Abstract Summary
A wind engineering database specifically for bridges has been built by using Access Database software and Java programming language based on the previous wind tunnel test results of long span bridges. The connection between the underlying database and the foreground application is driven by a local protocol to achieve platform independence and high execution efficiency. All the data can be summarized into three modules: basic information, dynamic characteristics, and aerodynamic parameters. The machine learning models for identifying flutter derivatives of closed box girders are trained and developed via gradient boosting decision tree based on this database. 20 sets of wind tunnel test data from this database are used for the machine learning modeling, but it is difficult to obtain a good training effect for flutter derivatives under 20 sets of data because the potential distribution characteristics of flutter derivatives are not very clear and the wind tunnel test data usually fluctuates greatly. Therefore, another data pattern was used: the flutter derivatives of these 20 sets of cross sections were re calculated by CFD numerical simulation, combined with the additional 35 sets of numerical simulation data as mixed datasets to jointly drive the training process of machine learning. The machine learning models can explore the underlying distribution of dataset. The trained models have good fitting and generalization ability under the current data conditions. In this way, the present research work can make the identification of flutter derivatives separated from tedious wind tunnel tests and complex numerical simulations to some extent. It can also provide a convenient and feasible option for expanding data sets of aerodynamic parameters. In addition, it can help determine the appropriate shape of the box girder cross section in the preliminary design stage of long span bridges and provide the necessary reference for the aerodynamic shape optimization by modifying local geometric features of the cross section to evaluate the influence of the aerodynamic shape on flutter performance. Before that, it was analyzed which flutter derivatives have the major effect on the flutter critical wind velocity so that the flutter performance analysis process can be simplified. Then the relationship between the shape of cross section and the flutter critical wind velocity can be analyzed. For closed box girder, there are not many factors affecting the aerodynamic shape without considering the influence of the auxiliary facilities on the flutter critical wind velocity in the construction stage. It is time consuming and may not lead to better calculation results if every detail of closed box girder is taken into account. Therefore, this study only discusses three important parameters: width to height ratio, wind fairing angle and inclined web slope. The calculation results show that flutter critical wind velocity decreases with the increase of width to height ratio. It first increases and then decreases with the change of wind fairing angle, and it decreases with the increase of inclined web slope, which is almost linear.
