Crossing fault zones is a challenging task during tunnel construction and can lead to failure if not properly accounted for during design and construction. These failures can be caused by high water inflow or stability loss of the rock mass surrounding the tunnel (examples can be found in CEDD, 2015). The prediction of the magnitude and development of displacement has a considerable impact on issues during tunnel construction. Changes of rock mass stiffness or structure ahead of the face, to a great extent influences stresses in the vicinity of the tunnel and thus deformations (Steindorfer, 1998). Further research with numerical simulations was carried out, and the qualitative and quantitative evaluation of displacement vector is presented in Grossauer et al, (2005). The concern associated with crossing fault zones was considered during the design of the Bátaapáti deep underground radioactive waste repository complex (NRWR) in Hungary. In the case of the NRWR project, the fault zones have an impact on maintaining the stability of the surrounding rock mass as well as the final sealing of the repository. As part of the final closure design, a complex numerical modelling program was completed to gain a better understanding of the mechanical behaviour of the rock mass and the monitoring observations during construction. Even though the need for the presented study is closely related to the NRWR project, the results can be used in any project related to fault zone crossing. The methodology used to better understand the impact of crossing fault zones for the NRWR project is presented in this paper. First the project background is introduced including an overview of the underground facilities, the aim of the repository, and the construction experience gained from the first crossings of the fault zones. After that the in-situ and laboratory testing is summarized. The numerical modelling methodology is detailed in the next section. Not only the outcome of the model is presented, but the results are compared with the experience and observations made on site. Since a good agreement is found between the actual and the modelled behaviour, the conclusions are based on the validated result. To fulfil the aim of the paper, the stress reflection concept is introduced to explain the mechanical behaviour of the fault zone and the rock support; furthermore straightforward suggestions are made for rock support design.
PROJECT BACKGROUND
Overview
The repository facilitates the disposal of low and medium activity radioactive waste from the Paks Nuclear Power plant (NPP) and it is located close to the village of Bátaapáti. Geologically, the Bátaapáti Site is part of the Mórágy Block, which is composed of granitic rocks of the Palaeozoic Mórágy Granite Formation. The rock mass consists of porphyritic monzogranite along with darker and more fine-grained, equigranular monzonitic and lighter aplitic rocks. Within the prevailing monzogranitic rocks, monzonitic rocks form inclusions and bodies of size up to 1-2km and are always enclosed (Balla, 2004).
The repository consists of several different facilities such as tunnels, shafts, caverns, access roads and portals. The access tunnels have an inclination of 10 per cent and arrive at the reference base level of ±0.00m aBSl.
The analysed tunnel where the full scale test is to be carried out is a side drift driven from one of the access tunnels. Steel fibre reinforced sprayed concrete lining with lattice girders were applied in the side drift as permanent support, supplemented with systematic rock bolting (3m in length) according to the defined rock class (support for rock class IV is shown in Figure 1). Full-face excavation was used and the tunnel diameter is approximately 7m.
Radioactive waste emplacement and final closure
The primary purpose of the facility is the final disposal of the low and intermediate level radioactive waste. Low-level waste (LLW) is generated from the nuclear fuel cycle. It contains small amounts of mostly short-lived radioactivity. Intermediate-level waste (ILW) contains higher amounts of radioactivity and may require shielding (WNA, 2014).
The rock mass is fractured and there are aquifer zones across the repository. According to the results of an installed hydrogeological monitoring system, however, some younger, regional clayey fault zones proved to be the most important confining elements of geological barrier. Since the fault zones are penetrated by the access tunnels, they need to be sealed when the facility is closed. Assuming that the tunnels are sealed, the travel time of a particle from the repository to the surface is 10,000- 100,000 years (assuming transfer by groundwater). If the critical fault zones are not sealed, the travel time is only 1-100 years (Mezo, 2010). The rock mass around the emplacement chambers has only minor significance as a geological barrier (Kovács et al., 2010).
The emplacement of radioactive waste has been started at the end of 2012 in the first disposal chamber. The radioactive waste will be stored in steel barrels and placed in concrete canisters measuring 2.25m x 2.25m x 1.55m. The voids in the canister are filled with concrete and clay or bentonite. The neck of the emplacement chamber will be sealed with a plug and the tunnels will be backfilled with crushed rock. Several plugs will be constructed along the tunnels as well as at the fault zones in order to seal them. Further information on possible plug options can be found in Megyeri et al. (2014).
Excavation through fault zones
The hydrogeological and geotechnical properties of the Bátaapáti site are strongly influenced by large-scale tectonic zones with complex inner structure and extensive development of fault gouge and breccia. The gouge – due to its clay mineral content – mainly responds to mechanical stresses with plastic deformations. During the excavation of the inclined access tunnels, two major fault zones (Péter and Klára fault zone) were penetrated. In these sections, 200-250mm steel fibre reinforced concrete and lattice girders were installed. The tunnel needed to be rescaled due to excessive displacements of the gouge clay (Szebenyi et al. 2009).
After the access tunnels were completed the shotcrete liner was cracked. Cracks were associated with the gouge plane of the fault zones (see left and middle pictures of Figure 6). Considering that such local failure of the shotcrete support was not experienced anywhere along the access tunnels, it was concluded that the cracks were caused by deformations characteristic to this location (Szebenyi et al. 2009). These deformations occur along the Péter fault zone, are perpendicular to the maximal principle stresses and relate to the tectonic characteristics and gouge material of the fault zone.
Despite an attempt to strengthen the shotcrete liner in this area, the cracks developed into an annular discontinuity in the liner a few months after they first appeared. Strains in the liner and crack opening displacements were continuously monitored, revealing that displacements continued to occur even after a year.
Monitoring chambers
In order to gain better understanding of the rock mass, monitoring chambers were built. One of these chambers, called the HGM chamber is traversed by a nearly vertical fault zone filled with clays and breccia. This fault zone greatly influenced the longitudinal arching effect of the tunnel excavation.
In this particular section, CSIRO-cell stress measurement, extensometer, and convergence measurement were carried out. Deformations were measured on the near side of the fault zone. When tectonic features such as minor faults (with thickness significantly lower than the tunnel diameter) dominate the behaviour of the rock mass, the so-called stress reflection phenomenon occurs. This means that the fault zone behaves similarly to a mirror, with the stress redistribution pattern “reflected” on the fault zone and the same pattern “shadowed” on the far side of the fault. Stress reflection is caused by the low shear strength of the fault zone and allows displacement along the fault (see Section Results) and the stiffness difference between the fault gouge and the rock mass (Steindorfer, 1998). One of the consequences of this phenomenon is that the effects of the far side excavation are mostly eliminated by the fault zone.
Consequently the tunnel face cannot support the excavation, resulting in extremely high deformation in front of the tunnel face, and low deformations on the near side after the tunnel face advances (Kovács et al. 2009b).
TESTING
During the excavation of the underground facilities, face mapping was performed, and the Q (Barton, 2002), RMR (Bieniawski, 1989) and GSI (Hoek, 1994) values were determined.
Systematic rock sampling and laboratory tests were carried out during construction to determine the properties of intact rock. The systematic sampling was in accordance with the relevant ISRM recommendations and the MSZ-EN codes. Traditional triaxial tests with various confinement stresses (0 -15MPa) as well as Multiple Failure State Tests were carried out to determine the Hoek-Brown parameters (Vásárhelyi et al., 2013). The shear strength of the joints had been measured by laboratory tests (Buocz et al., 2010) and had been verified by distinct element models (Horváth, 2011 and Borbély, 2013).
Two-dimensional Doorstopper-cell and 3D CSIRO HI-cell measurements were carried out to determine the in-situ stresses. These are widely used overcoring techniques. According to field measurements K0 varies between 1.34 and 1.5. CSIRO HI cells are also used for measuring the stress changes of rock wall caused by the tunnelling.
In the underground facilities at the NRWR convergence measurement arrays were installed in 19 sections so far, two of them were measured in test drift. In these sections relative displacement of the rock mass surrounding the excavation has been measured repeatedly in six radial directions that enclose an angle of 30° with each other. Thirteen borehole extensometer sections and two Modular Reverse Head extensometers are also installed for measuring the radial and axial displacements of rocks around the tunnels including the entire Longitudinal Displacements Profile (LDP). The capacities of 10 per cent of the rock bolts were tested at the NRWR Project. In case of some tests the displacements were measured, thus the grouting properties were determined. Automatic measuring sections are installed for controlling the loads acting on the supporting elements.
NUMERICAL MODELLING
To gain better understanding of the rock mass behaviour around the test drift driven through Peter-fault zone, 3D finite difference modelling was carried out using Itasca FLAC3D (model geometry shown in Figure 2). Using 3D instead of 2D modelling allowed for the following:
¦ Major horizontal in situ stresses are not parallel or
¦ Tunnel advance was modelled step by step; and
¦ Fault zone was not perpendicular to the tunnel.
As Figure 3 shows, the fault gouge is directly considered in the model (shown in red). The fault zone is accompanied by a highly fractured rock mass, found between the fault zone and the fairly good quality rock mass.
Each excavation sequence was modelled separately using the following steps: Excavate rock, install liner and rock bolts, increase concrete age by one cycle time – set time-dependent concrete properties.
The Hoek-Brown failure criterion and Geological Strength Index (GSI) system was chosen for describing the rock mass as it provides a practical means to estimate rock mass strength from a combination of laboratory test values and field observations (Hoek, 1994). Intact rock properties were determined separately for monzogranite and monzonite. The GSI value of the rock mass is determined during tunnel excavation through continuous face mapping. The encountered rock mass is categorized into one of the five pre-defined rock classes (rock class I to V). The rock mass properties along the tunnel axis were represented using rock classes III and V, based on exploratory drilling data (blue and green zones in Figure 3, respectively).
The steel fibre reinforced sprayed concrete lining is modelled with the built in shell elements of FLAC3D. In the case of sprayed concrete lined tunnels, the lining is subjected to loads long before it reaches its final strength and stiffness (in fact the concrete lining is usually loaded a few hours after the installation). The early age properties of the concrete were determined according to Chang and Stille (1993). The support provided by the rock bolts is taken into account with global reinforcement element (cable element). The rock bolt properties applied in the model are summarized in Table 2.
RESULTS
Displacements are reliable indicators of the model behaviour; therefore they can be used to verify the model by comparing to actual measurements. Displacements were extracted from the model in two sections. The first is sufficiently far from the fault zone, where the fault does not influence displacements. In this section the displacement was compared to the site specific RMR-displacement function, which is based on seven convergence monitoring sections in rock masses with a RMR value from 45-65. The modelled displacements were 5.9-9.7mm, while the actual measured displacements were 5.5-9.9mm (in case of sections with similar RMR value). As such, the displacement in the model shows very good agreement with the monitoring result. The second control section was in the middle of the fault zone. The modelled displacements in this area were 40cm (Figure 3). No measurements were carried out in the fault zone due to the difficulties of the crossing; however the construction logs indicate that the tunnel section needed to be re-scaled a few times before the sprayed concrete was installed due to the high displacements.
The high (>10cm) displacements were limited to the approximately 0.5m thick fault zone. Based on this observation, the modelled displacement is considered to be reasonable. According to the model and field observation, the clay gouge is in plastic state. This indicates that the stress redistribution is governed by both the low strength of fault gouge and the stiffness difference between the rock mass and the fault gouge. The longitudinal displacement profile (LDP) was also extracted from the model (Figure 4). A LDP relates tunnel wall deformations at successive stages to the physical location along the tunnel axis (i.e. the distance from the tunnel face). The observed LDP in the model is uncommon to rock tunnels. Near the fault zone, almost all displacements occur close to the tunnel face. Further away from the fault zone, the LDP in the model is close to the theoretical profile corresponding to circular tunnels in homogenous rock mass (Vlachopoulos and Diederichs, 2009). At Section 1, the effect of the fault is less significant, so the difference between the theoretical and modelled profile is minimal.
The LDP was not measured near the fault zone, however relaxation was measured in the HGM chamber where a minor fault zone was penetrated. In Kon-6 section (installed 1-2m before the fault zone) the measured convergence is significantly less than what is calculated by using the RMR value for the rock mass (Figure 5). This is potentially caused by high relaxation near the fault zone.
According to the model, high stresses and internal forces occurred in the liner. Both bending moments and shear forces were above the capacity of the lining. This result is in good agreement with the construction experience as it was observed that the liner cracked. The peak stresses in the lining coincide with the crack location and orientation, indicating that the internal forces in the tunnel lining are modelled with a good agreement with the reality (Figure 6).
The general model behaviour is considered to be in close agreement with the actual rock behaviour and can be used to facilitate understanding the observed behaviour. To demonstrate the stress reflection effect, the stress redistribution during tunnel drive was considered (Figure 7). One can see that the stress does not change on the far side of the fault zone before the fault zone is penetrated. In contrast, high stress concentration can be observed between the fault zone and the tunnel face. This finding is in agreement with the results presented in Grossauer et all (2005): tunnelling through a rock mass with frequently changing stiffness at short distances, stresses will concentrate in the “stiff ” sections. The stress redistribution on the far side of the fault zone takes place only after the fault zone is penetrated.
SUMMARY AND CONCLUSION
Detailed analysis of a fault zone was presented here in order to gain better understanding of the effects associated with tunnelling through thin fault zones. One of the most important findings of the paper is the confirmation of the stress reflection concept. It has been shown that the rock mass on the near side of the thin fault zone is fully or almost fully relaxed (all displacements occur) before the tunnel lining is installed; while the rock mass on the far side of the fault zone is not relaxed at all (has in-situ stress conditions and negligible displacements).
Based on the results, it is recommended that the rock support is designed to be able to handle stress and strains from the full overburden pressure on the far side, or at least be flexible enough to allow relaxation of stress after the rock support installation (Figure 3). Typically fast tunnel advance can be helpful, due to the early age creep of the concrete.
Most of the stress redistribution occurs before or at the same time as the rock support is installed on the near side of the fault crossing. Immediate support is recommended to provide support and avoid overbreak. Lattice girders were a good and reasonably simple solution. They provided immediate support to some extent while allowing movements on the other side of the fault.
It is recommended not to rely on the longitudinal load-bearing capacity of the tunnel lining once the fault zone is penetrated. The displacement after lining installation is significantly different on the two sides of the fault zone, therefore significant shearing and bending is expected. As it was pointed out earlier, a crack formed due to this effect. The crack did not pose a risk in the presented example as the tunnel is drained and the liner does not have any water insulation function, however, a developing crack might cause difficulties in case of a gasketed segmental tunnel lining design. The stress reflection effect can be modelled using 2D modelling for design purposes. Two separate models are needed to consider the near side and the far side of the fault zone. On the near side it is a reasonable estimation to use 90 per cent -100 per cent relaxation, i.e., the majority of the rock mass movements occur prior to tunnel installation. To model the far side of the fault 0 per cent-10 per cent can be used. Hence the suggested support method allows some movement on the far side of the fault zone, and provides immediate support on the near side. The adequacy of the designed tunnel support can be verified with simple 2D numerical modelling, while the three-dimensional effect of a fault zone is considered. If high overbreak, ravelling or squeezing of the rock mass is expected (which was not the case in NRWR project) spiling and forepoling can be used to prevent overbreak on the near side of the fault zone. Spiling or forepoling can provide support, and bridge the fault zone to reduce the stress reflection effect (as it was stated earlier the longitudinal elements are subjected to high loads once the fault zone is penetrated). This study was carried out during the detailed design of the side drift excavation where the full scale test is to take place.
Since then, the side drift has been successfully excavated. It is of interest to note that the fault zone has been found to be significantly different from the prognosis on which this study is based. As described above, the fault gouge mainly responds to mechanical stresses with plastic deformations at the location where the two inclined access tunnels penetrated it. However, at the location of the future full scale test (the side drift) the fault zone is significantly thicker and exhibits blocky behaviour as it is mainly composed of the fault breccia, and clay is only to be found in its discontinuities. Further studies will be carried out to back-analyse the side drift excavation in a similar manner as this paper and to summarise the experience of crossing the thicker fault zone
