A sound-vibration physical-information fusion constraint-guided deep learning method for rolling bearing fault diagnosis

• Bearing failure rules are learned and guided by multi-physical fusion constraints. • Real physical information is ensured by introducing a 15-DOF kinetic physical model. • Self-calibration of sound-vibrational physical-information margins is met by PF. • Dual learning of physical and data features for fault diagnosis is enabled by PFCG. • Model explanability is analyzed by feature visualization and sensitivity analysis. Although current deep learning models for bearing fault diagnosis have achieved excellent accuracy, the lack of constraint-guided learning of the physical mechanisms of real bearing failures and a physically scientific training paradigm leads to low interpretability and unreliability of…

• NN
  National Natural Science Foundation of China

  Awards: 61963018, 52165012
• NS
  Natural Science Foundation of Jiangxi Province

  Awards: 20202BAB214017, 20212ACB202004