Hence, it is apparent that the tolerable input model
error is closely related to the singular values of the
inverse input complementary sensitivity function.
The analysis for the closed-loop modeling error tolerance
is carried out for each controller and each process
type separately. The tolerances to output model
errors are given in Figures 20–23.
According to the preceding results, it is clear that
when only single disturbance tone is present the IHC,
FSCF and SFSCF methods have the greatest tolerance
to output model errors with almost identical performance.
This is an expected result: since the process
dynamics are not included in the design, it is natural
that a change in such dynamics has no impact. The
DOFC method has the worst tolerance to modeling
errors, while still maintaining an adequate margin of
40 dB, followed closely by the IDC method. As
expected, the DOFC method has the best tolerance to
static modeling errors with infinite margin. In the case
with multiple disturbance tones, the overall performance
remains the same. However, for the filter-based
methods the tolerance to errors between the disturbance
tones is lost, causing potential problems in the
final application. The tolerances to input model errors
are given in Figures 24–26.
Again the overall trend is similar to the one perceived
in the sensitivity to output model errors.
However, for MIMO systems, the IDC and DOFC