Multimodel Knowledge Base Fault Detection and Isolation System

A.R. Campos (Germany) and R.N. Silva (Portugal)


Fault detection, fault isolation, multimodels, rule basereasoning, fuzzy sets.


In the current market competition it is necessary to continuously supervise the production and guarantee the adequate functioning of plants, so that possible faults in the systems neither decrease the products' quality, nor cause breaks in the manufacturing that have natural consequences in costs. This paper proposes a method for Fault Detection and Isolation based on the combination of techniques from the adaptive multimodels theory with the knowledge base methods. The method consists in using on-line identified models for the plant in normal and faulty situations, and in storing them in a models case base. From this collected information, rules can be built (consolidation of cases into rules) with an associated confidence level established by the relative frequency of occurrence in the case base. A case studied in a model of a gas-liquid separation industrial-scale unit is included illustrating the potentialities of the proposed solution.

Important Links:

Go Back