Multi-fidelity shape optimization methodology for pedestrian-level wind environment

Abstract

In this study, a multi-fidelity shape optimization framework is proposed for the pedestrian-level wind environment (PLWE). In the proposed framework, low-fidelity computational fluid dynamics (CFD) models based on steady Reynolds-averaged Navier–Stokes equations (RANS) models and high-fidelity CFD models based on large-eddy simulation (LES) are efficiently integrated into the optimization process to improve the optimization reliability while maintaining its computational speed in an affordable range for practical engineering applications. The optimization solver is coupled with an approximation model generated by low-fidelity CFD samples obtained using a design of experiments (DOE) technique. The optimal candidates are then evaluated according to the degree of improvement of the objective function compared to the reference case. If the degree of improvement shows significant deviations between the low-fidelity and high-fidelity models, suitable corrections and modifications are applied to improve the reliability of the optimization process. The applicability of the proposed method was investigated in terms of minimizing the high-wind-speed area, as the optimization objective, around a high-rise building considering (a) uniform urban blocks and (b) real urban blocks with different frequency distributions of wind directions associated with two different local wind climates. In summary, a significant reduction in the critical strong-wind area around the target building was realized using the proposed optimization framework. Furthermore, the application of the proposed multi-fidelity optimization framework highlighted the importance of considering the local wind climate in architectural design.