NGnet-Based Sequential Optimization Technique of Variable Flux Memory Machines Considering Multimagnetization State Ripple Suppression

Abstract

Since the existence of multiple magnetization states (MSs), variable flux memory machines (VFMMs) are difficult to optimize, especially their torque ripples under multiple states fail to be suppressed simultaneously. This article proposes a novel NGnet-based sequential (NGBS) optimization technique, which can significantly improve the optimization efficiency and topological freedom of the machine. First, the key design parameters of the investigated VFMM are identified by employing the magnetic equivalent circuit (MEC) method. The parametric modeling with comprehensive sensitivity analysis of the investigated VFMM is employed to reveal the high sensitivity parameters of the key electromagnetic characteristics. Second, the multiobjective genetic algorithm is employed to optimize the machine with multiple MSs, which satisfy the initial electromagnetic characteristics. Third, based on the candidate case from the previous optimization, the local refined topology optimization is implemented using the NGnet method, which can significantly reduce the torque ripple. Additionally, a comprehensive comparison of the key electromagnetic characteristics of the initial and optimal machines is carried out using the finite element (FE) to verify the effectiveness of the proposed optimization method. Finally, an optimized VFMM prototype is fabricated and tested to validate the feasibility of the proposed NGBS optimization technique.