ODE-CNN: A Fault Diagnosis Model for Complex Data of Rolling Bearings

To address the limitations of traditional convolutional neural networks (CNNs) in managing complex rolling bearing data, this paper proposes an Ordinary Differential Equation-Convolutional Neural Network (ODE-CNN) model to improve the capture of dynamic fault signal characteristics. The ODE-CNN model integrates ordinary differential equations (ODEs) into the CNN structure, enabling continuous dynamic modeling of the local features extracted by the convolutional layers to effectively capture the signal's global dynamic behavior and long-term dependencies. Experimental results demonstrate that the ODE-CNN model achieves superior classification accuracy and feature representation capability in rolling bearing fault diagnosis tasks compared to traditional CNNs, confirming its effectiveness in processing nonlinear and non-stationary signals.