Remaining Useful Life Estimation in Predictive Maintenance

Remaining Useful Life (RUL) estimation lies at the heart of predictive maintenance, offering a transition from scheduled servicing to condition-based interventions. By forecasting the time until a component or system reaches failure, RUL estimation supports cost-effective resource allocation, reduces unplanned downtime and enhances safety across sectors such as aerospace, energy and manufacturing. Approaches range from physics-based degradation models to purely data-driven techniques, with hybrid strategies increasingly common. Data-driven methods exploit large volumes of sensor measurements, employing machine learning and deep learning to infer degradation patterns. Advances in semi-supervised and domain-adaptation techniques address the challenge of limited labelled data and variable operating conditions. Graph-based and spatio-temporal models have further enriched the field by capturing complex sensor interdependencies. Key challenges include uncertainty quantification, model interpretability and transferability across platforms. Practical implementations demonstrate substantial benefits, from extending aircraft engine life through sensor network analysis to optimising wind-turbine maintenance via adaptive prognostics. The global significance of RUL estimation continues to grow as Industry 4.0 and the Internet of Things proliferate, driving demand for robust, explainable and scalable prognostic solutions.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Remaining Useful Life Estimation in Predictive Maintenance publication trend

The graph below shows the total number of articles in remaining useful life estimation in predictive maintenance across all publications each year (not limited to Nature Index journals).

Technical terms

Remaining Useful Life (RUL): The predicted time span before a system or component reaches a predefined failure threshold.

Predictive Maintenance: A strategy that schedules service actions based on real-time condition monitoring and failure prognosis.

Prognostics and Health Management (PHM): The interdisciplinary field that integrates monitoring, diagnostics and prognostics to manage system health.

Convolutional Neural Network (CNN): A deep learning architecture specialised in extracting spatial hierarchies of features from data.

Long Short-Term Memory (LSTM): A recurrent neural network unit designed to capture long-range dependencies in sequential data.

Semi-supervised Learning: A machine learning paradigm that utilises both labelled and unlabelled data to improve model training.

References

1. Remaining useful life predictions for turbofan engine degradation using semi-supervised deep architecture. Reliability Engineering & System Safety (2019).
2. A Weighted Deep Domain Adaptation Method for Industrial Fault Prognostics According to Prior Distribution of Complex Working Conditions. IEEE Access (2019).
3. A Directed Acyclic Graph Network Combined With CNN and LSTM for Remaining Useful Life Prediction. IEEE Access (2019).
4. Spatio-temporal graph convolutional neural network for remaining useful life estimation of aircraft engines. Aerospace Systems (2020).