摘要
This paper proposes an AI-assisted motor design framework that integrates machine learning-based design evaluation with reinforcement learning-based optimization. A supervised learning model is first employed to rapidly classify candidate motor designs and filter out inferior solutions, thereby confining the search space to feasible design regions. Based on the classification results, a reinforcement learning strategy using the SARSA algorithm is applied to optimize cogging torque within the effective design domain. Experimental studies on motors with multiple power ratings demonstrate consistent classification performance and stable cogging torque reduction, confirming the practicality and effectiveness of the proposed framework.