An Intelligent Bearing Fault Diagnosis Model with Physics-Information Fusion: Design and Interpretability Mechanism Research
TL;DR
A convolutional neural network model incorporating spectral physical constraints is proposed that enhances the spectral interpretability of the model’s decisions, achieving a transition from a “black box” to a “white box” and effectively improving the reliability of deep diagnostic models.
Abstract
To address the issue of limited interpretability in current deep learning-based bearing fault diagnosis methods, this study proposes a convolutional neural network model incorporating spectral physical constraints. The model integrates data-driven learning with physical constraints by embedding the spectral features of bearing faults as prior knowledge into the network architecture during training. This guides the model to adaptively learn spectral features closely associated with fault mechanisms. Experimental results on both public datasets and self-collected data show that the proposed model not only maintains high diagnostic accuracy but also provides intuitive and credible justification for fault classification through the visualization of spectral responses at the network output layer. This enhances the spectral interpretability of the model’s decisions, achieving a transition from a “black box” to a “white box” and effectively improving the reliability of deep diagnostic models.