Improvement of Fast Kurtogram Combined with PCA for Multiple Weak Fault Features Extraction

Yongxing Song,Linhua Zhang,Dazhuan Wu,Jingting Liu

doi:10.3390/pr8091059

Yongxing Song, Linhua Zhang + Show 2 more

Open Access

https://doi.org/10.3390/pr8091059

Copy DOI

Abstract

Demodulation plays an important role in fault feature extraction for rotating machinery. The fast kurtogram method was proved to be effective for rotating machinery demodulation. However, the demodulation effectiveness of fast kurtogram was poor for multiple fault features extraction under low signal-to-noise ratio. In this paper, an improved method of fast kurtogram, called P-kurtogram, is presented. The proposed method extracted the multiple weak fault features from multiple envelope signals-based principal component analysis. Compared with extracting features from one envelope signal of fast kurtogram, P-kurtogram showed a better demodulation performance for multiple faults. Combined with principal component analysis method, the proposed method also showed a good performance under low signal-to-noise ratio(SNR). By simulation analysis, the P-kurtogram method showed good performance for multiple modulation features extraction and robust performance in demodulation under low SNR. Then, the proposed method was demonstrated by applications of bearing faults detection and propeller detection. The results verified that the P-kurtogram has a better demodulation performance than fast kurtogram for multiple weak fault features extraction, especially under low signal-to-noise ratio. The proposed method provides a reliable basis for multiple weak fault features extraction of rotating machinery.

Highlights

Condition-based maintenance [1] is an effective maintenance policy in industrial enterprises.Robust fault diagnosis technique is a growing necessity in condition-based maintenance [2].Fault features extraction methods are the key techniques for fault diagnosis of rotating machinery [3].According to the mechanism of rotating machinery, amplitude modulation is the main modulation type
A brief overview of spectral kurtosis and fast kurtogram is presented in this subsection
The results indicate the proposed P-kurtogram method has a better performance than fast kurtogram for multiple features extraction

Summary

Introduction

Condition-based maintenance [1] is an effective maintenance policy in industrial enterprises.Robust fault diagnosis technique is a growing necessity in condition-based maintenance [2].Fault features extraction methods are the key techniques for fault diagnosis of rotating machinery [3].According to the mechanism of rotating machinery, amplitude modulation is the main modulation type. Fault features extraction methods are the key techniques for fault diagnosis of rotating machinery [3]. Demodulation is the best way to extract fault features in monitoring signals of rotating machinery. Various methods have been proposed for demodulation of rotating machinery, such as envelope demodulation [4], resonance demodulation [5], cyclostationary analysis [6,7], and spectrum kurtosis [8]. Among these demodulation techniques, spectral kurtosis is an effective and usual method for rotating machinery

Methods

Results

Conclusion