Similarity-based prediction method for machinery remaining useful life: A review

Determining the remaining useful life (RUL) of increasingly complex machines provides the decision basis for the predictive maintenance process, which effectively ensures equipment safety, improves the utilization rate, and reduces the maintenance cost. Similarity-based prediction (SBP) methods are one type of RUL prediction technique, generally divided into four steps: condition monitoring data collection, degradation information fusion, similarity evaluation, and model prediction aggregation. SBP methods have advantages which include strong interpretability and a simple implementation process. Intensive studies and wide applications based on the SBP methods exist in both academia and industry. SBP methods have been included in numerous reviews, but they mainly focus on the first two steps or just one of the steps. Existing reviews lack recent advances of SBP methods and discussions of the four steps in detail. To fill the above gaps, this paper reviewed the whole procedure of SBP methods. Firstly, the prognostics industrial scenarios with limited failure data and sufficient failure data are introduced. Then, the degradation indicators (DIs) of the machines are constructed through a fusion of degradation information. Later, similarity calculation and similarity matching rule are utilized to evaluate the similarity of the DIs segments. After that, point estimation and uncertainty management are acquired by integrating the referential DIs segments. Finally, the effectiveness of the SBP methods in different industrial scenarios, the limitations, and future challenges are discussed.

Automated summary

This paper reviews the entire procedure of Similarity-based prediction (SBP) methods, including the industrial scenarios with limited failure data and sufficient failure data, the construction of degradation indicators (DIs), the utilization of similarity calculation and matching rule, the acquisition of point estimation and uncertainty management, and the discussion of the effectiveness, limitations, and future challenges of SBP methods.