《机电工程》杂志社科技期刊核心期刊论文投稿机电一体化机械、仪表电工自动化、计算机浙江大学主办

Introduction

International Standard Serial Number:

ISSN 1001-4551

Sponsor:

Zhejiang University;

Zhejiang Machinery and Electrical Group

Edited by:

Editorial of Journal of Mechanical & Electrical Engineering

Chief Editor:

ZHAO Qun

Vice Chief Editor:

TANG ren-zhong,

LUO Xiang-yang

Tel:

86-571-87041360,87239525

Fax:

86-571-87239571

Add:

No.9 Gaoguannong,Daxue Road,Hangzhou,China

P.C:

310009

E-mail:

meem_contribute@163.com

Rolling bearing fault diagnosis based on ICNN-BiGRU

Published:2023-01-30 author:YANG Hui, ZHANG Rui-jun, CHEN Guo-liang Browse: 1192 Check PDF documents

Rolling bearing fault diagnosis based on ICNN-BiGRU

YANG Hui1, ZHANG Rui-jun2, CHEN Guo-liang3

(1.School of Intelligent Manufacturing, Anhui Wenda University of Information Engineering, Hefei 230001, China;

2.School of Mechanical Engineering, Xi'an Jiaotong University, Xi'an 710049, China;

3.Joint Transmission and Bearing Technology Research Center, Shizuishan 753000, China)

Abstract: Deep learning-based rolling bearing fault diagnosis methods were vulnerable to environmental noise in actual use. To solve this problem, a model(method)based on improved convolutional neural network-bi-directional gated recurrent unit (ICNN-BiGRU) was proposed. Firstly, the collected rolling bearing signals were denoised by Laplace wavelet correlation filtering method and then the filtered vibration signals were transformed to power spectral domain. Secondly, the ICNN-BiGRU model was employed to extract the power spectrum characteristics during the rolling bearing failure, and the dynamic selection layer (DS) and self-attention layer (SA) were introduced on the basis of convolutional neural network to realize accurate and effective fault feature extraction and fault identification based on relevant feature information of different bearing fault states. Finally, the effectiveness of the proposed ICNN-BiGRU model and other deep learning models were compared with the Xi'an Jiaotong University-Changxing Sumyoung Technology (XJTU-SY) rolling bearing accelerating life testing data set,to verify the superiority of the ICNN-BiGRU model. The results show that the fault identification accuracy of the proposed ICNN-BiGRU model is higher compared to other deep learning models, and the accuracy can reach 99.65%. Under the interference of different background noises, compared with other deep learning models, the feature learning ability of the ICNN-BiGRU model is stronger,and have a certain engineering reference value.

Key words: deep learning model; feature learning ability; improved convolutional neural network (ICNN); bi-directional gated recurrent unit (BiGRU); Laplace wavelet; dynamic selection(DS); self-attention(SA)

Advanced Search

Online Office System

Chinese Core Periodicals
Chinese Sci-tech Core Periodicals
SA, INSPEC Indexed
CSA: T Indexed
UPD:Indexed