A Novel Residual Domain Adaptation Network for Intelligent Transfer Diagnosis

  • Jinyang Jiao
  • , Ming Zhao
  • , Jing Lin
  • , Kaixuan Liang
  • , Chuancang Ding

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

Abstract

Deep neural networks based intelligent diagnosis methods are able to learn powerful features for accurate fault classification, however they cannot always generalize well across changes in data distributions. To address this issue, a novel residual domain adaptation network is proposed for transfer diagnosis of machinery in this paper. In the proposed framework, one-dimensional residual network is designed as the feature generator, then a mixed moment matching strategy, including first-order statistics and second-order statistics, is proposed to minimize the distribution discrepancy across domains. The comprehensive experiments on rolling bearing fault dataset are constructed to evaluate the proposed method. The results show the effectiveness of the proposed method.

Original languageEnglish
Title of host publicationAdvances in Asset Management and Condition Monitoring, COMADEM 2019
EditorsAndrew Ball, Len Gelman, B.K.N. Rao
PublisherSpringer Science and Business Media Deutschland GmbH
Pages827-839
Number of pages13
ISBN (Print)9783030577445
DOIs
StatePublished - 2020
Event32nd International Congress and Exhibition on Condition Monitoring and Diagnostic Engineering Management, COMADEM 2019 - Huddersfield, United Kingdom
Duration: 3 Sep 20195 Sep 2019

Publication series

NameSmart Innovation, Systems and Technologies
Volume166
ISSN (Print)2190-3018
ISSN (Electronic)2190-3026

Conference

Conference32nd International Congress and Exhibition on Condition Monitoring and Diagnostic Engineering Management, COMADEM 2019
Country/TerritoryUnited Kingdom
CityHuddersfield
Period3/09/195/09/19

Keywords

  • Domain adaptation
  • Intelligent diagnosis
  • Residual network
  • Transfer learning

Fingerprint

Dive into the research topics of 'A Novel Residual Domain Adaptation Network for Intelligent Transfer Diagnosis'. Together they form a unique fingerprint.

Cite this