Residual Useful Life Estimation by a Data-Driven Similarity-Based Approach

This paper is to provide a general data‐driven, similarity‐based approach for residual useful life estimation for industrial components or systems. Degradation signal data from failure dynamic systems are used to create a database of reference trajectory patterns. Variable weights are assigned to associated points before pointwise similarity computation. Distance score and weight are computed by calculating the fuzzy similarity between test trajectory pattern and reference trajectory patterns. Remaining useful life then could be predicted through aggregating those failure references' residual lifetimes in a weighted sum and updated as time goes by. The potentiality of this approach is illustrated on a problem of contact resistance degradation data. The results indicate that there is no significant difference between Gaussian function and General bell shaped function.