A knowledge dynamic matching unit-guided multi-source domain adaptation network with attention mechanism for rolling bearing fault diagnosis

Wu, Zhenghong; Jiang, Hongkai; Zhu, Hongxuan; Wang, Xin

doi:10.1016/j.ymssp.2023.110098

Cited by 46 publications

(7 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…x n } to the interval [−1,1], and the values of each pair of vectors after the normalization are stored in the Gram matrix. Subsequently, transform the normalized value into polar coordinates, the values of radius and angle are shown in formula (1). The polar coordinate transformation encoding preserves numerical relationships while introducing temporal relationships,…”

Section: Gramian Angular Field (Gaf)mentioning

confidence: 99%

“…At this stage, researchers are becoming increasingly interested in the intelligent fault diagnosis of bearings due to advancements in industrial big data and inspection technology [1][2][3][4][5]. Deep learning-based techniques have made a number of noteworthy strides in the domain of bearing fault diagnosis * Author to whom any correspondence should be addressed.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A semi-supervised learning method based on pseudo-label iterative purification for intelligent fault diagnosis of rolling bearing

Yue,

Zhang,

Yang

et al. 2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

Bearing fault diagnosis based on semi-supervised learning shows good results in solving the scarcity of data in industrial scenarios. Pseudo-label samples can be used to expand labeled datasets, thereby to solve the shortage of training samples. However, when pseudo-labeling is employed to improve the performance of the model, the low confidence level of pseudo-labels will have a negative impact on the model. To solve this problem, the pseudo-label iterative purification (PLIP) method is proposed. Firstly, the graph neural network is employed to aggregate samples in order to increase the degree of dissimilarity among target domain data belonging to distinct categories. Subsequently, the pseudo-labeled samples are purified, while multiple positive-negative discriminators are trained on particular data from the source domain to accurately identify the target domain samples containing the pseudo-labels. Finally, the backbone network is fine-tuned by the purified pseudo-labeled samples. Results of experiments on the self-built bearing dataset and the Paderborn University bearing dataset validate that the method proposed in this article exhibits outstanding performance.

show abstract

Section: Gramian Angular Field (Gaf)mentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A semi-supervised learning method based on pseudo-label iterative purification for intelligent fault diagnosis of rolling bearing

Yue,

Zhang,

Yang

et al. 2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

show abstract

“…Inspired by WDCNN (Zhang et al 2017), this paper proposes multihead WDCNN. Using multiple convolution kernels of different sizes on the same feature map is an effective way of improving the modelʼs ability to extract features (Szegedy et al 2015;Das et al 2020;Wu et al 2023). The network structure is shown in Figure 2.…”

Section: Multihead Wdcnn Modelmentioning

confidence: 99%

Searching for Galactic H ii Regions from the LAMOST Database Based on the Multihead WDCNN Model

Wang

Bin

et al. 2023

ApJS

View full text Add to dashboard Cite

A H ii region is a kind of emission nebula, and more definite samples of H ii regions can help study the formation and evolution of galaxies. Hence, a systematic search for H ii regions is necessary. The Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) conducts medium-resolution spectroscopic surveys and provides abundant valuable spectra for unique and rare celestial body research. Therefore, the medium-resolution spectra of LAMOST are an ideal data source for searching for Galactic H ii regions. This study uses the LAMOST spectra to expand the current spectral sample of Galactic H ii regions through machine learning. Inspired by deep convolutional neural networks with wide first-layer kernels (WDCNN), a new spectral-screening method, multihead WDCNN, is proposed and implemented. Infrared criteria are further used for the identification of Galactic H ii region candidates. Experimental results show that the multihead WDCNN model is superior to other machine-learning methods and it can effectively extract spectral features and identify H ii regions from the massive spectral database. In the end, among all candidates, 57 H ii regions are identified and known in SIMBAD, and four objects are identified as “to be confirmed” Galactic H ii region candidates. The known H ii regions and H ii region candidates can be retrieved from the LAMOST website.

show abstract

“…Wang et al designed both intra-domain alignment and inter-domain alignment strategies to reduce biases between each pair of domains [8] . Wu et al introduced knowledge dynamic matching units into the feature extractor and embedded an attention mechanism into the classifier to effectively align the source and target domain distributions [9] . Although MUDA methods can fully utilize data from multiple source domains, when the distribution discrepancies between the source and target domains are substantial, the performance of these models may be limited, leading to negative transfer issues.…”

Section: Introductionmentioning

confidence: 99%

Multisource weighted domain adaptive network for bearing fault diagnosis

Huang,

2024

Third International Conference on Testing Technology and Automation Engineering (TTAE 2023)

View full text Add to dashboard Cite

Bearing failure is one of the common faults in rotating mechanical systems, significantly impacting the reliability and performance of the machinery. However, due to variations in operating conditions and environmental factors, bearing failure data exhibits a multi-source domain issue, posing challenges for traditional fault diagnosis methods to be effectively applied. In this paper, a new multi-source weighted domain adaptive (MSWDA) framework is proposed, which is implemented by feature weighted and multi-layer feature extraction structure. To achieve feature-level distribution alignment, the minimum slice Wasserstein distance is used as a measure to measure the difference in feature distribution. Extensive experiments are conducted on the CWRU bearing datasets to validate the effectiveness of the proposed model.

show abstract

A knowledge dynamic matching unit-guided multi-source domain adaptation network with attention mechanism for rolling bearing fault diagnosis

Cited by 46 publications

References 37 publications

A semi-supervised learning method based on pseudo-label iterative purification for intelligent fault diagnosis of rolling bearing

A semi-supervised learning method based on pseudo-label iterative purification for intelligent fault diagnosis of rolling bearing

Searching for Galactic H ii Regions from the LAMOST Database Based on the Multihead WDCNN Model

Multisource weighted domain adaptive network for bearing fault diagnosis

Contact Info

Product

Resources

About