Eulerian-Eulerian two-fluid computational fluid dynamic (CFD) models are increasingly used to predict bubbly flows at an industrial scale. In these approaches, interface transfer is modelled with closure models and correlations. Normally, the lateral void fraction distribution is considered to mainly result from a balance between the lift and wall lubrication forces. However, and despite the numerous models available that achieve, at least in pipe flows, a reasonable predictive accuracy, agreement on a broadly applicable and accurate modelling approach has not yet been reached. Additionally, the impact of turbulence modelling on the lateral void fraction distribution has not, in general, been examined in detail. In this work, an elliptic blending Reynolds stress model (EB-RSM), capable of resolving the turbulence field in the near-wall region and improved to account for the contribution of bubble-induced turbulence, is evaluated against best-practice k-ε and high-Reynolds second-moment turbulence closures. Lift and wall lubrication forces are initially deliberately neglected in the EB-RSM. Comparisons for flows in pipes and a square duct show that the EB-RSM reproduces the lateral void fraction distribution, including the peak in the void fraction in the near-wall region, and reaches an accuracy comparable to the other two models noted above. In rod bundles, even if none of the models considered performs with sufficient accuracy, the EB-RSM detects features of the flow that are not predicted by the other two approaches. Overall, the results demonstrate a much more prominent role of the turbulence structure and the induced cross-sectional pressure field on the lateral void fraction distribution than is normally considered. These effects need to be accounted for if more physically-consistent modelling of bubbly flows is to be achieved. The lift force is added to the EB-RSM in the final part of the paper, to provide a two-fluid formulation that can be used as the basis for additional developments aimed at improving the accuracy and general applicability of two-fluid CFD models.
CORE (COnnecting REpositories)
CORE (COnnecting REpositories)