A physical-numerical approach to predicting tension-induced rock mass collapse using reciprocal tilt rate.

Tension-induced rock masses at steep slope fronts undergo progressive failure under self-weight tensile stress, characterized by deep rear-edge fractures that propagate subcritically over decades before catastrophic collapse. Conventional monitoring systems frequently fail to provide adequate warning because the measurable deformation phase typically lasts only hours to days, despite decades-long crack incubation. We develop an integrated framework combining physical similarity modeling, smoothed particle hydrodynamics (SPH) simulation, and field validation to systematically investigate failure evolution. High-frequency MEMS accelerometers (1000 Hz) capture dynamic tilt progression through spatial vector analysis of gravitational acceleration components. Water-induced weakening at fracture tips simulates natural stress corrosion, accelerating time-dependent deformation under controlled conditions. Results reveal distinct two-stage evolution: prolonged quasi-static deformation followed by rapid acceleration, where reciprocal tilt rate exhibits linear decay toward zero before failure, enabling quantitative prediction. SPH simulations with an elastoplastic-mixed damage model demonstrate that fracture geometry significantly influences prediction window duration, with central horizontal fractures providing optimal early warning conditions. Field validation confirms method robustness: reciprocal tilt rate approached zero linearly across all scenarios, with prediction accuracy within 0.9% and a 38-hour acceleration phase in prototype scale, providing practical lead time for risk mitigation. This integrated physical-numerical approach advances understanding of brittle rock mass failure mechanisms and provides reliable tools for implementing effective early warning systems in rockfall-prone regions.