Lightweight bearing fault diagnosis based on improved MobileNetV2
In response to the high computing cost,time cost and communication cost of existing deep learning fault diagnosis models,as well as the low diagnosis accuracy and weak anti-noise ability of lightweight models,which are unsuitable for deployment in edge environments,a lightweight bearing fault diagnosis method based on continuous wavelet transform(CWT)and improved MobileNetV2 was proposed.By converting the original signal into a time-frequency graph as an input sample by CWT,and combining the efficient channel attention(ECA)mechanism and MobileNetV2,the fault diagnosis accuracy and anti-noise ability of the model were improved,with low parameters and calculational load.Experimental results based on the bearing dataset of Case Western Reserve University show that,the proposed method can achieve an accuracy of 98.27％under a signal-to-noise ratio of 2 dB,which has better fault diagnosis accuracy and noise immunity while being lightweight compared with other models,and is more suitable for fault diagnosis in edge scenarios.
万方数据
维普
