A GPU-accelerated solver for turbulent flow and scalar transport based on the Lattice Boltzmann method

In this article, we apply the Vreman model with a dynamic procedure for subgrid scale modeling of turbulent flow and scalar transport under the Lattice Boltzmann framework. Numerical results of two types of typical turbulent flow systems are presented and discussed, i.e., the channel Poiseuille flow with heat transfer, and the Rayleigh-Taylor convection. Due to their intrinsic differences, analyses of the two turbulent phenomena are carried out mainly according to the statistical averaging property and the temporal evolution process, respectively. The feasibility and accuracy of the present turbulent solver is validated via comparisons with direct numerical simulation results and theoretical predictions. In addition, the GPU parallel technique is applied and yields a very efficient solver for turbulent flow and scalar transport, and the algorithm is introduced in detail. Overall, our work offers a useful extension of the current Lattice Boltzmann method and a powerful tool for turbulent modeling.