IPAC'26 - the 17th International Particle Accelerator Conference

Hardware Aware Artificial Intelligence (HAAI): Progress on Realtime Beam Tomography Reconstruction for the Fermilab Recycler using Machine Learning and Edge Processing

MOP6316

18 May 2026, 16:00

2h

C.I.D

Poster Presentation MC6.D13: Instrumentation: Artificial Intelligence Poster session

Speaker

Kyle Hazelwood (Fermi National Accelerator Laboratory)

Description

The resistive wall current monitor (RWCM) data from the Fermilab Recycler Ring (RR) is used to re-construct the longitudinal profile of proton beam circulating in the machine. This procedure, commonly referred to as tomography, has proved to be invaluable in tuning the machine. In 2013 Recycler was re-purposed as a proton stacker and charged with implementing slip-stacking to double the intensity for the Main Injector (MI). With two slipping beams, the RWCM data from Recycler is difficult to differentiate and the tomography procedure is slow to compute using traditional means. Building upon past efforts, the Hardware Aware Artificial Intelligence (HAAI) project aims to develop a Machine Learning (ML) model to reconstruct the Recycler beam tomography in realtime and deploy this model on edge hardware. Once developed, this new streaming virtual diagnostic would be used to better track the beam parameters over larger time spans, attribute settings to beam effects, tune the machine, and provide an input into other future automations of the machines.