Peripheral coarse grain prediction in extruded AA6082: Combining finite element simulations with neural networks

Peripheral Coarse Grain (PCG) is a critical defect that affects the mechanical and crash performance of extruded AA6XXX aluminum profiles, particularly in automotive applications. Traditional methods to address this issue rely on extensive experimental campaigns, which are resource-intensive and often lead to conservative process parameters, reducing production efficiency. This study develops and validates a predictive model for PCG formation, combining finite element method (FEM) simulations and machine learning (ML) techniques. Data from FEM simulations and experiments were used to train and test a model employing artificial neural networks (ANNs) for PCG prediction. The proposed approach enables accurate PCG forecasting, providing a robust tool for optimizing process parameters, reducing reliance on empirical methods and advancing smart manufacturing solutions.