LAUSR

Vibration signal always performs poorly in the fault diagnosis of low-speed rolling bearings. The fact that rolling bearings running under different speed conditions further increases the difficulty of fault diagnosis on low-speed bearing. To address the above problems, this article proposes a fault diagnosis method for low-speed rolling bearings based on acoustic emission (AE) signal and subspace embedded feature distribution alignment (SADA). First, the AE signal of low-speed rolling bearing is collected and the spectral dataset is constructed. Second, subspace alignment is used to align the basis vectors for both domains in order to prevent feature distortion. Then, a base classifier is trained to predict the pseudolabels of the target domain, which is used to quantitatively estimate the weight of the edge distribution and conditional distribution of the two domains for adaption. Finally, following the structural risk minimization (SRM) framework, a kernel function is constructed to establish the classifier f, which iteratively updates the pseudolabels in the target domain and obtains the coefficient matrix of the final framework to complete the identification task. The feasibility and effectiveness of the proposed method are verified by two AE datasets of low-speed rolling bearing.