Optimization of RANS turbulence models using genetic algorithms to improve the prediction of wind pressure coefficients on low-rise buildings

Abstract

Being associated with natural ventilation, the pressure distribution on surfaces is relevant for energy consumption, thermal comfort, and air quality in buildings. The aim of this work is to present a simulation-based optimization methodology to recalibrate the closure coefficients of Reynolds-averaged Navier-Stokes (RANS) turbulence models in order to improve the prediction of wind surface-averaged pressure coefficients on a wide range of isolated low-rise buildings. To accomplish this, genetic algorithms and Computational Fluid Dynamics (CFD) simulations are dynamically coupled to find the closure coefficients set which minimize the CFD prediction error regarding wind-tunnel experimental data. The methodology is applied to two turbulence models, the renormalization group k-epsilon model (RNG) and the Spalart-Allmaras model (SA), considering as target cases buildings with different roof types (flat, gable and hip) and wind incidence angles. In order to show the strength of the novel optimal sets of closure coefficients obtained, an exhaustive validation is performed over other low-rise buildings (52 new cases) which were not calibrated against. Results validate using the optimal sets because the recalibrated RNG and SA models decrease the prediction error between 11-64% and 8?45%, respectively, regarding using the standard ones.