Knowledge based diagnosis: dealing with fault modes and temporal constraints

An expert system performing tasks of supervisory control and fault detection and diagnosis was developed for a beet-sugar factory. Even when it can manage most usual alarm situations, the diagnosis method can not be easily extended to manage problems requiring a more complex analysis, such as phenomena involving nontrivial time evolution or multiple fault diagnosis. This is mainly due to two facts: the rule based approach to the diagnosis and the simplified treatment of some temporal aspects. A diagnosis strategy based on fault modes, including temporal knowledge, is now proposed. The paper describes the outstanding ideas supporting the design, some comments about the system architecture, simulation results over a real plant, conclusions, and further research