FSF closed-loop identification issues and optimal input signal design for identification experiments

grantor: University of TorontoIn this thesis, closed-loop system identification issues and optimal input signal design using the Frequency Sampling Filter (FSF) model are studied. It is found that the FSF model when combined with a least squares estimator produces inconsistent FSF parameter estimates under closed-loop conditions. The reason for this is presented and a data shifting technique is suggested for the purpose of eliminating this problem. As well, it is found that the assumption of more timei delay reduces bias error callsed by prefilter mismatch for FIR and FSF models under closed-loop conditions. The optimal input signal energy distributionl with rctspect to minimizing a measure of thF quality of the FSF-generated step response model. is computed. This optimal energy distribution is indepeIldeIlt of the process parameters being identified. Finally, a new input signal design strategy is dpveloped for thc purpose of realizing binary, input signals having the optimal energy distriblltion via a practical, on-line methodology.M.A.Sc