Research On Wind Alarm System For Traffic Safety On Bridges In Complex Mountainous Terrain

In recent years,the highway transportation industry in the Midwest has flourished,with abundant large-span bridges spanning mountainous valleys.The wind field environment around long-span bridges is so changeable that the driving environment is complex,where traffic accidents frequently occur.In order to reduce the adverse impact of strong winds on the safe driving on bridges in complex mountainous areas,it is essential to establish a reliable wind alarm system for traffic safety.Aimed on the wind alarm system for traffic safety on bridges in the complex mountainous areas,this paper has carried out the following researches.Selecting the Dadu River Bridge in Luding,Sichuan Province as the engineering background,the wind characteristics of a typical U-shaped deep-cut dry-hot valley are analyzed.Firstly,mean wind characteristics such as the wind speed,the wind direction and the wind attack angle are investigated among several measurement points along the bridge.Furthermore,the influence of the local topography on the mean wind characteristics is clarified.Then,the stability analysis of the wind speed time history is conducted,where the non-stationary nature of the wind speed records in the complex mountainous area is identified.Lastly,based on the traditional steady wind speed model and the non-stationary wind speed model,turbulent wind characteristics such as the gust factor,the power spectrum and the turbulent integral scale are obtained,respectively.The differences of calculation results derived from different models are compared.Additionally,the variation of turbulent wind characteristics along the bridge and the corresponding influencing factors are discussed.Affected by the terrain,the average wind speed of the main beam along large-span bridges located in the complex mountainous area is obviously uneven.For the simulation of such nonuniform random wind speed field,the traditional spectral solution is inefficient.In this study,a new method for accelerating the simulation of non-uniform random wind speed field is proposed.By rounding the coherence function matrix,and then using the new efficient Cholesky decomposition,coupled with the strategy of eliminating a large number of zero elements after decomposition,the non-uniform random wind speed field simulation is successfully accelerated.Through numerical examples,the simulation results of the new method are verified to be accurate and fast.In order to improve the accuracy of very-short wind speed prediction,a novel hybrid model based on the wavelet packet decomposition,the Elman neural network,the gradient boost regression tree and the density-based spatial clustering of applications with noise is proposed.In this model,the original wind speed time history is decomposed by the wavelet packet decomposition at first.The structure of the Elman neural network is determined by using the gradient boost regression tree for each sub-sequence after decomposition,and representative sample data are selected by the density-based spatial clustering of applications with noise.Through actual examples,the key parametes in this hybrid model are systematically discussed.Moreover,the performance of the proposed model is validated so superior that the prediction accuracy is high.The signal decomposition technique,including the empirical mode decomposition with its improved algorithms and the wavelet transform,can effectively raise the wind speed prediction accuracy.However,the prediction error is large in the highest frequency subsequence generated after decomposition.In order to improve the short-term wind speed prediction accuracy and reduce the adverse effect of the highest frequency subsequence on the final prediction result,a novel method based on the singular spectrum analysis for advanced processing the highest frequency sequence is proposed.Combined with the Elman neural network with a good prediction performance,two kinds of prediction models consisting of the empirical mode decomposition with its improved versions and the wavelet transform are proposed,respectively.Through the practical examples,the key parameters in each hybrid model are evaluated,and the superior performance of the hybrid model is verified.In this study,the WINDINGS system as the wind alarm system for traffic safety on bridges in the complex mountainous terrain,is established.The WINDINGS system is based on the C/S and B/S framework and consists of softwares and hardwares such as the on-site data collection station,the alarm system center server,and the WEB client.The on-site data collection station is responsible for the collection and transmission of wind data on the bridge.The alarm system central server is the brain center of the WINDINGS system,employed for the data collection,the wind speed prediction,the road surface status recognition,the establishment of vehicle speed limit-rules,and the release of vehicle warning information.The WEB client is used to receive and query a series of information such as the wind speed level,driving opinions,weather conditions and the historical data in real time.