Isotope Separator On-Line system tuning: Bayesian optimization applied to the transport beamline case

Optimizing Isotope Separation On-Line (ISOL) systems requires tuning to find the best values for many correlated parameters, traditionally performed by experienced operators. This process is time-consuming and often suboptimal due to the large number of parameters involved. Optimization algorithms have emerged as valuable tools to support the tuning process, although their application has primarily focused on accelerators. This paper presents experimental results on optimizing the transport beamline of the ISOLDE Offline 2 mass separator system at CERN. Instead of formulating beamline tuning as a multi-objective optimization problem, performance objectives are modeled as constraints, thereby reducing the problem to a single-objective constrained optimization. The results indicate that Bayesian optimization-based algorithms successfully identified beamline parameters that meet mass separation requirements at the specified resolution. Additionally, the findings validate the use of a Bayesian optimization algorithm with a data-informed Gaussian process, which consistently improves convergence and outperforms benchmark algorithms.