Bearing Fault Diagnosis Method Based on Multi-Scale Fusion Denoising Capsule Network

Rolling bearings are key components in many mechanical equipment, and their performance directly affects the operating status of the entire equipment. However, in actual operating environments, bearings are often disturbed by various noises, which makes it extremely difficult to extract bearing fault characteristics. In response to these challenges, a capsule network model based on multi-scale fusion denoising mechanism is proposed. Multi-scale decomposition of vibration signals is performed to extract feature information at different frequencies, and the signal is denoised using attention thresholds to optimize feature expression. At the same time, the integration of capsule networks enhances the ability of model identification and expression of intrinsic data structures, thereby enhancing the effect of fault feature extraction. Through experimental verification, the model can extract fault features in complex environments, thereby offering a robust methodology for the detection of rolling bearing faults.