Research on performance degradation assessment of rolling bearings

The performance degradation assessment (PDA) in rolling bearings through vibration signals is crucial for the successful implementation of preventive maintenance strategies. However, traditional indicators in existing methodologies often fail to effectively characterize the degradation trend, and feature extraction is heavily reliant on specific experience and operational parameters. Addressing these limitations, this paper introduces a novel PDA method. Initially, the selected features under normal operating conditions are selected based on the correlation coefficient method. Subsequently, the dimensionality of the feature space is then reduced using the Kernel Principal Component Analysis (KPCA) algorithm. The refined comprehensive feature set is then input into the degradation assessment model, employing the novel method Support Vector Data Description (SVDD) to integrate these features into a coherent degradation indicator (DI). To fine-tune the SVDD parameters, including the penalty parameter and kernel parameter, the Hippopotamus Optimization (HO) algorithm is incorporated into the optimization process. Experimental results validate the method in degradation detection and in accurately reflecting the overall performance degradation of the bearings, showcasing distinct advantages over other existing methods.