A distance-aware approach for reliable out-of-distribution detection of wind turbine gearbox fault diagnosis DOI Creative Commons

Junli Zhou,

Yao Zhao

Frontiers in Energy Research, Journal Year: 2024, Volume and Issue: 12

Published: Nov. 28, 2024

Fault diagnosis of wind turbine gearbox is essential to ensure operational efficiency and prevent costly downtime. However, conventional deep learning models often struggle with domain shift, where the distribution testing data differs from that training data. This issue more pronounced out-of-distribution inputs—data outside conditions model was trained on. These challenges can lead unreliable diagnostic results potentially hazardous situations. To address this, we introduce Spectral Normalization Gaussian Process methods into Res2Net framework enhance its ability detect improve model’s assess distance between test handle due both epistemic aleatory uncertainty. The experiment collected raw vibration signals under varied conditions. Unknown faults simulated uncertainty, while noisy samples resulted in were converted images using Gramian Angular Difference Field transformation. resulting then fed model, enhanced Process. outputs include classification corresponding uncertainty values based on awareness. With quantified values, reflect trustworthiness results. By comparing these predefined thresholds, it possible distinguish whether are or not. Experiments have proven superiority Distance-Aware detection fault diagnosis.

Language: Английский

Interpretable modulated differentiable STFT and physics-informed balanced spectrum metric for freight train wheelset bearing cross-machine transfer fault diagnosis under speed fluctuations DOI
Chao He, Hongmei Shi, Ruixin Li

et al.

Advanced Engineering Informatics, Journal Year: 2024, Volume and Issue: 62, P. 102568 - 102568

Published: May 6, 2024

Language: Английский

Citations

18
Research on roller bearing fault diagnosis based on robust smooth constrained matrix machine under imbalanced data DOI
Haiyang Pan, Bingxin Li, Jinde Zheng

et al.

Advanced Engineering Informatics, Journal Year: 2024, Volume and Issue: 62, P. 102667 - 102667

Published: June 25, 2024

Language: Английский

Citations

6
Few-shot bearing fault diagnosis by semi-supervised meta-learning with graph convolutional neural network under variable working conditions DOI

Zhen Liu,

Zhenrui Peng

Measurement, Journal Year: 2024, Volume and Issue: 240, P. 115402 - 115402

Published: July 30, 2024

Language: Английский

Citations

6
Transformer-based intelligent fault diagnosis methods of mechanical equipment: A survey DOI Creative Commons

Rongcai Wang,

Enzhi Dong, Zhonghua Cheng

et al.

Open Physics, Journal Year: 2024, Volume and Issue: 22(1)

Published: Jan. 1, 2024

Abstract Transformer is extensively employed in natural language processing, and computer vision (CV), with the self-attention structure. Due to its outstanding long-range dependency modeling parallel computing capability, some leading researchers have recently attempted apply intelligent fault diagnosis tasks for mechanical equipment, achieved remarkable results. Physical phenomena such as changes vibration, sound, heat play a crucial role research of equipment diagnosis, which directly reflects operational status potential faults equipment. Currently, based on monitoring signals temperature using Transformer-based models remains popular topic. While review literature has explored related principles application scenarios Transformer, there still lack Therefore, this work begins by examining current methods This study first provides brief overview development history outlines basic structure principles, analyzes characteristics advantages model Next it focuses three variants that generated significant impact field CV. Following that, progress challenges are discussed. Finally, future direction proposed.

Language: Английский

Citations

5
Domain Adaptation for Bearing Fault Diagnosis Based on SimAM and Adaptive Weighting Strategy DOI Creative Commons
Ziyi Tang,

Xinhao Hou,

Xinheng Huang

et al.

Sensors, Journal Year: 2024, Volume and Issue: 24(13), P. 4251 - 4251

Published: June 30, 2024

Domain adaptation techniques are crucial for addressing the discrepancies between training and testing data distributions caused by varying operational conditions in practical bearing fault diagnosis. However, transfer diagnosis faces significant challenges under complex with dispersed distinct distribution differences. Hence, this paper proposes CWT-SimAM-DAMS, a domain method based on SimAM an adaptive weighting strategy. The proposed scheme first uses Continuous Wavelet Transform (CWT) Unsharp Masking (USM) preprocessing, then feature extraction is performed using Residual Network (ResNet) integrated module. This combined strategy Joint Maximum Mean Discrepancy (JMMD) Conditional Adversarial Adaption (CDAN) algorithms, which minimizes differences source target domains more effectively, thus enhancing adaptability. validated two datasets, experimental results show that it improves accuracy of

Language: Английский

Citations

5
An enhanced digital twin-driven fault detection and isolation method based on sensor series imaging mechanism for gas turbine engine DOI

Zexi Jin,

Jinxin Liu,

Maojun Xu

et al.

Applied Thermal Engineering, Journal Year: 2024, Volume and Issue: 257, P. 124308 - 124308

Published: Sept. 2, 2024

Language: Английский

Citations

5
Bearing Fault Diagnosis Based on Image Information Fusion and Vision Transformer Transfer Learning Model DOI Creative Commons
Zichen Zhang, Jing Li, Chaozhi Cai

et al.

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(7), P. 2706 - 2706

Published: March 23, 2024

In order to improve the accuracy of bearing fault diagnosis under a small sample, variable load, and noise conditions, new method based on an image information fusion Vision Transformer (ViT) transfer learning model is proposed in this paper. Firstly, applies continuous wavelet transform (CWT), Gramian angular summation field (GASF), difference (GADF) time series data, generates three grayscale images. Then, generated images are merged into (IFI) using processing techniques. Finally, obtained IFIs fed advanced ViT trained learning. verify effectiveness superiority method, rolling dataset from Case Western Reserve University (CWRU) used carry out experimental studies different working conditions. Experimental results show that paper superior other traditional methods terms accuracy, effect (TLViT) training better than Resnet50 (TLResnet50) loads sample addition, also prove IFI with multiple has anti-noise ability single image. Therefore, can load provide for diagnosis.

Language: Английский

Citations

4
A comprehensive approach with DTW-driven IMF selection, multi-domain fusion, and TSA-based feature selection for compound fault diagnosis DOI

A. Andrews,

Manisekar Kondal

Measurement, Journal Year: 2024, Volume and Issue: 242, P. 115974 - 115974

Published: Oct. 12, 2024

Language: Английский

Citations

4
Managing chaos in chemical reactions with uncertain system parameters: exploring 4-D hyperchaotic system DOI
Marcio Demetrius Martinez, Fábio Roberto Chavarette

SeMA Journal, Journal Year: 2025, Volume and Issue: unknown

Published: April 23, 2025

Language: Английский

Citations

0
Electronic eye and electronic tongue data fusion combined with a GETNet model for the traceability and detection of Astragalus DOI
Xinning Jin, Zhiqiang Wang, Jingyu Ma

et al.

Journal of the Science of Food and Agriculture, Journal Year: 2024, Volume and Issue: 104(10), P. 5930 - 5943

Published: March 9, 2024

Abstract BACKGROUND Astragalus is a widely used traditional Chinese medicine material that easily confused due to its quality, price and other factors derived from different origins. This article describes novel method for the rapid tracing detection of via joint application an electronic tongue (ET) eye (EE) combined with lightweight convoluted neural network (CNN)–transformer model. First, ET EE systems were employed measure taste fingerprints appearance images, respectively, samples. Three spectral transform methods – Markov transition field, short‐time Fourier recurrence plot utilized convert signals into 2D spectrograms. Then, obtained spectrograms fused image obtain multimodal information. A hybrid model, termed GETNet, was designed achieve pattern recognition fusion The proposed model improved transformer module Ghost bottleneck as backbone network, complementarily utilizing benefits CNN architectures local global feature representation. Furthermore, further optimized using channel attention technique, which boosted model's extraction effectiveness. RESULTS experiments indicate data strategy based on devices has better accuracy than attained independent sensing devices. CONCLUSION achieved high precision (99.1%) recall values, providing approach rapidly identifying origin Astragalus, it holds great promise applications involving types herbal medicines. © 2024 Society Chemical Industry.

Language: Английский

Citations

3