A data-driven method for life prediction based on performance degradation data under complicated stress

This paper presents a novel life prediction approach for degraded components under complicated stress. The approach mainly consists of two parts: a data discretization method for computing degradation rates and a clustering analysis rule to divide the complicated stress into several categories. An empirical model is built to describe the relationship between the center of stress and the degradation rate. Finally, life prediction under any given stress level can be conducted based on the empirical model. The data-driven method in this paper can avoid the difficulty of degradation modeling under complicated stress profiles. The effectiveness of the proposed method is demonstrated by a case study.