Thermal large-eddy simulation methods to model highly anisothermal and turbulent flows

Thermal large-eddy simulations (T-LES) of highly anisothermal and turbulent channel flows are assessed using direct numerical simulations (DNS). The investigated conditions are representative of solar receivers used in concentrated solar power towers. Four thermal operating conditions are considered. They aim to study several locations in the solar receiver. They are distinguished by different temperature profiles and thus different wall heat fluxes. The mean friction Reynolds number is close to 800 for all the simulations. The Navier-Stokes equations are solved under the low-Mach-number approximation. The nonlinear terms corresponding to the velocity-velocity and the velocity-temperature correlations are modeled. Functional, structural, and mixed models are investigated. An extension of the anisotropic minimum dissipation (AMD) model to compressible case and two-layer mixed models are proposed and assessed. Fourth-order and second-order centered schemes are tested for the discretization of the momentum convection term. First, a global assessment of 16T-LES approaches on mean quantities and correlations for three different meshes is performed in reference conditions. Then, three of the T-LES are selected for more detailed analyses. The mesh effect and the influence of the thermal conditions on the model accuracy are investigated. These detailed studies consist of the comparison of the relative error of the T-LES on mean quantities and correlations and the visualization of the normalized profiles as functions of the wall-normal distance. The results highlight the good agreement of two-layer mixed models consisting of the combination of the Bardina and the AMD models with the DNS for the three tested meshes.

Automated summary

Thermal large-eddy simulations (T-LES) of anisothermal and turbulent channel flows in solar receivers for concentrated solar power towers were evaluated using direct numerical simulations (DNS). Four different thermal operating conditions were considered to study the performance of various models, and the accuracy of two-layer mixed models combining the Bardina and AMD models was demonstrated. The results showed good agreement between the T-LES and DNS for the tested meshes.