Optimization and Designing of SPOKE ‐Type BLDC Motors Using Reinforcement Learning and Finite Element Analysis

This paper presents a novel optimization approach for Brushless DC (BLDC) motors, specifically focusing on the wing‐shaped SPOKE‐type rotor configuration. The study addresses critical performance challenges, including high cogging torque, torque ripple, and reduced air‐gap flux density, which hinder the efficiency and stability of conventional motor designs. Traditional optimization techniques, primarily based on Finite Element Analysis (FEA) and manual parameter tuning, suffer from high computational costs, lack of adaptability, and often result in suboptimal designs. To overcome these limitations, this research proposes a Reinforcement Learning (RL)‐based optimization framework that autonomously adjusts design parameters, such as magnet geometry, rotor topology, and pole arc ratio, employing high‐fidelity electromagnetic simulations. The framework integrates advanced RL algorithms, specifically Deep Deterministic Policy Gradient (DDPG) and Twin Delayed DDPG (TD3), to iteratively fine‐tune the motor's design. The RL‐driven optimization significantly improves motor performance, achieving enhanced average torque, reduced torque ripple, minimized cogging torque, and increased flux density and efficiency. The proposed RL‐based approach outperforms traditional static and heuristic optimization techniques, offering superior computational efficiency, adaptability, and scalability.