Application Of Data Mining Technology In Predicting Surface Settlement Of Tunnels Constructed By Soil Pressure Balance Shield Method

It is difficult for e arth pressure balance shield construction method to avoid the disturbance and surface settlement,resulting in stratum deformation and surface settlement,and impact on surrounding buildings and other infrastructure,so the settlement must be accurately predicted and strictly controlled.The existing prediction theories and models of surface subsidenc e are difficult to overcome the uncertainties of rock and soil materials,the high-dimensional nonlinearity of influencing factors,and the difficult problem of spatial and temporal variability of surface settlement development.In response to the above problems,this thesis uses data mining techniques to predict the maximum surface settlement generated by earth pressure balanced shield tunneling.The main contents and research results of this thesis are as follows:(1)The data pre-processing method is use d to pre-process the data within the influence range of surface subsidence;the feature engineering technology is used to select and extract features from the high-dimensional influence factor data of surface subsidence,so as to fully explore the importan t information contained in the influence factor data within the influence range to predict the surface subsidence while avoiding the dimensional disaster,and then to establish the data model that can best characterize the surface subsidence engineering pr oblem.(2)Logistic function and Boltzmann function were used to analyze the spatial and temporal variability of surface settlement generated by the earth pressure balance shield tunnel construction respectively,and the results showed that the generation of maximum surface settlement is the accumulation process of soil spatial displacement accompanying the shield boring process,which has spatial and temporal variability,and when conducting the prediction study of maximum surface settlement,it is necessa ry to determine the significant influence range of surface settlement and to predict the surface settlement based on the data of the influence factors within the significant influence range.(3)Based on the fitting results of the logistic function,the derivation obtains the settlement rate-time function,which shows that the settlement rate during shield tunneling is symmetrically distributed based on the peak rate,and proposes the identificati on method of the influence range of surface settlement with the settlement rate as the index based on this law;based on the two-dimensional spatial coordinate data of surface settlement on the engineering GIS map,a confidence ellipse calculation method o f the covariance matrix using two-dimensional interval estimation is proposed to obtain the significant influence range of surface settlement caused by earth pressure balance shield tunneling at a specific settlement rate level.(4)The machine learning al gorithm regression prediction model DE-SVR and the classification prediction model GV-XGBoost with adaptive selection of parameters of the optimization algorithm were established respectively to compare and predict the maximum surface settlement generated by earth pressure balance shield tunnel boring based on three models of surface settlement influence factors,and the prediction results all show that:the estimation method of the influence range ellipse of the covariance matrix is used to obtain the influence of surface settlement The prediction accuracy of the prediction model can be significantly improved by using machine learning methods to predict surface settlement using operational parameter data within the influence range;the overall prediction pe rformance of the prediction algorithm is enhanced with the expansion of the influence range of surface settlement within a specific range;the statistical characteristics of the surface settlement influence factor data contain important information reflect ing the prediction of surface settlement,and the full exploitation of the information is conducive to The statistical characteristics of surface settlement factors contain important information reflecting surface settlement prediction.The prediction resu lts of the engineering case in this paper show that the regression model prediction accuracy R~2=0.899and the classification model prediction accuracy 0.814,both of which achieve the highest prediction effect of the model,are predicted by using the data of seven statistical characteristics of the influencing factors in the influence range of surface settlement at the settlement rate of 1mm/d.(5)The PCA algorithm was used to reduce the dimensionality of engineering geological data,and based on the reduced data,the K-Means algorithm was used to cluster the geological categories of the tunnel interval,and the clustering effect was highly consistent with the analysis results of the geological cross-sectional map of the interval,and the clustering me thod was applied to the geological category identification with high applicability.(6)Apriori algorithm was used to mine the correlation between shield operation parameters data and ground settlement data in similar engineering geological intervals.The mining results provide a reference basis for the setting intervals of shield construction parameters under similar engineering geological conditions,and further guide the safe and efficient shield tunnel construction.