Question 1 of 5: Design strategy for uncertainty in a bedded/jointed highway tunnel

Question 1: Design strategy for uncertainty in a bedded/jointed highway tunnel (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.

A risk-based design strategy for this tunnel has two linked phases — a pre-construction phase that sets the initial ground model and support classes, and a construction-phase program that updates both as real ground is exposed. Both are organized around the eight issues raised.

(i) Uncertainty in material parameters. Strength and deformability of the siltstone and sandstone (and of the bedding-plane/joint discontinuities) are known only from a sparse drillhole program, so every design parameter is treated as a distribution, not a single number: report UCS, RMR/Q rating components, and joint shear strength as a mean ± range (or a characteristic lower-bound percentile) rather than a single “best-estimate” value, and carry that range through the support design (e.g. bound the Hoek–Brown m, s parameters, not just their mean). Point-load index testing on recovered core, supplemented by laboratory triaxial/direct-shear testing on the weaker bedding-plane interfaces, is the practical way to populate the distribution cheaply and at high sample count.

(ii) Probability of various “events” over the construction life. A simple event tree/point-form risk register is developed pre-construction: (a) chimney/block-fall from the horizontal bedding intersecting steep joints in the crown, (b) squeezing/overstress at the 210 m overburden if a weaker interbed is encountered, (c) water inflow along a permeable sandstone bed, (d) face instability during top-heading advance. Each event is assigned a qualitative likelihood×consequence rating (a standard risk matrix) from the sparse borehole data, and the highest-ranked events (typically crown block-fall, given the horizontal bedding directly overhead) drive where the construction-phase monitoring effort is concentrated.

(iii) Uncertainty in the initial ground state with only one drillhole per 100 m. A single centreline hole every 100 m cannot resolve bedding thickness or joint spacing variability between holes, so the pre-construction ground model is built as discrete domains bounded by the known holes, each assigned the RMR/Q class of its nearest control point with an explicitly wider uncertainty band mid-way between holes. The design carries at least two support classes (a base case and a contingency, weaker case) for every domain so a change in ground condition between holes does not stop the heading while a new support class is designed from scratch.

(iv) Geophysical techniques to reduce uncertainty. Between-hole geophysics (seismic refraction/cross-hole tomography, or ground-penetrating radar for the near-surface bedding contacts) is used to interpolate bedding-plane continuity and locate anomalies (e.g. a weathered zone or a fault) between the widely spaced boreholes without the cost of additional drilling; a probe-ahead technique (seismic-while-drilling or horizontal core drilling from the face) is added during construction specifically to de-risk the next 20–30 m of advance.

(v) Adequacy of rock mechanics design in large openings. A horseshoe highway tunnel is a large span relative to the bedding/joint spacing, so classification-based empirical design (RMR/Q) is supplemented, not replaced, by numerical modelling (e.g. a 2D FEM/DEM section through the worst-case domain) to check crown deflection and support-load compatibility, and by wedge-stability analysis of the specific joint sets identified from the core — empirical charts are calibrated on smaller openings and are known to be less reliable as span grows toward the tunnel's own scale.

(vi) Construction sequencing to reduce uncertainty. A staged top-heading-and-bench sequence (rather than full-face) is adopted specifically so each heading acts as a probe: the top heading exposes and characterizes the crown ground ahead of the bench, and the bench excavation is not started in a given domain until the heading's mapped condition confirms (or revises) the pre-construction support class for that domain. Advance length per round is shortened in any domain flagged higher-risk from (ii).

(vii) Rock support strategies and their use. Support is specified per RMR/Q class as a menu (systematic rock bolting with mesh in fair-to-good ground, escalating to shotcrete and, in the weakest domains, steel sets) exactly as tabulated in Bieniawski's RMR support guidelines (used quantitatively in Question 5 below); support is installed close to the face (a fixed maximum unsupported span/exposure time per class) rather than at a fixed round length, so that support response scales with the ground actually exposed, not with a pre-construction average.

(viii) Other factors. Convergence/extensometer monitoring at instrumented sections (the observational method) closes the loop between the pre-construction model and construction reality — measured convergence trends are compared against the design's predicted range at each domain boundary, and a pre-agreed trigger-and-response plan (add bolts, add shotcrete thickness, slow advance) is invoked automatically if convergence exceeds the predicted band, rather than being decided ad hoc underground.