Development of the Forecasting Model for the Complex Technical Systems' Failures Time During the Proactive Maintenance Using the Recurrent Neural Networks' Technology

The forecasting problem of the remaining useful life, which is regularly solved during the preventive maintenance of the complex technical systems, has been considered. The possibilities of effective solution of this problem using the neural networks, including the RNN, the LSTM and the GRU networks, have been investigated. A herewith, to develop the forecasting models based on the neural networks, the datasets based on the multidimensional time series, which are formed from the results of the readings of the corresponding sensors have been used. The approaches to improving the forecasting accuracy by engineering new features on the basis of the original dataset have been considered. The examples of forecasting the remaining useful life using the considered approaches in the case of solving the problem of detecting aircraft engine failures have been given.