Speaker
Description
The Generative phase space reconstruction (GPSR) method, based on neural networks and differentiable simulations, has been actively investigated to obtain comprehensive information of 6-dimensional phase space of particle beams. At the Pohang Accelerator Laboratory X-ray Free Electron Laser (PAL-XFEL) facility, 6-dimensional phase spaces along the first bunch compressor section were successfully obtained using the GPSR method. Through the successful AI/ML-based beam diagnostics, we plan to optimize the beamline to preserve the beam quality for high-brightness FEL. By using the reconstructed phase space as an input to simulations, we perform optimizations to achieve optimal settings such as quadrupole strengths. Even though we found the optimal condition, it is essential to calibrate the simulation model to accurately represent the real machine. Therefore, we present simulation model calibration based on the reconstructed phase space and experimental datasets with a variety of beamline settings. We further show the uncertainty of the model associated with the measurement. Through this model calibration, we finally present the optimal beamline setting that preserves the beam parameters such as emittances along the undulator section. We expect that the calibrated model can be incorporated into the digital twin frameworks for real-time virtual diagnostics and robust beam optimization for different FEL modes.
| I have read and accept the Privacy Policy Statement | Yes |
|---|