Exploratory modelling of collections effects in the FCC-ee High-Energy Booster using Machine Learning
Santiago Martinez, Ganil, CEA
Vendredi 21 aout à 10h30, CEA-Saclay, Orme-des merisiers, salle Cassini
Modern particle accelerators such as the Future Circular Collider (FCC-ee) require increasingly precise modeling of beam dynamics in regimes where nonlinear collective effects become significant. These effects, including space-charge forces, wakefields, and beam-induced fields, introduce complex correlations in the six-dimensional phase space distribution that are computationally expensive to simulate using traditional tracking methods. This internship focuses on the development of machine learning-based surrogate models for the efficient and physically consistent modeling of collective beam dynamics in the FCC-ee booster. In particular, the goal is to construct data-driven representations of the six dimensional beam distribution evolution under both single-particle Hamiltonian transport and collective effects. The proposed approach builds upon recent advances in operator learning and physics-informed neural networks. Neural operator architectures, such as Fourier Neural Operators, provide a natural framework for learning mappings between function spaces, making them well-suited to model the evolution of phase space densities rather than individual particle trajectories. In parallel, structure-preserving neural architectures inspired by symplectic maps (e.g. Henon-type networks) are considered to ensure consistency with the underlying Hamiltonian dynamics. In future developments, the dynamics will be decomposed into a Hamiltonian transport operator and a collective interaction operator, allowing for efficient composition of learned evolution steps. This hybrid approach aims to provide fast, accurate, and physically interpretable surrogate models for beam dynamics, enabling improved simulation, control, and optimization of next-generation accelerator systems.
Unsupervised Machine Learning for Fault detection in turn by turn Beam Position Monitors
Jakub Potaczala, IUT Orsay
Jeudi 16 juillet à 13h30, CEA-Saclay, Bat 130, pièce 102
Reporté
Turn-by-turn beam position monitors (TbT-BPMs) are essential for reconstructing the magnetic optics of high-luminosity colliders, yet faulty sensors and measurement noise routinely compromise data quality. As accelerators scale toward facilities like the Future Circular Collider (FCC), with thousands of BPMs distributed across a 91-km ring, manual inspection and hardware redundancy no longer suffice, motivating fully automated, unsupervised diagnostic pipelines. This seminar presents the improvements developed during an internship to a tool under development at CEA in collaboration with LISN at CentraleSupelec. The tool is used to study and compare complementary unsupervised machine learning approaches to TbT-BPM data from SuperKEKB's High and Low Energy Rings (HER/LER). A new pipeline including Isolation Forest, DBSCAN-based clustering coupled with two feature-engineering philosophies (a fully data-driven approach, Time2Feat, and a physics-informed fixed-feature approach) and a 1D convolutional autoencoder. Overall, these methods converge on a shared conclusion: unsupervised, physically interpretable pipelines can match or exceed manual fault curation, offering a scalable foundation for autonomous BPM diagnostics in next-generation accelerators.