Modified Method of Identification Potential Defects in Helicopters Turboshaft Engines Units Based on Prediction its Operational Status

Abstract

The work is devoted to the development of a method for identification potential defects in helicopters turboshaft engines units, which based on predicting their operational status in flight modes. This method is based on the use of GRNN neural network architecture. The use of neural networks is advisable in cases where it is necessary to overcome difficulties associated with non-stationarity, incompleteness, unknown data distribution, or when statistical methods are not entirely satisfactory. As a result of the experiments, the efficiency of the proposed method and sustainable training of the neural network were proved. The results of the comparative analysis showed that the use of the GRNN neural network architecture compared to AutoEncoder and LSTM AutoEncoder made it possible to obtain the best values of the neural network training quality assessment metrics.