Stress Monitoring Potential of Ambient Noise Interferometry in Deep Mine Environments

Thursday, 17 December 2015
Poster Hall (Moscone South)
Philippe Dales1, Pascal Audet2, Jean-Philippe Mercier3, Willem de Beer4 and Andrei Pascu3, (1)University of Ottawa, Earth Sciences, Ottawa, ON, Canada, (2)University of Ottawa, Ottawa, ON, Canada, (3)Golder Associés Ltée, Montreal, QC, Canada, (4)Golder Associates Canada, Mississauga, ON, Canada
Understanding the response of the rock mass to mining is of key importance for the planning of mine operations as well as assessing and mitigating the seismic risk. For decades, studies have shown that passive source tomography, also called local earthquake tomography, can provide information on the rock mass response through the estimation of the temporal variation and 3D distribution (spatio-temporal variations) of stress. The spatio-temporal resolution afforded by passive source tomography depends on the seismicity rate and the location of microseismic events. In a mine, seismicity is not stationary, i.e. the locus and rate of seismicity vary with time, thus limiting the spatio-temporal resolution of this technique. Recent developments in the field of ambient noise seismic interferometry (Green’s function retrieval from ambient noise) provide hints that continuous recordings of ambient vibrations collected around mines could be used to obtain information on the evolution and 3D distribution of the stress in the rock mass by providing measures of seismic travel times between pairs of sensors. In contrast to passive source tomography that relies on the distribution of seismic events, the resolution afforded by ambient noise interferometry tomography depends solely on the locations of sensors and the frequency content of the ambient noise. We present preliminary results which focus on the temporal stability of the estimated Green's functions, the effect of mine infrastructure on signal quality and preliminary methods to quantify stress changes in the rock mass. In addition, we present the adopted processing scheme built on the Apache Spark engine and demonstrate its effectiveness in parallelizing the computationally intensive cross-correlation routines.