Study On Surface Subsidence Deformation Analysis And Prediction Model Of Metro Shield Construction

With the development of the city,the subway becomes more and more important,the small construction space of the city and the numerous structures on the surface have put forward higher requirements for the shield construction,so the monitoring and data processing of the subway construction are also becoming more and more important.Based on the practice of ground settlement monitoring in a shield construction section of Hefei Metro Line 1,the scheme of surface uplift monitoring is analyzed,and the measured data of transverse and longitudinal settlement of the ground surface are analyzed.The maximum subsidence value of the surface is-20.33mm,which is fitted by the Peck empirical formula method,and the correlation coefficient is more than 0.9,which is in line with the law of Gauss curve.The longitudinal settlement is divided into different stages of excavation,the distance from the excavated surface 25m,surface disturbance effect,through the shield tunnel face is about 15m,the surface subsidence tends to be stable.According to the engineering geological conditions,the excavation model of the subway shield construction is established by using the finite difference software FLAC3D,and the numerical simulation is carried out.By monitoring the points on the model surface,we get the data of stratum change and surface subsidence with different excavation distances,and do some lateral analysis for the simulation data of several sections.The simulated values are in agreement with the measured values,which proves the feasibility of the FLAC3D numerical simulation in the shield construction.The method of wavelet transform is used to denoise the measured data of the settlement monitoring point H122-6.The measured data are processed by different wavelet basis functions,different thresholds,different decomposition levels and the selection of Scal.Finally,it is concluded that when the soft threshold,DB2 wavelet,rigrsure threshold,level1 and scal=sln are selected,the denoising effect is the best,when the root mean square error is RAMSE=0.2118,and the signal-to-noise ratio SNR=36.0610.In the basis of wavelet denoising,respectively using GM(1,1)model,BP neural network model and GM-BP neural network model to predict the surface subsidence after denoising by time series,three predict models make the prediction results,GM(1,1)model and BP neural network model has a good performance in the example for the forecast,GM(1,1)model performance is not obvious in the random data.Generally speaking,the prediction accuracy of GM-BP neural network combined model is the highest and the prediction effect is very good.The combination model can be used to predict the surface deformation of shield tunneling.