Optimal robust linear controller design for chemical processes using an extended regional mapping approach

Abstract

Chemical process control is subjected to uncertainty due to uncertain model parameters, input disturbances or process nolinearity. If one derives an approximate linear model from a physically meaningful nonlinear model, it is well-understood that the linear model is usually subject to “strcutured uncertainty”. An extended regional mapping representation of structured uncertainty for a nonlinear process was derived. With the help of “possibility theory”, in which the importance of each element in a set of possible linear models could be determined, an extended Nyquist plot was developed. Optimal robust controller design was formulated into a nonlinear programming problem. This design guaranteed the optimality of the so-called “most possible even set”, but also ensured that stability of the “unlikely event set” and minimum control quality for the “possible event set” requirements were satisfied. Model parameters of the model in a model-based controller were allowed to deviate from the nominal values so that superior controller design could be obtained.