104 / 2023-09-19 18:35:03

Phase Space Smooth Mode Decomposition for Bearing Fault Diagnosis

Phase space reconstruction,Smooth mode decomposition,Rolling bearing features,Convolutional neural network

Special Sessions > SS13 - Signal Decomposition-Integrated Mechanical Fault Detection and Remaining Useful Life Prediction

In the field of rolling bearing fault diagnosis, significant attention has been devoted to modeling and identifying parameters for the vibration dynamic system. Aiming at the problem that rolling bearings are difficult to model. In this study, we employ a novel approach by extending the Smooth Mode Decomposition (SMD) algorithm into the phase space (PS) domain to perform a qualitative analysis of rolling bearing dynamic systems. Firstly, the vibration signals recorded during rolling bearing operation are embedded into reconstructed PS. Subsequently, the high-dimensional PS trajectories are subjected to smoothness constraints to extract the modal structure and exclude disturbances. Finally, the convolutional neural network (CNN) model is constructed to learn the PS modes and diagnose intelligently. The experiment validates the efficacy of this approach, demonstrating its capability to accurately discern PS modes corresponding to various fault types in rolling bearings. The proposed approach introduces a promising avenue for the intelligent diagnosis of rolling bearing faults.