Linear Control For Hybrid Railway Curved Bridge With Rigid Frame And Continuous Girder

In recent years, the construction of high speed railway networks and passenger dedicated lines is at the peak period, and the long-span continuous rigid-frame curved bridges also get special attentions for adapting to rough terrains, ensuring uninterrupted transportation, minimizing the environmental damage to the existing topography, and important buildings, and meeting the requirements of water conservation and flood control. Fewer studies have been carried out about the linear control of this kind bridge by using the cantilever construction method. Based on the existing linear control of curved bridges and hybrid-bridge with rigid frame and continuous girder, the Shengquan No.l Bridge located at Guilin-Guangzhou Railway Line is taken as the reference to study the linear control of curved bridges with small radius, multiple spans, high piers, piles with different lengths and other adverse effects.1) The development of railway curved girder bridges, the state of art for bridge construction control and the basic principle of the adaptive construction control are breifly reviewed;2) The BP neural network, the genetic algorithms, grey correlation analysis and principal component analysis are adopted to study the linear control of hybrid bridge with rigid frame and continuous girder. The basic process of BP neural prediction is realized aided by the neural network toolbox in MATLAB software, and the parameter sensitivity analysis is engaged by the grey correlation analysis and principal component analysis, then the BP neural network improved by the genetic algorithm is introuduced to carry out the linear control of the hybrid bridge with rigid frame and continuous girder;3) Based on the approach of the vertical model elevation for casting in site cantilever segments, a novelty approach of transverse model elevation is put forward to meet the specific requirement of the curved bridge;4) The proposed method is verified numerically by a certain railway prestressed concrete continuous rigid frame bridge. The predicted vertical model elevation obtained from the BP neural network optimized by the genetic algorithm is compared with the numerical value, and the predicted value is consistent with the numerical value;5) The BP neural network optimized by the genetic algorithmis adopted to predict the vertical model elevation of Shengquan No.l Bridge during the construction process, and the predicated value is consistent with measured value. The transverse model elevation is also predicted by the proposed method, and it is corresponding to the numerical value; 6) The variations of pile foundation and curve radius have siginificant effects on the linear control and bridge design thorugh the intensive discussion of different pile foundation and curve radius.