Declare a coupling target. The optimiser evolves coil geometries — amax, Nr, Nt and turn spacing — toward it using a real-valued micro-GA. Multi-objective mode returns a Pareto front over k, η and power.
Single-objective runs minimise |k_achieved − k_target|. Multi-objective runs maximise k while balancing efficiency and delivered power. Manufacturing rules from IPC-2221 (current density, trace spacing, wire diameter) are enforced as hard constraints.
A real-valued chromosome encodes geometry; BLX-α crossover blends parents inside an extended hyper-rectangle; Gaussian mutation perturbs offspring; tournament selection with k = 3 picks survivors. 80 generations of population 40 is the default budget.
Bounds: amax ∈ [5, 200] mm, Nr ∈ [1, 30], turn spacing δ ∈ [0, 10] mm, Nt ∈ [1, 30]. Multi-objective runs return an NSGA-II-style Pareto front of non-dominated geometries you can browse.
