Magnetic Optimization for Muon Collider H-Type Magnet

Conducted in collaboration with the IMCC (International Muon Collider Collaboration), this project targets the design of fast-ramping dipole and quadrupole magnets for muon acceleration. The methodology combines analytical modeling, finite element simulations (FEMM), thermal modeling, and multi-objective optimization (inductance, mass, cooling), with a focus on reducing computational cost without compromising physical fidelity. A key innovation lies in the hybrid modal method for loss estimation (proximity & skin effects) in water-cooled, hollow square-section copper conductors subjected to pulsed currents with high harmonic content. To further accelerate optimization, the precomputed loss matrixes are used to train lightweight ANN surrogate models, reducing solve time by up to an order of magnitude compared to direct FEM computation.

Application: The methodology is validated on an H-type magnet for Rapid Cycling Synchrotrons (RCS), integrating dynamic and hysteresis losses with nonlinear frequency-domain equivalent modeling. The hybrid modal+ANN approach enables fast and accurate evaluation of design candidates within global optimization loops, supporting efficient magnet-converter co-design for future muon accelerators. To further accelerate the optimization process, precomputed loss matrices are used to train surrogate models based on artificial neural networks (ANN), significantly reducing computation time. The methodology is validated on an H-type magnet, incorporating both copper and iron losses, including skin and proximity effects, which are particularly significant in this application. The hybrid approach (modal + ANN) enables fast and accurate evaluation of prototype candidates, supporting efficient magnet-converter co-design processes for the future muon accelerator.