Speaker
SHI Defranco Andrea
Description
This talk explores a novel approach to optimizing particle accelerators by integrating machine learning techniques with physics-informed simulations. Faced with the challenge of predicting beam halo behavior at extreme precision, the strategy shifts from seeking better simulation to smart, constrained optimization of real machine parameters. Case studies from DESY, LIPAc/IFMIF, and SACLA demonstrate fast, robust tuning and loss minimization, highlighting the potential for safe, efficient control in 1A-class accelerators and industrial applications.