Numerical studies on industrial flows

The present thesis belongs to the field of applied physics, more precisely numerical simulation of industrial fluid flows. Computational fluid dynamics (CFD) simulations are applied to investigate four different flow scenarios. Particular focus is on usage of scale-resolving methods in CFD. Each of the studied flow scenarios is motivated by collaboration with the Finnish industry. In Publication I the gas flow phenomena are studied in a commercial atomic layer deposition (ALD) reactor to better understand the fluid flow effects on the surface growth. Both the numerical and experimental results in Publication I indicate that the surface growth depends on the mixing of the precursors and the carrier gas. However, comparison of the numerical results to the experiments reveals that the numerically obtained precursor concentrations are not sufficient to predict the surface growth rate. In Publication II the performance of a plate fin and a square pin fin heat exchangers is investigated in a constrained pipe flow. The numerical results show that for the planar fins the analytical Nusselt number correlation for channel flow applies in the interior fins. For the pin fins a large variation in the fin-averaged Nusselt number is observed. The downstream wake mixing characteristics are shown to match with experimental data. In Publication III, the ghost fluid method (GFM) for coupling the volume of fluid scalar equation and momentum equation is implemented and thoroughly tested in marine context. The results of the test cases indicate that GFM has moderate benefits over the standard volume fluid formulation in OpenFOAM but it does not clearly outperform the standard formulation either. Further, alternative formulations for the viscous term treatment are studied. In Publication IV, the implemented GFM is applied to study the Reynolds number scaling effect for a free surface backward facing step geometry for the first time using large-eddy simulation. The forming wave shape is shown to depend on the Reynolds number indicating that care should be taken when interpreting model scale simulations or experiments.