Sujin Seol

Ms. Sujin Seol


Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
12 October 2023 in Machine Learning: Science and Technology
Verified
Journal Article
Open access
SSSujin Seol
Sujin Seol
JLJungEun Lee
JungEun Lee
JYJaewoo Yoon
Jaewoo Yoon
BKByeong-Woo Kim
Byeong-Woo Kim
Recently, the xEV market has been expanding by strengthening regulations on fossil fuel vehicles. It is essential to ensure the safety and reliability of batteries, one of the core components of xEVs. Furthermore, estimating the battery’s state of health (SOH) is critical. There are model-based and data-based methods for SOH estimation. Model-based methods have limitations in linearly modeling the nonlinear internal state changes of batteries. In data-based methods, high-quality datasets containing large quantities of data are crucial. Since obtaining battery datasets through measurement is difficult, this paper supplements insufficient battery datasets using time-series generative adversarial network and compares the improvement rate in SOH estimation accuracy through long short-term memory and gated recurrent unit based on recurrent neural networks. According to the results, the average root mean square error of battery SOH estimation improved by approximately 25%, and the learning stability improved by approximately 40%.

12 July 2024 in Applied Sciences
Verified
Journal Article
Open access
PDF file
SSSujin Seol
Sujin Seol
JYJaewoo Yoon
Jaewoo Yoon
JLJungEun Lee
JungEun Lee
BKByeong-Woo Kim
Byeong-Woo Kim
The advancements in artificial intelligence have encouraged the application of deep learning in various fields. However, the accuracy of deep learning algorithms is influenced by the quality of the dataset used. Therefore, a high-quality dataset is critical for deep learning. Data augmentation algorithms can generate large, high-quality datasets. The dataset quality is mainly assessed through qualitative and quantitative evaluations. However, conventional qualitative evaluation methods lack the objective and quantitative parameters necessary for battery synthetic datasets. Therefore, this study proposes the application of the rate of change in linear regression correlation coefficients, Dunn index, and silhouette coefficient as clustering indices for quantitatively evaluating the quality of synthetic time-series datasets of batteries. To verify the reliability of the proposed method, we first applied the TimeGAN algorithm to an open-source battery dataset, generated a synthetic battery dataset, and then compared its similarity to the original dataset using the proposed evaluation method. The silhouette coefficient was confirmed as the most reliable index. Furthermore, the similarity of datasets increased as the silhouette index decreased from 0.1053 to 0.0073 based on the number of learning iterations. The results demonstrate that the insufficient quality of datasets used for deep learning can be overcome and supplemented. Furthermore, data similarity can be efficiently evaluated regardless of the learning environment. In conclusion, we present a new synthetic time-series dataset evaluation method that is more reliable than the conventional representative evaluation method (the training loss rate).

20 January 2026 in IEEE Access
Verified
Journal Article
Open access
JYJae Woo Yoon
Jae Woo Yoon
SSSujin Seol
Sujin Seol
BKByeong-Woo Kim
Byeong-Woo Kim
Intelligent fault diagnosis requires models that remain robust under diverse and noisy industrial environments. Although vibration signals are widely used for monitoring rotating machinery, they are often collected under varying noise conditions, requiring strong diagnostic capability even in the presence of severe noise. This study proposes a one-dimensional Swin transformer-based model that directly processes raw vibration signals without preprocessing, enabling effective use by non-experts. The proposed model extracts local fault features through self-attention among patches within each window and progressively captures global features via a shift-window mechanism and hierarchical structure, providing a comprehensive representation of fault characteristics. Ablation studies were conducted to design a lightweight yet high-performance model. Extensive evaluations were performed on three public benchmark datasets using comparative models under various noise conditions. Experimental results indicated that the proposed model achieved the highest classification accuracy and superior robustness and computational efficiency in both noise-free and strong-noise environments.

29 November 2023 in Research Square
Verified
Preprint
Open access
PDF file
SSSujin Seol
Sujin Seol
JYJaewoo Yoon
Jaewoo Yoon
JLJungEun Lee
JungEun Lee
BKByeong-Woo Kim
Byeong-Woo Kim
Advancements in artificial intelligence have encouraged the application of deep learning in various fields. A critical element of deep learning is the dataset, with the accuracy of deep learning algorithms depending on the quality of the data. Data augmentation algorithms can secure large, high-quality datasets, with qual- ity assessment of the generated datasets mostly performed through qualitative and quantitative evaluation. However, the conventional qualitative evaluation has no objective and quantitative evaluation suitable for battery synthetic datasets. Therefore, this study proposes applying the rate of change in linear regression correlation coefficients, the Dunn Index, and the silhouette coefficient as cluster- ing indices for quantitatively evaluating the quality of the synthetic time-series datasets of batteries. To verify the reliability of the proposed method, we applied the TimeGAN algorithm to an open-source battery dataset and then generated a synthetic battery dataset and compared its similarity to the original dataset using the proposed evaluation method. Consequently, we present a new synthetic time- series dataset evaluation method that has higher reliability than the conventional representative evaluation method (i.e., the training loss rate).