Overcoming the bottlenecks of the VOF model: DNS of an extreme case of a plate with holes for a gas-liquid photobioreactor
Résumé
The goal of the research project Twins4GasPBR is to develop an open-source digital twin of a gas-fed photobioreactor, and to validate it experimentally. Since we are aiming for a 3D CFD-based model, one of the most important aspects with respect to realism, and that will allow better control over upscaling, is an appropriate reproduction of the geometry of the reactor prototype. As well, we intend to reproduce a somehow realistic bubbly flow, hence the choice of a (volume of fluid) VOF model.Our current experimental setup (Figure 1) is a cylinder with a liquid height of 0.64 m and a radius of 0.05 m. In order to facilitate the solubilization of the gas in the liquid, a micro-diffuser is used as gas inlet. This diffuser is an extreme version of a plate with holes: the pore diameter is only of a few µm, which means that a circular surface with r = 0.05 m is covered by an uncountable number of them. This poses two exceptional challenges that question the state of the art of modeling and simulation. As of today, and specially on places that work with applications of mathematical modelling, rather than on the development of modeling tools, it is common practice to use the built-in meshers that are shipped with the software. This means that the geometry must be drawn either point-by-point (e.g. comsol), or with an external CAD tool, and then imported into the simulation environment. Although current CAD packages allow for parametric drawing, both solutions are too manual, and hence not adapted to deal with hundreds or thousands of inlet patches.In the case all the inlets could be drawn to some extent, simulation of such a system would require a prohibitive amount of computing power. The VOF model relays on a diffuse interface approach, in which the interface (in this case between the liquid and the gas) spans several volume elements. Consequently, the quality of the numerical results will be extremely dependent not only on the general coarseness of the mesh, but on the ratio of faces per inlet patch. Additionally, in order to avoid artificial deformation of the bubbles, the cells should as close to a cubic shape as possible. The present work intends to show how to overcome these difficulties in a way that the numerical results are of practical use, using a reasonable amount of computing power, while keeping an appropriate representation of the bubbly flow provided by the VOF model.The mesh generation problem has been dealt by developing the tool blockMeshScript, that assists on the generation of a mesh that represents the reactor and the diffuser only requiring as inputs the pore size, their spacing, the radius and the height of the liquid. The output of this tool is a specification file that is compatible with the mesh generation tool blockMesh, that ships with openFoam.This tool was then used for a mesh consistency test that allowed to establish the number of faces per inlet patch that is necessary in order to reproduce the bubbly flow numerically without the need of turbulence models. Finally, a number of meshes that increase the amount of pores and decrease their radius were generated, and used to establish the liquid height at which bubble coalescence occurs. These two numerical experiments will allow to find a compromise between physical realism and practicality. Finally, the numerical bubble distributions were compared with experimental data.