Mine development imposes stress changes in the surrounding rock that typically induce or trigger seismicity with a wide range of magnitudes. Seismic monitoring can provide insight into the rock deformation and give critical feedback to the on-going operations. We have developed a passive seismic imaging algorithm based on earthquake seismology to jointly locate induced microseismic events and update the velocity of the rock model illuminated by the seismicity. We calculate travel-time based on the fast sweeping method to account for complex 3D distribution of velocity and use the adjoint method to transform the inverse problem to a forward problem which can also be solved by the fast sweeping method. In this paper, we demonstrate the efficiency and robustness of the fast sweeping and adjoint method based travel-time tomography in monitoring fracture evolution and dynamic velocity variations.
During the development of a mine site, monitoring of induced seismicity is generally used to provide insight into the rock fracturing and deformation processes and to give critical feedback to the on-going operations. Passive microseismic (MS) monitoring applications are often restricted to imaging induced fractures using source locations assuming a velocity model and ignoring the time-lapse variation of velocity with complex structures. We have adapted the seismic tomography technique from earthquake seismology (e.g. Thurber, 1983) and developed a passive seismic inverse algorithm to jointly locate induced MS events and update rock velocity model traveled by the ray paths. Honoring the coupling between the event location and the velocity model, passive imaging can provide a more realistic image of the rock velocity structure, its time-lapse variation and more accurately locate events. Recent advancement includes the double-difference tomography using S-P times in addition to both absolute and relative P- and S-wave arrival times (Zhang et al., 2009), and velocity tomography only using the absolute arrival time of P- and S-wave to derive 3D vp, vp/vs as well as quality factor QP (Tselentis et al., 2011). However, both methods calculate the travel times using conventional ray tracing which cannot treat complex velocity shapes and arbitrary discontinuities commonly observed in mines. We propose to use the fast sweeping method (Zhao, 2004) to account for complex 3D distribution of velocity. Furthermore, poorly covered region included in the model introduces additional unstable issues to the conventional linearized travel-time inversion. Damped least square method can stabilize the inversion but scarifies the model resolution. The adjoint method-based tomography (Huang & Bellefleur, 2012) formulates an inverse problem as a forward problem and only updates the region with adequate coverage. In the following section, we first briefly introduce the background of the fast sweeping and the adjoint method. We then demonstrate the performance of the fast sweeping and adjoint method-based travel-time tomography in locating MS events and recovering velocity variations using synthetic data based on field well logs and a sample of the seismicity induced during the caving development of Lift 2 of Northparkes mine. We demonstrate that the fracture evolution and dynamic velocity variations at Northparkes mine is well detected by the tomographic velocity model.