Unlike for deep tunnels, prediction of displacements represents a key factor when analyzing tunnel excavation with low overburden, not only for defining the influence of the excavation on preexisting structures but also for providing a basis to put the observational method in place. The accurate assessment of the deformations induced by tunneling is strongly dependent on the choice of the appropriate constitutive model and on the parameter calibration against experimental data. The case study presented in this paper deals with the construction works of a new highway in Southern Italy, including five twin tube shallow tunnels. Considering that two-dimensional analyses are usually not able to reproduce simultaneously all the aspects of the stress-strain response due to tunneling, three-dimensional finite element analyses were performed in order to compare the results of the numerical analysis with the monitoring data. The available database includes a wide range of displacement and deformation measurements.
In standard practice, numerical modelling of tunnel construction relies in most cases on two-dimensional analyses, since three-dimensional calculations are still too time consuming with respect to design and construction deadlines. Especially when the observational method has to be applied and back-analyses have to be performed in order to promptly update the original design, it is usually not possible to run very time-consuming calculations and a simple 2D analysis is therefore the only possibility. The most common way to simulate such a 3D process in 2D is the stress reduction method (Panet & Guenot 1983). According to this approach the radial stress acting on the tunnel boundary is reduced in order to simulate the arrival of the excavation face at the analyzed section. The reduction factor of the initial stress is usually referred to as the relaxation factor λ.
If the main goal is to obtain a reliable estimate of surface settlements, a combination of a 2D and a fast 3D analysis (so called all-in-one installation) represents also a relatively fast alternative as suggested by Vermeer et al. (2002). This procedure leads to an estimate of the relaxation factor, which allows to reproduce the 3D subsidence phenomena using a 2D analysis. However, the so calibrated λ-value does not guarantee the match of lining forces and displacements as calculated with a full 3D analysis. Moreover, relaxation factors are strongly dependent on the soil constitutive model adopted as well as on the geometry of the excavated section, the excavation sequence and the overburden (Möller 2006). This means that the λ value obtained from a certain back analysis would not be anymore valid if, for instance, ground conditions remain the same but one of the other conditions changes. For example, the λ value found by back-analyzing (with a plane-strain analysis) the monitoring data of an exploratory adit could not be reliably used for the prediction of the real tunnel displacements (Svoboda & Masin 2011). However, if the stress reduction factor is carefully chosen, the stress reduction method can reproduce some aspects of a full 3D simulation but a small variation of λ value can strongly affect the results (Galli et al. 2004).