Optimization of Stator Structure for Improved Accuracy in Variable Reluctance Resolvers Using Advanced Machine Learning Techniques

Abstract

This study presents an optimized design for a Segmented Sinusoidal Parameter Winding with Magnetic Wedge Variable Reluctance Resolver (SSPWMW-VRR), addressing challenges like winding asymmetry and harmonic distortion in conventional designs. By integrating particle swarm optimization (PSO) for winding design, magnetic equivalent circuit (MEC) analysis for leakage flux, and machine learning techniques (XGBoost and Multi-Layer Perceptron), the stator slot shape was fine-tuned for improved accuracy. XGBoost outperformed MLP in prediction accuracy with a mean absolute error (MAE) of 0.1172. Finite element analysis (FEA) simulations and experimental validation demonstrated a reduction in position errors from ±30′ in conventional VRRs to ±5′ in the optimized design, along with significant harmonic reduction.

Keywords

• variable reluctance resolver
• machine learning
• electromagnetic field simulation