VCM Five-Factor Multi-Objective Optimization Algorithm Based On Neural Network-Guided NSGA-II/PSO

Abstract

To improve the comprehensive performance of voice coil motor (VCM) planar coils under limited spatial constraints, this paper proposes a neural-network-guided NSGA-II/PSO multi-objective optimization method. A five-factor parameterized model is established by selecting coil outer diameter, trace width, trace spacing, copper thickness, and resistance as design variables, while average thrust, power consumption, and linearity error are defined as optimization objectives. A neural network surrogate model is introduced to predict electromagnetic performance and reduce the cost of repeated high-fidelity evaluation. Based on this model, NSGA-II is used for global Pareto search, and PSO is combined to enhance local search ability. The obtained Pareto front provides feasible design choices for balancing force output, energy consumption, and motion accuracy. Simulation and physical test results show that the optimized VCM planar coil achieves higher thrust, lower power consumption, and reduced linearity error, verifying the effectiveness and feasibility of the proposed method.