Rock crack type identification by Gaussian process learning on acoustic emission

A good understanding of crack development helps reveal the mechanism of different rock failures. Identification of crack types (tensile or shear) can be a useful tool for characterizing the crack development process. In this work, we employed AE (acoustic emission) pattern identification to detect the crack type during the granite failure. First, three-point bend tests and direct shear tests were performed on granite to capture the characteristics of the AE signal corresponding to the tensile cracks and shear cracks, respectively. Analysis of the acquired AE signals shows that the VMD-based SampEn (the sample entropy of AE signals decomposed using variational mode decomposition) for tensile cracks differs considerably from that for the shear cracks. Then, we developed a recognition model based on the Gaussian process to detect the crack type during rock failure using VMD-based SampEn and amplitude as feature indexes. The proposed approach was then used to identify the crack type during rock failure in a three-point bend test. The results demonstrate that the fraction of tensile cracks in three-point bend tests is greater than 90%, comparable with other research findings, demonstrating that the suggested recognition model is feasible for capturing tensile and shear crack types during rock failure. In addition, the proposed technique is used to characterize crack development during the rockburst process, and the findings show that rockburst involves a shift from outward tensile cracks to interior shear cracks. This finding is consistent with those found in prior research, indicating that the proposed strategy is feasible.