Machine Learning for Predictive Maintenance: A Literature Review

Purpose: The advancement of Industry 4.0 has driven the adoption of predictive maintenance as a strategic step in optimizing equipment performance. Today, machine learning plays a key role in predicting potential equipment failures to minimize equipment downtime. However, assessing the accuracy of the predictive maintenance model presents a challenge for researchers in creating a reliable predictive maintenance model. Therefore, this review specifically focuses on the evaluation of the outputs of various related literatures. Design/methodology/approach: The approach used in making this review focuses on exploring the method of developing a predictive maintenance model with machine learning algorithms. This search is also carried out with the help of special software to search the literature database and screen the suitability of the research results, which is the main purpose of this literature review. Findings: Predictive maintenance models with machine learning algorithms such as support vector machines and random forests, including their integration with time-series classification and AdaBoost, demonstrate the ability of real-time abnormality prediction and equipment failure classification, as well as improve operational efficiency and resource optimization with high stability on complex data. Originality value: The results of this literature review are expected to be a guide for researchers and practitioners in exploring and validating ML algorithms to develop predictive maintenance models of equipment that can be accessed in real-time.