ABSTRACT
An important problem facing the designers of control systems is their ever-increasing complexity, which stems from demands for better process yield, system performance, and robustness to changes within the operating environment. To meet these demands, significant efforts have been undertaken in control engineering research to devise algorithms that deliver adequate performance over an extended period, even when the controlled process is subject to varying operating conditions, parametric drifts, and structural changes. These algorithms are units of computation that collaborate with one another to perform functions that contribute to some system objectives. They originate from different branches of control system research and apply techniques within the realm of robust control,1-3 adaptive control,4-7 fault diagnosis,8-10 and, more recently, intelligent control11-16 and automatic reconfiguration.17 However, none of these algorithms is capable of single-handedly coping with the wide variety of situations that may occur in practice. Consequently, extending the operating range ofa control system requires employing many of these algorithms together with a supervisory mechanism to select and activate the appropriate algorithms.
