Y. Frayman and B.F. Rolfe (Australia)
neurocontrol, loop-shaping, persistency of ex
citation, process uncertainty
In this paper, we propose a loop-shaping approach to in
telligent control with dynamically constructed neurocon
troller. In the proposed control scheme, the process uncer
tainly is reduced in the controller rather than in the process,
without explicit identification of the process under control.
The inherent noise/distrurbances in the process are utilized
to satisfy persistency of excitation condition. The use of
a reference model in form of a filter allow the frequency
response of the closed-loop to be adapted in line with the
changes in frequency response of the filter. The approach
is evaluated on the example of control of polymerization
reactor with promising results.