2021
DOI: 10.1088/1361-6501/ac0a0c
|View full text |Cite
|
Sign up to set email alerts
|

Wasserstein distance based Asymmetric Adversarial Domain Adaptation in intelligent bearing fault diagnosis

Abstract: Addressing the phenomenon of data sparsity in hostile working conditions, which leads to performance degradation in traditional machine learning-based fault diagnosis methods, a novel Wasserstein distance-based asymmetric adversarial domain adaptation is proposed for unsupervised domain adaptation in bearing fault diagnosis. A generative adversarial network-based loss and asymmetric mapping are integrated to alleviate the difficulty of the training process in adversarial transfer learning, especially when the … Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
4
1

Citation Types

0
9
0

Year Published

2022
2022
2024
2024

Publication Types

Select...
7

Relationship

0
7

Authors

Journals

citations
Cited by 14 publications
(9 citation statements)
references
References 31 publications
0
9
0
Order By: Relevance
“…With this adaptation of the DANN structure, the conditional distribution discrepancy of the domains can also be minimized. Yu et al [259] separated the training process into two stages. In the first stage, a classifier and a source feature extractor are trained with source data.…”
Section: Adversarial Approachesmentioning
confidence: 99%
“…With this adaptation of the DANN structure, the conditional distribution discrepancy of the domains can also be minimized. Yu et al [259] separated the training process into two stages. In the first stage, a classifier and a source feature extractor are trained with source data.…”
Section: Adversarial Approachesmentioning
confidence: 99%
“…A promising technique was presented by Jin et al [16] that made use of mini-max entropy adversarial learning and multi-adversarial training. Furthermore, asymmetric adversarial domain adaptation, a unique Wasserstein distancebased strategy, was also introduced to reduce negative domain migration [17].…”
Section: Introductionmentioning
confidence: 99%
“…Su et al [14] improved the generative adversarial network (GAN) to measure the distribution distances among different fault data using Mahalanobis distance, thus completing the gearbox fault diagnosis with small sample sets. Yu and Cheng et al [15,16] used Wasserstein distance in adversarial adaptation training to measure data distribution differences among different tasks, thus learning more domain-invariant feature information to recognize the health status of bearings. Meanwhile, many adaptation methods also utilize other distance metrics, such as maximum mean discrepancy (MMD) [17] and central moment discrepancy (CMD) [18,19], to calculate the distribution distance between the training and diagnosis tasks for feature difference evaluation.…”
Section: Introductionmentioning
confidence: 99%