A fault diagnosis method based on label-wise density-domain space learning

Abstract

Nonlinear space learning of fault samples is a category of common fault diagnosis methods, which usually use Euclidean distances to describe manifold structures among fault samples. However, in the nonlinear space learning, Euclidean distances lead to a potential manifold loss problem, and density structures that can constrain relationships from different viewpoints. Aiming these issues, we propose a novel fault diagnosis method based on label-wise density-domain space learning, which can learn a label-wise density-domain space with more intrinsic manifold structures. Density-constrained order graphs constructed by our method integrates different discriminative relationships from original fault samples with the help of density-domain information, and the density-domain information can effectively capture potential density information and global structure between fault samples. By density Laplacian of the graphs, we further construct a label-wise density-domain manifold space learning model, and the analytical solutions of space projections can be obtained by solving the model. Fault features directly obtained by the space projections possess well class separability. Extensive experiments on the Case Western Reserve University fault dataset and a roll bearing fault dataset from our roll bearing test platform shows the effectiveness and robustness of our method.