一种基于自适应提升小波与卷积KAN网络的轴承故障诊断方法
By combining adaptive lifting wavelet and convolutional KAN network, the problems of weak noise resistance, large feature extraction deviation and loss of high frequency details in bearing fault diagnosis are solved, and high-precision and robust fault identification is achieved, which has significant advantages, especially in complex working conditions.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NORTHEASTERN UNIV CHINA
- Filing Date
- 2026-05-22
- Publication Date
- 2026-07-17
AI Technical Summary
Existing deep learning methods for bearing fault diagnosis suffer from problems such as weak noise resistance, large feature extraction bias, insufficient nonlinear expression ability, loss of high-frequency detail information, and non-optimal feature space distribution, which limit the recognition accuracy and generalization ability under complex working conditions.
We employ a combination of adaptive lifting wavelet and convolutional KAN network. Initial features are extracted using large convolutional kernels, and local convolutional and spline nonlinear KAN residual blocks are fused. A learnable soft threshold denoising wavelet module is used for feature decomposition and dynamic denoising. Furthermore, a prototype-aware contrastive learning mechanism is introduced to optimize the loss function, thereby achieving multi-scale feature fusion and classification.
It significantly improves the robustness and accuracy of the model under complex working conditions, effectively identifies weak fault features, and enhances classification accuracy and generalization ability, especially maintaining high recognition sensitivity and stability in strong noise environments.
Smart Images
Figure CN122241387B_ABST