Closed-loop system identification with operator intervention

grantor: University of TorontoDuring a process identification experiment, it often occurs that the process output variable drifts outside the acceptable operating region due to process disturbances. In this situation, the operator will attempt to bring the process output variable back inside the desired operating region by adjusting the manipulated input variable. This thesis studies the effects of such operator intervention on the accuracy of the identification results and proposes the use of correct noise models in the design of a process input-output data prefilter to remove the resulting biasing effects. In addition, this thesis develops a modified generalized least squares algorithm to simultaneously estimate the process model and the noise model.M.A.Sc