Rolling Bearing Fault Diagnosis Method Based on Eigenvalue Selection and Dimension Reduction Algorithm

Abstract

The bearings used in the mechanical equipment that bear and transfer the load are vulnerable parts. In this paper, a rolling bearing fault diagnosis method based on eigenvalue selection and dimensionality reduction is presented. This is suitable for analyzing fault signals with nonstationary characteristics, and it has a good recognition rate. The characteristic quantity of vibration signals in the time domain and the frequency domain is calculated, and the characteristic quantity is selected by calculating the degree of difference. A dimension reduction algorithm is proposed, which is based on a neural network and ISOMAP. Its performance is compared using PCA, LTSA, and ISOMAP algorithms. Fault diagnosis is carried out by using KNN and SVM classification algorithms, and good recognition results are obtained.