基于CWT-MDFA的轴承故障诊断方法OA

In response to the difficulty in capturing complex features of weak early fault signals in roll-ing bearings comprehensively,and the limitations of traditional fault diagnosis methods,such as constrain-ed feature extraction and suboptimal fault detection accuracy,a new intelligent fault diagnosis method for bearings is proposed,which is based on wavelet transform and multi-scale dilated convolution for bearing fault diagnosis.The original one-dimensional vibration signals are converted into two-dimensional time-frequency representations via wavelet transform using a wavelet function.An improved multi-scale dilated convolutional network is then constructed to enlarge the receptive field of the neural network,mitigating the potential loss of local information during feature extraction.Additionally,spatial and channel attention mechanisms are introduced to address sensitivity issues in time-frequency representations under different damage severities of the same fault type.Experimental validations are conducted on the CWRU and JNU bearing datasets,and the generalization experiments are performed on the Southeast University bearing dataset.The results show that the proposed method can accurately identify operational information of bear-ings under different fault conditions and severity levels,achieving an accuracy of up to 99.07%,and exhib-its strong generalization capability and robustness.