Analysis And Prediction Of Surface Deformation Of Highway Tunnel

With the development of the times and the increase of China’s highway construction,the amount of tunnel engineering has increased.With the construction of such a large-scale tunnel project,accidents during tunnel construction have also increased,which is caused by complex geological conditions during tunnel excavation.Therefore,it is very necessary to carry out safety monitoring of the tunnel and an effective method to avoid dangerous accidents in advance,which ensures the safe construction and operation of the tunnel.According to the actual situation,the project overview,monitoring methods and research objects of the Yangjiabian Tunnel are monitored and analyzed.In order to monitor the surface subsidence,the DSZ3 precision level and total station are used to measure the elevation of the measuring point,and the change of the elevation is calculated.Through the research and analysis of the surface deformation of highway tunnels,the prediction model is used to analyze and predict the surface deformation.The specific work is as follows:（1）The data needs to be processed before the time series data analysis and prediction,and the outliers and gross errors in the surface deformation data are removed.Wavelet analysis and singular spectrum analysis can be used as two methods of data processing.Both methods can effectively obtain the deformation trend and period,and remove the outliers and noise.Therefore,both methods can be applied to the processing of tunnel monitoring data and compared with the measured data.Several types of time series models,the selection and testing of parameters,and the selection of the most suitable type of engineering project by judgment.After understanding the principle and training process of RBF and BP neural network,the matlab software is used to predict the tunnel data,and the prediction accuracy of the two neural networks is compared to determine the most suitable model for tunnel data.Compared with BP neural network,RBF neural network is more suitable for predicting settlement,but these two models have their own characteristics when predicting settlement.Introduce data processing and processing methods and experimental results to highlight your workload and contribution.（2）Combine the above models to obtain the prediction accuracy of wavelet ARMA combination model,wavelet RBF combination model,SSA-ARMA model and SSA-RBF combination model,and compare with ARMA-RBF to find time series model and RBF.The neural network combined model has the highest prediction accuracy,and exerts the characteristics of the two models.It can be applied to the prediction of surface settlement of the highway tunnel.FLAC^3D finite difference software can be used in the tunnel excavation support process to analyze the settlement process of the tunnel surface.After analysis by FLAC^3D software,the maximum surface subsidence is at the centerline of the tunnel,which is-40mm,and the sinking trend is normally distributed.The closer to the centerline,the larger the sinking amount.Finally,the actual data is collected and compared with the numerical simulation values to verify the feasibility of the numerical model.（3）Using the combined model to analyze the surface deformation data of highway tunnels,the results show that the prediction accuracy is in the centimeter level,which is in line with the expected forecasting accuracy,and satisfies the actual requirements of engineering surveying.At the same time,the singular spectrum analysis method in the field of digital signal is applied to the actual engineering measurement for the first time.From the actual results,the singular spectrum analysis method has good applicability,and provides for foundation pit settlement,ground deformation and urban area plate research.A good research idea.