A Novel Fault Diagnosis method for Rotating Machinery of Imbalanced Data

In this paper, a novel classification approach for imbalanced data with high-dimensional and intra-class imbalance is proposed, and they applied to fault diagnosis of rotating machinery. It is noted that the most of existed work on imbalanced learning focus on the inter-class imbalance, and ignore the intra-class imbalance. To solve the classification of imbalanced data with high-dimensional and intra-class imbalance, we proposed an integrated data-based and feature-based algorithm, which combines hybrid feature dimensionality reduction with a varied density based safe level synthetic minority oversampling technique (VDB-SLSMOTE), transforming the imbalanced data into balanced data. The balanced data is classified by random forest, and the final experimental result verified the effectiveness of the algorithm.