Speaker
Description
We propose to develop advanced ML models, such as physics informed neural network (PINN) based surrogate models, to accurately represent accelerator phase space transport. These surrogate models will enable precise diagnosis and prediction of beam phase space evolution along the beamline, facilitating real-time control and optimization. The developed models will be tested using the Upgraded Injector Test Facility (UITF) at Thomas Jefferson National Accelerator Facility (JLab), providing a pathway toward ML-driven enhanced diagnostics and beamline control in operational accelerator environments. The primary aim will be to facilitate this by developing machine learning models that outperform traditional simulations in speed and precision. We will build a virtual beamline, train a reinforcement learning (RL) controller across varied calibration scenarios, and then transfer it to the real machine. Beyond operation, fast and accurate models are also essential for design optimization workflows using machine learning methods that iterate through design parameters. A long-term goal of this work will be to establish such workflows and apply them to the design of a compact accelerator at Old Dominion University (ODU).
| In which format do you inted to submit your paper? | LaTeX |
|---|---|
| Preprint marking on your proceeding paper | I do not wish my paper to be marked as preprint. |