Digital twin for gear contact fatigue with contact affine parametrization

Gear contact fatigue, initiating from subsurface microcracks, evades vibration-based detection due to its lack of distinctive signal patterns, creating an urgent need for virtual sensing solutions. To address the computational bottleneck of high-fidelity simulations in digital twin applications, this study proposes an efficient framework for digital twin. At the core of the framework is the contact affine parameterization method (CAPM) utilizing perturbation decomposition and affine parameterization to decouple offline and online computations. CAPM achieves a computational efficiency approximately 2000 times higher than that of conventional FEM while maintaining comparable accuracy. A probabilistic failure criterion is further introduced to quantify contact fatigue crack states under dynamic loads, accounting for size effects and material uncertainties. Based on that, a physics-embedded digital twin model is developed for virtual sensing of gear contact fatigue, enabling accurate condition monitoring under data-scarce scenarios. Validation via rough-surface contact analysis and gear pitting tests demonstrates the model's high computational efficiency and prediction accuracy across varied operational conditions. This physics-embedded digital twin approach offers a robust solution for lifecycle management of gear systems, proving particularly valuable in data-scarce conditions.