Real-time forecasting of brittle rock failure: a robust acoustic emission approach validated on a large experimental dataset
Fuqiang Gao, Lei Yang, Zhiguo Lu, Rui Wu, Wenju Liu, Chaodong Xi
Abstract
Abstract Accurate and real-time forecasting of brittle rock failure remains a major challenge in geotechnical engineering and mining safety. In this study, a novel failure forecasting method based on acoustic emission (AE) monitoring was proposed and validated using a large experimental dataset of 403 sandstone specimens subjected to uniaxial compression. A new early warning indicator, WI b —the b-value change rate per unit time—was developed to capture the frequency and amplitude of b-value fluctuations, which intensify as microcracking evolves toward macroscopic failure. WI b was computed continuously during loading, and a statistically adaptive threshold was established using the evolving mean and standard deviation of WI b . The results demonstrated that WI b consistently escalated before failure in all specimens, with warning signals occurring at loads averaging 76.7% of the ultimate compressive strength. Sensitivity analyses confirmed the robustness of the method across variations in b-value calculation parameters and AE sensor configurations. Importantly, WI b and its threshold were calculated solely from real-time data without requiring knowledge of the full loading path, ensuring practical applicability. The proposed approach offers a physically meaningful and operationally viable solution for real-time rock failure forecasting and has potential for integration into field-scale AE or microseismic monitoring systems.