Research On Deformation Prediction Of Subway Foundation Pit Based On BP Neural Network And Its Improved Algorithms

With the rapid development of China’s national economy,in order to relieve traffic congestion and improve the competitiveness of cities,subway construction has become the first choice,but in the subway construction,it is inevitable to face the deformation problem of station foundation pit excavation.Due to the influence of complex physical properties of the soil around the foundation pit and the fact that the foundation pit is generally located in the dense area of buildings and traffic lines,there will be great safety risks in the process of foundation pit construction.Based on the complex deformation mechanism of subway foundation pit,BP neural network and its improved algorithm are used to predict the deformation trend of foundation pit and put forward effective preventive measures,which is of great significance to the safe construction of subway foundation pit.Taking Zhengzhou city rail transit line 4 station foundation pit engineering as the background,first established scientific and reasonable deformation monitoring scheme,and then based on the measured data of a variety of deformation monitoring method modeling analysis of the deformation of the soil physical properties and construction factors,and from retaining structure,the bottom soil and surrounding surface three deformation data analysis of the mechanism of deformation of foundation pit.In the research of foundation pit deformation prediction,the BP neural network model based on time series,the PSO-BP model based on the influencing factors of foundation pit deformation and the improved PSO-BP neural network model are established to predict the foundation pit deformation trend.Finally,the validity of the model is verified by comparing the prediction results of several models with the measured data.The results show that in the BP neural network model based on the monitoring time series of foundation pit deformation,the prediction effect of the monitoring data series of steady change and the obvious monitoring data series of trend item is better than that of the sequence containing the mutation data.In the BP neural network model established based on the influencing factors of foundation pit deformation,the monitoring data of support axial force were predicted.Among the60 groups of prediction results,the relative error was more than 1%,which accounted for 18 groups.In addition,the convergence rate of this model was slow and it was easy to fall into the local extreme value.Based on strut axial forces,deep horizontal displacement and retaining pile top settlement data of three kinds of monitoring data and predicted results contrast analysis,the improved PSO and BP neural network algorithm(the introduction of random weighting adaptive factor)fastest,the best prediction,PSO and BP findings and the convergence speed and effectiveness of convergence speed of BP neural network model and the prediction effect of the worst;Finally,based on the deformation prediction trend of the foundation pit,the scientific and reasonable measures to prevent the deformation of the foundation pit are put forward to ensure the construction safety of the foundation pit engineering.