Research Article
Application of Improved ConvNext and Fused Images of Gramian Difference Field and Sum Field in Gearbox Health Assessment
@INPROCEEDINGS{10.1007/978-3-032-33662-0_40, author={YunTao Zhao and Zengyi Ma and WeiGang Li}, title={Application of Improved ConvNext and Fused Images of Gramian Difference Field and Sum Field in Gearbox Health Assessment}, proceedings={Communications and Networking. 20th EAI International Conference, ChinaCom 2025, Xi’an, China, November 8--9, 2025, Proceedings}, proceedings_a={CHINACOM}, year={2026}, month={10}, keywords={Attention Mechanism Convolutional Networks Fault Diagnosis Gram Angle Field Gears}, doi={10.1007/978-3-032-33662-0_40} }- YunTao Zhao
Zengyi Ma
WeiGang Li
Year: 2026
Application of Improved ConvNext and Fused Images of Gramian Difference Field and Sum Field in Gearbox Health Assessment
CHINACOM
Springer
DOI: 10.1007/978-3-032-33662-0_40
Abstract
To address the challenges of extracting gearbox fault features and ensuring stable fault identification accuracy, this paper presents a gear fault diagnosis method based on the fusion of Gram difference field (GADF) and Gram sum field (GASF) images, combined with an improved CCB-ConvNeXt convolutional neural network. First, the original one-dimensional vibration signals are transformed into GADF and GASF images, and their timing and amplitude information are fully integrated through weighted fusion. Then, the ConvNeXt network is enhanced by introducing ContMIX dynamic convolution, a CBAM attention mechanism, and a bidirectional attention gate (BAG) module, thereby improving context modeling and key feature extraction capabilities. Experimental results on the Southeast University (SEU) gear fault dataset demonstrate that the proposed method achieves a classification accuracy of 99.9%, outperforming mainstream convolutional neural network models. These results indicate that the method offers strong robustness, good generalization, and provides an efficient and stable reference for mechanical fault diagnosis.