Rock Fracture Monitoring Based on High-Precision Microseismic Event Location Using 3D Multiscale Waveform Inversion

Abstract

Microseismic (MS) source location can help us obtain the fracture characteristics of a rock mass under a thermal-hydraulic-mechanical (THM) coupling environment. However, the commonly used ray-tracing-based location methods are easily affected by the large picking errors, multipath effects of travel time, and focusing and defocusing effects of rays in wavefield propagation, which are caused by the strong three-dimensional (3D) heterogeneity in mining areas. In this paper, we will introduce the rapidly developed waveform inversion-based location method into a mine MS field study. On this basis, the wavefields were modeled by utilizing a high-resolution 3D velocity model, a fractional-order Gaussian wavelet source-time function, and spectral element method (SEM) wavefield modeling. In order to reduce the computation cost of wavefield modeling, the 3D ray-shooting method based on a coarse grid was adopted to obtain an approximate MS event location. Based on this initial location, the multiscale waveform inversion strategy (from coarse to fine grids) and the L-BFGS iteration optimization algorithm were separately jointly selected to improve wavefield modeling speed efficiency and iterative convergence rate. Then, the IMS MS monitoring system set in the Yongshaba mine (China) and its tomographic 3D velocity model were used to conduct the synthetic test and application study. Results show that the source-time function based on the fractional-order Gaussian function wavelet can better fit complex recording waveforms compared with the conventional Ricker wavelet-based source-time function, and the waveform misfit during the L-BFGS iteration decreased rapidly. Furthermore, the multiscale waveform inversion method can obtain a similar location accuracy compared with the waveform inversion based on a single fine grid, and it can significantly decrease the iteration times and wavefield modeling computational cost. The average location error of the eight premeasured blasting events by the proposed approach is only 17.6 m, which can provide a good data research basis for analyzing MS event location and rock mass fracture characteristics in a mine.