Large-eddy simulation of turbulent natural convection in a cylindrical cavity using an off-lattice Boltzmann method

Abstract

In the present work, a characteristic-based off-lattice Boltzmann method with the large-eddy simulation (LES) as the turbulence model is developed for the simulation of turbulent natural convection. A double-distribution-function approach is used to resolve flow and thermal fields, and the proposed framework is developed, in three-dimensional curvilinear coordinates. The solver is verified using three benchmark cases, namely, the turbulent Taylor–Green vortex flow, natural convection in a periodic tall cavity, and Rayleigh–Bénard convection. Due to the absence of an inlet in this kind of closed cavity flow, initial perturbations are proposed and verified, which accelerate transition to a turbulent state. The turbulent natural convection in a cylindrical cavity is simulated for a Rayleigh number of [Formula: see text], and the flow and thermal characteristics are analyzed. A grid sensitivity study is conducted and an appropriate mesh resolution is selected, that is, further verified using the LES index of quality-of-resolution. The resulting turbulent flow and the associated thermal plume are analyzed using instantaneous and time-averaged mean and second-order statistics, vortical structures, turbulence anisotropy maps, energy budgets, frequency spectra, and the mean and root mean square of temperature and Nusselt numbers. The results indicate that the thermal plume region is highly anisotropic, whereas the rest of the annulus contains single-component axisymmetric turbulence. The production and convection of turbulence are dominant on top of the inner cylinder in the thermal plume region, whereas diffusion is dominant closer to the outer cylinder. The azimuthal profiles of mean Nusselt number for the inner and the outer cylinders are observed to be negatively correlated. Furthermore, natural convection in the cylindrical cavity is simulated for [Formula: see text] to [Formula: see text] and the effect of the Rayleigh number on the mean Nusselt number and flow patterns is studied.