《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

Cross domain diagnosis method for bearing faults with small samples based on 1D-DCGAN and 1D-CAE

Published:2023-05-25 author:LIN Pei, XU Yang-jian, FU Jun-ping, et al. Browse: 4417 Check PDF documents

Cross domain diagnosis method for bearing faults with small samples

based on 1D-DCGAN and 1D-CAE

LIN Pei1, XU Yang-jian1, FU Jun-ping2, CHEN Dong-dong2, JU Xiao-zhe1, LIANG Li hua1

(1.College of Mechanical Engineering,Zhejiang University of Technology, Hangzhou 310014, China;

2.Zhejiang Academy of Special Equipment Science, Hangzhou 310020, China)

Abstract: Sufficient fault samples are the guarantee for the fault diagnosis method based on deep learning to achieve good results. However, data imbalance is a typical feature of industrial big data. In order to reduce the dependence of intelligent diagnosis methods on the number of samples, and to solve the problem of fault diagnosis between the same device and different devices under small samples,a fault diagnosis method combining one-dimensional convolutional generative adversarial network (1D-DCGAN) and one-dimensional convolutional auto-encoder (1D-CAE) was proposed. Firstly, a 1D-DCGAN network was constructed by one-dimensional convolution layer, and the fault data set was expanded by its powerful data generation capability. Secondly, the 1D-CAE network was constructed by using the one-dimensional convolutional layer, and the potential features in the fault samples were effectively extracted by unsupervised learning to realize the fault diagnosis of the equipment. Based on the idea of transfer learning, the parameters of 1D-CAE model were transferred to further realize crossdomain diagnosis under limited samples. Finally, in order to verify the effectiveness of the proposed method, the bearing datasets of Case Western Reserve University（CWRU） and Xi‘’an Jiaotong University(XJTU) were used for the experiment. The experimental results show that the proposed method is superior to other comparative model, the fault identification accuracy of same device reached 99.21%, the cross domain fault identification accuracy between different equipments reach 98.87%. The results show that,even in the proposed procedure under the scenario of less sample size,the method based on 1D-DCGAN and 1DCAE can effectively achieve the same equipment fault diagnosis and the cross-domain diagnosis between different devices.

Key words: fault diagnosis of rotating machinery; one-dimensional convolutional generative adversarial network (1D-DCGAN); one-dimensional convolutional autoencoder (1D-CAE); transfer learning; deep learning; sample size

Advanced Search

Online Office System

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