Multiobjective process controllability analysis

A new approach for process controllability analysis by using multiobjective optimisation techniques is proposed. Within the approach, a set of performance specifications, such as minimum control error and input effort with closed-loop pole placement are represented as a set of linear matrix inequalities (LMI). The solution to the LMI conditions can be identified as feasible or infeasible. If the solution is feasible there is at least one controller that can make the closed-loop system satisfy all performance specifications simultaneously. Therefore, for the process plant, these performance specifications are achievable. Otherwise, they are unachievable. There is a Pareto-optimal set or a trade-off curve in the performance space to separate these two areas. The paper shows that such trade-off curves can be used for process controllability analysis, and therefore, can be applied to control structure selection problems. (C) 2003 Elsevier Ltd. All rights reserved.