The Application of a Knowledge-Based Expert System to Chemical Plant Fault Diagnosis

The architecture of a system for rapid, accurate diagnosis of the cause or causes of upsets in an operating chemical plant is developed. The system combines algorithmic and expert system techniques. An algorithm identifies the primary process variable deviation and an expert system lists the most probable faults that could produce that upset and recommends corrective action. The algorithm is insensitive to alarm level settings, does not depend on pre-defined alarm patterns, utilizes a modular architecture for easy adaptability to new plant settings, accounts for the dynamics of fault propagation, and can diagnose multiple faults occurring simultaneously in the plant. The expert system allows knowledge about the equipment and process to be utilized in diagnosis.