Prediction Of Ground Settlement In Subway Construction Based On IGA-BP Neural Network And Realization Of Information System

At present,many cities in China are greatly developing urban rail transit.As a new,efficient and green public transport tool,subway plays an increasingly important role in the city.Urban subway construction is a difficult project.Although the construction technology is improving,the ground settlement caused by construction is still inevitable,which may cause major safety accidents.Therefore,it is of great practical significance to effectively monitor and accurately predict the ground settlement.In order to accurately predict the ground settlement caused by subway construction,the genetic algorithm was improved,and the weight and threshold of BP neural network model were determined by the improved genetic algorithm.The model of BP neural network based on improved genetic algorithm(IGA-BP neural network)was constructed,and the effectiveness of the improvement of genetic algorithm and the model were verified by the actual monitoring data of the project.The results show that IGA-BP neural network model has the highest prediction accuracy for ground settlement.In order to improve the monitoring efficiency,a ground settlement information system for subway construction was built based on BIM and GIS integration technology,which made the ground settlement monitoring information and environmental information visualized in the three-dimensional scene.At the same time,the IGA-BP neural network model was used as the underlying technical support to realize the prediction of ground settlement in the system.The specific work is as follows.(1)The genetic algorithm was improved.Based on the idea of tournament method,the population of each generation was ranked according to the increasing fitness,and a “grade selection method” was proposed.According to the fitness,individuals in the population were divided into four grades: excellent,good,medium and poor.In the next generation selection operation,the individuals of the four levels were selected according to the specific ratio.This method can not only ensure theselection of better individuals in the population,but also maintain the diversity of the population selection.The global search ability of genetic algorithm has not been destroyed.The ability of genetic algorithm to get the optimal solution is improved.When the improved genetic algorithm is combined with neural network to predict the ground settlement,it can achieve higher prediction accuracy.(2)The BP neural network model based on the improved genetic algorithm was constructed.The BP neural network model was optimized by using the genetic algorithm with the “grade selection method” as the selection operator.Although BP neural network model has a good optimization ability,it has poor search ability in the global scope,slow convergence speed,easy to fall into local extreme value,and the network structure is not easy to determine,weight and threshold have a great impact on the training results,so if this method is used alone,the effect is not very ideal.The improved genetic algorithm has strong global search ability,easy to get the global optimal solution,and can overcome the defects of BP neural network model.Therefore,the improved genetic algorithm is used to determine the weight and threshold of the neural network model and optimize the BP neural network model.IGA-BP neural network model is constructed.The validity of IGA-BP neural network model was verified by the ground settlement data during the construction of Beijing subway.Using Matlab simulation,the prediction results of IGA-BP neural network model are compared with those of GA-BP and BP neural network model.The results show that the prediction accuracy of IGA-BP neural network model is 43.18% higher than that of GA-BP neural network model.The prediction accuracy of IGA-BP neural network model is 49.89%higher than that of BP neural network model.Therefore,the IGA-BP neural network model has better prediction accuracy for ground settlement prediction,that is,the genetic algorithm using “grade selection method” as the selection operator has better optimization ability than the genetic algorithm using roulette method as the selection operator,and can obtain more accurate prediction results when optimizing the BP neural network for ground settlement prediction.(3)The ground settlement information system for subway construction was constructed.There are two problems in the traditional two-dimensional groundsettlement monitoring system: on the one hand,the system only has the monitoring function but not the prediction function.The accuracy of the management personnel to predict the future ground settlement based on the settlement value and experience is not high;on the other hand,the monitoring information is based on CAD drawing,which can't express the state of surrounding environment and the spatial relationship between construction and environment directly.In order to solve these two problems,a ground settlement information system for subway construction was constructed.The IGA-BP neural network model,which has been studied and verified,was used as the technical support of the bottom layer of the system to improve the accuracy of ground settlement prediction.At the same time,the integration technology of BIM and GIS was introduced.The multi-source data in subway construction,such as ground settlement monitoring data,model data and environmental data were integrated.The three-dimensional visualization of monitoring information,engineering information,geospatial information and settlement warning were realized.