Digital Twin-Based Defect Size Updating for Bearing Quantitative Diagnosis

Bearing quantitative diagnosis is indispensable for the reliability assessment of rotating machinery, facilitating informed maintenance decisions. Existing research typically estimates bearing defect size using signal processing techniques or machine learning tools. However, these approaches rely on elaborate hand-crafted feature extraction or require extensive training data from various fault states, making them difficult to apply in practice. In this article, a novel digital twin-based defect size updating (DT-DSU) scheme is proposed to address these challenges. As a combination of simulation and virtual-reality interaction, digital twin (DT) technology provides a fresh perspective for bearing quantitative diagnosis. A modeling-calibration-updating strategy within the DT framework is adopted to enable unsupervised defect size estimation. First, DT-DSU constructs a DT prototype by embedding fine-grained defects into a simplified dynamic model of bearing. Time-frequency similarity calibration is conducted to obtain a high-fidelity DT model. Then, unlabeled vibration measurements are exploited to build the virtual-reality connection, enabling fault state updating of the DT model. The defect size is determined through the update process. Finally, the effectiveness and superiority of DT-DSU are validated on two different bench tests.