ProCEED: an expert system for multivariate process control systems design

Abstract

Design and implementation of effective control systems is vital to the competitiveness of the process industry. Process control system design requires knowledge based systems which are capable of efficient numeric computing along with the symbolic reasoning. A brief survey presented in this paper shows that many industrial and academic research groups, are actively developing such knowledge based tools, in the U.S. and other countries.ProCEED, the expert system described in this paper has been developed on Digital's A.I. work-station and using Intellicorp's Knowledge Engineering Environment (KEE). This ideally combines the KEE and LISP based reasoning and Fortran based number crunching. The modular structure and the user interface permit easier integration of additional knowledge about existing systems, as well as incorporation of completely new controllers and processes.A controller design session begins with process modeling. Then the extent of interaction in the multivariable process is determined using a variety of techniques, including relative gain analysis and singular value decomposition, and conflicts among them are resolved. Based on the apparent interaction and controller implementation resources available to the user, multi-loop or completely multivariable forms of one or more control strategies, including proportional-integra-derivative mode control and simplified model predictive control, are chosen. The control systems are designed to achieve the specified performance criteria, utilizing the available numeric software in Fortran. The expert system presents the user with the design results and the simulation of the controlled system, in a graphical form.