A Deep Domain-Adversarial Transfer Fault Diagnosis Method for Rolling Bearing Based on Ensemble Empirical Mode Decomposition

In recent years, the deep learning-based fault diagnosis methods for rotating mechanical equipment have attracted great concern. However, because data feature distributions present differences in applications with varying working conditions, learning models cannot provide satisfactory performance of prediction such scenarios. To address this problem, paper proposes a domain adversarial-based rolling bearing transfer model EMBRNDNMD. First all, an EEMD-based time-frequency graph (EEMD-TFFG) construction method is proposed, and information nonlinear nonstationary vibration signal extracted; secondly, multi-branch ResNet (MBRN) structure designed, which used to extract features representing state from EEMD-TFFG; finally, solve adaptation problem under adversarial network module MK-MMD distribution difference evaluation are introduced optimize MBRN, so as reduce probability between source target domain, improve accuracy EMBRNDNMD domain. The results experiments carried out on two test platforms prove that can maintain average above 97% tasks, also has high stability strong ability scene adaptation.