LBM for two-phase (bio-)reactors

Studying the hydrodynamics in industrial stirred tanks via Computational Fluid Dynamics has gained a lot of attention in recent years, since the increasing computational capacities allow for sophisticated numerical simulations for a lot of engineering applications. Modern graphic cards are extensively developed for parallel calculations for scientific purposes. As a highly parallelizable algorithm, the lattice Boltzmann method has been applied successfully for efficient calculation of single-phase, as well as multiphase simulations in various flavors. A Lagrangian approach for the calculation of the gas phase in industrial bioreactors, specifically stirred tanks, is presented on top of the standard lattice Boltzmann method with the BGK collision operator for single phase flow. Different fluid phase boundary conditions are compared to model the moving walls inside the bioreactor, e.g., the stirrer blades. Turbulence is modeled with the Large Eddy Simulation approach. Production scale reactors may range up to several hundred cubic meters with internal structures such as heat exchanger bundles inside the fluid domain, therefore the size of individual tubes can be significantly smaller than the vessel diameter. In order to resolve the exact fluid flow field around these tubes the resolution must be significantly higher than in the bulk of the fluid. Increasing the resolution in the whole domain makes the simulation unfeasible due to the higher computational cost. Alternatively, local grid refinement can be applied or, if the exact flow inside the tube bundle is of minor interest, a porous media model can be applied that recovers the correct flow around the tube bundles. In order to correctly predict the fluid flow field with the porous media approach the flow rate and pressure gradient through the modeled porous zone must be known prior to the full-scale simulation. Modeling the distribution of soluble species in the liquid phase is crucial for the simulation of mixing times in stirred tank reactors. Therefore, a mass conserving and stable LBM-approach, recently presented by Osmanlic and Körner (2016) is applied for the simulation of advective transport. This approach does not produce unphysical oscillations, only models advective transport and is therefore applied in the case of high Peclet numbers where diffusion is negligible. For the simulation of thermal energy distribution inside stirred tanks, the transport model of Osmanlic is adapted and studied extensively regarding numerical diffusion.