基于物理引导双分支网络的轴承故障诊断研究OA

To address the challenge in current rolling bearing fault diagnosis where single vibration signal representations struggle to stably extract mechanism-related features,this paper proposes a Physics-Based Guided Bi-Branch Network(PBINet)to enhance fault feature recognition performance during fault diagnosis.The proposed method takes the raw time-domain signal and a transformed envelope spectrum as inputs,uses two convolutional branches to extract impact features and band-structure features,respectively,and introduces a gated fusion mechanism to achieve channel-adaptive weighting,thereby completing complementary feature fusion and discriminative learning.Experiments conducted on the CWRU and uOttawa datasets show that the proposed method achieves diagnostic accuracies of 99.2%and 100%,respectively.The results indicate that the proposed method can realize high-accuracy fault diagnosis under complex operating conditions.