Coal-rock instability risk prediction based on coupled microseismic spectral, energy, and fracture evolution characteristics

With increasing mining depth, coal-rock failure and instability under high in situ stress exhibit complex behaviours, heightening the risk of dynamic disasters. Although microseismic (MS) monitoring is widely used, current methods focus mainly on single-dimensional indicators, such as event energy and frequency, making it difficult to accurately capture the evolution and precursors of coal-rock fracturing instability. This study analyses nearly 5 months of MS data from the Ji15-21050 working face of Pingmei No. 8 mine to explore spatiotemporal coupling among spectral characteristics, energy responses, and fracture structures. Results show that: (1) MS events spectra are mainly low-frequency, single-peak patterns, indicating that fracturing is dominated by large-scale, long-duration events, consistent with the progressive stress-release mechanism in longwall mining. (2) Although high-frequency, complex spectra are rare, they concentrate in structurally active zones and are highly sensitive to localised stress anomalies and structural activation. (3) The spectral complexity index (SCI) peaks during the instability incubation stage, well before energy accumulation and fracture connectivity, demonstrating strong early-warning capability and sensitivity to microcrack activation. (4) The fracture direction concentration index ( D c ) increases notably prior to intense energy release, indicating its effectiveness in capturing geometric responses to energy release and serving as a physical precursor to strong instability. An entropy-weighted multidimensional risk prediction model was developed. The on-site coal-rock instability risk score rose from 0.27 in December 2023 to 0.76 in February 2024, closely aligning with the spatiotemporal distribution of coal cannon events, roof deformation, and high-energy MS in March, validating the model's predictive accuracy and engineering applicability.