Rolling bearing fault diagnosis based on t-SNE-GA-LSSVM

In order to improve the rolling bearing fault diagnosis accuracy, a rolling bearing fault diagnosis method with Variational Modal Decomposition (VMD), t-distributed domain embedding (t-SNE) and genetic algorithm optimized least squares support vector machine (GA-LSSVM) is proposed. Firstly, the bearing signal is decomposed into several intrinsic modal components (IMFs) by applying VMD; secondly, the time-frequency eigen indices of each IMF are calculated to compose the high-dimensional fault features, and the t-SNE algorithm is utilized for the secondary feature extraction, which reduces the influence of redundant features on fault diagnosis. Finally, the low-dimensional sensitive features are used as inputs to the GA-LSSVM model. Experimental analysis shows that the model based on t-SNE-GA-LSSVM has high fault diagnosis accuracy, good stability