Vortex-Induced Vibration Performance Prediction Of Flat Streamlined Box Girders Based On Neural Network

The streamlined box girder section is a commonly used section in modern long-span bridges,unreasonable aerodynamic shape of the section will cause vortex-induced vibration(VIV)of the bridge at low wind speeds.Due to the complicated mechanism of VIV,the most commonly used method for studying vortex vibration is to conduct wind tunnel tests based on engineering experience,but the selection of measures is somewhat blind.In order to quickly estimate whether VIV will occur in the section,wind tunnel experiments and CFD simulations are performed on the two sections respectively,a torsional VIV database is established,and the VIV performance of the sections is automatically identified through neural network.The main research contents of this paper are as follows:(1)Collected the data when VIV occurred on the rectangular section with the aspect ratio of 6,and used the idea of cross-validation to select the thin-plate spline interpolation with smaller error to expand the sample.The database is divided into two parts: training set and test set.The training set is used to train the BP neural network and the RBF neural network,and the test set is used to test the trained network.The results prove that both of these networks can be used to predict the VIV performance of a rectangular section,of which the aspect ratio is 6.(2)In order to predict the VIV characteristics of the streamlined box girder section,the VIV data of the two streamlined box girder sections of bridge A and bridge B were collected.Design the segment model of bridge A and conduct wind tunnel tests,change the section’s wind fairing angles,torsional mass moments of inertia,.and damping ratio,and obtain the VIV data of bridge A under different conditions.Perform VIV numerical simulation on bridge B and verify the accuracy of the numerical simulation.Change the wind fairing angle,wind fairing position,mass moment of inertia,and damping ratio of the section of bridge B,calculate the VIV data of bridge B under different conditions.(3)The VIV data of bridge A and bridge B are integrated to provide data support for neural network learning.Six parameters of section aspect ratio,wind fairing angle,wind fairing position,web angle,damping ratio,and mass moment of inertia are selected as the characteristic parameters of vortex vibration,and matlab is used to verify that the characteristic parameters are independent of each other.The thin-plate spline interpolation is used to expand the existing VIV samples,and the expanded samples are divided into training set and test set.Use the training set to train the neural network and optimize the best parameters of different networks.The error evaluation of the learned network is carried out through the test set.Finally,the following three networks are obtained: The BP neural network and the RBF neural network that can predict the VIV characteristics of the streamlined box girder section,and the probabilistic neural network that can determine whether vortex vibration occurs in the bridge.