Abstract：With the increasing application of additively manufactured Ti-6Al-4V alloys in aerospace and high-performance structural components, understanding their fatigue behavior under complex loading conditions has become essential. This study proposes a data-physics hybrid fatigue life prediction model that integrates Mises equivalent stress as prior physical information to enhance the modeling of multiaxial fatigue mechanisms. Focusing on L-PBF Ti-6Al-4V material, three representative data-driven methods—artificial neural networks, random forests, and support vector machines—are compared. Furthermore, the effects of different physics-informed strategies, including feature engineering, loss function design, and residual connections, are systematically evaluated. Results show that the proposed physics-informed residual network achieves higher predictive accuracy and improved physical consistency, particularly in low-cycle regimes and under multiaxial loading. The findings offer valuable insights for developing physically reliable fatigue life prediction models for advanced structural materials.
2025, Vol. 46 
