Filtered Modelling Approaches for Industrial Gas-Particle Suspensions: Basics and Future Developments

This course focuses on numerical methods for simulating large-scale environmental and industrial gas-particle flows, such as fluidized beds used in biomass combustion, olefin polymerization, and oil cracking. Numerical simulations are vital for optimizing these processes to reduce CO2 emissions, energy consumption, and pollutant output. However, a major challenge lies in the wide range of scales involved: while large flow structures span meters, the smallest particle structures are only a few diameters wide. Simulating such systems requires solving trillions of particle trajectories and their interactions with the surrounding fluid, which is computationally infeasible even with state-of-the-art supercomputers. To overcome this, continuum approaches are often employed, treating fluid and particle phases as interpenetrating continua. These methods, however, demand fine grid resolutions to capture heterogeneous structures like bubbles in fluidized beds or particle clusters in risers. As a result, their application is limited to a handful of high-performance computing (HPC) centers worldwide. Filtered (coarse-grained) models offer a solution by using coarser grids, but this can lead to inaccuracies, such as overestimating bed expansion in fluidized beds or misestimating bedload transport in deserts. The course delves into the implications of coarse graining for numerical methods. For continuum models, the two-fluid model (TFM) is widely used in engineering applications. Coarse graining leads to the filtered two-fluid model (FTFM), which incorporates sub-grid effects. Without proper resolution, TFM predictions can be highly inaccurate, such as overestimating dense fluidized bed expansion or solid mass flow rates. FTFM addresses this by filtering TFM equations, introducing sub-grid terms that require closures. These closures are typically derived from analogies to single-phase turbulence or through machine learning/artificial intelligence (AI). FTFM has been successfully applied in industries like iron ore reduction and polyolefin production. The course also examines Euler-Lagrangian methods, where the fluid phase is treated as a continuum and particles are tracked individually. These methods are computationally demanding, so parcel-based approaches are used, grouping particles into computational molecules (parcels) to reduce the number of equations. However, this introduces challenges similar to those in FTFM. A notable example is the MPPIC (multi-phase particle in cell) method. Lecturers from leading research teams will provide an overview of state-of-the-art numerical methods for large-scale particulate flows, addressing challenges across different length scales. They will discuss current bottlenecks, innovative approaches (e.g., AI), and the advantages of coarse-graining models in practical applications. Additionally, the course will highlight how fluid and particle properties influence inter-particle and fluid-particle forces, ultimately shaping flow regimes. This course especially addresses doctoral students, young researchers and practical engineers. Participants are encouraged to bring a poster covering their current research topic. Some of the lectures will also be accompanied by practical training examples.