New Procedures of Pattern Classification for Vibration-Based Diagnostics via Neural Network

In this paper, the new distance-based embedding procedures of pattern classification for vibration-based diagnostics of gas turbine engines via neural network are proposed. Diagnostics of gas turbine engines is important because of the high cost of engine failure and the possible loss of human life. Engine monitoring is performed using either 'on-line' systems, mounted within the aircraft, that perform analysis of engine data during flight, or 'off-line' ground-based systems, to which engine data is downloaded from the aircraft at the end of a flight. Typically, the health of a rotating system such as a gas turbine is manifested by its vibration level. Efficiency of gas turbine monitoring systems primarily depends on the accuracy of employed algorithms, in particular, pattern recognition techniques to diagnose gas path faults. For pattern recognition of vibration signals, the characteristics usually used are: (1) amplitude, (2) frequency, and (3) space. In investigations, many techniques were applied to recognize gas path faults, but recommendations on selecting the best technique for real monitoring systems are still insufficient and often contradictory. In this paper, the new distance-based embedding procedures for pattern classification (recognition) are presented. These procedures do not require the arbitrary selection of priors as in the Bayesian classifier and allow one to take into account the cases which are not adequate for Fisher's Linear Discriminant Analysis (FLDA). The results obtained in this paper agree with the simulation results, which confirm the validity of the theoretical predictions of performance of the presented procedures. The computer simulation results are promising.