Laboratory earthquake prediction of granite

Abstract

Earthquake prediction refers to predicting the magnitude, location, and time of earthquakes and is challenging. Four attempts to predict the magnitude, location, and time of the laboratory earthquake of granite were made. First, the stick-slip failure process of granite samples was studied by a homemade test setup, and the whole process was monitored in real-time by strip strain gauges, rosette strain gauges, and acoustic emission (AE) transducers. Second, six parameters related to the stress drop were selected: dry density, Young's modulus, loading rate, static friction coefficient of the interface, normal stress, and shear stress. The prediction model of stress drop was established based on a backpropagation neural network (BPNN) considering the above six parameters. The stress drop in this paper is positively correlated with the released energy (magnitude) when the simulated fault slips suddenly. Third, the simulated fault interface was scanned by a high-precision 3D topography scanner before the stick-slip tests, and then the 3D topography scanning image was quantitatively measured based on digital image processing (DIP). According to the measurement results, the relationship between the nucleation position and the microstructure of the simulated fault interface was discussed. Fourth, the characteristics of AE parameters (absolute energy, ringing count rate, amplitude, b1 value) in the stick-slip process were studied. This paper aims to determine the precursory factors of sudden slip. This study helps explain the mechanism of shallow-focus earthquakes.