A Continual Learning Approach for Failure Prediction under Non-Stationary Conditions: Application to Condition Monitoring Data Streams

Accurate estimation of the remaining useful life (RUL) of critical assets plays a pivotal role in predictive maintenance strategies. Traditional RUL estimation approaches often face challenges in handling evolving operating conditions and data drift, which can result in degraded performance and reduced robustness. In this study, we propose a novel paradigm for RUL estimation, leveraging the power of continual learning. Continual learning enables the model to incrementally acquire new knowledge while retaining previously learned information, facilitating adaptation to changing operating conditions. By employing a continual learning framework, our approach demonstrates enhanced performance and robustness in RUL estimation tasks. Experimental results on real-world datasets showcase the effectiveness of the proposed method, outperforming traditional approaches in terms of accuracy and adaptability. This research contributes to the field of prognostics by providing a viable solution for improved RUL estimation, offering valuable insights for optimizing maintenance strategies and minimizing downtime in critical systems.