Machine learning applications for improving accelerator operations at the hadron injector complex at BNL

ORAL

Abstract

After decades of successful operation, human operators at the Relativistic Heavy Ion Collider (RHIC) complex at Brookhaven National Laboratory (BNL) have become highly skilled at finding effective operating points. However, because the accelerator chain is a dynamic, nonlinear system with many degrees of freedom and competing objectives, maintaining truly optimal conditions remains a demanding challenge. Moreover, no single metric exists to quantify whether an operating point is globally optimal. In this work, we demonstrate how machine learning and improved physics-based digital twins can be combined to optimize beam quality and streamline tuning in the hadron injector complex at BNL. We present experimental results on the application of automatic optimization algorithms (e.g., Bayesian optimization) and real-time digital twin modeling at the AGS synchrotron and its Booster.

*Work supported by Brookhaven Science Associates, LLC under Contract No. DE-SC0012704 and No. DE-SC0024287 with the U.S. Department of Energy and by NASA (Contract No. T570X).

–

Publication: Y. Gao et al., "Applying Bayesian Optimization to Achieve Optimum Cooling at the Low Energy RHIC Electron Cooling System", Physical Review Accelerators and Beams 25, 014601 (2022).
W. Lin et al., "Simulation Studies and Machine Learning Applications at the Coherent electron Cooling experiment at RHIC", in Proc. IPAC'22, Bangkok, Thailand, Jun. 2022, pp. 2387-2390.
X. Gu et al., "Enhancing beam intensity in RHIC EBIS beamline via GPTune machine learning-driven optimization", in Proc. IPAC'24, Nashville, TN, May 2024, pp. 118-120.
T. Balasooriya et al., "Reinforcement Learning for Charged Particle Beam Control to Minimize Injection Mismatch in Particle Accelerators", ICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Hyderabad, India, 2025, pp. 1-5.
W. Lin, "Maintaining optimal beam brightness and luminosity using machine learning", 16th International Conference on Heavy Ion Accelerator Technology, Michigan State University, East Lansing, Jun. 2025.
W. Lin et al., "Improve beam brightness with bayesian optimization at the AGS booster injection at BNL", in Proc. NAPAC2025, Sacramento, CA, Aug. 2025, pp. 157-159.
E. Hamwi et al., "Application of Bayesian optimization to BtA injection at BNL", in Proc. NAPAC2025, Sacramento, CA, Aug. 2025, pp. 58-60.
W. Lin et al., "Machine learning assisted Bayesian calibration of accelerator digital twin from orbit response data", in Proc. NAPAC2025, Sacramento, CA, Aug. 2025, pp. 177-180.
W. Lin et al., "Digital Twin Development for the NASA Space Radiation Laboratory", submitted to Physical Review Accelerators and Beams.

Presenters

  • Weijian Lin

    • Brookhaven National Laboratory (BNL)

Authors

  • Weijian Lin

    • Brookhaven National Laboratory (BNL)

  • Kevin A. Brown

    • Brookhaven National Laboratory (BNL)

  • Georg H Hoffstaetter de Torquat

    • Cornell Laboratory for Accelerator-based Sciences and Education
    • Cornell University

  • Eiad Hamwi

    • Cornell University

  • Vincent Schoefer

    • Brookhaven National Laboratory (BNL)

  • Yuan Gao

    • Brookhaven National Laboratory (BNL)

  • Trevor Olsen

    • Brookhaven National Laboratory (BNL)

  • Levente Hajdu

    • Brookhaven National Laboratory (BNL)

  • Christopher Kelly

    • Brookhaven National Laboratory (BNL)

  • Nathan Urban

    • Brookhaven National Laboratory (BNL)

  • Xiaofeng Gu

    • Brookhaven National Laboratory (BNL)

  • Ryan Roussel

    • SLAC National Accelerator Laboratory

  • Auralee Edelen

    • SLAC National Accelerator Laboratory

  • Armen Kasparian

    • Jefferson Lab

  • Malachi Schram

    • Jefferson Lab

  • Yinan Wang

    • Rensselaer Polytechnic Institute (RPI)

  • Tia Miceli

    • Fermi National Accelerator Laboratory (FNAL)

  • Christopher Hall

    • RadiaSoft LLC

  • David Sagan

    • Cornell University