Study on a Regression Model for the Electromagnetic Characteristics of SPMSM Based on Convolutional Neural Network with Attention Mechanism

Surface-mounted permanent magnet synchronous motors (SPMSMs) are widely used in high-performance system. However, they suffer from torque ripple and cogging torque, which degrade output quality and cause noise and vibration, necessitating optimized design. While finite element analysis offers high accuracy for optimization, it comes with significant computational cost. To address this limitation, this paper proposes an attention mechanism-based convolutional neural network (CNN) regression model to predict SPMSM electromagnetic performance. CNNs capture spatial structures of motor designs, while attention mechanisms highlight key design features, boosting prediction accuracy and efficiency. This study analyzes the effects of attention mechanisms and proposes a CNN-based regression model incorporating them, confirming the effectiveness of the attention mechanism through comparisons with conventional CNN models.