Welcome to the Digital Transformation Lab!
At the Digital Transformation Lab, we develop advanced methodologies for modeling, analyzing, optimizing, and calibrating large-scale production and material handling systems. Our research integrates model-based decision-making with statistical approaches to address challenging problems in modern industrial operations. In particular, we focus on surrogate modeling and active design of experiments to support efficient decision-making and enable practical digital transformation. Our application domains include semiconductor and display manufacturing, distribution centers, and container terminals.
Our research interests include the following:
Topics: Computer modeling, analysis, optimization, and calibration
Methodologies: Statistical surrogate modeling and design of experiments
Application areas: Semiconductor/display fabs, distribution centers, container terminals, and distribution centers
Research Topic I: System Modeling & Analysis
We develop computer models that capture the underlying mechanisms of real-world operations and generate quantitative insights for complex production and logistics systems. These models provide a rigorous foundation for digital twin research, enabling studies in tactical facility design, operational planning, and real-time control. Across these applications, we integrate mathematical optimization and machine learning to support model validation and decision analysis.
Small Size FAB Simulation.mp43D SVSRS.mp4반도체생산시스템.mp4Research Topic II: Model Optimization
In semiconductor manufacturing, efficient material handling is essential for sustaining high throughput in complex fab operations. We developed a detailed AMHS model and used it to evaluate performance across configurations and scenarios, identifying bottlenecks in material handling operations. Building on these insights, we proposed a model-based optimization framework for proactive vehicle dispatching for future transport demand. Extensive computer experiments demonstrated improved throughput and reduced vehicle travel time, confirming robustness under stochastic conditions. This work has been published in IEEE Transactions on Automation Science and Engineering and presented at the 2023 INFORMS Annual Meeting (Phoenix, AZ, USA).
[3D] FAB AMHS Animation 3.mp4[3D] FAB AMHS Animation 2.mp4Research Topic III: Model Calibration
In modern semiconductor manufacturing, system model increasingly support planning and control, yet their predictive accuracy can deteriorate under uncertainty, creating gaps between simulated and realized performance and biasing decisions. To mitigate this issue, we developed a Bayesian calibration framework that learns and corrects model discrepancy using limited real-world data, substantially improving predictive fidelity even under congested operating regimes. This work has been published in the Journal of Manufacturing Systems and presented at the 2024 INFORMS Annual Meeting, with ongoing efforts to extend the framework to broader smart manufacturing applications.
