A systematic review of time series algorithms and analytics in predictive maintenance

The evolution of Industry 5.0, along with its predecessor Industry 4.0, has significantly boosted the adoption of predictive maintenance through integrating Internet of Things (IoT) sensors and real-time big data analysis, enabling the identification and prevention of equipment failures. This integration has also facilitated the development of time series-based predictive maintenance methods, addressing univariate and multivariate problems to capture temporal relationships and predict future equipment conditions. These approaches encompass prognostic tasks such as Remaining Useful Life (RUL) estimation, anomaly detection, failure classification, and clustering. Despite the extensive application of time series techniques in predictive maintenance, a comprehensive review focusing specifically on integrating time series methods with traditional and advanced machine learning and deep learning models is still missing. This study aims to fill that gap by systematically reviewing the literature on using time series algorithms in predictive maintenance. Using the PRISMA framework, we extracted and analyzed relevant literature from two major scientific databases, SCOPUS and Web of Science (WOS). The focus is on peer-reviewed journal papers on predictive maintenance and time series algorithms published since 2018. The review identified 55 peer-reviewed papers that utilized time series algorithms in predictive maintenance. This study systematically analyzed the most commonly used time series algorithms in predictive maintenance, including benchmark datasets and implementation methods. It highlighted common preprocessing steps for time series analysis and provides a comparative analysis of these algorithms and their performance metrics. The study also explored the challenges in utilizing time series algorithms for predictive maintenance and suggested potential research areas and future directions.