A Rolling Bearing Fault Diagnosis Method Based on Complementary Fusion of Multidomain Features

Feature extraction is a critical step in fault diagnosis. In order to address the limitations of fault diagnosis methods based on single-domain feature extraction, which rely on the quality and quantity of data samples and suffer from insufficient information extraction and limited generalization capabilities, a fault diagnosis method for rolling bearings based on multidomain feature complementary fusion is proposed. First, recursive, time-domain, and frequency-domain features are extracted from the vibration signals, and the three domain features are fused to construct the original feature set. Considering that the fused feature set contains numerous irrelevant and redundant features, an improved distance evaluation (IDE) criterion is introduced to select relevant features from the original set, forming a sensitive feature subset. Finally, this sensitive feature subset is inputted into a classifier for fault diagnosis. This method is applied to rolling bearing datasets provided by Paderborn University in Germany and Jiangnan University. Fault diagnosis was performed on these datasets using common classifiers, such as support vector machine (SVM) and random forest (RF). The results indicate that multidomain fused features not only outperform single-domain features but also maintain robust diagnostic performance across different classifiers and datasets.