Citation: Zhang, R.; Gu, Y. A Transfer
Learning Framework with a
One-Dimensional Deep Subdomain
Adaptation Network for Bearing
Fault Diagnosis under Different
Working Conditions. Sensors 2022, 22,
1624. https://doi.org/10.3390/
s22041624
Academic Editors:
Athanasios Rakitzis, Khanh T.
P. Nguyen and Kim Phuc Tran
Received: 28 January 2022
Accepted: 16 February 2022
Published: 18 February 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Article
A Transfer Learning Framework with a One-Dimensional Deep
Subdomain Adaptation Network for Bearing Fault Diagnosis
under Different Working Conditions
Ruixin Zhang
1
and Yu Gu
2,3,4,
*
1
College of Information Science and Technology, Beijing University of Chemical Technology, Beijing 100029,
China; rxzhang1@mail.buct.edu.cn
2
Guangdong Province Key Laboratory of Petrochemical Equipment Fault Diagnosis, Maoming 525000, China
3
Beijing Advanced Innovation Center for Soft Matter Science and Engineering, Beijing University of Chemical
Technology, Beijing 100029, China
4
Department of Chemistry, Institute of Inorganic and Analytical Chemistry, Goethe-University,
Max-von-Laue-Str. 9, 60438 Frankfurt, Germany
* Correspondence: guyu@mail.buct.edu.cn
Abstract:
Accurate and fast rolling bearing fault diagnosis is required for the normal operation of
rotating machinery and equipment. Although deep learning methods have achieved excellent results
for rolling bearing fault diagnosis, the performance of most methods declines sharply when the
working conditions change. To address this issue, we propose a one-dimensional lightweight deep
subdomain adaptation network (1D-LDSAN) for faster and more accurate rolling bearing fault diag-
nosis. The framework uses a one-dimensional lightweight convolutional neural network backbone
for the rapid extraction of advanced features from raw vibration signals. The local maximum mean
discrepancy (LMMD) is employed to match the probability distribution between the source domain
and the target domain data, and a fully connected neural network is used to identify the fault classes.
Bearing data from the Case Western Reserve University (CWRU) datasets were used to validate
the performance of the proposed framework under different working conditions. The experimental
results show that the classification accuracy for 12 tasks was higher for the 1D-LDSAN than for
mainstream transfer learning methods. Moreover, the proposed framework provides satisfactory
results when a small proportion of the unlabeled target domain data is used for training.
Keywords: fault diagnosis; deep learning; rolling bearing; domain adaptation; transfer learning
1. Introduction
Due to advances in industrial technology, rotating machinery is increasingly used in
many fields, such as electric power generation, chemical production, and aerospace [
1
,
2
].
Rolling bearings are indispensable elements in rotating machines [
3
] and are the main
source of faults in this equipment [
4
]. Rotating machines may operate under unfavorable
conditions, such as high ambient temperatures, high humidity, and overload conditions,
resulting in bearing malfunctions [
5
]. Bearing faults can cause significant damage to
mechanical equipment [
6
]. Therefore, accurate and rapid methods for rolling bearing fault
diagnosis are required to ensure the normal operation of rotating machinery.
In recent years, artificial intelligence methods, such as heuristic algorithm [
7
], expert
knowledge-based methods [
8
], and deep learning (DL) models [
9
], have gained increasing
attention in diverse fields. In particular, DL models have been broadly employed for
machinery fault detection and diagnosis systems [
10
]. Most DL models, such as the long
short-term memory network (LSTM) [
11
], deep belief network (DBN) [
12
], and convolu-
tional neural network (CNN) [
13
–
15
], perform well if the datasets of the source domain
and target domain tasks have the same distribution [
16
]. However, this assumption is
Sensors 2022, 22, 1624. https://doi.org/10.3390/s22041624 https://www.mdpi.com/journal/sensors
