Identifying Acoustic Emission Precursors for Spalling in Rock Tunnels using a Scale-Model Test

Rock spalling is a type of brittle instability that occurs during tunnel excavation. Spalling degrades the rock mass, reducing its load-bearing capacity and compromising tunnel safety. Identifying precursors allows for timely mitigation measures and helps manage emerging changes during excavation and long-term tunnel stability. This study identifies acoustic emission (AE) signals that precede brittle spalling onset in a 51 mm diameter circular tunnel model. The tunnel is excavated in a 300 × 300 × 300 mm 3 cubical analog rock specimen placed in the true-triaxial apparatus. The test specimen was then incrementally and anisotropically loaded under three independent-controlled major principal stress until spalling failure was observed. Six wideband AE transducers were attached to the specimen surfaces to monitor AE activity throughout the excavation and incremental loading stage. This research aims to address two key questions: (1) What are possible precursors for spalling onset that can be identified from AE data? and (2) Is there any strategy to improve the reliability and resolution of the precursors? The anisotropic stress change within the specimen resulted in a V-shaped notch failure progression at the tunnel wall characterized by extensional and shear fracturing. AE monitoring provided insights into the spalling onset, with specific signal patterns identified as precursors through frequency domain analysis (peak amplitude, average frequency, and rise angle) and -value analysis. Optimization of temporal deconstruction emerged as the key strategy for AE parametric analysis, enhancing resolution into visible spalling onset and damage progression compared to physical observation. In frequency domain analysis, clustering signals using amplitude and frequency thresholds, combined with AF/RA classification proved more effective when integrated with temporal deconstruction. For -value analysis, results indicate that applying a moving number of AE hits with 25% overlap between adjacent windows provides more robust spalling precursors than either discrete or smooth time-interval approaches. The optimization framework developed through back-analysis, as demonstrated in this study, should serve as a methodological guide rather than fixed protocol. This approach enables adaptation to varying boundary conditions, sensor configurations, and system capabilities, ensuring more reliable precursor detection and improved safety in tunneling and underground excavation projects.