4.3. Stability and robustness to modeling errors
Probably the most important, and unfortunately often
overlooked, feature of any control system is its stability
and capability to tolerate modeling errors: in essence
the robustness. The model errors may be due to the
failed identification of the model used in the design or
to some actuator-related problems; for example, a voltage
signal fed into a system may not be the same as the
one assumed by the control law. Regardless of the error
source, they exist in any real process and should not be
overlooked. From the vibration control point of view,
it would be desirable that the control law is capable of
handling very large model errors outside the frequency
range of the controlled vibrations tones, allowing a very
crude simplification of the process, significantly reducing
the controller complexity and the required computation
power thereof. The model truncation is made all
the time for high-frequency components, where the
closed-loop stability can be expected, but not usually
for low-frequency dynamics. Take, for example, a process
with vibration problems at 400 Hz. There is no
reason, besides the stability-driven issues, to model
the low-frequency dynamics. Hence, by guaranteeing
high tolerance to modeling errors at these frequencies,
the model can be truncated, thus avoiding many of the
identification issues. The process stability is usually studied
by applying the Nyquist stability criterion, which