Remaining useful life prediction of degradation equipment considering multiple performance indexes correlation

Abstract

Aiming at a type of multivariate degradation equipment with high-dimensional monitoring data and the coupling relationship between performance indices, a remaining useful life prediction method of multivariate degradation equipment is proposed, considering the correlation of multiple performance indices. First, using information entropy and mutual information methods to select performance indices that are rich in information and can effectively characterize the change in the health state of the equipment. Then, the Copula function is used to describe the interdependence of multiple performance indices. Through stochastic simulation, health indices that consider the correlation of multiple performance indices are obtained. Finally, the health indices are inputted to the BiLSTM network that integrates the attention mechanism. The weights of hidden states at different times are adjusted by the attention mechanism to optimize the prediction network. The validity and practicability of the proposed method are verified by the aeroengine data set.