PREDICTIVE MAINTENANCE OF PRODUCTION EQUIPMENT USING ARTIFICIAL INTELLIGENCE METHODS

Abstract

The work provides a comprehensive analysis of modern approaches to predictive maintenance of production equipment using artificial intelligence methods. It examines the features of reactive, preventive, and predictive maintenance strategies, as well as the principles for building predictive maintenance systems. Modern machine learning methods are analyzed, including Random Forest, LSTM, GRU, Isolation Forest, One-Class SVM, and autoencoders, highlighting their advantages, limitations, and suitability for tasks such as classifying equipment condition, predicting remaining equipment life, and detecting anomalies. The typical architecture of predictive maintenance systems has been summarized, and a comparative analysis of modern industrial predictive maintenance platforms has been carried out. Based on a review of the literature, an original qualitative comparison of machine learning methods and a generalized scheme of how predictive maintenance systems function have been developed. It was found that the effectiveness of such systems depends on a combination of quality data preparation, the choice of machine learning algorithms, and architectural decisions. The results obtained can be used as a methodological basis for designing intelligent predictive maintenance systems for production equipment in various industrial sectors.