一种抗噪声的自注意力神经网络轴承故障诊断方法

Fault Diagnosis Method for Bearings Based on Anti-Noise Self-Attention Neural Network

为抑制噪声信号对轴承故障诊断的不良影响,应对轴承诊断方法在生产现场噪声干扰下精度下滑的问题,提出了一种对噪声具有强鲁棒性的故障诊断方法。首先,以自注意力神经网络为主要研究对象,对模型的深度、宽度进行分析,确定模型的宽度为16维,深度为8层时诊断效果最佳;然后,通过对比试验验证了位置编码模块在模型中的必要性;最后,与CNN,LSTM,MLP,SIM等模型的对比结果表明,基于自注意力神经网络的模型能够在低强度噪声环境下取得接近100%的诊断精度,在高强度噪声环境下的诊断精度也优于其他模型。

In order to suppress the negative influence of noise signal on fault diagnosis of bearings and solve the problem of declining accuracy of diagnosis method for bearings under noise interference in production site,a fault diagnosis method with strong robustness to noise is proposed.Firstly,the self-attention neural network is taken as main research object,the depth and width of model are analyzed,and the diagnosis effect is best when the width of model is 16 dimensions and the depth is 8 layers.Then,the necessity of position coding module in model is verified by comparative experiments.Finally,compared with CNN,LSTM,MLP,SIM and other models,the results show that the model based on self-attention neural network achieves nearly 100%diagnostic accuracy under low-intensity noise environments,and the diagnostic accuracy of the proposed model is better than that of other models under high-intensity noise environments.