Research On Pretreatment And Prediction Methods Of Subgrade Settlement Monitoring Data Of Yinxi High-speed Railway

The high-speed operation of high-speed railway trains requires a track system with high stability and high smoothness.The subgrade is the basis of the track structure,and its settlement deformation needs to be controlled within a reasonable range,therefore,accurate prediction of subgrade settlement is an important task.In this paper,the subgrade of a certain section of Yinxi high-speed railway is taken as the research object,the pre-treatment analysis of the subgrade settlement monitoring data is carried out,the accuracy of the subgrade settlement prediction method is studied,the main research contents and conclusions are as follows:(1)It mainly introduces the basic engineering overview and geological overview of the subgrade of the study section,the specifications and requirements for the settlement monitoring,the layout of the observation targets and the working base points,and the implementation of observation and the accuracy and frequency of observations.Finally,the overall settlement of the section subgrade is analyzed.The results show that the total settlement value of each observation point of the subgrade settlement board and the shoulder observation pile meets the requirements of the specification,and the representative observation signs of two typical sections are selected for monomer analysis,and the analysis results show that the accumulated settlement amount is within the allowable range,and no abnormal settlement occurs.(2)The main methods of error checking for the observation data are mainly studied.The gross error test of the observed data is analyzed.Firstly,the statistical test method,the correlation analysis method and the 3σ criteria are introduced.Then,based on the 3σ criterion method,the gross error test is performed on the measured settlement data of the two observation targets.The test results show that there is no gross error.The systematic error test of the measured data is studied.The U-test method,the t-test method and the mean square error test method are analyzed.Then,the U-test method is used to test the systematic error of the two measured data of an observation target.The results show that there is no systematic error;the random error test of the measured data is studied,the F-test method and the x~2-test method are analyzed,and then the F-test is used to test the random error of the measured data of an observation target.The test results show that it accords with the propagation law of accidental error.Finally,the interpolation methods for non-holonomic monitoring data are studied.The Lagrange interpolation method,the curve fitting method based on algebraic polynomial and triangular periodic function,and the multi-face function fitting method are summarized.Lagrange interpolation method is used to carry out interpolation experiments on a certain monitoring,and the result example shows that this interpolation method has higher precision.(3)The basic principles of hyperbolic method,exponential curve method,Asaoka method,three-point method and GM(1,1)model are mainly studied,and two observation indexes 0587363G1 and 0785552G2 are taken as engineering examples,subgrade settlement prediction results and accuracy of these methods are respectively analyzed in detail.Finally,the same accuracy evaluation criteria are used to compare and analyze the prediction accuracy of the five subgrade settlement prediction methods.The results show that the prediction curves of the hyperbolic method,the exponential curve method,the Asaoka method and the three-point method have the same trend as measured curve.The correlation coefficients are both 0.98～0.99,which meets the requirements of the specification and the prediction accuracy is high.The prediction curve of the GM(1,1)model is inconsistent with the measured curve,the prediction accuracy is poor,and the correlation coefficient is lower than the minimum value of 0.92 of the specification,which does not meet the accuracy and specification requirements,thus this model is not applicable to the subgrade settlement prediction of the observations 0587736G1 and 0785552G2.