Speaker
Description
Muons have broad applications in muon spin rotation (μSR), muon tomography, precision measurements, and muon colliders, driving increasing demand for higher beam intensity and quality. This report presents our research progress on optimizing muon beamline design using AI methods. Two intelligent optimization approaches were employed: Genetic Algorithms (GA) and the Center-Evolving Algorithm (CE). The GA method was validated on the μE4 beamline, achieving over 20% improvement in surface muon intensity compared to the original optics design, with a 10% gain in online beam tuning. It also supported the feasibility design of the MELODY surface and decay muon beamlines. To address GA's weak convergence precision in small beam spot optimization, the CE method was developed — incorporating parallel simulation, parabolic-fitting gradient descent, and random jump strategies — successfully achieving a small beam spot of 7.8×10⁵ μ/s within ∅20 mm at 20 kW for the MELODY surface muon beamline.
