Towards automatic decentralized control structure selection

A subtask in integration of design and control of chemical processes is the selection of a control structure. Automating the selection of the control structure enables sequential integration of process and control design. As soon as the process is specified or computed, a structure for decentralized control is determined automatically, and the resulting decentralized control structure is automatically tuned using standard techniques. Dynamic simulation of the resulting process system gives immediate feedback to the process design engineer regarding practical operability of the process. The control structure selection problem is formulated as a special MILP employing cost coefficients which are computed using Parseval's theorem combined with RGA and IMC concepts. This approach enables selection and tuning of large-scale plant-wide decentralized controllers through efficient combination of model formulation and mathematical programming. (C) 2000 Elsevier Science Ltd. All rights reserved.