measures the
low for two orthogonal modal vectors. degree of proportion between two modal vectors in the form
of a correlation coefficient. MAC is defined in equation 1.
However how sensitive is the MAC value between two ortho-
gonal modal vectors if the diagonal elements in the mass ma-
trix are unequal?
1320
Therefore if the mass (or stiffness) matrix is introduced as a 2.1 Mass matrix with equal masses
weighting matrix into the definition of the MAC value as in
equation 3, the MAC value will be zero. Assume that all masses have the same weight: M1, ..... ,Ms =
2kg. The stiffness of the springs are also the same: K1, ..... ,Ks
= 1 N/m. This will yield the following mass and stiffness matri-
ces;
2 0 0 0 0 2 -1 0 0 0
If the measured or calculated points represent equal amount 0 2 0 0 0 -1 2 -1 0 0
of mass the mass matrix will be proportional to an identity ma- [M] [K]= 0 -1
0 0 2 0 0 2 -1 0
trix and the definition of MAC with the weighting matrix will be
identical with the correct definition given in equation 1. Hence 0 0 0 2 0 0 0 -1 2 -1
for two orthogonal modal vectors the MAC value will be zero. 0 0 0 0 2 0 0 0 -1 2
However if the points do not represent equal amount of mass
the mass matrix will not be proportional to an identity matrix
and the MAC value will not be zero even if the two modal vec- The result from calculation of MAC matrix and modal mass
tors are orthogonal. matrix is given below;
Ms
10 0 0 0 0 2 -1 0 0 0
Figure 1: Mass-spring model with 5 degrees of freedom. 0 2 0 0 0 -1 2 -1 0 0
[M] 0 0 2 0 0 [K]= 0 -1 2 -1 0
0 0 0 2 0 0 0 -1 2 -1
0 0 0 0 2 0 0 0 -1 2
1321
3. MAC and mass orthogonality for a beam with non-uni-
3,60 0 0 0 0 form properties
t 0 2,79 0 0 0
['1'] [M] ['1'] 0 0 2, 12 0 0 A beam with non-uniform properties is investigated for in-pla-
0 0 0 2,03 0 ne flexural modes. The beam is an Euler-Bernoulli beam with
0 0 0 rectangular cross-section and free boundary conditions. The
0 2,00
beam is divided into 100 linear finite elements (4], i.e. 101 no-
des, equally spaced,with 2 dofs (z, 8 ). This gives a mass and
The modal mass matrix will still have off-diagonal terms that stiffness matrix of size 202 *202. Al in the previous example
are zero. The modal masses on the diagonal have changed with the mass-spring system MATLAB c is used to calculate
values which reflect the new system. eigenvalues and eigenvectors. For the calculation of MAC
only translational degrees of freedom are used, the rotational
degrees of freedom are not included.
0, 32 0, 05 0, 01 0 In the first example the mass per unit length of the material is
0,32 1 0, 02 0, 01 0 varied, this only affect the mass matrix and not the stiffness
MAC= 0,05 0,02 1 0 0 matrix for the beam. In the second example the cross-section
0,01 0,01 0 0 height is varied, this affect both the mass and stiffness matrix
0 for the beam. The variation is carried out in two dimensions;
0 0 0 1
variable ratio of mass/height as a function of the beam length.
The MAC values for all configurations is calculated for the first
three modes.
For the MAC matrix the off-diagonal terms are ll.Q1 zero as in
the previous case with a mass matrix with equal diagonal The considered beam has the following basic input data:
terms. It is seen that the highest value for these terms is 0.32 Youngs modulus= 210 GPa, density=7800 kg/m 3 , thickness=
and is between the two lowest modes: mode 1 and 2. 2 mm and length = 1meter.
Further calculation of MAC values for just mode 1 and 2 is 3.1 A beam with non-uniform mass-distribution
carried out on the mass-spring model. The weight of mass M1
is continuously varied between 2 kg and 20 kg, i.e. by a factor In the first example the ratio of the mass per unit length:m2/
of 1 to 10. The result is given in figure 2. It is seen that the m1 of the beam is varied as a function of beam length, see fi-
MAC value is zero at factor 1 (M 1 =2kg) ) which is the case gure 3. The initial mass per unit length of the beam is m 1 . The
where all masses in the system have equal weight. As the mass per unit length m2(x) is a factor of 1 to 10 greater than
weight of mass M 1 increases the MAC value will deviate from m 1 and will cover a certain percentage part of the beam
zero and gradually increase. At a factor of 5 (M 1=10kg) the length: x in the range 0% to 100%. The stiffness of the cross-
MAC value is 0.32 as calculated in the previous example. At section is kept constant.
a factor of 10 the MAC value is 0.5
m 1 (l-x)
05
C\1
"0
c:
Cll
0 45
0 4
-=======:=::::::J
I ~X
Tz
(/) 0 35 L
(])
"0 03 ... 1. .
0
E 0 25
...
E 02
Figure 3: Beam with non-uniform mass distribution.
"iii
(])
::J 0 15
./ . . . . . . . . 1. . .
~7-
0 1
> The MAC values between mode 1 and 2, i.e. the first and se-
0
<{
() 05
cond free bending modes, is given in figure 4. It is seen that
~ 0
1 the MAC value will in general increase for increased mass ra-
10
1322
se plot edges are zero since these configurations represent 3.2 A beam with non-uniform cross-section
beams that have uniform mass distribution.
In this second example the beam cross-section is varied as
Figure 5 shows MAC values between modes 1 and 3 and fi- function of beam length, see figure 7. Both the mass and stiff-
gure 6 between modes 2 and 3.The increase of MAC value is ness matrix are now affected compared to the previous case,
somewhat less than between modes 1 and 2 in figure 4 and where a beam with non-uniform mass distribution only affects
mostly at the configuration when 90% of the beam has in- the mass matrix.
creased mass per unit length.
In this example the ratio of the height: h1/h 2 of the beam is va-
ried as a function of beam length, see figure 7. The initial
height of the beam is 2 mm. The height: h2 is a factor of 1 to
10 greater than h1 and will cover a certain percentage part of
the beam length: x in the range 0% to 100%.
h 1 (L-x)
-=======:::J
I .. X
Tz
L
mass per unit l&ngth ratio[-)
1 ... ~I
p9rcentage part of the beoam [%]
Figure 4: MAC value for mode 1 and 2 as function of mass per Figure 7: Beam with non-uniform cross-section
unit length ratio and percentage part of the beam length
The mass matrices in the examples are identical: a beam with
non-uniform height and, as in previous example, non-uniform
mass ratio for a given parameter of property ratio (mass/
height) and percentage part of beam The stiffness matrices
will only be identical for property ratio equal one, as the ratio
increases the stiffness matrix for a beam with non-uniform
height will deviate.
0.6
0.6
Figure 6: MAC value for mode 2 and 3 as function of mass per Figure 8: MAC value for mode 1 and 2 as function of height
unit length ratio and percentage part of the beam length ratio and percentage part of the beam length
1323
(4)
1'~'*11 'I' 21 + '1'*12 'I' 22 + ... + '1'*15 'I' 251
2
Equation 4 will not be zero since the mass matrix does lli21
have equal diagonal masses. If a weighting matrix is introdu-
ced which is similar to the mass matrix a zero value will be ac-
hieved independent of the ratio between the diagonal
elements
2
Figure 9: MAC value for mode 1 and 3 as function of height '1'21
ratio and percentage part of the beam length
'1'22
['~'*11 '1'*12 '1'*13 '1'*14 '1'*15] [MJ '~'23
'1'24
'1'25
2
'1'21
'I' 22
'I' 23
'1'24
M55 '1'25
Figure 10: MAC value for mode 2 and 3 as function of height
ratio and percentage part of the beam length
1324
4. Concluding remarks
2
_ M 1( 4 ( '* 11 ' 21) + '* 12' 22 + . + '* 15' 25 t
- ({IV} i {IV},)( {IV}~ {IV} 2)
It has been shown that if MAC is calculated from two modal
vectors that are orthogonal the value could deviate greatly
from zero. This is so if the mass matrix exhibits large ratio dif-
ferences along the diagonal in conjunction with certain geo-
metry of the structure.
5. References
(8) [1) Allemang R.J and Brown, "A Correlation Coefficient for
Modal Vector Analysis", Proceedings 1st IMAC, pp 110-116,
1983.
06
<'<I 0.5
~
E 0.4
~
-0.3
~E 0.2
b
~0.1
height ratio(-) 0 0
percentage part of the beam {%)
Figure 11: MAC value for mode 1 and 2 for the case when the
modal vectors is extended.
1325