Artificial Intelligence and Machine Learning in Predictive Maintenance of Rotary Machinery: A Systematic Literature Review

The rising complexity of modern mechanical systems has made predictive maintenance (PdM) essential for ensuring reliability, reducing costs, and avoiding unplanned downtime. Conventional time-based and failure-based maintenance practices often fall short of these demands, creating the need for data-driven solutions. This research aims to bridge the fragmented understanding of Machine Learning (ML) applications in predictive maintenance (PdM), with a particular focus on rotary machine components such as bearings and gears. To provide clarity, a systematic review of over 550 research papers published between 1985 and 2023 is presented. The analysis categorizes methods into supervised, unsupervised, and ensemble learning approaches, with particular focus on techniques such as Support Vector Machines, Decision Trees, K-Nearest Neighbors, Naïve Bayes, Random Forests, and XGBoost. Outcomes of this review include a structured synthesis of existing methods, identification of research gaps, and recognition of emerging trends across sectors such as power generation, chemical processing, nuclear facilities, continuous manufacturing, and the oil and gas industry. The results serve as a holistic guide for academics and practitioners seeking to advance the design and execution of intelligent maintenance strategies in the context of mechanical systems.