GNGTS 2024 - Atti del 42° Convegno Nazionale

Session 3.2 GNGTS 2024 Figure 1 a) Example of tme-frequency analysis for a precursory event. b) Amplitude vs tme of precursory seismic signals Preliminary results indicate a slow initaton process, potentally linked to a precursory nucleaton phase driven by an intermitent preparatory process before the large stck–slip acceleraton accompanying the rock mass failure. Despite their low amplitude and absence of direct visibility in raw seismic records, these signals were efectvely discerned through the deep learning analysis. References Seydoux, L., Balestriero, R., Poli, P., Hoop, M. D., Campillo, M., & Baraniuk, R.; 2020: Clustering earthquake signals and background noises in contnuous seismic data with unsupervised deep learning. Nature communicatons, 11(1), 3972. Corresponding author: guidomaria.adinolf@unito.it