Speaker
Description
Optimization of beam transport systems in linear accelerators is a nonlinear and computationally intensive task due to the complex interdependencies between beam properties and machine parameters. In this work, a physics-motivated surrogate modeling approach is developed to optimize the processes of the best accelerator parameters exploration. The proposed framework approximates particle-tracking simulations using a neural surrogate architecture designed to reflect the desired skeleton of the beam transport system. The model predicts beam evolution at several longitudinal locations corresponding to key regions of the beamline, allowing the surrogate to capture the sequential snapshots of beam dynamics cases. The trained model is then used within a constrained optimization procedure to minimize normalized transverse emittance while maintaining acceptable beam size along the beamline. The AREAL linear accelerator (at CANDLE SRI) is used as a validation case to demonstrate the effectiveness of the method.
| I have read and accept the Privacy Policy Statement | Yes |
|---|