Prediction Of Segment Floating In Shield Tunnel Based On Support Vector Machine

In recent years,the rapid development of urban rail transit construction has led to the continuous increase of river sea crossing tunnel projects.National key large-scale projects,such as South-to-North Water Transfer and west-to-east gas transmission,involve the problem of river crossing and sea crossing.Shield construction is usually used in tunnel engineering.Shield construction has become one of the important construction methods suitable for modern tunnel engineering and underground engineering construction.Segment floating often occurs in shield construction,resulting in dislocation,crack,leakage,axis offset and so on.In recent years,domestic and foreign experts and scholars have predicted the segment floating problem in the process of shield tunnel construction through theoretical analysis,experimental research and numerical simulation.However,no matter which research method,the process is more complex.Compared with these methods,machine learning,as one of the statistical and data mining methods,with its convenience and efficiency,can quickly predict the floating amount of segments in the actual project,greatly improve the work efficiency and ensure the smooth progress of the project.Therefore,based on the Yuji intercity railway project,this paper selects the support vector machine method in machine learning to predict the segment floating.The main work contents and conclusions are as follows:(1)Combined with the on-site construction tunneling parameters of Yuji intercity railway,the influence of five tunneling parameters on the segment floating of shield tunnel is analyzed.The analysis results show that the five tunneling parameters have a great influence on the segment floating,which must be considered when predicting the segment floating.Pearson correlation coefficient is used to analyze the linear correlation between the five tunneling parameters and the segment floating volume.The results show that the linear correlation between the five tunneling parameters and the segment floating volume is relatively low.The segment floating volume cannot be judged only from a single tunneling parameter,which should be considered at the same time when predicting the segment floating volume.(2)In the actual shield tunneling process,the excavation bin pressure,the amount of mud entering and discharging are also the tunneling parameters of the shield machine.However,the influence of these parameters on the segment floating is not clear.In view of this situation,based on the on-site construction data of Yuji intercity railway,four different prediction models are established by using support vector machine algorithm and taking the average absolute error(MAE),mean square error(MSE)and correlation coefficient(R~2)as the model evaluation criteria.Different combinations of tunneling parameters are considered in the model.The results show that The buried depth and tunneling angle as the model input value have the best effect on the prediction of segment floating.At this time,the average absolute value error(MAE)of the model prediction result is 2.63,the root mean square error(MSE)is 5.40,and the correlation coefficient(R~2)is 0.81.(3)The BP neural network model is used for comparative prediction analysis,and the same input and output values as the support vector machine model are selected.The average absolute error(MAE)of the prediction results of the BP neural network model is 5.29,the mean square error(MSE)is 9.02,and the correlation coefficient(R~2)is 0.69.Compared with the prediction results of the support vector machine model,the support vector machine model has small error and high prediction accuracy,which is more suitable for the prediction of segment floating volume.The results of this paper can effectively predict the floating amount of segments in the process of shield tunneling,which is conducive to the safe and smooth development of shield construction,and provide a certain theoretical basis for the application of shield machine in tunnel construction.