Multifractional and long-range dependent characteristics for remaining useful life prediction of cracking gas compressor

Abstract

Cracking gas compressor (CGC) is a complex equipment used in ethylene production facilities. For the reliable and safe operation of CGC, the prediction of its remaining useful life (RUL) of relevance. The degradation process of a CGC from a normal state to a failure state has long-range dependence (LRD) with nonlinear and multifractal features. Concurrently, the increment of the degradation process obeys a non-Gaussian distribution. In this study, a degradation model for RUL prediction of CGC is developed. The model is based on a nonlinear drift function and Linear Multifractional Levy Stable Motion (LMSM). The drift function describes the nonlinear characteristics of the degradation process, whereas the LMSM allows accounting for its LRD, multifractal and non-Gaussian characteristics. The LRD features reflect the slowness of the degradation process, the multifractional features allow capturing local irregularities due to degenerate data fluctuations, and can specifically describe degenerate sequences. Finally, a RUL prediction framework for CGC is proposed and, then, verified with real observation data collected from an operating CGC.