Intelligent Process Control

It is often observed that human experts can tune the parameters of a controller based on their knowledge and experience, rather than on complicated algorithms. In fact, more often than not, they have only a vague idea of the process model. An attempt is made here to create a fuzzy-logic based expert which would emulate such behavior. The expert is, specifically designed to tune the gains of a Proportional-Integral-Derivative (PID) controller, applied to stable dominant pole plants having large rise times. It is observed, that a number of plants found in the chemical process industry can be suitable modeled as such systems. a rule base for the expert was developed after analysis and simulation studies. Attempts have been made to keep the rules as few and simple as possible. At no point is any attempt made to estimate the parameters of the plant model. The expert observes only the output from the plant. Results of the application of the expert to a second order plant, to the separator temperature control loop of the Tennessee Eastman problem, and to a third order plant are presented. The expert is found to successfully tune the PID gains, and the results provide encouragement for the creation of such experts which can handle a class of plants.