METHOD FOR MAINTENANCE MANAGEMENT OF INDUSTRIAL EQUIPMENT BASED ON CONDITION PREDICTION USING THE KALMAN FILTER, MARKOV CHAINS, AND GENETIC SEARCH

Abstract

This paper proposes a maintenance management method for industrial equipment that combines Kalman filter-based condition estimation, Markov chain-based degradation prediction, and a genetic search algorithm for maintenance planning. The study is motivated by the increasing demand for improving equipment availability, reducing unexpected failures, and minimizing maintenance costs in modern manufacturing systems. An analysis of existing predictive maintenance approaches shows that most available solutions address condition estimation, degradation prediction, or maintenance optimization separately, while limited attention has been paid to their integration into a unified decision-support framework. The proposed method establishes a closed-loop maintenance strategy in which the estimated equipment condition obtained from noisy sensor measurements is used to predict transitions between discrete technical states, and the predicted state probabilities serve as input for a genetic algorithm that generates an efficient maintenance schedule. The developed mathematical model includes a Kalman filter for hidden-state estimation, a Markov-based degradation prediction model, and a risk-oriented objective function that simultaneously considers preventive maintenance costs, expected failure losses, production downtime, and maintenance resource constraints. Experimental validation was performed through computer simulation of a manufacturing system consisting of twenty industrial machines operating under different degradation conditions. The proposed method was compared with reactive maintenance, periodic preventive maintenance, and threshold-based predictive maintenance strategies. The experimental results demonstrated improved condition estimation accuracy, more reliable degradation prediction, a reduction in emergency failures and equipment downtime, and lower overall maintenance costs. The practical significance of the proposed approach lies in its applicability to automated predictive maintenance systems for industrial enterprises, where maintenance schedules can be generated according to both the current technical condition of equipment and its predicted future degradation.

Keywords: predictive maintenance, industrial equipment, Kalman filter, Markov chains, genetic algorithm, mathematical modeling, degradation prediction