Design of an embedded expert system for process model-based real-time control
An expert system has been embedded within an environment representative of a process model-based controller (PMBC) which performs as an on-line real-time diagnostic consultant utilizing forward chaining and backward chaining. The system runs in a multitasking environment on a PC, and has the capability to perform corrective actions in real-time, or provide corrective action recommendations to the operator.
An information translation module was developed to translate the sensor readings from the numerically intensive computation environment of the PMBC to the symbolically intensive computation environment of the expert system. The embedded expert system, written in CLIPS, is then spawned as a separate process. The PMBC is put to sleep awaiting an appropriate response from the expert system. When the first appropriate response is delivered, corrective action is taken (if possible), and the PMBC is awakened in order to resume control. The expert system then continues searching for alternative responses as the PMBC is running concurrently.