High in situ stress and its effects on tunnel design: An update based on recent project experience from WestConnex tunnels

The high virgin horizontal in situ stress field in the Sydney Basin and its impact on civil engineering projects is a well-known and accepted phenomenon found in significant literature (e.g., Pells, 2013). The prevailing high-stress effects depend on many factors — rock quality, tunnel orientation, proximity of geological features to tunnel crown, size and shape of opening, depth of excavation and stress magnitude. The behavior of the rock mass under high in situ stress conditions can cause stress fracturing and consequent dilation of the tunnel periphery, resulting in rock spalling at the tunnel crown/ invert or raveling of rock blocks on the tunnel sidewall. This type of failure is of a brittle nature and may create construction and safety risks during tunnel excavation, if it occurs behind the excavation face where ground support has already been installed. Therefore, the associated risks need to be managed during construction. This paper presents the design strategy adopted to mitigate these adverse tunneling conditions for the WestConnex Project to-date.

Overview of the WestConnex M4 East and new M5 tunnel projects

WestConnex is one of the New South Wales government’s key infrastructure projects. This 33 km (20.5 mile) project aims to ease congestion and connect communities and is the largest integrated transport and urban revitalization project in Australia. It was a key recommendation of the State Infrastructure Strategy released in October 2012. It brings together a number of important road projects, which, together, form a vital link in Sydney’s Orbital Network. These road projects include a widening of the M4 east of Parramatta, a duplication of the M5 East, and new sections of motorway to provide a connection between these two key corridors. The WestConnex project includes a number of stages: Stage 1a: M4 Widening; Stage 1b: M4 East; Stage 2: New M5; and Stage 3: M4–M5 Link. The tunnel design referenced in this paper relates to Stage 1b and Stage 2 of the WestConnex Project (Fig. 1).

What are the underlying causes of high horizontal stress?

The effects of high horizontal stress in tunnels and underground excavations have been well-documented in Australian literature (Pells, 1993; Oliveira and Diederichs 2017). As discussed by Oliveira and Diederichs (2017), two rock failure mechanisms may be observed at stress levels lower than the rock unconfined compressive strength (UCS):

• Brittle failures involving crushing, spalling and or slabbing of intact rock blocks, more often associated with buckling of thin sandstone beds in Sydney.
• Shear failures associated with planes of weakness, either pre-existing or induced by the excavation process, such as faults, cross bedding partings and bedding shears.

The primary cause of such failures is associated with removal of confining rock during tunnel excavation, which causes stress redistribution and results in stress concentration around the excavated periphery. This induced stress condition can be estimated using classical solutions for stress distribution around an elastic circular opening, such as that given in Fig. 2. As shown, the horizontal stresses within the tunnel crown increase proportionally with the horizontal-to-vertical stress ratio (λ). Therefore, the effects of induced high horizontal stresses can be expected to be more pronounced with increasing depth.

Considering that rocks like sandstone may fail at stress levels of approximately 50 percent of the UCS (Oliveira and Diederichs, 2017), it can be easily demonstrated (using the classical solution of Fig. 2) that the risk of stressinduced failure increases beyond depths of 40 to 50 m (131to164 ft) in Sydney. Therefore, the excavation induced stresses estimated with recent approximations of in situ stress magnitude (Table 1) approach values of approximately 12.5 MPa, which are equivalent to 50 percent for an average Hawkesbury Sandstone UCS = 25 MPa (Table 2).

The effect of such high horizontal stresses is often altered locally by the presence of major geological features, such as valleys, fault zones, and dykes, but also varies with orientation and with respect to stronger and/or weaker bands of rock — all making for variability and unpredictability. In addition, another important factor is the presence of planes of weakness, particularly bedding partings near an excavated tunnel, which magnify the induced stresses. For example, the simplified solution provided in Fig. 3 (Asche and Cooper, 2002), indicates that the stress concentration factor of SCF = 3 given in Fig. 2 may increase to approximately SCF = 7 for a low friction bedding parting located at about 0.5 m (1.6 ft) above a 7 m diameter (23 ft) tunnel.

What are the issues associated with high horizontal stress?

The immediate effect of elevated excavation induced horizontal stresses near a tunnel is the occurrence of significant shear displacements on subhorizontal discontinuities. Such an effect may cause damage to rock bolt corrosion protection, thus affecting durability, and, in more severe cases, tensile rupture of the bolts (Oliveira and Diederichs, 2017). The shear displacements may also induce local loosening of rock wedges or blocks near excavation shoulders. Because of the way the excavation stress release occurs, this can be a nuisance behind the face. In addition, such stress concentrations cause localized rock mass failures in the crown (as previously discussed), which require appropriate ground support with rock bolts and shotcrete.

An important aspect for the design of such a ground support is that the staged release of excavation induced stress means that the effect does not necessarily occur entirely at the excavation face. This becomes more pronounced when taking into account 3D effects, such as the orientation of the excavation in relation to subhorizontal discontinuities that affect stress concentration, as discussed. For instance, driving downhill means that bedding partings would typically rise into the roof, transitioning from a low stress concentration factor (SCF) to an extreme or high SCF environment where brittle failure is more likely, thus giving rise to the feeling that this is “unpredictable.” This condition is shown in Fig. 4, where the SCF transitions from low at section A-A to extreme at section B-B. A real example of this case is presented in Fig. 5, where the roadheader marks indicate failure occurred close to the face. On the other hand, driving uphill can mean that rock spalling is triggered backward from a failure initiating closer to the face when the stresses exceed the applicable spalling limit — for example, at section B-B and propagating to C-C (Fig. 6). This also gives rise to the feeling that it is unpredictable. Oliveira and Diederichs (2017) presented a simplified numerical figure to illustrate initiation of a brittle failure at a distance of 2 to 2.5 times the height of the excavation (Fig. 7).

The major implication of potential stress induced failure post excavation relates to safety risks for construction personnel who require access within the tunnels. Current tunnel construction practice is to have no personnel entry under unsupported ground because of legislative requirements. Supported ground is currently defined, within competent Sandstone on the New M5 and M4 East, as when both rock bolts and shotcrete have been installed, with the shotcrete having gained a certain minimum strength. Stress-induced spalling is therefore an issue if it occurs within supported ground conditions.

Given that rock bolts and shotcrete are installed soon after excavation, confinement is provided to the tunnel periphery. The level of confinement provided by the rock bolts and shotcrete is low. However, rock bolts provide sufficient retaining capacity should the spalled rock be wide enough to be captured by the rock bolts. Analysis of the results presented herein suggest that the plausible stress-induced spalled rock size rarely exceeds the extent of one rock bolt spacing. Any residual spalled rock not retained by rock bolts must, therefore, be accounted for in the design of the applied shotcrete.

However, the process of stress fracturing is complex and dependent on multiple factors. Hence, the groundsupport interaction is also complex and so is the design, particularly with shotcrete applied early to the excavated rock (due to supported ground requirements). Such complexity can only be addressed by an observational approach during construction in an attempt to manage such risks.

Design strategy

There is no practical way to avoid stress-induced spalling and other related consequences, and the associated risk of such an event occurring after support installation (i.e., under supported ground) cannot be ignored. The consequences of rock spalling on permanent shotcrete lining (particularly post installation) are that relatively large volumes of broken rock may build up behind the permanent shotcrete lining some distance behind the excavation face. This buildup of broken or spalled rock could induce fallout of shotcrete and rock. This raises a safety risk for the construction personnel as well as for the long-term end users of tunnels.

The large shear movement associated with stressinduced spalling may also impact the longevity of installed rock bolt reinforcement. The large shear movement may exceed the allowable shear limit of the rock bolts’ corrosion protection (plastic sheath). The damage of the corrosion protection means that the residual design life of the rock bolts is reduced and does not meet the design durability requirement. Rebolting of sheared rock bolts will be necessary to maintain the design life requirement in this case.

The design aim is to implement controls to reduce the probability of ground support damage should stressinduced spalling occur. Given that stress-induced spalling cannot be fully avoided with reasonably practical means, the consequences are unlikely to change. The associated risks detailed above therefore remain. Based on practical risk analysis, control measures must be applied so that the likelihood of ground-support damage is reduced, and the overall risk level of a stress-induced spalling event of shotcrete and bolt shearing is thereby reduced to an acceptable level.

The viability of any proposed control measure depends on its ability to meet the construction constraints where applicable/possible such that it must perform the following:

• Integrate and be compatible as much as possible with the already developed typical tunnel excavation and support installation sequence for tunnels without a stress-induced spalling problem.
• Incorporate either rock bolts or shotcrete as tunnel support. Support elements other than rock bolts and shotcrete may yield procurement difficulty and increase construction complexity.
• Conform to the protocols of the project-wide instrumentation and monitoring plan in the context of identification and transition in and out of elevated stress conditions.
• Yield tangible triggers to facilitate site-based observations. This follows, in principle, the observational approach of conventional tunneling with sequential excavation.

Site feedback, guided by in situ observational triggers, are adopted to make adjustment(s) to the already developed tunnel support design. Control measures for elevated stress conditions are thus developed as adjustments to already developed typical tunnel support with associated monitoring triggers.

Selected control measures for elevated stress condition

The selected risk control measures adopted for elevated stress conditions are:

• Mandatory split headings with minimum lag distance between headings: Typical tunnel cross sections applicable to these WestConnex projects are excavated using multiple headings. Each heading is approximately 5.5-m to 7-m wide by 6-m high (18-ft to 23-ft wide by 20-ft high). A 10-m (33-ft) minimum lag distance (approximately 1.5 times the single heading span of 7 m (23 ft)) between headings has been adopted. The main objective of the split heading (Fig. 8) is to induce the brittle failure near the face of the second heading and reduce the shear displacements within the second heading.
• Staggered rock bolt pattern: For the New M5 tunnels, the typical rock bolt pattern was adopted as square. However, for ground support against an elevated stress condition, the rock bolt pattern is altered to staggered. For the M4 East tunnels, the staggered rock bolt pattern is typical, and thus this control is less sensitive. Rock bolt spacing is also tightened where applicable. The benefits of a staggered pattern is its increased ability to contain fracture propagation within a bolt spacing. This prevents fractures from extending over multiple bolt spacings, with associated increased displacements, as would be observed for a square pattern (Fig. 9).
• Increased shotcrete thickness: The primary shotcrete developed for these projects is generally quite thin and relies on adhesion to the substrate. The thinnest crown primary shotcrete thickness is 55 mm (2.2 in.) for Sandstone Class I in the New M5 tunnels and 60 mm (2.4 in.) in the M4E tunnels. The minimum shotcrete thickness adopted for elevated stress condition is 125 mm (4.9 in.) targeting flexural capacity.

Numerical analysis

Numerical analyses were undertaken to assess responses of the rock mass and installed tunnel support subjected to elevated stress conditions. These numerical analyses include continuum (utilizing FLAC2D and FLAC3D) and discontinuum analyses (utilizing UDEC and RS2). The studies undertaken utilizing these analyses aided development of the adopted control measures for elevated stress conditions.

To assess the tunnel support performance when the support is subjected to elevated stress condition, additional rock material modelling was undertaken. Recent research development has shown that to better capture the stressinduced spalling zone and extent around a tunnel, a modified failure criterion for the rock mass should be used (Diederichs et al., 2010; Oliveira and Diederichs, 2017). Table 3 presents a set of modified Hoek-Brown rock mass material parameters adopted for these projects.

Figure 10 presents the analysis results as part of the evaluation of the stress-induced spalling effects. Two heading excavations followed by bench excavation have been adopted for the analysis results shown. A single bedding parting set at 1 m (3.3 ft) from the tunnel crown has been analyzed, with and without support installed. Rock spalling is likely to extend toward the full depth of rock bounded by the tunnel excavated periphery and bedding parting. With support installed, the rock spall size is much reduced. This confirmed the beneficial confinement effects provided by the rock bolts and shotcrete.

By adopting the same excavation sequence and material model, the effect of rock bolt shearing was assessed. Twin mainline tunnels were analyzed within the same cross section to capture the effects on the adjacent tunnel (Fig. 11). The results showed that potential rock bolt shearing is only likely to be confined to the lead heading of the first excavated tunnel. The magnitude of shear displacements is likely to exceed the shearing limit for corrosion protection, but does not exceed the ultimate structural capacity of rock bolts. That is, if elevated stress conditions are observed, rebolting for long-term reinstatement will be required and is likely to be limited to the rock bolts installed within the lead heading of the first excavated tunnel. This concludes that if there is sufficient lag distance between the lead and trailing headings of the first tunnel, the extent of rebolting required will likely decrease. Confirmation of the extent of rebolting typically is done using endoscope observations.

Based on the spalling analysis results, it was assessed that the most likely rock spall depth ranges from 0.5 m to 0.7 m (1.6 to 2.3 ft); subject to different ground conditions. Rock spall extent is slightly greater than a one-bolt spacing (Fig. 10). Staggered rock bolt patterns were then adopted/confirmed to better arrest and contain spalling rock as discussed above. With the rock bolts arranged in a staggered manner, the longitudinal strip of spall rock is arrested and contained. The staggered rock bolt pattern is therefore considered more suitable for elevated stress conditions.

Additional analyses were undertaken to evaluate the performance of the primary shotcrete lining. The prevalence of rock spall within a one-rock-bolt spacing suggests that the shotcrete needs to provide retaining capacity to retain the rock spall. Stress-induced spalling is a fracture process that likely induces multiple fractures within the rock spall, especially when the spall rock size is relatively small (Fig. 12). The mechanism involves a load transfer of the spall rock weight to the shotcrete, which in turn transfers to the rock bolts through the connection between the shotcrete and the rock bolt. A typical rock bolt–shotcrete connection is facilitated using handle bar plates. The primary shotcrete lining is therefore critical for containment of the rock spall.

Deterministic structural analysis based on Barret and McCreath (Oliveira and Diederichs, 2017) was then undertaken to assess the mechanical response of the shotcrete. This analysis assumes that the process of stress fracturing is associated with ground stress dissipation (i.e., stress transfer from the rock mass movement to the shotcrete is reduced to a negligible magnitude). However, it was also realized that some level of in situ stress remains within the spall rock boundary, providing frictional restraint (Fig. 12). The overall rock spall weight was then adjusted accordingly when applied to the deterministic analysis to determine an appropriate shotcrete thickness.

The assumption of stress dissipation used in the deterministic analysis is not definite. There is limited evidence or in situ experience available to confirm this assumption. The complexity of the stress fracturing process and the associated ground stress transfer/ dissipation to the shotcrete is not completely understood. The implication is that the risk of shotcrete failure is further increased, regardless of shotcrete thickness (reasonably practical thickness), because of the magnitude of plastic strain experienced during stress fracture. The primary shotcrete was analyzed using numerical analyses to confirm adequacy of the adopted shotcrete thickness.

Design outcomes

The different analyses undertaken indicate the adequacy of the control measures adopted for the elevated stress conditions in these projects. To facilitate implementation of these control measures, a set of observational criteria was developed:

Elevated stress condition susceptibility definition.

Tunnel sections defined below shall be classified as areas subject to potential rock spalling risk:

• First bedding ≤ 1.5 m (4.9 ft) above crown.
• Presence of any seams or shears ≤ 1.5 m (4.9 ft) above crown.
• Presence of shale lens or mudstone facies ≤ 1.5 m (4.9 ft) above crown.

Note the first bedding distance of 1.5 m (4.9 ft) from tunnel crown was set to provide an early alarm to initiate in situ observations earlier. This was to better react and to implement control measures in advance.

Elevated stress condition observational triggers/ identifiers. High stress conditions are evidenced by a combination of (but not limited to):

• Ground cover greater than 45 m (148 ft).
• Higher horizontal ground movement measurements (above 75 percent of the amber level of the typical conditions).
• Spalling or cracking of shotcrete within two weeks of application.
• Higher horizontal movements detected in endoscopes (more than 9 mm or 0.4 in.).
• Presence (or predicted presence) of a subhorizontal discontinuity or adverse feature(s) within 1.5 m (4.9 ft) of the tunnel crown. This can be predicted:
  • If the bedding is rising in the face — from geological face mapping.
  • If the bedding is falling toward the crown — from geotechnical endoscope mapping.

The above observational triggers/identifiers were also adopted to aid removal of elevated stress control measures.

The strategy of the elevated stress control measures is:

• To promote stress-induced spalling to occur at the lead and trailing heading interface. This is to permit/aid removal of spall rock during trailing heading excavation, and constrain spall rock to unsupported or less trafficked supported ground.
• To reduce the uncertainty related to rock bolt shearing. Endoscope installation density is also increased (doubled) to better define the affected extent of rock bolts.
• To give time for shotcrete strength gain to provide support to spall rock.
• To control spall rock failure after support installation with increased shotcrete thickness and tightening of rock bolt spacing such that tangible damage (i.e., not sudden brittle failure fallout) can be observed and minimized.

It should also be noted that, although the M4 East project incorporated a few of these key control measures, the majority were developed as part of the New M5 project given that, on average, the tunnels are located at greater depths.

Construction observations to date

Limited construction data are available to date from these projects that may indicate stress concertation. However, anecdotal observations to date across both projects indicate larger overbreak in sections that are deeper than 45 m (148 ft). Additionally, lead heading shotcrete near the central temporary haunch has exhibited a slightly higher frequency of cracking than in other parts of the tunnel. Although this correlates well with spalling analysis results, a new lap detail at the lead and trailing heading interface was developed to mitigate risks associated with overstress of the primary shotcrete lining (Fig. 13).

Conclusions

Limited construction data are available to date to provide sufficient validation to the design implemented for high stress condition for the WestConnex Project. Construction observations are ongoing to provide feedback to the performance of the selection system. However, the construction data collected to date show that the expectation for elevated stress conditions perceived from the design analysis presented in this paper holds. This concludes that an observational based approach allows for selection of adequate tunnel support design to manage the risks associated with elevated stress conditions. Coupled with engineering judgment, the design principles discussed in this papers although developed for Sydney Sandstone, are applicable to other conventionally excavated tunnels in horizontally bedded strata of similar stress states.

References

Asche H.R. and Cooper D.N., 2002. Estimation of tunnel support requirements for TBM driven rock tunnels. ITA World Conference, Sydney. Barrett, S.V.L. and McCreath, D.R., 1995. Shotcrete Support design in Blocky Ground: Towards A Deterministic Approach. Tunneling and Underground Space Technology. 10(1): 79–89.

Bertuzzi, R. 2014. Sydney Sandstone and shale parameters for tunnel design. Australian Geomechanics. 49(2): 95–104.

Diederichs M.S., Carter T., and Martin C.D., 2010. Practical rock spall prediction in tunnel. Proceedings of World Tunnelling Congress ‘10 – Vancouver.

McQueen L.B., Bewick R.P., Sutton J., and Morrow A., 2017. Stressinduced brittle failure of the Hawkesbury Sandstone – Case study from crack initiation to tunnel support. 16th Australasian Tunnelling Conference 2017.

Oliveira D., and Diederichs, M., 2017. Tunnel support for stress induced failures in Hawkesbury Sandstone. Tunnelling and Underground Space Technology. 64: 10–23.

Oliveira, D.A.F., and Parker, C.J., 2014. An alternative approach for assessing in-situ stresses in Sydney. 15th Australasian Tunnelling Conference 2014.

Pells, P.J.N., 1993. Rock mechanics and engineering geology in the design of underground works. 1993 E.H. Davis Lecture, Australian Geomechanics Society.

Transport Sydney, 2013. WestConnex – Building for the Future https:// transportsydney.files.wordpress.com/2013/09/2013-09-19-westconnexmap. png.

Villaescusa E., Kusui A., and Drover C., 2016. Ground support design for sudden and violent failures in hard rock tunnels. 9th Asian Rock Mechanics Symposium, 18–20 October 2016, Bali, Indonesia.