Application of Grey Correlation Analysis in Effective Utilization of Similarity-based Remaining Useful Life Prediction Methods

Abstract

Similarity-based remaining useful life (RUL) prediction methods are useful tools in prognostics, which are capable of making a long-term RUL prediction in a high accuracy by comparing signals from the test instance and references. In order to utilize the similarity-based methods effectively in practice, a uniform grey similarity measure was proposed based on grey correlation analysis method after data preprocessing. First, a grey time series was generated to represent the degradation of the test instance based on the monitoring data to ensure its size are same as the references both in length and time dimension. Second, a uniform grey similarity measure was developed improve the accuracy. It can not only measure the local similarity but also the whole degradation trend of the time series. Finally, the RUL of the current degradation process can be predicted using a weighted average method. The board-level package degradation data under random vibration loadings was used to evaluate the performance of this method and the results show that the proposed method is more practical with a better prediction performance in comparison with the existing methods.