2022
DOI: 10.1109/tim.2021.3127636
|View full text |Cite
|
Sign up to set email alerts
|

Improved Generative Adversarial Network for Rotating Component Fault Diagnosis in Scenarios With Extremely Limited Data

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
5

Citation Types

0
11
0

Year Published

2022
2022
2024
2024

Publication Types

Select...
8

Relationship

0
8

Authors

Journals

citations
Cited by 34 publications
(11 citation statements)
references
References 26 publications
0
11
0
Order By: Relevance
“…This method yielded positive outcomes for fault diagnosis with limited data. Miao et al [29] introduced an enhanced variational self-coding GAN-based data augmentation technique. This method generates diverse synthetic samples while maintaining high similarity.…”
Section: Introductionmentioning
confidence: 99%
“…This method yielded positive outcomes for fault diagnosis with limited data. Miao et al [29] introduced an enhanced variational self-coding GAN-based data augmentation technique. This method generates diverse synthetic samples while maintaining high similarity.…”
Section: Introductionmentioning
confidence: 99%
“…Dixit et al [28] proposed a novel conditional auxiliary classifier ACGAN framework combined with model agnostic meta-learning, which is validated by bearing and air compressor datasets. Miao et al [29] proposed an improved VAEGAN to complete the data augmentation task under the background of a few samples.…”
Section: Introductionmentioning
confidence: 99%
“…However, in the presence of extremely limited fault samples, the following problems still need to be addressed to further improve the diagnosis performance. (1) The inputs of generators applied in the GANs, as mentioned above, are mostly random noise or random noise with some labels, which fails to fully capture each training sample's information and make it hard to extract diverse and representative feature representation [24,27,29].…”
Section: Introductionmentioning
confidence: 99%
“…Regarding the loss function, terms were incorporated to enforce the similarity of generated data to real data and the similarity of generated features to real features [47]. Moreover, researchers have employed sample similarity measures based on the Euclidean distance [48], Mahalanobis distance [48], and Pearson correlation coefficient (PCC) [49] to select high-quality generated data. However, it is important to note that current researches hava placed excessive emphasis on evaluating the quality of generated data solely in terms of their similarity to real data, often neglecting the importance of diversity in the generated data.…”
Section: Introductionmentioning
confidence: 99%