Fast prediction of the flutter critical wind speed of streamlined box girders by using aerostatic force coefficients and artificial neural networks

Abstract

In the preliminary design stage of long-span bridges, there are many bridge deck schemes that need to check flutter stability through wind tunnel tests, and therefore it will take a lot of time and money. So, the authors are committed to investigating a way to quickly predict the flutter critical wind speed (FCWS) of bridge decks, and then select the schemes with better aerodynamic performance for further wind tunnel tests. Firstly, taking the streamlined box girder as the research object, the relationship between aerostatic force coefficients and flutter stability was derived, which proves that aerostatic force coefficients can be used to estimate the flutter stability of streamlined box girders. Secondly, through wind tunnel tests, a sample database with 315 sets of test data was obtained to establish the artificial neural networks (ANNs). Finally, by combining the ANNs and the CFD software, a procedure for quickly evaluating the flutter stability of streamlined box girders was proposed. Moreover, taking one long-span bridge as an example, the proposed procedure was used to predict the FCWS of each scheme and select the schemes with good aerodynamic performance. To check the prediction accuracy, the wind tunnel tests corresponding to the bridge deck schemes were performed. By comparing the prediction and test results, it is found that: 1) the proposed procedure can reasonably select the streamlined box girder schemes with good aerodynamic performance for further wind tunnel tests, thus avoiding unnecessary tests and improving work efficiency; 2) by combining with ANNs, aerostatic force coefficients can be used to evaluate the aerodynamic performance of streamlined box girders.