基于代理模型的方形凹角截面超高层建筑气动外形优化

Aerodynamic shape optimization of a square supertall building with corner recession based on surrogate model

采用CFD数值模拟和基于代理模型的优化方法,研究方形凹角截面超高层建筑在多参数设计空间内的平面外形气动优化问题。首先,选取凹角率、凹角角度和凹角个数作为设计参数并考虑风向角的影响,基于最优拉丁超立方设计和非定常RANS方法获得初始样本点的气动力系数;其次,基于遗传算法(GA)和广义回归神经网络(GRNN)方法构建能准确预测设计参数与气动力系数之间复杂关系的代理模型;再次,以最小气动力系数为目标,利用GA-GRNN代理模型及其更新策略和非支配排序遗传算法在整个设计空间内寻优,确定全风向角下方形凹角截面的最优气动外形;最后,探讨凹角修正对截面气动性能的影响机理。结果表明,基于GA-GRNN代理模型的优化方法在气动外形优化方面具有较高的预测精度和计算效率;全风向角下的最优气动外形截面的凹角率为18.5%、凹角角度为70.4°、凹角个数为5,相比方形截面,该截面可使平均阻力系数和均方根升力系数最多分别减小53.4%和70.5%。该方法可为建筑结构气动外形优化及相关研究提供重要参考。

The CFD numerical simulation and surrogate modeling-based optimization method are adopted to investigate the aerodynamic shape optimization of a square-sectional supertall building with corner recession in multi-parameter design space.Firstly,the corner recession ratio,corner recession angle,number of corner recessions and wind direction angle are selected as the design parameters,and the optimal Latin hypercube design method is adopted to obtain the original sampling points of the design parameters uniformly distributed in the design space.the unsteady RANS method is utilized to calculate the aerodynamic force coefficients of each sampling point.Secondly,the genetic algorithm(GA)and general regression neural network method(GRNN)are adopted to construct a surrogate model to accurately predict the complex relation between the design parameters and the aerodynamic force coefficients.Thirdly,based on the objective function of minimum aerodynamic force coefficients,the non-dominated sorting genetic algorithm is utilized to search the multi-objective optimization in the whole design space,and the most optimal aerodynamic shape of the corner recessed square section in all the wind directions is determined.Finally,the influence mechanism of corner recession on the aerodynamic performance are discussed.The results show that the GA-GRNN surrogate modeling-based optimization method has good prediction accuracy and computational efficiency in the field of aerodynamic shape optimization.The design parameters of the optimal aerodynamic shape section are 18.5%for the corner recession ratio,70.4°for the corner recession angle,and 5 for the number of corner recessions.The maximum reductions of the mean drag and RMS lift coefficients of this section can reach 53.4%and 70.5%respectively when compared to the square section.The proposed method can provide reference for the aerodynamic shape optimization of building structures and relevant studies.