An Edge-AI Powered Micro-Seismic Digital Twin for Predicting Dynamic Rockburst Risks in Deep Hard-Rock Excavations

Abstract: As mining operations aggressively push past depths of 2 km to 3 km, underground excavations encounter extreme, non-linear geo-stresses. At these depths, traditional microseismic (MS) systems fail to predict catastrophic, sudden rockburst events due to data transmission latency, manual waveform processing bottlenecks, and static geological models. This paper reviews the emerging paradigm of shifting from centralized cloud processing to decentralized, real- time edge computing integrated into high-fidelity Digital Twins (DTs). By processing massive, high-frequency waveform data directly at the underground sensor level (Edge-AI) and feeding these localized dynamic metrics into a continuously updating 3D physical-virtual model, modern deep mining can achieve automated, low-latency, and predictive hazard zoning. This review contextualizes the transition from legacy empirical seismic monitoring to dynamic, event-driven Edge-AI systems, identifying critical bottlenecks in micro-seismic wave classification, edge-hardware survival, and real- time bidirectional data synchronization.