Study On Prediction Accuracy By The Response Surface Model With Random Parameters For Train-induced Vibration And Its Environmental Influence

In recent years,many cities in China are vigorously developing subway transportation which brings a series of environmental impact problems,such as the affects to nearby residents,precision instruments and ancient buildings.Therefore the efficient and accurate prediction of the environmental vibration in the subway construction stage has become the key technology link of subway line selection optimization and track design.At present,the prediction of subway traffic environment vibration is still in the stage of deterministic prediction.no systematic research has been carried out for various random factors,and the accuracy of prediction results is also lack of quantitative estimation.Based on the above-mentioned background,the accomplishement of this paper is supported by the National Natural Science Foundation Project "Study of prediction accuracy and reliability of train-induced vibration and its environmental influence"(51778049)and "Study on the parameter uncertainty of metro train-induced environmental vibrations based on hybrid prediction method"(51978043).Combined response surface design method,considering random parameters in "train track tunnel stratum building" system,focusing on the key scientific issues of the accuracy evaluation of the standard prediction method,a systematic study is carried out by combining analytical calculation,numerical simulation and field test.The main work is as follows:1.Based on response surface method,the model considering the soil parameters(shear wave velocity,poisson ratio,damping ratio,density),tunnel depth,horizontal distance from the surface prediction point to the tunnel centerline and train speed as variables is proposed.It realizes the fast prediction of the maximum Z vibration level and the model of divided frequency vibration level on the surface points.Based on the global sensitivity analysis of the random prediction model of the response surface of the maximum Z vibration level,it is found that the sensitivity order of each parameter is: horizontal distance,buried depth,damping ratio,shear wave velocity,train speed,density,poisson ratio.2.Based on the ground vibration response measurement project of a certain section of Beijing Metro,The realization process and results of deterministic prediction and probability prediction of response surface random prediction model are given.In the deterministic prediction,the maximum Z vibration level and divided frequency vibration level of a particular point can be obtained by substituting the actual parameters into the prediction models;in the probabilistic prediction,by assuming that the inputting parameters as independent normal random variables,the probabilitic distribution of the prediction results can be obtained by using the first-order second moment method and the Monte Carlo method.The predicted results are in good agreement with the measured results.3.Based on a large number of field measurement cases accumulated by the research group,the vibration response data of different working conditions are collected and sorted out,forming a data set with more than 160 points.The prediction results of the code and the response surface model at all the measured points are calculated respectively.The definition of the accuracy of the prediction results is,taking the maximum Z vibration level as the index,assuming that the measured value is the true value,using various calculation methods to quantitatively calculate the degree of difference between the predicted measured values,so as to evaluate the accuracy of the prediction results of the chain prediction formula,and gives a conclusion that the response surface random prediction model is more accurate than the chain empirical prediction method.