变转速下L_1,1,2范数与张量核范数联合约束的TRPCA滚动轴承故障特征提取方法

As one of important components of rotating mechanical equipment, rolling bearing's working status directly affects the operational safety of rotating equipment. Therefore, its fault feature effective extraction is of great significance for ensuring normal operation of mechanical equipment. In practical applications, rolling bearings usually operate at varying rotating speed, and non-stationary signals collected by a single sensor are often covered by severe background noise to make it very difficult to extract fault features. Here, to solve this problem, a tensor robust principal component analysis (TRPCA) rolling bearing fault feature extraction method with joint constraints of Lx, 2 norm and tensor kernel norm under variable rotating speed was proposed. Firstly, the time frequency representation (TFR) was taken as a forward slice to construct tensors and explore the tube sparsity of time-varying fault features of rolling bearing in tensor domain and the low tube rank property of background noise in tensor domain. Then, the TRPCA with joint constraints of Lx j 2 norm and tensor kernel norm was used to extract fault feature tensors and obtain fault feature tensors with tube sparsity. Finally, the extracted fault feature tensors were fused in channel index to obtain a TFR which could effectively represent fault features. Simulation and experimental results verified the effectiveness of the proposed method in bearing fault feature extraction.