| With the gradual degradation of bridge structure performance and the increase of load capacity,the number of dangerous bridges has increased year by year.To ensure the long-term performance of the bridge structure,it is very urgent to determine its load-limiting capacity based on the performance of existing bridges.Online real-time identification of load information is a new research direction in the future.The acquisition of vehicle load information is one of the important research foundations to determine the design load of the new bridge,the prediction of the remaining life of the existing bridge,and the early warning of overshoot.In the early stage of vehicle load information acquisition,a dynamic weighing system is generally used.In application,the original road surface needs to be dug,which not only destroys the original road surface structure,but also has a high cost of construction.The annual equipment maintenance and maintenance costs are expensive,which makes the system popular and Use is restricted.Therefore,it is very important to carry out mobile load identification.Based on the investigation of the current status of research at home and abroad,this paper first elaborated on the theory and method of moving load identification,and proposed a method of load parameter identification based on BP neural network model through comprehensive analysis.Secondly,the dynamic analysis of the large finite element program ANSYS was used to study the dynamic response of the bridge structure under multi-operating conditions through parametric modeling.The BP neural network structure was established based on MATLAB,and the dynamic response generated in ANSYS was used as the input of the BP neural network.The data completes the load identification process.Then,a simple supported T-beam analysis model was established to identify the parameters of the velocity and lateral position of the moving load.The dynamic response data such as the displacement,velocity and acceleration of the structure were selected as the input parameters of the BP neural network,and the load identification of different input parameters was studied.The influence of accuracy,on this basis,the multi-parameter input optimization design was carried out for the BP neural network structure,and the influencing factors of load identification were analyzed.The moving load speed,load,location of measuring point,unevenness of bridge deck and Influence of noise level and other factors on load identification.Finally,taking a 4-span prestressed concrete continuous T-beam bridge project as the analysis background,the finite element model is modified by the static measured deflection data,and the BP neural network structure is built and trained based on a large number of numerical simulation analysis data to take the vertical direction under heavy vehicles The measured acceleration data validates the feasibility of this BP neural network structure for identifying simple moving loads in real bridges.The research results show that for the same BP neural network structure,the acceleration input parameter recognition accuracy is the highest in the single input parameter recognition,and the speed and displacement single parameter input recognition accuracy is sequentially reduced;after the multi-input parameter combination optimization design,the speed and acceleration input parameter combination can be used Obviously optimize the recognition process and improve the recognition accuracy of moving loads.Among the influencing factors of moving load recognition,the smaller the moving load speed,the larger the axle load,the closer the acquisition point is to the mid-span,the smoother the bridge deck,and the lower the noise level,the recognition The higher the accuracy,the lower the recognition accuracy.The research results can provide a reference for bridge load identification and bridge structure performance evaluation,and can also provide technical support for bridge structure health monitoring. |
