Creep-Fatigue Life Prediction of FGH96 Based on Advanced Physical-Information Neural Network

A creep-fatigue life prediction model based on a Physics-Informed Neural Network (PINN) is proposed to accurately evaluate the service life of FGH96 turbine disc materials under complex loading conditions. This method combines the powerful nonlinear fitting ability of neural networks with physical boundary constraints, such as frequency correction models, to improve the predictive performance of the model while ensuring that it conforms to the actual physical mechanisms. The experimental results show that the predicted lives by this method all fall within 1.8 times the scatter band of the experimental life, and the life prediction of FGH96 alloy by this method is more consistent with the experimental results compared with the traditional creep-based fatigue life model and the purely data-driven model.