With terrestrial resource depletion, submarine resource exploitation is vital for national resource security. As China's first undersea hard-rock gold mine, Xinli Mine in San Shan Island faces progressive direct shear failure risks in its bedrock weathered aquifer due to fault movements and mining disturbances. Instability could cause seawater inrush. This research investigates direct shear failure mechanisms and early warning methods for water-saturated weathered granite in fault zones. Direct shear tests with multi-field monitoring combined with critical slowing-down variance analysis reveal: (1) Strain evolution undergoes four phases: random distribution, localization, band initiation, and accelerated propagation. The first strain acceleration point (42%–49% peak stress) is a failure precursor. (2) Infrared evolution shows three phases: Compaction cooling, Thermal fluctuation, Thermal peaking. "Cooling-reheating" fluctuations and hotspots at 70%–78% peak stress constitute precursor phenomena. (3) Acoustic emission evolution exhibits three phases: Micro-damage initiation, Damage accumulation, and Accelerated destabilization. Event rate plummeting to 30%–40% peak value with ordered-to-chaotic phase-space transition predicts failure. Frequency bands broaden pre-failure: F1(0–55 kHz), F2(55–80 kHz), and F3(> 80 kHz). (4) Through optimization with the critical slowing down variance method, the precursor failure time was advanced on average by 83.3 s for the infrared field, 51.6 s for the acoustic emission field, and 18.3 s for the strain field. Highlights A multi-field integrated monitoring system combining acoustic emission, full-field strain, and infrared thermography was established. This system uncovers the interlinked evolution laws of acoustic energy dissipation, strain localization evolution, and thermal differentiation phenomena during progressive direct shear failure in submarine-weathered granite across multi-scale perspectives. Multi-dimensional reconstruction of the event rate parameter was implemented via phase-space reconstruction, thereby quantitatively characterizing the nonlinear dynamics—including chaotic attributes and periodicity transitions—throughout progressive failure stages. Characteristic curves of the acoustic, strain, and temperature fields were optimized through critical slowing-down variance-based adaptation, achieving an average precursor response time advancement of 56 s.
Li et al. (Mon,) studied this question.