A-priori evaluation of data-driven models for large-eddy simulations in natural convection

Abstract

Natural convection is a commonly occurring heat-transfer problem in many industrial flows and its prediction with conventional large eddy simulations (LES) at higher Rayleigh numbers using progressively coarser grids leads to increasingly inaccurate estimates of the Nusselt number (Nu). Thus, to improve the Nu number predictions, we utilise Gene Expression Programming (GEP) to develop sub-grid scale (SGS) stress and heat-flux models for LES. The models' development is performed using reference direct numerical simulation data of a Rayleigh-Bernard convection case. The developed models are then also evaluated on a geometrically different case to test the re-usability of the models. The a-priori results demonstrate a significant improvement in both mean and instantaneous quantities, using metrics such as alignment and predicted values of the SGS stress and heat-flux. In the future, the models will be tested a-posteriori by conducting LES with the GEP developed models.