Dynamic Reweighted Domain Adaption for Cross-Domain Bearing Fault Diagnosis

Certain progress has been made in fault diagnosis under cross-domain scenarios recently. Most researchers have paid almost all their attention to promoting domain adaptation a common space. However, several challenges that will cause negative transfer ignored. In this paper, reweighting method is proposed overcome difficulty from two aspects. First, extracted features differ greatly one another positive transfer, and measuring the difference important. Measured by conditional entropy, weight of adversarial losses for those well aligned are reduced. Second, balance between class discrimination influences transferring task. Here, dynamic strategy adopted compute factor. Consideration perspective maximum mean discrepancy multiclass linear discriminant analysis. The first item supposed measure degree source target domain, second show classification performance classifier on learned current training epoch. Finally, extensive experiments bearing datasets conducted. shows our model an obvious advantage compared with other algorithms.