Enhanced Fault Localization Using Probabilistic Fusion With Gas Path Analysis Algorithms

Abstract

A method for gas turbine fault identification from gas path data, in situations with a limited number of measurements, is presented. The method consists of a two stage process: (a) localization of the component or group of components with a fault and (b) fault identification by determining the precise location and magnitude of component performance deviations. The paper focuses on methods that allow improved localization of the faulty components. Gas path analysis (GPA) algorithms are applied to diagnostic sets comprising different combinations of engine components. The results are used to derive fault probabilities, which are then fused to derive a conclusion as to the location of a fault. Once the set of possible faulty components is determined, a well defined diagnostic problem is formulated and the faulty parameters are determined by means of a suitable algorithm. It is demonstrated that the method has an improved effectiveness when compared with previous GPA based methods.