Assessmentand Analysis Of Settlement Deformation Based On Nonparametric Estimation Of Local Polynomial

New railway crossing existing lines will inevitably lead to settlement deformation of the related structures.Through the monitoring of settlement deformation and analyzing of monitoring data,the safety status of deformation bodies can be evaluated,the causes of deformation can be identified,and the trend of future deformation can be forecasted.It is important to prevent construction accidents and ensure construction safe.Based on a project example,this dissertation shows the shortage of the existing methodology about settlement deformation assessment analysis and puts forward a new method.Settlement deformation is often a complex process which is non-linear,indefinite and changeable trend.The existing models which analyze complex and changeable settlement deformation have its own limitations or deficiencies.With the popularization of automatic settlement deformation monitoring technology,it has been being a trend to obtain massive settlement monitoring data in real time and at high frequencies.This provides ample data samples for statistical analysis of settlement deformation.The focus of this paper is the using of advanced non-parametric estimation of local polynomial methods to analyze settlement deformation which is non-linear,complex,and changeable trend.The research results and specific contents are as follows:(1)Existing assessment analysis methods assessing and analyzing settlement deformation which is complex and changeable has limitations.The classical parametric regression model is suitable for the analysis of the settlement deformation whose process following a certain functional relationship which usually cannot be accurately obtained.The chi2 test method is used to evaluate the rationality of the stability of the state of settlement and deformation on the premise that the parametric regression method cannot be successful.The project examples shows that: it is a suitable method for the stability analysis of settlement deformation bodies.The premise of the ARMA model assessment and analysis of non-stationary deformation observation sequences is whether can accurately extract trend items,and its scope of use has obvious limitations.When analyzing the process of settlement deformation of nonlinear,uncertain and trend changes,the Kalman filter model and BP neural network model also have obvious defects and deficiencies.(2)To select the kernel function,the degree of local polynomial of the model according to the theory of non-parametric estimation of local polynomial and the characteristics of settlement deformation is what should be done firstly.And then,the model accuracy under different bandwidth h is analyzed.Accordingly,the appropriate bandwidth h should be selected carefully to establish the non-parametric local polynomial estimation model which is more suitable for the settlement deformation evaluation analysis.(3)The regression analysis and prediction accuracy of nonparametric estimation of local polynomial is compared with the BP neural network model and the Kalman filter model.The results show that the non-parametric estimation of local polynomial analyze the settlement deformation with higher accuracy and better robustness,and it can weaken and eliminate the disturbance due to the inconsistency of the front and rear trends.The non-parametric local polynomial estimation method,which is simple and has a wide range of applicability,is much suitable for analyzing nonlinear,complex,and variable trend settlement deformation.