Speaker
Description
Precise measurement of the 4D phase-space distribution (x, x', y, y') is essential for optimizing accelerator performance and ensuring stable operation. However, the prolonged measurement time of traditional scanning-based diagnostics limits their application in real-time tuning. In this study, we propose a deep learning-based surrogate model for the rapid reconstruction of heavy-ion beam distributions in the Low Energy Beam Transport (LEBT) section at RAON.
Deep learning integrated with physics simulations was adopted in this study to achieve physically reasonable prediction. This method enables the reconstruction of the 4D phase-space distribution at the exit of the ion source, providing a robust framework for real-time diagnostics and autonomous tuning for accelerator systems. The model is trained and validated using high-fidelity particle tracking simulations to ensure numerical stability and minimize statistical bias. This approach significantly reduces the computational overhead compared to conventional tracking codes, allowing for instantaneous beam characterization required for dynamic machine protection and optimization.
| I have read and accept the Privacy Policy Statement | Yes |
|---|