Analysis of gas-solid flow using particle-resolved direct numerical simulation: Flow physics and modeling.

摘要

Gas-solid flows are encountered in many industrial processes such as pneumatic conveying, fluid catalytic cracking, CO2 capture and fast pyrolysis process. In spite of several experimental and numerical studies performed to understand the physics governing observed phenomena in gas-solid flows, and to propose accurate closure models for computational fluid dynamics (CFD) simulations using the averaged conservation equations, there are several challenges in gas-solid flows that yet need to be addressed. In many of the industrial processes, the solid-to-fluid density ratio is of the order of 100 to 1000, and the particle diameter ranges from 50 to 500 mu m. The interaction of heavy and large particles with the carrier phase leads to the formation of a boundary layer around each particle that in turn gives rise to interphase momentum transfer at the fluid-solid interface. The rate of work done by the carrier flow to sustain the interphase transfer of momentum leads to generation of velocity fluctuations in both the gas phase and the solid phase. Gas-phase velocity fluctuations enhance gas-particle heat transfer and the mixing of chemical species. Additionally, fluctuating motion of solid particles together with microscale hydrodynamic instabilities give rise to formation of mesoscopic particle clusters in gas-solid flows. The particle clusters then modify the hydrodynamic field and then the interconnected phenomena mentioned above dynamically modify the response of the system. Furthermore, if there exists a particle size distribution in the dispersed phase, the differences in the gas-particle and particle-particle drag forces lead to the segregation phenomenon.;In this study, particle-resolved direct numerical simulation (PR-DNS) is used to address some aspects of the challenges noted above, and to propose closure models for device-scale CFD calculations. First, the level of gas-phase velocity fluctuations is quantified, and its dependence on flow parameters is explained. An algebraic Reynolds stress model is proposed by decomposing the Reynolds stress into isotropic and deviatoric parts. Also the influence of solid particles with isotropic turbulent flow has been addressed using PR-DNS. In addition, in this study the slip velocity between two particle size classes in a bidisperse mixture is quantified, which is the key signature of segregation of particle size classes. The predictive capability of two-fluid closure models in predicting the slip velocity between particle size classes is also assessed. PR-DNS is used to propose a bidisperse gas-particle drag model that improves the prediction of the mean slip velocity between the two particle size classes. In addition, the mechanism of transfer of kinetic energy from the mean flow to fluid-phase and particle velocity fluctuations in a homogeneous bidisperse suspension is explained. This mechanism of transfer of energy is important because particle velocity fluctuations affect the particle-particle drag, which jointly with the gas-particle drag on each particle class determines the mean slip velocity between the two particle classes. In this study we have also used PR-DNS to quantify the mean drag force on particle clusters that are statistically consistent with those observed in experiments. A clustered particle drag model has been proposed based on our PR-DNS results. To address the effect of filtering the hydrodynamic field on flow statistics, which is used in LES of gas-solid flows, we have shown that the source and sink of kinetic energy in particle velocity fluctuations obtained from the PR-DNS are different from those predicted by the LES approach. These differences lead to a different level of kinetic energy in the solid phase obtained from the two approaches, and thus the flow characteristics that depend on solid-phase kinetic energy, such as formation and evolution of particle clusters, may not be comparable between the PR-DNS and LES approaches. In this study we have also used PR-DNS to quantify the growth rate of mixing length in a particle-laden mixing layer, and the corresponding mechanism is identified by using a scaling analysis.