Flutter aerodynamic shape optimization method for streamlined box girders based on support vector regression with Bayesian optimization
To solve the problems of time-consuming and laborious wind tunnel tests and large computational cost in computational fluid dynamics(CFD),a new aerodynamic shape optimization method for the flutter performance of streamlined box girder sections was proposed.The wind fairing parameters were selected as the design variables.The sectional aerostatic force coefficients were obtained through CFD,and the flutter critical wind speed estimated by quasi-steady theory was used as the optimization objective.A surrogate model was constructed by the support vector regression with Bayesian optimization,and updated by the mixed approach for adding data points.The optimal section was determined by the optimization algorithm.Taking the Humen Bridge as an example,the optimal section scheme for the flutter performance of the bridge within the feasible domain was obtained.The results show that the flutter critical wind speed increases firstly and then decreases with the increase of the wind fairing height.The overall performance is optimal with the relative height of 0.6,and the optimal section can be obtained with the relative height of 0.7.The flutter performance decreases significantly with the increase of the bottom flange width.When the inclination angle of the lower inclined web is between 14° and 16°,the flutter performance is optimal.After optimization,the estimated flutter critical wind speed of the bridge increases by about 31％compared with that of the original section.
streamlined box girderaerodynamic optimizationflutter performancesupport vector regres-sionBayesian optimizationquasi-steady theory
万方数据
维普
