Low-voltage electric mobility applications impose strict voltage limits that considerably limit the design of electromagnetic motors (Zhu and Howe, 2007; Huang et al., 2022). This study aims to focus on a 48V axial-flux dual-stack (AFDS) permanent magnet motor for direct drive urban electric mobility, with particular emphasis on enforcing voltage limitations during the early optimisation stage to ensure practical feasibility (Yang et al., 2023).
A voltage-constrained Non-dominated Sorting Genetic Algorithm II (NSGA-II) multi-objective optimisation framework is developed for a 48V AFDS motor, with back-electromotive force (EMF) enforced as a hard constraint. An adaptive Kriging surrogate model coupled with three-dimensional (3D) finite element analysis (FEA) improves computational efficiency. Performance is validated analytically, by JMAG FEA, and experimentally across 357 operating points. Global Sobol sensitivity analysis (Jansen, 1999) identifies dominant design variables.
The optimisation generates a Pareto front under strict 48V back-EMF constraint, identifying a 16-pole/24-slot AFDS design achieving 15.09 Nm peak torque, 6.29 Nm/kg torque density and 94.39% peak efficiency. Sobol analysis identifies g as the primary torque density driver (Si = 0.299), diameter ratio as the torque ripple governor (Si = 0.800) and Do as the cost controller (Si = 0.801).
The back-EMF voltage constraint is directly integrated into the NSGA-II loop for 48V AFDS motor design, eliminating gear-shifting for urban mobility. A multi-fidelity validation framework combining analytical optimisation, 3D FEA and 357 experimental points is delivered, supported by variance-based Sobol sensitivity analysis with the Jansen estimator.
