Application Of BP Neural Network On Construction Control Of Long Span Steel Box Girder Cable-stayed Bridge

With the employment of high strength material, improvement of structural analysis methods, development of construction methods, it is a new era of rapid development for the cable-stayed bridge in recent years. But the cable-stayed bridge is a high-order statically indeterminate and flexible structure, the design objective (the permanent load internal force and the final bridge line) has relativity with the process of construction. During construction, along with the variation of the load status and the structure system of the cable-stayed bridge, the structure internal force and the structure deformation change continuously. A specific ordain to the construction sequence must be established and effectively control in the construction must be actualized.In the thesis, the development status of cable-stayed bridge and the cable-stayed bridge construction control are summarized at home and abroad. The basic theory, principle, content, method and influence factor of the cable-stayed bridge construction control are expatiated. The emulation analysis methods of cable-stayed bridge construction control (forward-calculation method, back-running method, no-stress method) are amplified. Least square method, grey system theory, Kalman filter method, artificial neural network method are discussed and analyzed from the point of construction control methods that were used in bridge project. BP neural networks algorithm is derived in detail to prove the possibility of the BP neural network method used to construction control from theory. At the foundation of analyzing the blemish of BP neural networks algorithm in physically applied, a homologous improvement method is proposed. Based on the practical engineering of the Huangpu Bridge over north branch of Pear River in Guangzhou, a long-span cable-stayed bridge elevation prediction procedure is designed with the MATLAB procedure language and BP neural networks algorithm and a favourable effect is achieved. Depending on the parameters analysis of the long-span cable-stayed bridge elevation prediction procedure, some valuable conclusions and experiences are obtained.