Strategy for Bayesian Optimized Beam Steering at TRIUMF-ISAC's MEBT and HEBT Beamlines

TL;DR

This paper develops a Bayesian optimization strategy for beam steering at TRIUMF's ISAC facility, integrating machine learning to enhance semi-automated tuning of complex particle accelerators.

Contribution

It introduces a novel Bayesian optimization approach tailored for beamline tuning, improving automation and precision in particle accelerator operations.

Findings

Successful implementation of Bayesian optimization for beam steering.

Enhanced tuning efficiency compared to traditional methods.

Validation through multiple machine development experiments.

Abstract

In preparation for operation of multiple Rare Isotope Beams (RIBs) when the Advanced Rare Isotope Laboratory (ARIEL) becomes operational, TRIUMF embarked on a program of advanced beam tuning applications and machine learning tools. The strategy for operationalizing Bayesian Optimization for beam steering purposes is being developed. A previously reported centroid correction algorithm is used to tune accelerated charged particle beams at TRIUMF's ISAC postaccelerator facility. We present findings and results from multiple machine development experiments conducted between October and November 2024, as part of a pivot toward semi-automated machine tuning methods. These findings were instrumental in shaping the tuning strategy for the medium and high energy beam transport (MEBT, HEBT) lines at ISAC, by sequentially optimizing sub-sections of the beamlines.