Weakness zones or faults containing swelling clay represent a challenging situation in hard rock tunnelling. When excavating in such zones, failures have occurred occasionally even though particular precautions have been taken. Instability has been encountered during tunnel excavation and sometimes also long after tunnel completion. One of the most recent cases of failure, the rock fall at the Hanekleiv road tunnel in southeast Norway, occurred ten years after tunnel completion. The rock fall was in a fault zone where swelling clay had been identified, about 1.1 km from the northern tunnel entrance of the southbound tube. Cracks had been detected in the applied shotcrete at the fault zone during tunnel excavation. In one major weakness zone containing swelling clay at the Finnfast subsea tunnel, monitoring was carried out by the Norwegian Geotechnical Institute (NGI). Strain gauges and load cells were installed in the reinforced shotcrete rib near the crown and at springlines of the tunnel.
The rock mass quality was estimated for the weakness zones of the two cases based on the Q classification system. However, it is hardly possible to characterize the complex conditions of weakness zones containing swelling clay with an empirical classification system. Except the important feature of the size (thickness) of zones, effects of swelling clay in weakness zones can not be fully accounted for in the system.
Gouge materials were collected from the zones and mineral composition identified with X-ray diffraction analysis. Laboratory testing based on measuring the swelling pressure of remoulded specimens, and free swelling test were used to quantify the swelling potential.
Numerical modelling is a powerful tool in rock engineering planning and design, particularly when difficulties and uncertainties are expected in the underground excavation. A three-dimensional program is normally preferred for the complicated engineering mechanical computation of tunnelling through weakness zones/faults. Inappropriate use of two-dimensional modelling may induce a large deviation in simulation results. The extent of deviation is illustrated with numerical simulation and comparison based on the two-dimensional finite element program Phase2 and the threedimensional finite difference program FLAC3D.
The instrumented weakness zone at the Finnfast subsea tunnel provides the opportunity to quantitatively evaluate the loading effects on reinforced shotcrete ribs. FLAC3D was used for this modelling, where the pore pressure distribution around the tunnel periphery after each blasting round was determined by using the ground water flow analysis. The full rock support (spiling bolts, shotcrete, radial bolts, face bolts, reinforced shotcrete ribs) applied during tunnel excavation was included in the model. The weakness zone, side rock and the rock support of sprayed concrete were simulated with the Mohr-Coulomb model. The spiling bolts, radial bots, face bolts and steel bars in the reinforced ribs of shotcrete were simulated with various built-in structural elements according to the loading characteristics. Different swelling pressures were applied on the rock support during simulation, ranging from zero to 0.20 MPa at an interval of 0.04 MPa. Though swelling pressure considerably increases the loading on rock support, all instrumentation data and simulation results show that the loading on the sprayed concrete is far less than its compressive strength. The loading on rock support was found to be much higher close to the excavation face, and this area is more critical in terms of tunnel stability.
It was concluded that the rock fall at the Hanekleiv road tunnel was caused by a combination of swelling and gravitational collapse due to the very low internal friction. The swelling process most likely was caused by both the water from joints and accumulation of moisture behind the water/frost shielding, when ventilation has had little effect after tunnel completion. The strength of the rock mass was also reduced with absorption of water during the swelling process. Swelling and strength reduction gradually developed till the collapse suddenly took place. In the numerical simulation with FLAC3D, the fault zone and side rock were simulated based on the Mohr-Coulomb model. The strain softening model was used for shotcrete in order to simulate its post failure behaviour in consideration of the cracks detected during tunnel excavation. Three stages of mechanical states have been focused in the analysis. The first stage represented tunnel excavation and detection of cracks in the shotcrete. The swelling pressure on the rock support of shotcrete was considered at the second stage. Combined effects of strength reduction of the fault zone and swelling were simulated at the last stage. Simulation results verify both the detected cracks during tunnel excavation and the tunnel collapse. The swelling pressure, according to the analysis, had a limited influence on the shotcrete, while the strength reduction played an important role in the development of instability.
Based on the analysis of swelling effects on rock support of the two cases, a flow chart of practical procedures for rock support estimation in weakness zones/faults containing swelling clay is recommended, in which the site investigation, laboratory testing and numerical simulation are integrated. Instrumentation and regular inspection for signs of instability are important for the stability control and back analysis. Further research on this issue is recommended to enrich the knowledge on tunnelling through such ground.