The Application Of Improved Adaptive Genetic Algorithm And BP Neural Network In Bridge Construction Control

With the rapid development of national economy and integrative strength,the highway and bridge construction in our country also increase very fast.Pre-stressed concrete continuous girder bridges have many advantages,such as easy construction, small deformation,reasonable cost and strong spanning capacity,and that's why it's widely used.During the construction of the pre-stressed concrete continuous bridges by using the cast-in-place cantilever method,the internal force and deformation of the structure will be changed greatly and complicatedly for the effect of creep and shrinkage of concrete,temporary load,temperature effects and some others uncertain factors. Besides,the key of structure computation is to ensure the veracity of its result;the key of error adjustment is to confirm the elevation of the formwork according to the deflections that are collected during the construction.In order to guarantee the quality and safety of the construction,the construction control is indispensable.The construction control can also ensure the line shape and internal force of the structure, after the construction is completed,in accordance with the design.At the present time,a large number of literatures indicate that it's feasible and effectual to use the artificial neural networks in actual projects.However,the application of bridge construction control is still at the exploratory stage.BP neural network has obvious advantages in function approximation.So it can be used to forecast the elevation(error) in bridge construction control.But the traditional BP neural network has some problems,such as slow convergence rate,local optimization and sensitive to the initial value.The probability is used in genetic algorithm,so it can avoid falling of the local minimum point.But it lacks of the local search capabilities.Thus,combining the genetic algorithm and BP neural network will be beneficial to have complementary advantages.And it can also speed up the algorithmic convergence speed on the premise of ensuring the overall characteristics of the genetic algorithm.Firstly,this paper generalizes the research status of bridge construction control in our State,and introduces the purpose,the content and implementing measures of the bridge construction control.Then analyzes the characteristics and deficiencies of the Neural Network and Genetic Algorithm,discusses the confluent ways of these two algorithms and potential application.Secondly,For the purpose of getting stress and deflection during construction,a model is made to simulate the construction course of the approach spans of Danjiangkou bridge for South-to-north water diversion project,which can be provided a theory base for the initialize heights of girders as well.Finally,an adaptive Genetic Algorithm is put forward to train artificial Neural Network,which is based on the maximum evolutionary step of mini-zone.And the algorithm is programmed by Matlab R2007.According to the research,the method is practical,which is used in the adjustment of the deflection errors for the girder during the construction control.Meanwhile,the method used in this paper also has some reference value for using the GANN in bridge construction control.