Recognition of rolling bearing life status based on GI performance degradation indicator and AGNCN

Abstract

To address the complexities of bearing performance degradation assessment, the challenges of life status identification in small sample scenarios, and the issue of unclear diagnostic mechanisms, a rolling bearing life status recognition method is proposed based on the Gini Index (GI) and physical features weighted Adaptive Gate Neurons Convolutional Network (AGNCN). Firstly, the Gini Index (GI) is employed to assess the degradation process of bearing performance and categorize the bearing life status. Secondly, an Adaptive Gate Neurons model with physical features weighted is proposed. Specifically, we have designed the Adaptive Gate Neuron with physical features weighted that adaptively adjusts the combination mode of quadratic convolutions. Furthermore, we have established the mathematical relationship between the learnable weighted autocorrelation and the physically features weighted Adaptive Gate Neuron, theoretically proving its superiority in feature extraction capabilities, thereby which enables it to assign higher weights to the periodic features in the time-domain signals of bearings at different life status. Based on this, we propose the physically features weighted Adaptive Gating Neurons Convolutional Network (AGNCN) combined with a residual structure. Then, we conducted a thorough analysis of the internal mechanisms of the physically features weighted Adaptive Gate Neuron, revealing an embedded attention mechanism that endows the AGNCN model with inherent interpretability. Finally, experiments were carried out using the full-life dataset of rolling bearings. The results demonstrate that the proposed method can effectively and reliably identify the life status of rolling bearings.