Direct numerical simulation of reactor two-phase flows enabled by high performance computing
The presence of near wall bubbles may reduce the skin friction drag. It can be simulated by the two-fluid model (TFM) coupled with the population balance model (PBM). One disadvantage of the TFM-PBM coupling is that the bubbles with different sizes are transported with identical velocities. In this work, the phenomenon of drag reduction by the injection of micro-bubbles into turbulent boundary layer was investigated by the four-way coupled Eulerian quadrature-based moments method (E-QBMM). In the E-QBMM, bubbles movement is controlled by the generalized population balance equation (GPBE). The GPBE is transformed to the moment transport equations and solved by the higher-order realizable finite volume scheme. Simulation results predicted by the E-QBMM are compared against experimental data in the literature. Our results show that the E-QBMM
Since I have the raw results already. Manuscript should be completed in three months after you undertake this project. One month extension is tolerated, if necessary.
A rapidly increasing number of studies have been made on the modeling of bubble columns using computational fluid dynamics (CFD). In this work, we present a hybrid Eulerian-Eulerian (E-E) volume of fluid (E-E-VOF) method to simulate laboratory rectangle bubble columns with different superficial velocities. This method treats each phase under the E-E framework. Meanwhile, an interface compression term is added in the phase fraction equation to capture the large-scale interface. Quasi direct numerical simulation (quasi-DNS) with interface tracking and the classical E-E method are also employed as benchmark models. Our results show that the macroscopic features in the investigated laboratory bubble columns, such as the averaged phase fraction at different cross sections, the liquid velocities and the bubble plume period, predicted by the E-E-VOF, E-E method and quasi-DNS agree well with experimental data. Both the E-E-VOF and the quasi-DNS can predict the bubble swarm with large-scale interface. However, the quasi-DNS is highly computational resources demanding compared with the E-E-VOF method, which limits its application to industrial simulations.
Left: Prediction of the horizental liquid velocity by the quasi-DNS (not finished yet!). Right: Prediction by the E-E method in the literature.
Since I have the raw results already. Manuscript should be completed in two months after you undertake this project. One month extension is tolerated, if necessary.