Application of the Machine Learning Tools in the Problem of Classifying Failures in the Work of the Complex Technical Systems

Abstract

The problem of classifying equipment failures in the complex technical systems using the machine learning tools has been considered. The aim of the study is to develop the intelligent classifier which will allow to classify efficiently and quickly the probable class of error in the work of equipment of the complex technical systems in the context of proactive maintenance activities. The prospects of applying the random forest algorithm and the algorithms based on the artificial neural networks to solve the set problem have been analyzed. The intelligent classifiers based on the recurrent neural networks of the LSTM and GRU type have been developed. The training has been performed with the experimental dataset containing information about the work of aircraft engines and hosted by NASA Ames Research Center in the public domain. Based on the learning results, the most effective classifier has been highlighted, in addition, the recommendations on its further modification to improve the quality of classification have been made.