Sigmoid-Based Refined Composite Multiscale Fuzzy Entropy and t-Distributed Stochastic Neighbor Embedding Based Fault Diagnosis of Rolling Bearing

TL;DR

This paper introduces a robust entropy measure called SRCMFE for short time series analysis and combines it with t-SNE and VPMCD for effective fault diagnosis in rolling bearings, demonstrating high accuracy in experiments.

Contribution

The paper proposes SRCMFE, a new entropy measure that improves robustness for short time series, and integrates it with t-SNE and VPMCD for enhanced fault diagnosis of rolling bearings.

Findings

SRCMFE effectively captures complexity in short time series.

The combined method accurately classifies different bearing faults.

Experimental results show high diagnostic accuracy.

Abstract

Multiscale fuzzy entropy (MFE) has been a prevalent tool to quantify the complexity of time series. However, it is extremely sensitive to the predetermined parameters and length of time series and it may yield an inaccurate estimation of entropy or cause undefined entropy when the length of time series is too short. In this paper the Sigmoid-based refined composite multiscale fuzzy entropy (SRCMFE) is introduced to improve the robustness of complexity measurement of MFE for short time series analysis. Also SRCMFE is used to quantify the dynamical properties of mechanical vibration signals and based on that a new rolling bearing fault diagnosis approach is proposed by combining SRCMFE with t-distributed stochastic neighbor embedding (t-SNE) for feature dimension and variable predictive models based class discrimination (VPMCD) for mode classification. In the proposed method, SRCMFE firstly is employed to extract the complexity characteristic from vibration signals of rolling bearing and t-SNE for feature dimension reduction is utilized to obtain a low dimensional manifold characteristic. Then VPMCD is employed to construct a multi-fault classifier to fulfill an automatic fault diagnosis. Finally, the proposed approach is applied to experimental data of rolling bearing and the results indicate that the proposed method can effectively distinguish different fault categories of rolling bearings.