Intelligent methodology for turbine engine diagnosis and health management

This paper presents applications of A.I. in turbine engines fault diagnosis and health management. Self-organizing map and back-propagation neural networks supported with fuzzy-logic decision-making tool were developed and integrated together as diagnostics software for turbine engines. Two different neural network architectures were trained and used. An unsupervised network (SOM) was used to cluster sensors data to distinct locations on a two dimensional map. A supervised neural network was used to identify features representing different fault signatures and abnormalities. A Fuzzy logic module was used as decision-making tool to resolve any uncertainty in the decision made by the neural networks. The developed diagnostics approach was validated by computer simulation using Matlab-Simulink model of an advanced model of turbine engine. The testing results presented in this paper showed that proposed diagnostics method is of great potential. However, further investigation is necessary using actual engine data for experimental validation of the current findings.