Question 2 of 5: Pre-Construction and Construction-Phase Strategy for a Subway Tunnel Under Uncertainty

Question 2: Pre-Construction and Construction-Phase Strategy for a Subway Tunnel Under Uncertainty (20 marks)

Question text not reproduced: the examination questions are © Engineers and Geoscientists BC. Open the official past paper (linked at the top of this page) to read the question, then follow the worked solution below.

The governing philosophy for a 30 m-deep horseshoe subway tunnel in horizontally bedded, jointed Ordovician limestone with interbedded non-swelling plastic shale beneath Toronto is the observational method (Peck 1969): treat every pre-construction parameter as a probability distribution rather than a single number, design for the most-probable case with a pre-planned contingency for the credible worst case, and use continuous monitoring during construction to decide, in real time, which design is actually being encountered. Each of the seven issues below maps onto one stage of that structured decision loop.

1. Uncertainty in material parameters. Report every strength/deformability input (limestone UCS and RQD/RMR, shale plasticity index and residual $\phi_r$, bedding-plane and joint shear strength) as a mean and a range from the drillhole/lab programme, not a single value. Run the tunnel support design at the mean, the lower-bound (weak-shale-controlled) and upper-bound (limestone-controlled) parameter sets to bound the required support — a simple three-point sensitivity study rather than a single deterministic run. Bedding-plane shear strength (limestone-on-shale contacts) and the shale's short-term-vs-long-term (undrained vs. drained, or plastic creep) strength are typically the two governing but least-certain parameters and deserve the largest lab-testing allocation.

2. Probability of adverse "events" over the construction life. Build a simple event tree for the excavation: (i) face instability / raveling at a bedding-plane or joint intersection, (ii) squeezing/creep of the plastic shale into the opening, (iii) unexpected groundwater inflow along a bedding-plane aquifer, (iv) a locally low-RMR zone (fault gouge, clay seam) between drillholes. Assign each a qualitative likelihood (from local case-history frequency in similar Toronto-area geology) and a consequence class (delay only / support-only / life-safety), and rank them — this ranking, not a single "worst case," drives where monitoring instrumentation and contingency support stocks are concentrated.

3. Uncertainty in initial ground state with only one drillhole per 100 m. A single centreline drillhole every 100 m constrains the horizontally-bedded stratigraphy well along the tunnel axis (bedding is, by the problem statement, sub-horizontal and therefore highly correlated between holes) but says very little about a discrete cross-cutting feature (a joint set, fault, or channel-fill) that could exist anywhere in the 100 m gap. Pre-construction strategy: treat the stratigraphic sequence (bed thicknesses, limestone/shale proportions) as reliably interpolated between holes, but treat local discontinuities and their orientation as a probabilistic hazard requiring ahead-of-face reconnaissance (probe holes, face mapping) rather than as something the drillhole programme can be relied on to have found.

4. Geophysical techniques to reduce uncertainty. Between the widely-spaced drillholes, run surface or borehole geophysics to fill the gap cheaply: seismic refraction or MASW along the alignment to track the limestone/shale interface and locate low-velocity (weathered, fractured, or shale-rich) zones; cross-hole seismic tomography or electrical resistivity tomography between adjacent 100 m holes to image anomalies directly ahead of design; ground-penetrating radar for near-surface utility/void conflicts. Geophysics does not replace the drillholes (it needs them for calibration) but converts point data into a continuous along-alignment profile, sharply reducing the "unknown between holes" uncertainty at a fraction of the cost of closer-spaced drilling.

5. Adequacy of rock mechanics design methods in a large opening. A horseshoe transit tunnel is large enough (typically 6–10 m span) that closed-form elastic solutions (e.g. Kirsch) are useful only for a first-pass stress estimate around the opening; they cannot capture bedding-controlled block/wedge failure at the crown and shoulders, or time-dependent squeezing of the shale. Design must therefore combine (i) empirical rock-mass classification (RMR/Q) calibrated against local Toronto-area tunnel case histories for a first support estimate, (ii) discontinuum/block-theory or key-block analysis for the specific bedding+joint geometry to check crown and haunch wedge stability, and (iii) where the shale is thick or plastic enough to creep, a numerical (FEM/FDM) analysis with a suitable time-dependent constitutive model — no single method is adequate alone for this size of opening in this geology.

6. Construction sequencing to reduce uncertainty. Excavate top-heading-and-bench (rather than full-face) so each heading is a probe for the bench below and the sidewalls beyond; keep face advance short (1–1.5 m) with support installed close behind the face wherever plastic shale or an adverse bedding/joint intersection is exposed, lengthening rounds only in confirmed competent limestone. Pilot/probe drilling ahead of the face (every round, extended to 2–3 rounds ahead in a suspected hazard zone) converts the along-tunnel uncertainty from "one point per 100 m" to essentially continuous, letting the support class be selected just before each round is excavated rather than fixed in advance.

7. Rock support strategies and their use. Adopt a small family of pre-designed, RMR/Q-linked support classes (e.g. light spot bolting in good limestone; systematic rock bolts + wire mesh + thin shotcrete in jointed/bedded ground; heavier bolts + thicker steel-fibre shotcrete, closer spacing, and possibly light steel sets or an invert strut in plastic shale/squeezing ground) selected face-by-face from the mapped conditions — the observational method's core idea. Instrument every round with convergence pins and, in any zone flagged high-risk from steps 2–4, extensometers and piezometers, with pre-agreed trigger/action levels (e.g. convergence rate exceeding X mm/day → install the next-heavier support class immediately) so that the "program during construction" is a documented decision procedure, not an ad-hoc judgement call.