We develop AI technologies that innovate manufacturing processes, from novel material discovery through process design and control. Our ultimate goal is to realize AI-native manufacturing processes, in which AI agents autonomously learn, adapt, and reconfigure processes beyond the limits of human intuition. To this end, we pursue research across the four core areas of process systems engineering: process modeling, process design, process control, and process monitoring.

Process Modeling & Simulation

We develop process models that form the foundation of process design and control. We combine first-principles modeling based on governing equations from reaction engineering and thermodynamics with data-driven modeling using universal approximators such as artificial neural networks, to accurately describe steady-state and dynamic behavior. When physical properties required for process design are missing, we predict them with AI models based on molecular structure, such as graph neural networks. When there is not enough data to train AI models, we also predict properties using computational chemistry methods such as DFT/AIMD.

Process Modeling & Simulation
Process Modeling and SimulationFirst-principles ModelData-driven ModelSurrogate ModelProperty PredictionComputational Chemistry

Process Design & Optimization

Based on steady-state models, we design processes that produce target products from raw materials and conduct multi-objective optimization research that simultaneously maximizes economic and environmental performance. By applying various optimization algorithms such as Bayesian optimization and superstructure optimization, we design many types of chemical processes, including CO2 conversion (CCU) and bioprocesses. We also study agentic AI–based autonomous process design, developing methods in which AI performs process modeling and optimization on its own.

Process Design & Optimization
Process Design and OptimizationMulti-Objective OptimizationBayesian OptimizationSuperstructure OptimizationReaction Network ExplorationAgentic AI

Process Control & Operation

We develop optimal operating strategies for chemical processes based on dynamic models and data. We apply optimization-based control techniques such as model predictive control (MPC), dynamic optimization, and reinforcement learning to chemical, semiconductor, and bioprocesses. We build control loops using tools such as Python and Simulink Desktop Real-Time, and improve the economics and stability of processes through real-time control. We also use design of experiments (DoE) techniques to efficiently explore the uncertainties that arise in process operation.

Process Control & Operation
Process Control and OperationModel Predictive ControlDynamic OptimizationReinforcement LearningControl Loop DesignDesign of Experiments

Process Monitoring & Safety

We study early detection and prevention of process faults and accidents from both big data and sparse data. By combining multivariate statistical techniques with machine learning (e.g., data manifold learning and Bayesian neural networks), we perform anomaly detection and root-cause diagnosis, and implement soft sensing for unmeasurable variables. We also model sparse safety-related data with probability distribution functions to support conservative decision-making and predictive maintenance scheduling, and conduct knowledge graph–based Graph RAG research for the automatic generation of standard operating procedures (SOPs).

Process Monitoring & Safety
Process Monitoring and SafetyFault DetectionFault DiagnosisCausality AnalysisKnowledge Graph

Related Projects

View all →

No related projects yet. View all →