Abstract:
Lattice Boltzmann method (LBM), as a new computational fluid simulation method, has aroused widespread attention in recent decades within engineering practice. LBM with large eddy simulation (LBM-LES) model is commonly used in predicting high Reynolds flow, and is considered to have a prediction accuracy similar to traditional finite volume method (FVMLES). Nonetheless, a systematic discussion on the accuracy of LBM-LES, and its consistency with FVM-LES, in indoor turbulent flow situations, is still insufficient. In this study, simulations of an indoor isothermal forced convection benchmark case (from IEA Annex 20) are implemented by using both LBM-LES and FVM-LES, with the aim of comparing the accuracies of LBM-LES and FVM-LES, in indoor turbulent flow situations. A comparison of their relative computation speeds, and parallel computation performances, is also implemented. The results show that LBM-LES can achieve the same level of accuracy as FVM-LES, in indoor turbulent flow situations; however, more refined meshes are required. Compared with FVMLES, half size grids are required for LBM-LES to approach similar levels of accuracy, meaning that the meshes of LBM-LES are approximately eight times as large as FVM-LES. The computation speeds of both LBM-LES and FVM-LES scale well, with the increase in the number of computation cores in one node. Their computation speeds (with the same accuracy) approach a similar level; however, the parallel computation speed of the LBM-LES speed can be larger than FVM, owing to its superior parallel speedup performance.
