An autosynthesizing non‐linear control system using a rule‐based expert system

An advanced non-linear control system is obtained by combining recent developments in non-linear control system synthesis with a rule-based system approach to real-time control. The basic problem to be addressed is the control of a non-linear plant which is sufficiently sensitive to both operating point and input amplitude that the desired control system performance can be obtained only with a non-linear controller that is ‘retuned’ or ‘resynthesized’ whenever the operating point changes significantly. (The term ‘changes significantly’ in this context signifies an operating point change that causes the non-linear control system input/output behaviour to change substantially in an undesirable way.) This is accomplished by a hierarchically organized intelligent control system with a conventional reprogrammable controller under the direction of a rule-based expert system. The function of the expert system is to (i) monitor the behaviour of the non-linear control system to determine when retuning or resynthesis is required; if the behaviour is satisfactory, then continue passive monitoring; else (ii) when retuning or resynthesis is required: set up and execute experiments to derive the model information required to tune or synthesize a new non-linear controller in terms of a given structure and parametrization; execute the retuning/resynthesis procedure; reprogramme the controller (download the parametrization); and recommission the updated non-linear control system and return to normal operation. In essence, the rule-based system provides supervisory control (‘meta-control’) for a conventional reprogrammable non-linear controller that will autotune or autosynthesize as required. Autosynthesis takes place to accommodate variability in plant behaviour due to operating point changes, and the nonlinear controller thus synthesized corrects for amplitude sensitivity. This concept represents one way to combine artificial intelligence with control; it will be discussed and illustrated by example below.