Speaker
Description
We propose an AI-driven control architecture that sustains resonance among magnetrons and SRF cavi-ties, enabling efficient accelerator operation. Injection locking will be achieved by extracting pick-up signals from normal-conducting cavities integrated into a beamline with the main SRF cavities. A bunched elec-tron beam behind injector section will excite these resonators to deliver controllable extracted RF power to the magnetrons. This approach eliminates the need for external RF sources to obtain phase locking. AI will optimize cavity tuning to preserve the phase-locked condition, and AI-driven Ferro-Electric Tuners (FRTs) will compensate microphonics and other detun-ing effects in accelerating SRF cavities. We highlight the complexity of controlling FRTs, which require both fast (sub-millisecond) bias-voltage signals and slower temperature-driven adjustments from an integrated chiller. Fast control covers microphonics compensa-tion and magnetron power amplitude, while tempera-ture control expands tuning range to compensate for slow fluctuations. A human-in-the-loop approach is impractical for such a system. We therefore plan to develop and deploy high-fidelity AI-driven digital twins to enable dynamic, in-situ linac control and op-eration, ultimately reducing linac cost, improving beam quality, and increasing reliability by mitigating undesired trips.
| I have read and accept the Privacy Policy Statement | Yes |
|---|---|
| Paper submission software | MSWord |