Abstract

The safe, cost-effective advance of deep tunnels increasingly hinges on real-time forecasts of structurally controlled rock-block failure—capabilities that lie beyond classic Key-Block Theory (KBT) and remain computationally prohibitive for full-physics numerical models. We introduce the Universal Discontinuity Index (UDi), a single probabilistic score that fuses four physically grounded components—physical damage, effective?stress intensity, active-fracture ratio, and kinematic feasibility—into a scalable indicator of block instability. The index embodies a complete-system view: it treats the fracture network, local stress field, and block release kinematics as an integrated, dynamically updating subsystem within a Digital-Twin (DT) workflow. Three levels of validation are presented. (i) Theoretical equivalence: UDi's kinematic term reproduces Warburton's vector criteria while providing a continuous risk scale. (ii) Numerical benchmarks: in controlled wedge-formation tests, UDi captured progressive batch failure with the same temporal hierarchy recorded by Bonded-Particle DEM and hybrid FDEM, yet required milliseconds rather than hours. (iii) Field proof: 52 photogrammetry-mapped faces from Chile's El Teniente mine were analysed via a stochastic DFN–KBT engine (>5 300 realizations); the normalized UDi predicted the mean number of unstable blocks with R2 = 0.93 and 95 % accuracy. Leveraging its speed, we embed UDi in a Poisson-regression early-warning model that converts each index update into an expected block count per advance round, validated via jack-knife resampling. Fracture-propagation-informed refinement extends the method to blocks still in formation, enabling continuous hazard maps as excavation proceeds. UDi therefore bridges the gap between deterministic kinematic checks and high-fidelity simulations, delivering DT-ready, probabilistic, and progressive failure forecasts that unlock proactive support design and truly adaptive tunnel operations. © 2025 Elsevier Ltd