Machine learning enabled fast evaluation of dynamic aperture for storage ring accelerators

Abstract

For any storage ring-based large-scale scientific facility, one of the most important performance parameters is the dynamic aperture (DA), which measures the motion stability of charged particles in a global manner. To date, long-term tracking-based simulation is regarded as the most reliable method to calculate DA. However, numerical tracking may become a significant issue, especially when lots of candidate designs of a storage ring need to be evaluated. In this paper, we present a novel machine learning-based method, which can reduce the computation cost of DA tracking by approximately one order of magnitude, while keeping sufficiently high evaluation accuracy. Moreover, we demonstrate that this method is independent of concrete physical models of a storage ring. This method has the potential to be applied to similar problems of identifying irregular motions in other complex dynamical systems.

References in corpus (10)

• TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems
• Beam by design: laser manipulation of electrons in modern accelerators
• Modern and Future Colliders
• Nonlinear Accelerator Lattices with One and Two Analytic Invariants
• Nuclear Physics Experiments with Ion Storage Rings
• Machine learning for beam dynamics studies at the CERN Large Hadron Collider
• Optimizing the lattice design for a diffraction-limited storage ring with a rational combination of particle swarm and genetic algorithms
• Multiobjective optimization of the dynamic aperture for SLS 2.0 using surrogate models based on artificial neural networks
• Improving the nonlinear performance of the HEPS baseline design with genetic algorithm
• Machine learning for design optimization of storage ring nonlinear dynamics