Tools for numerical modelling of tunnelling interactions

Abstract

The ongoing development in the world's urban areas inevitably leads to the construction of structures in close proximity to already driven tunnels. Care should be taken to ensure that construction is carried out without damaging the tunnels or any other adjacent or overlying infrastructure. Considerable research has been undertaken for the case of a single tunnel where empirical methods for predicting tunnel induced deformations are applicable. For more complex geometries, however, empirical methods fail to make accurate predictions since they do not account for the soil- tunnel-structure interaction mechanism. The finite element method (FEM) appears to be a solution to this prediction problem, however many difficulties in its use remain. The aim of this thesis is to investigate and validate tools for numerical modelling of tunnelling related interactions in soft ground. The generation of a suitable mesh is a major overhead in the use of three dimensional (3-D) FE analyses. Preparing and checking a complex tunnelling mesh can be extremely time consuming. Thus, a parametric scheme for automated, efficient and robust 3-D mesh generation was part of this project. FE analyses of a single driven tunnel are made for comparison with empirical methods on the direction of the surface displacement vectors. Another parametric study of twin tunnelling schemes is carried out in both 2-D and 3-D using various FE packages. The objective is to focus on the effects of surface loading on the tunnels themselves in terms of deformations and bending moments and study the changing effect as tunnel layout is altered. The results obtained highly depend on the constitutive relations assumed, the soil properties and the discretization employed. Further to this the predictions show areas of general agreement and disagreement between the different types of analyses and FE packages used, indicating that accurate numerical modelling of this problem remains difficult and requires care.