Study On Long-term Settlement Prediction Methods Of Suzhou Metro Shield Tunnel

In recent years,many cities in China have started the metro construction work and put into operation successively.With its advantages of convenience,safety and speed,metro has become an important way of urban public transport.However,the metro shield tunnel,especially the metro shield tunnel in soft soil region,will cause long-term settlement during the operation.Excessive settlement will do harm to the normal operation and the safety of metro.And the accurate prediction of the long-term settlement of the metro shield tunnel can provide an important basis for analyzing and solving the security issues of the metro train operation.Based on the settlement monitoring data of Suzhou Metro Line One and Line Two,a variety of prediction models are set up to predict the longitudinal and the transverse long-term settlement of Suzhou metro shield tunnel.The main research contents are as follows:1?Based on the settlement monitoring data of Suzhou Metro Line One and Line Two,this paper analyzes the characteristics of the transverse and the longitudinal long-term settlement of Suzhou metro shield tunnel,and discusses the settlement mechanism of Suzhou metro shield tunnel.2?By using MATLAB,the hyperbolic model,power polynomial model,BP neural network model,support vector machine model,least squares support vector machine model,GM(1,1)grey model and grey relational analysis model are established to predict the transverse long-term settlement of Suzhou metro shield tunnel.Comparing the prediction results with the measured data,LSSVM is the most suitableprediction model for the transverse long-term settlement of Suzhou metro shield tunnel.It's standard deviation is 0.86.The power polynomial model has the lowest accuracy and the standard deviation has reached 3.19.3?The support vector machine model,the least squares support vector machine model and the wavelet neural network model are used in MATLAB to predict thelongitudinal long-term settlement of Suzhou metro shield tunnel in this paper.Comparing the prediction results with the measured data,LSSVM is the most suitable prediction model for the longitudinal long-term settlement of Suzhou metro shield tunnel.It's standard deviation is 1.29.The prediction accuracy of wavelet neural network is the lowest and the standard deviation is 1.53.