Identification of Magnetic Field Errors in Synchrotrons

This technical report discusses the identification of magnetic field errors in synchrotrons using deep Lie map networks (DLMN). Magnetic field errors can significantly impact synchrotron performance by causing beam loss and reducing dynamic aperture. Traditional methods for identifying these errors rely on orbit response matrices or resonance driving terms, which can be time-consuming. The report introduces a novel approach that utilizes machine learning to create an accelerator model that incorporates multipole components of magnetic field errors. By analyzing simulated beam position monitor readings from the SIS18 synchrotron at GSI, the DLMN method demonstrates the ability to infer the location and strength of gradient and sextupole errors across all accelerator sections simultaneously. This approach aims to enhance the setup of corrector magnets, leading to improved control over tunes, chromaticities, and resonance compensation. The report is structured to introduce the DLMN model, describe its training procedure, and present simulation results, concluding with insights into its effectiveness.