Automated process fault analysis

The FALCON (Fault AnaLysis CONsultant) Project was a joint venture between the University of Delaware, DuPont, and the Foxboro Company. Its main objective was to develop a knowledge based system capable of performing continuous, real-time process fault analysis. The system was designed to function on-line in a commercial scale adipic acid plant operated by DuPont in Victoria, Texas. The FALCON System was able to correctly detect and diagnose better than 95% of the actual process fault situations which occurred while it was being developed and field tested.;The motivation for initiating the FALCON Project was to identify and study the various general issues involved with developing knowledge-based programs for actual process applications. The major goal of the project was to develop a generalized approach for automating process fault analysis.;A generalized approach, known as the Method of Minimal Evidence, has been developed. This diagnostic strategy is based upon model-based reasoning. It produces structured diagnostic knowledge bases which optimizes the overall competence, diagnostic resolution and sensitivity of the resulting fault analyzer. This fault analyzer would be able to diagnose both single and multiple fault situations with diagnostic rules developed specifically for single fault situations. The strategy can also be used to determine the optimal placement of process sensors for improving fault analysis and to determine the optimal division of a large process system for effectively distributing process fault analyzers.