Prediction Of Ground Deformation And Assessment Of Constructions Risks Of Excavation Face During Shield Tunneling In Weathered Stratum

The stability of the excavation face and surface settlement have large influence on the shield construction,but the risk level of the excavation surface is difficult to evaluate,and too many factors affecting the surface settlement results much difficulty to predict it.Based on the actual project,this study uses the finite element model to predict surface settlement and deformation of the excavation face,the neural network algorithm and GBDT algorithm are also used to predict surface settlement,the AHP is used to evaluate the risk level of the excavation surface.The research mainly relies on the shield tunnel project from Shenzhen Airport North Station to the Airport Station of Shenzhen Suiguanshen Intercity Railway.Firstly,the finite element model is established by using Plaxis.The influence law of excavation on the surface settlement and the deformation of the soil in front of the excavation face were studied.Then the AHP is used to analyze the risk level of the excavation face.Finally,the multi-layer BP neural network and the gradient lifting algorithm model are trained by using engineering data,and the settlement of the predicted segment surface is calculated and the importance degree of each input parameter is extracted.The main conclusions of this study are as follows:1)During the construction of the shield,the working conditions of the excavation surface have a great influence on the surface settlement and the deformation of the soil in front of the excavation face.The finite element model was built using Plaxis and calculated separately under different support ratios.When the ratio is reduced from 1 to 0.3,the deformation model of the soil in front of the excavation face will gradually form the deformation of the wedge deformation zone from the overall smooth envelope,which can reasonably reflect the deformation of the soil under the excavation surface load.There is still a big gap between the ground deformation prediction and the measured values using the finite element model,which needs to be further improved.2)The risk level of the excavation surface from the Shenzhen Airport North Station to the Airport Station,which is dominated by granite weathered strata,is mainly three.Through the application of traditional analytic hierarchy process and interval fuzzy analytic hierarchy process combined with fuzzy comprehensive evaluation method,the risk analysis results of the left-line shield tunnel of Shenzhen Airport North Station to Airport Station of Suiguan-Dongcheng Intercity Railway show that the risk level is Grade III and III.The number of shield rings below the grade is more than 98% of the entire line.3)The multi-layer neural network model can be used to predict the surface deformation in shield construction.By establishing a 1/3/5/9-layer neural network structure model,377 sets of data such as tunnel design parameters,shield construction parameters and geological parameters are used to train the multi-layer neural network model,and finally 64 neurons are used in the first hidden layer,4 structural hidden layers,a total of 9 layers of hidden layer neural network as the optimal model,the ground deformation of the range of 82+760?82+910 is predicted.The linear fitting degree between the model prediction result and the measured surface deformation is 0.968.4)The GBDT algorithm can be used to predict the surface deformation in shield tunnel construction.By establishing a GBDT algorithm model consisting of different number of learners and learning rate,377 sets of data such as tunnel design parameters,shield construction parameters and geological parameters of the Sui-Guan-Shen Intercity Railway Shenzhen Airport North Station to the Airport Station are used.The two data of the deformed data were separately trained,and finally the obtained model was used to predict the ground deformation of the range of 82+760?82+910.The prediction results in the two cases were linearly fitted with the measured surface deformation and settlement,and the linear fitting coefficients were0.948 and 0.943,respectively.5)The buried depth,grouting volume and total thrust of the tunnel axis are three factors that have a great influence on the ground surface in the gradient lifting algorithm model.The weight of each input parameter is given by the GBDT algorithm model.The results of training of all 377 sets of data and the data with large deformation are shown.The tunnel depth,grouting volume and total thrust are the three factors that have a great influence on the surface deformation.Factors.Shield posture also has a relatively large impact on ground deformation.