一种基于复合生成对抗网络的轴承故障诊断方法
By constructing a composite generative adversarial network model, adversarial samples with fault characteristics are generated, which solves the problems of vulnerability of deep neural networks to attacks and data imbalance in bearing fault diagnosis, and achieves high-precision and low-cost fault diagnosis results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANDONG COMP SCI CENTNAT SUPERCOMP CENT IN JINAN
- Filing Date
- 2023-09-04
- Publication Date
- 2026-07-17
AI Technical Summary
Deep neural networks are vulnerable to adversarial attacks in bearing fault diagnosis, and data class imbalance leads to reduced classification accuracy, making it difficult for existing technologies to achieve accurate and efficient fault diagnosis.
A composite generative adversarial network model is constructed. Adversarial examples are generated through a first generator and a second generator. An attention mechanism and a two-dimensional convolutional layer are combined to expand the dataset for training, generating adversarial examples with fault features for training the fault classifier.
It improves the generalization ability of the fault diagnosis model, achieves high-precision and low-cost fault type classification, solves the problem of insufficient data samples, and has good diagnostic testing performance.
Smart Images
Figure CN117168814B_ABST