In urban area, constructions of soft ground tunnels are usually important in terms of prediction and control of surface settlement and gradient. Several approaches are readily used for prediction of the ground deformations associated with tunneling. This paper discusses the subsidence prediction using FEM analysis and Artificial Neural Network (ANN) with FEM database. This paper, firstly, investigates the application of FEM simulation using a proposed model to predict ground movement caused by tunneling of a shallow NATM tunnel in unconsolidated soil. The proposed model used here incorporates reduction of shear stiffness, as well as strain softening effects of given material strength parameters. Numerical simulation is performed with material property values, E, v, c, and φ, obtained from laboratory. Some additional parametric studies are performed. FEM results shows as agree well with compare of field data. Secondly, ANN modeling is performed for subsidence prediction. ANN studies database to provide an FEM analysis result. A learned (trained) ANN model has the potential to provide accurate desired output (true output) from input data. However, once the network is trained, its running speed is very high, thereby reducing the total time consumed in the analysis. The trained ANN model is further validated by carrying out parametric studies to assess whether the model gives logical and consistent trends and a case study to verify the application to the actual NATM tunnel in prediction problem. The two methods, FEM and ANN, offer a practical way for predicting final displacement of shallow NATM tunnel, enabling rational safety management scheme to be employed.
抄録全体を表示