Dynamic modelling of fine grinding process in wet-operated stirred media mills using a coupled CFD-DEM, AI, and PBM framework

• Applicant: Professor Dr.-Ing. Arno Kwade
• Subject Area: Mechanical Process Engineering
• Term: since 2026
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 571547750

Fine grinding in wet-operated stirred media mills (WSMM) is essential across various industries, including pharmaceuticals, chemicals, and minerals processing. However, the complexity of multiphase interactions among fluid, grinding media, and particles poses significant challenges for process design, energy efficiency, and scale-up. This project addresses the need for a comprehensive, predictive, and dynamically adaptable model for WSMM by integrating advanced simulation tools—Computational Fluid Dynamics (CFD), Discrete Element Method (DEM), and Population Balance Modelling (PBM)—with data-driven Artificial Intelligence (AI). The central aim is to develop a dynamic modelling framework that overcomes the computational limitations of fully coupled CFD-DEM-PBM simulations by using surrogate AI models trained via Genetic Reinforcement Learning (GRL). These models will approximate machine functions such as stress energy distribution and energy transfer factors based on a compiled simulation database. In order to ensure accurate and generalizable predictions, the project includes experimental and numerical investigation of particle capture between grinding media—an aspect critical to understanding how stress energy is actually imparted to product particles. Detailed CFD-DEM simulations and targeted laboratory experiments will be used to quantify this behavior under different process conditions. Moreover, the project includes experimental characterization of material-specific breakage behavior using specialized equipment developed in-house. Materials like quartz and calcium carbonate will serve as model systems, with attention paid to behavior at the lower micron scale. Furthermore, rheological behavior of particulate suspensions—strongly influencing grinding media motion and stress transfer—will be characterized using shear-dependent viscosity measurements. These will be modeled using a hybrid mechanistic-genetic algorithm framework that captures the effects of particle size, concentration, and interparticle forces. All these elements—machine and material functions, material capture, and rheology—will be systematically integrated into a dynamic PBM framework. This framework will be implemented within an open-source flowsheet simulation tool, providing a flexible platform for future users. The resulting model will be capable of simulating particle size evolution almost in real time and under varying process conditions, offering a powerful tool for optimization and scale-up. This multidisciplinary approach combines mechanistic understanding with AI-driven data modeling to establish a novel framework for dynamically predicting WSMM performance. It sets the foundation for highly energy-efficient, scalable, and sustainable comminution processes, contributing to advances in digital process engineering.

• DFG Programme: Research Grants
• Co-Investigator: Professor Dr.-Ing. Carsten Schilde