Performance Assessment of Nonminimum Phase Systems

Abstract

With a complete knowledge of process dynamics, any possible limitations on the achievable performance may be calculated via procedures such as convex optimization and linear programming (Boyd and Barratt 1991; Dahleh and Diaz-Bobillo 1995). This is generally not a very attractive approach to process performance monitoring since a typical plant can have hundreds or thousands of control loops. Identification of all loops, to obtain process models, is a very demanding requirement. Performance monitoring should be carried out in such a way that the normal operation of a process is affected as little as possible. In addition, process dynamics and disturbances may drift from time to time, and the initially identified model may not represent the true dynamics. Thus, on-line performance monitoring is necessary.