A step towards the prediction of a rock collapse: analysis of micro-acoustic bursts

Abstract

Forecasting collapses in a rock face is still an unresolved issue due to the lack of clearly noticeable forerunners. However, technological advances carried out in our research group have made possible the acquisition of micro-acoustic emissions induced by the enlargement of cracks in the rock. At the same time, both the geological literature and evidence propose an evolutionary model for the burst spectrum, which is expected to shrink towards low frequencies once getting closer to the rock collapse. Bursts, acquired through accelerometers sampled at 2kHz, need to be suitably processed to remove outliers and false positives, before any forecast action can be envisaged. For the first time we are in possess of a dataset of micro-acoustic emissions acquired by our systems on the Alps. These signals have been analyzed through a computational intelligence approach, to interpret and classify them in bursts associated with proper fractures in the rock or false alarms (e.g., due to stone falls, and outliers).