数据驱动下基于改进生成对抗网络的滚动轴承故障诊断方法OA

To avoid the impact of rolling bearing faults on the normal operation of large wind turbines,research is conducted on the diagnosis and analysis of rolling bearing faults in wind turbines.Considering the complex feature extraction of traditional wind turbine rolling fault diagnosis,which can only distinguish true and false,and cannot identify fault types,a rolling bearing fault diagnosis method based on improved generative adversarial network for fault data is proposed.On the basis of explaining the working principle of traditional generative adversarial network,the JS divergence in the traditional generative adversarial network is replaced by the Wasserstein distance,forming the Wasseratein GAN-GP.Based on this,a rolling bearing fault diagnosis model is constructed,and the diagnosis model is optimized using gradient penalty method to solve the problem of gradient vanishing during model training.On the basis of learning the distribution of raw data,the training sample set is further expanded.Finally,experiments are conducted under the condition of imbalanced data,and the results show that the proposed rolling bearing fault diagnosis method is practical and effective,which can effectively reduce the impact of imbalanced and insufficient training data on fault diagnosis.