Skip to main content
OSTI.GOVU.S. Department of Energy Office of Scientific and Technical Information
U.S. Department of Energy
Office of Scientific and Technical Information
 

Harnessing the power of gradient-based simulations for multi-objective optimization in particle accelerators

Journal Article · · Machine Learning: Science and Technology
 [1];  [2];  [3];  [4];  [5];  [3];  [5];  [5];  [5]
  1. Thomas Jefferson National Accelerator Facility (TJNAF), Newport News, VA (United States); Univ. of Houston, TX (United States)
  2. Thomas Jefferson National Accelerator Facility (TJNAF), Newport News, VA (United States); Old Dominion Univ., Norfolk, VA (United States)
  3. SLAC National Accelerator Laboratory (SLAC), Menlo Park, CA (United States)
  4. Old Dominion Univ., Norfolk, VA (United States); Hampton Roads Biomedical Research Consortium, Portsmouth, VA (United States)
  5. Thomas Jefferson National Accelerator Facility (TJNAF), Newport News, VA (United States
+ Show Author Affiliations
Particle accelerator operation requires simultaneous optimization of multiple objectives. Multi-objective optimization (MOO) is particularly challenging due to trade-offs between the objectives. Evolutionary algorithms, such as genetic algorithms (GAs), have been leveraged for many optimization problems, however, they do not apply to complex control problems by design. This paper demonstrates the power of differentiability for solving MOO problems in particle accelerators using a deep differentiable reinforcement learning (DDRL) algorithm. We compare the DDRL algorithm with model-free reinforcement learning (MFRL), GA, and Bayesian optimization (BO) for simultaneous optimization of heat load and trip rates in the continuous electron beam accelerator facility. The underlying problem enforces strict constraints on both individual states and actions as well as cumulative (global) constraints on energy requirements of the beam. Using historical accelerator data, we develop a physics-based surrogate model which is differentiable and allows for back-propagation of gradients. The results are evaluated in the form of a Pareto-front with two objectives. We show that the DDRL outperforms MFRL, BO, and GA on high dimensional problems.
Research Organization:
SLAC National Accelerator Laboratory (SLAC), Menlo Park, CA (United States); Thomas Jefferson National Accelerator Facility (TJNAF), Newport News, VA (United States)
Sponsoring Organization:
USDOE; USDOE Laboratory Directed Research and Development (LDRD) Program; USDOE Office of Science (SC), Advanced Scientific Computing Research (ASCR)
Grant/Contract Number:
AC02-76SF00515; AC05-06OR23177
OSTI ID:
2558155
Alternate ID(s):
OSTI ID: 2531256
OSTI ID: 2560753
Report Number(s):
DOE/OR/23177-7724; JLAB-CST--24-4230; arXiv:2411.04817
Journal Information:
Machine Learning: Science and Technology, Journal Name: Machine Learning: Science and Technology Journal Issue: 2 Vol. 6; ISSN 2632-2153
Publisher:
IOP PublishingCopyright Statement
Country of Publication:
United States
Language:
English

References (30)

A practical guide to multi-objective reinforcement learning and planning journal April 2022
Multi-module-based CVAE to predict HVCM faults in the SNS accelerator journal September 2023
CPOpt: A modular framework for genetic algorithm optimization and post-optimization analysis in complex charged particle optical design
  • Huber, Kilian; Wirtz, Tom; Hoang, Hung Quang
  • Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, Vol. 1067 https://doi.org/10.1016/j.nima.2024.169702
journal October 2024
Multi-objective Bayesian active learning for MeV-ultrafast electron diffraction journal June 2024
Reinforcement learning-trained optimisers and Bayesian optimisation for online particle accelerator tuning journal July 2024
Online multi-objective particle accelerator optimization of the AWAKE electron beam line for simultaneous emittance and orbit control journal May 2020
Deep differentiable reinforcement learning and optimal trading journal June 2022
Robust errant beam prognostics with conditional modeling for particle accelerators journal March 2024
Parallel general purpose multiobjective optimization framework with application to electron beam dynamics journal May 2019
Machine learning for orders of magnitude speedup in multiobjective optimization of particle accelerator systems journal April 2020
Multiobjective optimization of the dynamic aperture using surrogate models based on artificial neural networks journal January 2021
Multiobjective Bayesian optimization for online accelerator tuning journal June 2021
Physics model-informed Gaussian process for online optimization of particle accelerators journal July 2021
Real-time artificial intelligence for accelerator control: A study at the Fermilab Booster journal October 2021
Uncertainty aware anomaly detection to predict errant beam pulses in the Oak Ridge Spallation Neutron Source accelerator journal December 2022
Uncertainty aware machine-learning-based surrogate models for particle accelerators: Study at the Fermilab Booster Accelerator Complex journal April 2023
Bayesian optimization of laser-plasma accelerators assisted by reduced physical models journal August 2023
Bridging the gap between machine learning and particle accelerator physics with high-speed, differentiable simulations journal May 2024
Bayesian optimization algorithms for accelerator physics journal August 2024
The Continuous Electron Beam Accelerator Facility at 12??GeV journal August 2024
Bayesian Optimization of a Free-Electron Laser journal March 2020
Simultaneous optimization of the cavity heat load and trip rates in linacs using a genetic algorithm journal October 2014
A fast and elitist multiobjective genetic algorithm: NSGA-II journal April 2002
Pymoo: Multi-Objective Optimization in Python journal January 2020
The Hypervolume Indicator journal July 2021
Comparison of Multiobjective Evolutionary Algorithms: Empirical Results journal June 2000
Online Accelerator Tuning with Adaptive Bayesian Optimization text January 2022
Differential Machine Learning preprint June 2020
Developing Robust Digital Twins and Reinforcement Learning for Accelerator Control Systems at the Fermilab Booster
  • Kafkes, Diana; Herwig, C.; Pellico, William
  • Developing Robust Digital Twins and Reinforcement Learning for Accelerator Control Systems at the Fermilab Booster https://doi.org/10.2172/1825276
conference January 2021
Deep Learning Based Superconducting Radio-Frequency Cavity Fault Classification at Jefferson Laboratory journal January 2022

Similar Records

Related Subjects

43 PARTICLE ACCELERATORS
Bayesian optimization
MOGA
MORL
differentiable simulation
genetic algorithm
multi objective
reinforcement learning