A novel hybrid framework coupling the discrete velocity based-method, the particle-based method and the Grad's distribution function-based method for supersonic rarefied flows

Abstract

Supersonic rarefied flow has an intrinsically multiple-scale nature due to the large variations of density and characteristic length of the flow structures. Although particle-based methods such as direct Monte Carlo simulation (DSMC) present efficient performance in supersonic rarefied flows, the statistical noise as the nature of stochastic method causes a large amount of averaging steps to be required. To minimise the contamination of statistical noise and preserve the efficiency of capturing the multiscale rarefied effects, a finite volume based hybrid framework is proposed to couple the discrete velocity method (DVM), the particle-based method and the Grad's distribution function-based moment gas kinetic method (G13-MGKS). Benefiting from the coupled treatment both in the spatial and velocity space, the DVM-based method is applied to low-velocity areas close to the wall, which greatly eases the statistical noise and numerous steps of statistical averages can be avoided. Besides, the partition of different sub-methods conveniently determined by the initial flow field from conventional CFD solver has been verified without compromising any accuracy. Further applying the Grad's distribution function-based method in the area with moderate rarefied effect, the present hybrid DVM-Particle-Grad (HDPG) method exhibits the high efficiency and achieves a tenfold speedup ratio compared to the DVM.