130 7. RECONCILIATION OF FINITE ELEMENT MODELS AND MODAL TEST DATA
"
k
O
C
X
i
p
i
k
i
#
Œ
'
a
D
"
m
O
C
X
i
p
i
m
i
#
Œ
'
a
Œ
a
;
Œ
ˆ
D
N
ˆ
OL
Œ
'
: (7.33)
e metric for evaluation of approximate solution convergence is the cross-orthogonality
matrix associated with exact and approximate (RMA) modal sets, specifically,
Œ
COR
ea
D
ˆ
T
e
Œ
M
O
Œ
ˆ
a
: (7.34)
Convergence of the approximate (RMA) modal set is therefore judged on the basis of how
close the (absolute value) cross-orthogonality matrix is to an identity matrix. In addition, the
difference between exact and approximate corresponding modal frequencies is also employed as
part of convergence evaluation.
Before engaging in the actual RMA convergence study, a preliminary evaluation of modal
sensitivities for each of the 28 parametric variations (parameter change set to a value of p
i
D 1)
was conducted, wherein the cross-orthogonality between the baseline modes and exact perturbed
modes,
Œ
COR
eO
D
ˆ
T
e
Œ
M
O
Œ
ˆ
O
; (7.35)
and corresponding modal frequencies were evaluated. Results of that exercise, summarized in
Table 7.1 (note DF represents modal frequency change), indicate that 13 of the total of 28
parametric variations were insignificant.
It should be noted that the numerical values provided in the above table are the peak
frequency and cross-orthogonality alterations associated with the lowest 60 normal modes of
each “unit” parametric variation. As a result of this finding, only the “sensitive” 16 parameters
were evaluated in the RMA convergence study.
Results associated with the RMA convergence study, which used “unit” parametric vari-
ations and values of tolerance .tol/ set to 1e-4, 1e-5, and 1e-6, respectively, are summarized in
Table 7.2.
It is clear from the above results that tol D 1e-6 produces highly converged RMA modes
for the ISPE.
7.1.7 CLOSURE
Alteration of a structural dynamic model for the purpose of reconciliation with respect to mea-
sured data typically requires moderate to large variation in (stiffness and/or mass) parameters.
Moreover, even when small parametric variations of parameters are required, close spacing of
system modes produces large variations in modal vectors. erefore, modal derivatives are not
well suited for tracking of parametric sensitivities of structural dynamic modes. A more robust
strategy for approximate modal sensitivity analysis employs the Ritz method. Specifically, SDM