A PNN fault diagnosis method for gas turbine

According to the complex fault diagnosis question of Gas turbine (GT), a probabilistic neural networks (PNN) fault diagnosis method for GT is presented. PNN can meet the needs of real time requirements for engineering practice due to its simple learning algorithm, and quick training and generalizing property. In addition, newly trained patterns can be easily supplemented to the already trained classifier, thus facilitating the improvement of the accuracy of diagnosis results. Considering the combinatorial and undefined faults problems, the PNN fault diagnosis program is put forward based on the practical fault model library of one GT. The classifying and generalization capabilities are checked, and the influence of the parameters normalization for diagnosis precision is analyzed too. The results show that the proposed PPN method is fast, accuracy, modified easily, and have good diagnosis robustness to measure noise, and can be easily applied to practical application.