Artificial Intelligence and Machine Learning push PIAVE-ALPI accelerator to record performance in 2024
May 2024: SPES Phase 1 completed
Visit of the Director of the Legnaro National Laboratories to the European Spallation Source
Artificial Intelligence and Machine Learning push PIAVE-ALPI accelerator to record performance in 2024
In June 2024, after concluding an extensive experimental campaign, PIAVE-ALPI superconductive accelerator reached outstanding performance levels. Main points for this achievement were the integration of advanced AI (Artificial Intelligence) with ML (Machine Learning) techniques, Bayesian optimization algorithms and particle swarm optimization algorithms. These complex algorithms played an essential role in parameters optimization of accelerator configuration such as cavities phases, longitudinal and transverse optics, resulting in remarkable accelerator performance improvements and in the halving of accelerator setting time.
ALPI overall transmission reached values of 85%, very near to 93% theoretical maximum transmission and significantly higher than 35% reached in the previous decades.
Essential for this record were improvements implemented on ECR ion source, on RF system, on alignment system, on cryostats, on control system and on beam diagnostics, but above all the huge teamwork carried out by three LNL divisions.