Author: Bogaert, M.P.
Paper Title Page
MOPOST047 Determination of the Phase-Space Stability Border with Machine Learning Techniques 183
 
  • F.F. Van der Veken, R. Akbari, M.P. Bogaert, E. Fol, M. Giovannozzi, A.L. Lowyck, C.E. Montanari, W. Van Goethem
    CERN, Meyrin, Switzerland
 
  The dynamic aperture (DA) of a hadron accelerator is represented by the volume in phase space that exhibits bounded motion, where we disregard any disconnected parts that could be due to stable islands. To estimate DA in numerical simulations, it is customary to sample a set of initial conditions using a polar grid in the transverse planes, featuring a limited number of angles and using evenly distributed radial amplitudes. This method becomes very CPU intensive when detailed scans in 4D, and even more in higher dimensions, are used to compute the dynamic aperture. In this paper, a new method is presented, in which the border of the phase-space stable region is identified using a machine learning (ML) model. This allows one to optimise the computational time by taking the complex geometry of the phase space into account, using adaptive sampling to increase the density of initial conditions along the border of stability.  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2022-MOPOST047  
About • Received ※ 06 June 2022 — Accepted ※ 17 June 2022 — Issue date ※ 20 June 2022  
Cite • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)