Advanced engine diagnostics using artificial neural networks

Abstract

Gas turbines are used for aero and marine propulsion, power generation and as mechanical drives for a wide range of industrial applications. Often, they are affected by gas path faults which have hitherto been diagnosed by techniques such as fault matrixes, fault trees and gas path analysis. In this paper, an artificial neural network system is applied. The system is trained to detect, isolate and assess faults in some of the components of a single spool gas turbine. The hierarchical diagnostic methodology adopted involves a number of decentralised networks trained to handle specific tasks. All sets of networks were tested with data not used for the training process. The results show that significant benefits can be derived from the actual application of this technique.