Stationary Wavelet-Fourier Entropy and Kernel Extreme Learning for Bearing Multi-Fault Diagnosis.

Nibaldo Rodriguez,Pablo Alvarez,Guillermo Cabrera-Guerrero,Lida Barba

doi:10.3390/e21060540

Abstract

Bearing fault diagnosis methods play an important role in rotating machine health monitoring. In recent years, various intelligent fault diagnosis methods have been proposed, which are mainly based on the features extraction method combined with either shallow or deep learning methods. During the last few years, Shannon entropy features have been widely used in machine health monitoring, improving the accuracy of the bearing fault diagnosis process. Therefore, in this paper, we consider the combination of multi-scale stationary wavelet packet analysis with the Fourier amplitude spectrum to obtain a new discriminative Shannon entropy feature that we call stationary wavelet packet Fourier entropy (SWPFE). Features extracted by our SWPFE method are then passed onto a shallow kernel extreme learning machine (KELM) classifier to diagnose bearing failure types with different severities. The proposed method was applied on two experimental vibration signal databases of a rolling element bearing and compared to two recently proposed methods called stationary wavelet packet permutation entropy (SWPPE) and stationary wavelet packet dispersion entropy (SWPPE). Based on our results, we can say that the proposed method is able to achieve better accuracy levels than both the SWPPE and SWPDE methods using fewer failure features. Further, as our method does not require any hyperparameter calibration step, it is less dependent on user experience/expertise.

Highlights

Health conditions of rolling-element bearings (REBs) play a vital role in the working performance of the rotation machine
The bearing vibration signal database was split into five folds, where four out of the five folds were used during the training phase and the remaining fold was used during the testing phase
The stationary wavelet packet permutation entropy (SWPPE)-kernel extreme learning machine (KELM) method achieved a 100% F-score value for all working conditions only for D = 128 input features

Summary

Introduction

Health conditions of rolling-element bearings (REBs) play a vital role in the working performance of the rotation machine. REBs’ fault diagnosis is a very important task to guarantee the availability and reliability of the rotation machines in industrial processes. During the last few years, several bearing health indicators such as Shannon entropy, spectral kurtosis (SK), the smoothness index [1], the Gini index [2,3], and the spectral Lp/Lq norm [4,5] have been used in bearing failure diagnosis. Other signals such as current signals [6,7], acoustic signals [8], and stray flux signals [9] have been used for fault diagnosis. We use vibration signals as they are easier to measure and can provide useful dynamic information that reflects bearings’ health condition

Objectives

Methods

Results

Conclusion