Research

Industrial processes transform raw materials into products with tailored functional, structural, and sensory properties. Designing such processes requires understanding how molecular interactions ultimately determine the performance of industrial production systems. Our research therefore bridges multiple length and time scales—from molecular phenomena to full-scale manufacturing—through an integrated process systems engineering framework that combines physics-based modeling, multiscale experimentation, and artificial intelligence.

At each scale, different physical mechanisms dominate material behavior. Molecular interactions govern chemical reactions and intermolecular forces, microstructure determines functional material properties, continuum mechanics describes macroscopic behavior, while process and plant scales capture transport phenomena, equipment performance, and manufacturing efficiency. Only by linking these scales can materials and processes be designed in a predictive rather than empirical manner.

Because many industrial processes occur inside enclosed equipment, direct observation is often impossible. Our research therefore combines advanced experiments with computational models to reveal the underlying process dynamics. Experimental observations continuously validate and improve numerical models, while simulations provide information that cannot be measured directly. The resulting multiscale datasets form the foundation for modern data-driven engineering.

Machine learning connects these heterogeneous data sources into predictive digital workflows. Experimental measurements, pilot-scale sensor data, material characterization, and three-dimensional CFD simulations are integrated to identify governing mechanisms, estimate model parameters, optimize operating conditions, detect process abnormalities, and develop digital twins. New experimental data continuously improve both the physical models and learning algorithms, creating a self-improving framework for intelligent manufacturing.

Our research is built upon three complementary pillars.

Physics-Based Modeling

We develop physics-based and microstructure-informed computational models that describe the coupled transport, reaction, and deformation phenomena governing industrial processes. Particular emphasis is placed on evolving material properties arising from thermal effects, chemical and enzymatic reactions, phase transformations, and structural evolution.

These models enable predictive simulations of complex engineering processes such as mixing, spray drying, encapsulation, coating technologies, filtration, wet steam cleaning, additive manufacturing, and bioprinting. Beyond reproducing experimental observations, they provide mechanistic insight into process behavior and support rational process design, optimization, and scale-up.

Multiscale Experimental Characterization

Reliable predictive models require high-quality experimental data. We therefore develop customized experimental platforms that enable in-situ and real-time characterization of materials under realistic processing conditions.

Our experimental toolbox combines rheological and mechanical testing with advanced characterization techniques—including SAXS, SANS, μ-CT, FBRM, Raman spectroscopy, and high-resolution optical imaging—to monitor structural evolution from the molecular to the macroscopic scale. At the process level, state-of-the-art sensors are integrated into laboratory and pilot-scale equipment to continuously capture material properties, process variables, and product quality throughout manufacturing.

These measurements provide the quantitative basis for constitutive model development, CFD validation, and machine learning.

Data-Driven Process Systems Engineering

The increasing availability of high-performance computing and industrial process data has transformed the way engineering systems are designed and operated. We integrate machine learning with physics-based models to develop predictive, interpretable, and transferable engineering solutions.

Artificial intelligence supports parameter estimation, surrogate modeling, process optimization, quality prediction, anomaly detection, soft sensing, model predictive control, and digital twin technologies. Rather than replacing physical understanding, data-driven methods complement first-principles models by identifying hidden relationships, accelerating simulations, and continuously improving predictive accuracy as new data become available.

This hybrid modeling philosophy combines the robustness of physical laws with the flexibility of modern artificial intelligence.

Vision

Our vision is to establish an integrated digital process systems engineering platform that seamlessly links experiments, simulations, sensing technologies, and machine learning across all relevant engineering scales. By connecting molecular phenomena with industrial manufacturing, we aim to accelerate process development, enable intelligent functional materials, improve resource efficiency, and contribute to the next generation of sustainable and autonomous production systems.