Hong Je-Gal

Mr. Hong Je-Gal


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
Verified
Journal Article
Open access
PDF file
HJHong Je-Gal
Hong Je-Gal
YPYoungseo Park
Youngseo Park
SPSeong-Ho Park
Seong-Ho Park
JKJi-Uk Kim
Ji-Uk Kim
JYJung-Hee Yang
Jung-Hee Yang
HLHyun-Suk Lee
Hyun-Suk Lee
As engine monitoring data has become more complex with an increasing number of sensors, fault prediction based on artificial intelligence (AI) has emerged. Existing fault prediction models using AI significantly improve the accuracy of predictions by effectively handling such complex data, but at the same time, the problem arises that the AI-based models cannot explain the rationale of their predictions to users. To address this issue, we propose a time-series explanatory fault prediction framework to provide an explainability even when using AI-based fault prediction models. It consists of a data feature reduction process, a fault prediction model training process using long short-term memory, and an interpretation process of the fault prediction model via an explainable AI method. In particular, the proposed framework can explain a fault prediction based on time-series data. Therefore, it indicates which part of the data was significant for the fault prediction not only in terms of sensor type but also in terms of time. Through extensive experiments, we evaluate the proposed framework using various fault data by comparing the prediction performance of fault prediction and by assessing how well the main pre-symptoms of the fault are extracted when predicting a fault.

Verified
Journal Article
Open access
PDF file
HJHong Je-Gal
Hong Je-Gal
SLSeung-Jin Lee
Seung-Jin Lee
JYJeong-Hyun Yoon
Jeong-Hyun Yoon
HLHyun-Suk Lee
Hyun-Suk Lee
JYJung-Hee Yang
Jung-Hee Yang
SKS.W. KIM
S.W. KIM
Ensuring operational reliability in machinery requires accurate fault detection. While time-domain vibration pulsation signals are intuitive for pattern recognition and feature extraction, downsampling can reduce analytical complexity, but may result in low-precision data, affecting fault detection performance. To address this, we propose time–frequency feature fusion, combining information from both the time and frequency domains for fault detection. Our approach transforms vibrational pulse data into instantaneous revolutions per minute (RPM) and employs statistical analysis for the time-domain features. For the frequency-domain features, we use the combined method of empirical mode decomposition and independent component analysis (EMD-ICA), along with the Wigner bispectrum method to capture the nonlinear characteristics and phase conjugation. Using a deep neural network (DNN), we classify the anomaly states, demonstrating the effectiveness and versatility of our approach in detecting anomalies and improving diagnostic precision. Compared to using time or frequency features alone, our time–frequency feature fusion model achieves higher accuracy, with 100% accuracy at lower downsampling rates and 96.3% accuracy at a downsampling rate of 100×.

12 December 2025 in arXiv
Verified
Preprint
Open access
PDF file
HJHong Je-Gal
Hong Je-Gal
CYChan-Bin Yi
Chan-Bin Yi
HLHyun-Suk Lee
Hyun-Suk Lee
Applying reinforcement learning (RL) to real-world tasks requires converting informal descriptions into a formal Markov decision process (MDP), implementing an executable environment, and training a policy agent. Automating this process is challenging due to modeling errors, fragile code, and misaligned objectives, which often impede policy training. We introduce an agentic large language model (LLM)-based framework for automated MDP modeling and policy generation (A-LAMP), that automatically translates free-form natural language task descriptions into an MDP formulation and trained policy. The framework decomposes modeling, coding, and training into verifiable stages, ensuring semantic alignment throughout the pipeline. Across both classic control and custom RL domains, A-LAMP consistently achieves higher policy generation capability than a single state-of-the-art LLM model. Notably, even its lightweight variant, which is built on smaller language models, approaches the performance of much larger models. Failure analysis reveals why these improvements occur. In addition, a case study also demonstrates that A-LAMP generates environments and policies that preserve the task's optimality, confirming its correctness and reliability.