Fuzzy classification for fault isolation in gas turbine engines

RSLExpert is a fuzzy expert classifier for fault identification that is based upon expert reasoning and diagnosis of trend case residuals that are formed during an airplane's first ten seconds of flight. The fuzzy classifier assigns a residual vector of possible fault symptoms (called a Trend Case) to the type(s) of faults and faulty characteristics that may have caused it. The implementation of the classifier allows an expert to modify the fuzzy rule base "on the fly" so that no further recompilations of the model are necessary. This paper will address the fuzzy diagnosis strategy in the context of building RSL's Total Health Usage and Monitoring Systems (THUMS) for the cost-effective safety monitoring of gas turbine jet engines. Alternative and complementary soft computing approaches are also reviewed in the context of building robust THUMS systems.