Asian Journal of Current Research

Predictive maintenance has become an important strategy for improving the reliability, safety, and operational efficiency of industrial systems. A central task in predictive maintenance is the estimation of Remaining Useful Life (RUL), which supports maintenance planning by predicting the time available before a component or system reaches a defined failure condition. This review provides a comparative discussion of major RUL modelling approaches used in predictive maintenance, with particular attention to physics-based, data-driven, and hybrid methods. Physics-based models use engineering knowledge and mathematical representations of degradation processes to generate interpretable predictions; however, their development depends on a detailed understanding of system behaviour and failure mechanisms. Data-driven models use historical and real-time sensor data to learn degradation patterns and have shown strong potential in complex industrial environments, particularly through machine learning and deep learning techniques. Their performance, however, often depends on the availability of sufficient labelled data and may be limited by poor interpretability. Hybrid models combine physical knowledge with data-driven learning to improve robustness, reliability, and practical applicability. The review also discusses transfer learning, explainable artificial intelligence, digital twins, and Industry 4.0 integration as emerging directions for RUL prediction. Key challenges identified include limited run-to-failure data, changing operating conditions, model scalability, uncertainty in predictions, and integration with industrial maintenance systems. Overall, the review indicates that no single modelling approach is universally suitable for all predictive maintenance applications. The selection of an appropriate RUL model should depend on data availability, domain knowledge, system complexity, interpretability requirements, and deployment conditions in practice.