Rock Crack Types Identification by Machine Learning on the Sound Signal

Sound signals generated during rock failure contain useful information about crack development. A sound-signal-based identification method for types is proposed. In this method, the sound of tensile cracks, using Brazilian splitting test, and those shear direct are collected to establish training samples. The spectrogram used characterize signal taken as input. To solve small sample problem, since only a amount can be obtained in our experimental pre-trained ResNet-18 feature extractor acquire deep characteristics spectrograms. Gaussian process classification (GPC) employed recognizing model classify extracted verify proposed development processes biaxial test identified. results show that feasible. Moreover, investigate rockburst process. consistent with previous research results, further confirming accuracy rationality method.