基于CFA参数优化DAE方法的滚动轴承故障诊断

Research on Rolling Bearing Fault Diagnosis Based on CFA Parameter Optimization DAE Method

为了进一步提高滚动轴承的故障诊断效率,设计了一种混沌萤火虫参数优化去噪自动编码器(Denosing auto-encoder,DAE)轴承故障诊断方法,标记为CFADAE。选择核去噪自动编码器构建深度神经网络,从初始信号中提取得到首层特征,经过自动编码器堆叠处理形成深度网络,实现深层特征参数的准确提取。研究结果表明:利用FA参数确定的准确率为92.82%,标准差为1.45,相对CFA方法95.94%准确率发生了降低,表明混沌改进萤火虫算法进行诊断的准确率更高,并且能够大幅减小波动幅度。相比较其他的方法,DAE网络的诊断平均准确率达到95.84%,表明去噪自动编码器与L2惩罚项能够促进模型鲁棒效果与泛化性的显著提升,由此获得更优的诊断精度。该研究可以拓宽到其它机械传动故障诊断领域,具有很好的推广应用价值。

In order to further improve the fault diagnosis efficiency of rolling bearings,a bearing fault diagnosis method of chaotic firefly parameter optimization denoising auto-encoder(DAE)is designed.The core denoising autoencoder is selected to construct the deep neural network,the first layer features are extracted from the initial signals,and the deep network is formed through the stacking processing of the auto-encoder to achieve the accurate extraction of the deep feature parameters.The results show that the accuracy and standard deviation of the standard FA parameter is 92.82%and 1.45,which is lower than that of the CFA method of 95.94%,indicating that the chaotic modified firefly algorithm has a higher diagnostic accuracy and can significantly reduce the amplitude of fluctuation.Compared with other methods,the average diagnostic accuracy of the DAE network reached 95.84%,indicating that the denoising auto-encoder and L2 penalty can significantly improve the robustness and generalization of the model,thus achieving better diagnostic accuracy.This research can be extended to other fields of mechanical transmission fault diagnosis and has good application value.