19–24 May 2024
Music City Center
US/Central timezone

Slow extracted spill ripple control in the CERN SPS using adaptive Bayesian optimisation

TUPS55
21 May 2024, 16:00
2h
Blues (MCC Exhibit Hall A)

Blues

MCC Exhibit Hall A

Poster Presentation MC6.D13 Machine Learning Tuesday Poster Session

Speaker

Nico Madysa (GSI Helmholtzzentrum für Schwerionenforschung GmbH)

Description

The CERN Super Proton Synchrotron (SPS) offers slow-extracted, high-intensity proton beams at 400 GeV/c for 3 fixed targets in the CERN North Experimental Area (NA) with a spill length of about 5 seconds. Since first commissioning in the late seventies, the NA has seen a steady increase in users, many of which requiring improved spill quality control. Slow extraction is sensitive to small perturbations with the effect of reduced spill quality. While some of these effects have been addressed in recent years, continuous compensation of intensity fluctuations at 50 Hz harmonics originating from power converter ripple has been particularly difficult to achieve. In 2023, the deployment of two techniques - "Empty-Bucket Channeling" and active control with Adaptive Bayesian Optimization – resulted in a significant suppression of these intensity modulations. This paper focuses on using Adaptive Bayesian Optimization for 50 Hz harmonic control. The chosen algorithm is described, together with details of integration in the CERN control system. The 2023 results are presented and complemented with an overview of the next steps.

Region represented Europe
Paper preparation format LaTeX

Primary author

Verena Kain (European Organization for Nuclear Research)

Co-authors

Ewald Effinger (European Organization for Nuclear Research) Fabio Follin (European Organization for Nuclear Research) Francesco Velotti (European Organization for Nuclear Research) Matthew Fraser (European Organization for Nuclear Research) Michael Schenk (Ecole Polytechnique Fédérale de Lausanne) Nico Madysa (GSI Helmholtzzentrum für Schwerionenforschung GmbH) Pablo Andreas Arrutia Sota (Oxford University)

Presentation materials

There are no materials yet.