Research On Selection And Optimization Of Regression Model In Tunnel Construction Monitoring Measurement Data Processing

One of the most important purposes of tunnel construction monitoringmeasurement is to ensure the safety of the tunnel construction, the stability ofsurrounding rock and the lining support structure after the forecast and the feedback bythe necessary computing and judgment with monitoring data processing analysis. Toachieve this purpose, the accurate feedback information is required. The effectivemonitoring data processing method is directly related to the accuracy and reliability ofthe feedback information. For its simple, understandable and easy operation,regression analysis becomes one of the most popular data processing methods fortunnel monitoring. However, the selection of regression model in regression analysismethod tends to be subjective, and the method to analyze the mutations of monitoringdata is still rare. Aimed at the aforementioned problems, based on the foregoneresearches, the "research on selection and optimization of the regression model intunnel construction monitoring measurement data processing" was proposed in thispaper, and the systematic deep-going study based upon a vast amount of monitoringdata of tunnel construction was carried out.At first, the method of monitoring data analyses for tunnel construction wassummarized, and the regression analysis, time series analysis, gray forecast methodand catastrophe theory were studied. Furthermore, the advantages and shortcomings ofregression analysis used for tunnel construction monitoring measurement dataprocessing were pointed out, and the solution of regression model selection andoptimization in the traditional regression analysis was creatively put forward bycomprehensively using the merits of kinds of analyses. By studying the tunnelmonitoring data, the regression models for different monitoring objects were obtainedand the changing law of each monitoring objects with time was discussed. To narrowthe selection range of the regression models for tunnel monitoring data, the tunnelmonitoring regression model set was established. Moreover, since the quantity ofevaluation indexes of this regression model was rare and the concept of this indexeswas unclear, the evaluation index set of the regression models for tunnel monitoringwas established.Using the gray situation decision, the evaluation standard was quantized and theregression model was selected. Considering the simplicity and applicability of the regression model, the defect of regression models selection in traditional regressionanalysis method was avoided. Taking the monitoring data of vault crown settlement asexample, the feasibility and advantages of this method was proved.According to the regression theory and exponential smoothing weighted theory,the Regression Trend Adding Method(RTAM) was proposed. Using the gray situationdecision to select regression model, the regression model was optimized. Moreover, bytaking the monitoring data of tunnel surface settlement as example, the monitoring datawas analyzed. This method cannot only determine the frequency of tunnel monitoringmeasurement, but can also predict the occurrence of the mutations. Comparing withother methods, Regression Trend Adding Method was simpler and the feedbackinformation was more accurate. Therefore, the advantage of this method which can beused to predict the monitoring data of tunnel is obvious.Based on the subject frontier, using the advanced mathematical method, the tunnelmonitoring data processing method was studied. The results have high theoretical andapplication value which can provide theoretical basis for dynamic feedback analysis oftunnel construction.