Journal of Instrumentation, vol. 15, no. 06, pp. P06002-P06002, June 2020, DOI: 10.1088/1748-0221/15/06/P06002

Wasserstein Distance Based Deep Multi-Feature Adversarial Transfer Diagnosis Approach Under Variable Working Conditions

Daoming She1,2, N. Peng1, Minping Jia1, and Michael G. Pecht2
1School of Mechanical Engineering, Southeast University, 79 Suyuan Avenue, Nanjing 211189, P.R. China
2CALCE, Center for Advanced Life Cycle Engineering, Department of Mechanical Engineering, University of Maryland, College Park, Maryland 20742, USA


Intelligent mechanical fault diagnosis is a crucial measure to ensure the safe operation of equipment. To solve the problem that network features is not fully utilized in the adversarial transfer learning, this paper develops a Wasserstein distance based deep multi-feature adversarial (WDDMA) transfer diagnosis approach under variable working conditions. Domain adaptation is realized by adapting multi-feature in adversarial training and reducing the Wasserstein distance between the two domains after the discriminator network. The multi-feature are adapted at the same time to improve the recognition ability of the transfer network to the target domain. Experimental datasets under variable working conditions support the value of our approach.

This article is available online here and to CALCE Consortium Members for personal review.

[Home Page] [Articles Page]
Copyright © 2020 by CALCE and the University of Maryland, All Rights Reserved