Acoustic emission statistical characteristics of composite coal rock and precursor warning based on k-means clustering

The analysis of failure mechanism and acquisition of failure precursor information of composite coal rock is crucial to study mine dynamic disasters. In this paper, the uniaxial compression acoustic emission (AE) tests of composite coal rock under different layer thickness ratios were carried out. Based on the theory of avalanche dynamics, the probability density distribution of AE energy, amplitude, duration, waiting time and other parameters of composite coal rock was statistically analyzed. The k-means clustering method was used to explore the failure precursors of composite coal rock. The results show that with the increase of coal thicknesses, the AE b-value in the failure process of composite coal rock decreases gradually. In addition, the energy distribution exponent, amplitude distribution exponent, and duration distribution exponent of composite coal rock specimens decrease gradually. The distribution exponent of waiting time shows robustness. Based on k-means clustering, five characteristic parameters in the AE signal are clustered, and the AE signals are successfully clustered into three categories. Time-series evolution analysis shows that the third type of signal is continuously dense in the critical failure stage, and this type of signals also has the characteristics of large energy proportion and small signal quantity proportion, which can reflect the information of the failure precursor of composite coal rock. The findings may contribute to the development of early warning methods for composite coal rock failure.