Frozen soil is a powerful tool for engineering purposes due to its increased strength, stiffness and decreased permeability. Water between the soil particles bonds them together, making it possible to use frozen soil bodies as impermeable barriers and load-carrying structures. Furthermore, during a freeze and thaw cycle different processes cause deformations in the frozen and unfrozen soil. For example; frost heave, consolidation of the unfrozen zone, creep and thaw settlements. These phenomena are often called frost actions. Artificial ground freezing (AGF) is regularly used during the construction of cross passages between bored tunnels. The frost actions are expected to increase the loads acting on the lining of the main bored tunnels. This thesis investigates if a quantitative measure of loads due to frost actions on the main tunnels lining can be given with a numerical model, supporting the physical understanding of frozen soil. The objective is to determine if loads due to AGF may be a governing load case on segments of the bored tunnel lining.

Load situations that influence the interaction between frozen soil and the tunnel lining have been identified for the construction of cross passages using AGF. These load situations are based on the principles of ground freezing, construction stages in cross passage construction with AGF, the behaviour of frozen soils and case studies. The following five load situations are identified: frost heave, enclosure of water in the frozen heart, excavation, construction of the lining and thaw weakening.

The load situations have been investigated for one of the cross passages of the Westerschelde tunnel. The studied cross passage was constructed with AGF at a depth of -28,5 m in boom clay. The monitoring program for the studied cross passage of the Westerschelde tunnel was very extensive. Different types of monitors have been used to measure the soil stresses, deformations, water pressures and temperatures in the soil near the cross passage during construction. Before construction, several frozen and unfrozen soil test were carried out on the boom clay. The constitutive model used in the numerical calculation is the frozen and unfrozen soil model of Plaxis. The model requires seventeen model parameters. Furthermore six thermal parameters and three parameters for the soil freezing characteristic curve are necessary. Not all these parameters could be determined directly from the laboratory test, therefore correlations and default values were used as well. The determined parameter set is optimized and validated with help of available laboratory tests. Simulating these simple soil tests gave the opportunity to explore the capabilities of the model. In later stages the optimisation and validation of the parameters turned out to be crucial to obtain a plausible soil response in the large scale models of the cross passage.

The frozen and unfrozen soil model is only available in a two-dimensional version. Therefore, two numerical models have been made representing the construction of the cross passage: one axisymmetric model and one plain strain model. The model results have been compared to the measured data and to each other. The frozen and unfrozen model is able to describe important features of frozen soil behaviour. For more complex engineering challenges, like cross passages, some assumptions in the model are made that influence the capability of the model to simulate certain load situations. The fact that the deformations are independent of the temperature gradient has a large influence on the lining displacements, but also on pore water pressures inside the frozen cylinder. Beforehand it was already known that the constitutive model is rate independent and thus not capable to take creep into account.

Four of the load situations could be qualitatively analysed with the two numerical models .The enclosure of water in the heart of the frozen cylinder could not be simulated with the numerical models. On the other hand, soil stresses due to frost heave and excavation gave a good quantitative measure. In this research one case is extensively investigation, therefore this research is non-statistical. Henceforward, the conclusion cannot be drawn that this quantitative measure of frost heave stresses can also be obtained for other cases. A qualitative measure of loads due to frost heave in construction with AGF can certainly be given with these numerical models. Although not all loads due to AGF could be taken into account (i.e. creep, enclosure of water in the frozen heart), one of the most important load situations (i.e. frost heave) could be quantitatively defined for the boom clay. This load situation is worth investigation in AGF projects, since stresses can become 2.5 times higher than initially measured soil stresses. At the start of the project the boom clay was given a frost-susceptibility index of negligible to low. Even with this mild index the stresses due to frost action increased significantly. This factor and index are probably not the same for other soil types. However, this study shows that such large stress increases are a real possibility during cross passage construction with AGF.