ISIn-Seok Song

In-Seok Song

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


In-Seok Song has a total of 5 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Attention Mechanisms-Based CNN Regression Model for Optimal Design of SPMSM

    by Tae-Hyuk Ji, Seok-Won Jung, In-Seok Song, Sang-Yong Jung

    This paper proposes an optimal design method for surface-mounted permanent magnet synchronous motors (SPMSMs) based on electromagnetic performance prediction using a convolutional neural network (CNN)-based regression model enhanced with attention mechanisms. Among various machine learning techniques aimed at improving design efficiency through prediction, CNNs are well-suited for performance prediction as they effectively capture spatial structures and local patterns, extracting key features from motor designs. Additionally, the integration of attention mechanisms, which we applied to the CNNs, further improves prediction performance by assigning greater weights to critical design features, thereby enhancing prediction accuracy and learning efficiency. Moreover, attention mechanisms enable the identification of relationships among features, overcoming the limitations of localized learning in conventional CNN models. In this study, the effectiveness of attention mechanisms is evaluated, and a CNN-based regression model incorporating these mechanisms is developed. The proposed regression model is then applied to optimize the design of an SPMSM.


  • Efficient Multi-Physics Optimization of High-Speed EV Motors: Employing a Hybrid, Adaptive Surrogate-Assisted Strategy

    by Taek-Hyo Nam, Tae-Hyuk Ji, Young-Ho Hwang, In-Seok Song, Seok-Won Jung, Sang-Yong Jung

    The optimal design of high-speed electric vehicle (EV) motors requires simultaneously addressing both electromagnetic (EM) performance and mechanical stress (MS). Reducing the significant computational time associated with this multi-physics optimization, however, is a fundamental challenge. To address this challenge, this paper presents a hybrid, two-step adaptive strategy. As performing finite element analysis (FEA) for all physics is computationally burdensome, the proposed strategy applies a surrogate model exclusively to the MS analysis, while continuing to use FEA for EM analysis. The proposed process first employs an initially imprecise surrogate model as a statistical filter to intelligently narrow the design space, and then transitions to a full surrogate replacement for the expensive MS FEA evaluations once the model achieves high fidelity. This strategy was applied to the design of a high-speed EV propulsion motor using a particle swarm optimization algorithm. To validate its efficacy, the method was benchmarked against a baseline method that omits the initial adaptive filtering step. The results demonstrate a substantial reduction in total computational time while successfully yielding a final design.