https://scholarworks.unist.ac.kr/handle/201301/47 2026-04-08T21:21:46Z https://scholarworks.unist.ac.kr/handle/201301/90982 Title: An iTransformer-based Approach for BTX Yield Forecasting in the Coke Oven Gas Purification Process Author(s): Han, GeonHee Abstract: The complexity of the BTX recovery process makes it challenging to accurately predict its yield. The nonlinear and intricate relationships between process variables (e.g., temperature, pressure, flow rate) and the BTX yield are not adequately captured by traditional modeling approaches. To overcome this limitation, this study applies the iTransformer model to predict the BTX yield. The iTransformer is a state-of-the-art machine learning architecture capable of effectively learning long-term dependencies and inter-variable interactions in multivariate time series data, making it well-suited for modeling the dynamic characteristics of the COG purification process. We propose a novel architecture combining GAT-Linformer with iTransformer as the backbone. This approach aims to maximize prediction accuracy by deeply learning complex latent patterns and interac- tions among variables within the data, independent of physical simulations. Experimental results using real-time process data demonstrate the superior performance of the proposed model compared to existing baselines. Major: Department of Industrial Engineering 2026-01-31T15:00:00Z https://scholarworks.unist.ac.kr/handle/201301/90981 Title: A Multi-Period Spatial Optimization Model for Equal and Equitable Access in Emergency Logistics Author(s): Song, Yujin Abstract: This study develops a multi-period location-allocation model for mobile Points of Distribution (PODs) to enhance both spatial equality and spatial equity in emergency relief distribution. Traditional fixed-site PODs often fail to ensure equitable access, particularly in low-density, underserved, or transportation-disadvantaged areas. To address these limitations, we propose a bi-objective mixed-integer linear programming model that determines optimal mobile POD placements and demand assignments across multiple time periods. The model simultaneously aims to (i) maximize population coverage within a defined travel distance threshold and (ii) minimize total travel burden by accounting for disparities in transportation access, such as vehicle ownership and income levels, across demand groups. The proposed model is validated using real-world data from Flint, Michigan, including road networks, demographics, and socioeconomic status. Compared to fixed-site PODs, our mobile POD strategy reduces average travel time by up to 59.03% and achieves full population coverage at 61.68% lower operational cost. Vulnerable communities, particularly those with limited transportation access, benefit the most from this approach. Our sensitivity analysis reveal that travel distance thresholds have a greater impact on accessibility than budget levels, highlighting the need for context-specific strategies reflecting the spatial and socioeconomic characteristics of each ward. Our results demonstrate that mobile PODs offer a scalable and policy-relevant solution for improving access to emergency supplies in disadvantaged areas. These findings underscore the importance of integrating both spatial equality and equity into disaster relief logistics and position mobile PODs as a viable strategy for strengthening community resilience in the face of increasing disaster risk. Major: Department of Industrial Engineering 2026-01-31T15:00:00Z https://scholarworks.unist.ac.kr/handle/201301/90980 Title: A Retrieval-Augmented Generation Framework for Root Cause Analysis in Maritime Accident Adjudication Reports Author(s): Kim, Seongjin Abstract: This thesis proposes a retrieval-augmented generation (RAG) framework for the root cause analysis (RCA) of maritime accidents, defining the task as the inference of detailed causal narratives by retrieving and analyzing historically similar adjudication precedents. To address the challenge of providing grounded evidence in specialized domains, we construct a domain-specific knowledge base by converting 13,329 Korea Maritime Safety Tribunal adjudication summaries (1971–2025) into structured Card units. We assign hierarchical cause labels to serve as structural metadata, optimizing the retrieval of relevant precedents across distinct document fields (summary, causes, disposition). By fusing sparse and dense re- trievers via Reciprocal Rank Fusion (RRF), the framework identifies pertinent precedents and synthesizes them into evidence-based explanations. We further design a quantitative evaluation framework using a Metadata Relevance Score to overcome the absence of a manually labeled gold standard. Experiments demonstrate that the proposed Multi-Field Retrieval strategy substantially outperforms baselines in rank- ing quality (nDCG) and coverage. The final framework generates structured RCA outputs comprising textual cause summaries and hierarchical tags, providing a systematic decision-support tool for maritime safety investigations. Major: Department of Industrial Engineering 2026-01-31T15:00:00Z https://scholarworks.unist.ac.kr/handle/201301/90979 Title: Complex Multi-Stage Production Scheduling with Unrelated Parallel Machines in High-Mix Low Volume Environments Balancing Cost-Efficiency and Carbon Emissions Author(s): Kim, Namgyo Abstract: This study investigates a complex production scheduling problem in high-mix low-volume (HMLV) manufacturing environments under carbon regulation, where multi-stage process dependencies and heterogeneous machine characteristics jointly shape cost and emission outcomes. We develop a multi- stage production scheduling framework that integrates sequential process dependencies with unrelated parallel machines, capturing heterogeneous processing times, operating costs, and carbon emission rates that substantially increase scheduling complexity. Within this framework, we formulate two mixed- integer linear programming models to represent alternative regulatory settings: a cost-carbon trade-off model and a carbon credit model. The cost-carbon trade-off model explicitly examines the trade-off between total production cost and carbon emissions under binding emission limits, whereas the carbon credit model minimizes total cost while allowing excess emissions to be offset through carbon credit purchases. The proposed models are validated using real-world data from a marine engine parts manufacturer in Ulsan, Republic of Korea. Numerical experiments reveal that, under strict emission limits, the cost-carbon trade-off model experiences severe shortages and high penalty costs, while relaxing the emission limit shifts production toward lower-cost, higher-emission machines and substantially reduces total cost. In contrast, the carbon credit model maintains production feasibility even under tight emission constraints by offsetting excess emissions through carbon credit purchases, resulting in only marginal changes in total cost as emission limits are relaxed. Sensitivity analyses further demonstrate how penalty costs, machine settings, and carbon credit prices reshape machine assignments, excess emissions, and cost structures. Overall, the proposed framework provides a unified decision-support tool for analyzing production scheduling decisions under alternative carbon regulation schemes and offers actionable insights for managing the cost-emission trade-off in environment- conscious HMLV manufacturing systems. Major: Department of Industrial Engineering 2026-01-31T15:00:00Z