Machine Learning-Driven Framework for Optimal Split Ratio Determination in PMSM Design

The split ratio, which is the ratio of rotor diameter to stator diameter, plays a pivotal role in determining the performance of permanent magnet (PM) motors. This study presents a machine learning-driven framework to optimize the split ratio from a cost-efficiency perspective. Extensive datasets generated through a meta-heuristic optimization algorithm are analyzed using regression techniques to clarify the intricate relationships between the split ratio and key motor design variables. From this analysis, a novel logarithmic formula is derived, capturing these interdependencies with remarkable precision and offering predictive robustness across previously unseen design scenarios. The proposed framework provides two major advancements: (1) it enables rapid and accurate computation of the optimal split ratio across major six pole/slot families without extensive iterative simulations, and (2) it establishes a scalable, data-driven methodology applicable to broader motor design optimization tasks. The validity and effectiveness of the proposed approach are rigorously verified through finite element analysis (FEA) and experimental validation.