Description
The extraction of a uniform and stable beam at a divergence below 7 mrad is a critical requirement for RF-driven negative ion sources used in neutral beam injection (NBI) systems planned for ITER, operating with both hydrogen and deuterium isotopes. Meeting this requirement calls for systematic assessment of the beam characteristics and its reproducibility under well-defined operating conditions. For that purpose, experiments are performed at the BATMAN Upgrade test facility, where beam footprints are diagnosed using a Carbon Fibre Composite (CFC) target and infrared imaging. The CFC images acquired during hydrogen and deuterium operation are analyzed using conventional fitting techniques, complemented by machine-learning-based methods. This combined approach allows correlations between beam parameters and their reproducibility to be thoroughly explored, beyond qualitative visual inspection of the beam footprints. Machine-learning-assisted analysis provides additional, subtle descriptors of the beam footprints, enabling robust comparisons across operating conditions.