Performance Optimization of Flywheel Motor by Using NSGA-2 and AKMMP

This paper proposes an efficient and accurate surrogate model, called adaptive Kriging model based on maximum projection (MaxPro) design (AKMMP), and establishes the design method by using the non-dominated sorting genetic algorithm with the elitism approach (NSGA-2) and AKMMP to optimize the mass and torque density as the performance parameters of flywheel motor with consideration of rotational inertia. The Pareto optimal solution set of flywheel motor is provided and used as the design border for the flywheel motor. The detailed design parameters are the armature diameter and axial length, rotor and stator split ratios, and permanent magnet width related to the mass and torque density. The brief optimization process includes five steps: A) getting the initial samples by MaxPro design; B) building the Kriging model based on the initial samples; C) optimization by NSGA-2 based on the Kriging model; D) selecting the corrected test point by the proposed sample strategy; and E) judging the convergence. The seven cases with different masses are analyzed, and the finite-element method is used to verify the optimization method.