Artificial Intelligence-based Modeling of Hot Deformation Behavior in Aa 5052-h32 Alloy

This study aims to model the hot plastic deformation behavior of the AA 5052-H32 aluminum alloy using both traditional and Artificial Intelligence (AI)-based approaches. Hot tensile tests were carried out using the Gleeble® 540 thermomechanical simulator at different temperatures (100–450°C) and strain rates (0.01 and 1 s⁻¹), enabling the acquisition of the material's flow curves. Initially, the experimental data were fitted using the Hensel-Spittel constitutive equation, which is widely employed to simulate hot forming processes. Subsequently, predictive models based on AI were developed using Artificial Neural Networks (ANN) and the eXtreme Gradient Boosting (XGBoost) algorithm. The ANN was optimized with different architectures and activation functions, achieving a mean absolute error (MAE) of 3.61 MPa and a mean squared error (MSE) of 19.09 MPa². The XGBoost model outperformed the ANN, with a MAE of 0.74 MPa, MSE of 2.46 MPa², and a coefficient of determination (R²) of 0.9988. The results indicate that both approaches are effective for predicting the hot flow behavior of the AA 5052-H32 alloy, with XGBoost showing particular efficiency in modeling complex nonlinear relationships between strain, strain rate, and temperature.