Fault Diagnosis Algorithm of Gearboxes Based on GWO-SCE Adaptive Multi-Threshold Segmentation and Subdomain Adaptation

The data distribution of the vibration signal under different speed conditions gearbox is different, which leads to reduced accuracy fault diagnosis. In this regard, paper proposes a deep transfer diagnosis algorithm combining adaptive multi-threshold segmentation and subdomain adaptation. First all, in acquisition stage, non-contact, easy-to-arrange, low-cost sound pressure sensor used collect equipment signals, effectively solves problems contact installation limitations increasingly strict layout requirements faced by traditional signal-based methods. continuous wavelet transform (CWT) then convert original device into time–frequency image samples. Further, highlight target characteristics samples, gray wolf optimization (GWO) combined with symmetric cross entropy (SCE) perform on A convolutional neural network (CNN) extract common features source domain samples Additionally, local maximum mean discrepancy (LMMD) introduced parameter space fully connected layer align sub-field edge so as reduce difference sub-class working improve diagnostic model. Finally, verify effectiveness proposed method, preset experiment variable carried out. results show that compared other methods, method has higher superiority.