Modeling the Remaining Useful Life of a Gas Turbine Engine using Neural Networks

The paper presents a variant of estimating the residual lifetime of a gas turbine aircraft engine. A data set for modeling commercial modular aircraft propulsion systems for gas turbine engines (C-MAPSS) has been selected as the data set used for training the system and monitoring the results. The formal statement of the classification problem was carried out and quality criteria for the algorithm were selected: the standard deviation (mse) and the mean absolute deviation (mae) of the forecast from the true value of the objective function. The analysis of the "heat map" reflecting the correlation between the variables of the model is performed. The general scheme of the algorithm for obtaining a regression estimate using the machine learning method is considered. A variant of the piecewise linear regression line is proposed and the achieved improvement in the quality of the prediction algorithm is shown.