Detection and Localization of Excavation-Disturbance-Related Near-Field Microseismic Events During TBM Tunneling

Tunnel boring machine (TBM) excavation generates continuous mechanical vibration that can obscure weak, short-duration propagating responses related to structural-plane disturbance. This study develops a Signal-Constrained Activation Detection and Localization (SCADL) framework using continuous three-component geophone-array records. SCADL first constructs an adaptive multi-station consistency trigger from synthesized three-component envelopes and rejects non-propagating mechanical disturbances using coherence and polarization constraints. First arrivals are picked by fusing statistical abrupt-change, local onset-gradient, and polarization-variation evidence, and event locations are estimated using an ahead-of-face layered velocity model and relative correction of similar event pairs. A multi-evidence activation index then integrates spatial clustering, coupling with the face-disturbance zone, shear/compression energy ratio, temporal evolution, and event quality to identify high-confidence candidate structural-plane activation events. The workflow was evaluated using one 16 h continuous field monitoring record acquired from a single TBM monitoring section and manually reviewed reference sets comprising 286 propagating events, 136 high-confidence events for arrival-time evaluation, and 96 events for activation-assessment review. SCADL identified 263 valid propagating events, achieved an event-level F1-score of 0.88, reduced the median arrival-time picking error to 2.4 ms, constrained the localization residual to 2.9 ms, and compressed the corrected cluster thickness to 0.82 m. Among the detected events, 86 high-confidence candidate activation events formed two clusters spatially consistent with the F04 and F02 structural zones confirmed by post-excavation geological validation. These results support the feasibility of SCADL for single-section TBM monitoring.