State-space modeling and novel entropy-based health indicator for dynamic degradation monitoring of rolling element bearing

This work is dedicated to the establishment of state-space modeling combined with a novel probabilistic entropy-based health indicator (HI), needed to assess the dynamic degradation monitoring and estimation of remaining useful life (RUL) of rolling element bearing. The classical statistical HI such as kurtosis exclusively fails to hold the understanding and steadiness for fault detection under multifaceted noisy situations. It is highly influenced by load and speed because of its sensitiveness towards deterministic vibrations (high probabilistic distribution data). Contemporary, the proposed probabilistic entropy-based HI is less sensitive to high probabilistic distribution data, which makes it capable of using it under different load and speed conditions. The proposed HI is skilled enough to be deployed for initializing the proposed state-space (SS) model, intended to predict futuristic values of HI of time horizon. The continuous updating of the model is done using predicted HI values to determine the futuristic failure time and RUL of bearing. The proposed methodology is deployed to two different data sets: Intelligent Maintenance Systems (IMS) and Xi'an Jiaotong University (XJTU). The experimental result suggests that our entropy-based State Space model is superior in comparison with the existing models General Regression Neural Network (GRNN) and Auto-Regressive Integrated Moving Average (ARIMA) for estimating RUL and carrying out the dynamic degradation monitoring of rolling element bearing.