Multi-source heterogeneous fusion entropy ratio distance feature of bearing performance degradation based on DTW

Guangbin Wang,Shangjun Hu,Jinhua Chen,Zhixian Zhong

doi:10.21595/vp.2021.22269

Guangbin Wang, Shangjun Hu + Show 2 more

Open Access

https://doi.org/10.21595/vp.2021.22269

Copy DOI

Abstract

The key to the evaluation of bearing performance degradation is to extract sensitive characteristic indicators that can reflect the degradation process. In order to solve the problem of insufficient sensitivity of the bearing multi-source signals feature, a fusion entropy ratio feature based on dynamic time warping is proposed. First, the segmented approximate entropy, sample entropy and fuzzy entropy of the bearing performance degradation process are extracted, and then the mean clustering is used to obtain the standard entropy vector of the health state. In order to eliminate the influence of working environment and entropy value on state evaluation, a concept of entropy ratio is defined by the ratio of the entropy vector of each state to the standard entropy vector of the health state. Finally, using dynamic time warping has the advantage of accurately describing the similarity between vectors from a global perspective, and the multi-source heterogeneous fusion entropy ratio distance feature is constructed to describe the degraded state of bearing performance. Experiments show that this feature is more sensitive to the bearing performance degradation than other features, such as fusion entropy ratio feature based on Euclidean distance, single entropy ratio feature based on dynamic time bending.

Highlights

The key to accurate evaluation of machinery equipment’s state is to find sensitive indicators that reflect the degradation process
We propose a multi-source heterogeneous fusion entropy ratio distance feature based on dynamic time bending for bearing performance degradation
According to the above analysis, Our algorithm steps are as follows: Step 1: Collect the multi-physical quantity monitoring data of the whole life cycle of bearing operation, here it is assumed that the sampling is vibration, strain and stress data, i.e. X(i) = {V, SE, SA }, the data is pre-processed, and the bearing status is divided into four stages of healthy state, early fault, mid-term fault and late fault by observation

Summary

Introduction

The key to accurate evaluation of machinery equipment’s state is to find sensitive indicators that reflect the degradation process. We propose a multi-source heterogeneous fusion entropy ratio distance feature based on dynamic time bending for bearing performance degradation.

Results

Conclusion