A Similarity-Based Prognostics Approach for Remaining Useful Life Prediction

Physics-based and data-driven models are the two major prognostic approaches in the literature with their own advantages and disadvantages. This paper presents a similarity-based data-driven prognostic methodology and efficiency analysis study on remaining useful life estimation results. A similarity-based prognostic model is modified to employ the most similar training samples for RUL estimations on each time instance. The presented model is tested on; Virkler’s fatigue crack growth dataset, a drilling process degradation dataset, and a sliding chair degradation of a turnout system dataset. Prediction performances are compared utilizing an evaluation metric. Efficiency analysis of optimization results show that the modified similarity-based model performs better than the original definition.