A new automatic noise elimination method for rock collapse warning system

Reliable, accurate, and timely warning of rock collapse is crucial for mitigating geotechnical hazards. However, due to the complexity of natural environments, existing monitoring methods often respond to transient environmental vibration rather than actual rock-bridge degradation, resulting in frequent false alarms. Previous studies have shown that when rock-bridge stiffness decreases, vibration amplitude ( VA ) tends to increase while gravity frequency ( GF ) declines. Building on this principle, this study proposes a noise-resistant early warning approach based on cross-validation between VA and GF to distinguish damage-related responses from environmentally induced fluctuations. The method was validated through a freeze–thaw laboratory experiment and a field monitoring case, both of which demonstrated a significant reduction in false alarms while maintaining sensitivity to actual stiffness degradation. The results confirm that the combined use of VA and GF enhances the interpretability, noise robustness, and reliability of rock collapse early warning systems. This study highlights the practical potential of integrating dynamic indices for long-term stability monitoring and contributes to the advancement of geotechnical hazard warning technologies.