Rockburst risk identification model based on microseismic fractal dimensions and its application

To address insufficient integration of multi-attribute microseismic information, predominantly qualitative extraction of precursor features, and limited quantification of risk levels in existing rockburst risk identification methods, this study proposed a rockburst risk identification model based on microseismic fractal dimensions and fuzzy comprehensive evaluation (FCE). Based on the temporal, spatial, and energy distribution characteristics of microseismic events, a multidimensional fractal indicator system comprising 3 fractal capacity dimensions and 6 fractal information dimensions was constructed to characterize the clustering degree, frequency distribution, and energy distribution of microseismic events at different scales. An anomaly index was introduced to characterize the deviation of each fractal dimension from its background value, and Gaussian membership functions were used to establish a mapping between the fractal indicators and 4 rockburst risk levels: no risk, weak risk, medium risk, and strong risk. A confusion matrix and F-scores were then used to determine the weight of each fractal indicator, and an FCE model was constructed to enable probabilistic representation of rockburst risk and comprehensive identification of risk levels. To validate the model, uniaxial compression tests with acoustic emission (AE) monitoring were conducted on coal samples, and the evolution of fractal dimensions during coal-rock fracture was analyzed. The results showed that, as coal samples approached instability and failure, AE events gradually evolved from a dispersed distribution to a locally concentrated distribution, fractal dimensions decreased markedly, and the risk levels predicted by the model progressively increased as failure developed, confirming that low-value anomalies in fractal dimensions could serve as precursors of coal-rock instability. Field validation was conducted using 15 months of microseismic monitoring data from the working face 13218 at Xiaojihan Coal Mine. The results showed that microseismic events at the working face 13218 were concentrated mainly near the section coal pillar, and periods of anomalous decreases in fractal dimensions corresponded well to strong mine tremor events. With a sliding time window of 10 d, the model achieved a hit rate of 85% and an average warning lead time of approximately 4.5 d. The results demonstrate that the model effectively captures precursory features of rockburst occurrence and that its assessment results are consistent with engineering practice.