A Risk-Driven Probabilistic Framework for Blast Vibrations in Twin Tunnels: Integrating Monte Carlo Simulation to Quantify Cavity Effects

Predicting blast-induced vibrations in twin tunnels is challenging due to complex wave-cavity interactions, which render conventional scaled-distance (PPV-SD) models inadequate. This study introduces a hybrid empirical-probabilistic framework to quantify the probability of exceeding regulatory vibration thresholds. Field data from the Northern Marmara Highway project first quantitatively confirm the severe degradation of the classical scaled-distance (PPV-SD) method in twin-tunnel geometry, reducing a strong correlation (R = 0.82) to insignificance. A Random Forest sensitivity analysis, applied to 123 blast records, ranked the governing parameters, guiding the development of a deterministic multi-parameter regression model (R = 0.72). The core innovation of this framework is the embedding of this deterministic model within a Monte Carlo Simulation (MCS) to propagate documented input uncertainties, thereby generating a full probability distribution for PPV. This represents a fundamental advance beyond deterministic point-estimates, as it enables the direct calculation of exceedance probabilities for risk-informed decision-making. For instance, for a regulatory threshold of 10 mm/s, the framework quantified the exceedance probability as P (PPV > 10 mm/s) = 5.2%. The framework’s robustness was demonstrated via validation against 100 independent blast records, which showed strong calibration with 94% of observed PPV values captured within the model’s 90% confidence interval. This computationally efficient framework (<10,000 iterations) provides engineers with a practical tool for moving from deterministic safety factors to quantifiable, risk-informed decision-making.