Exploring the Use of a Time-Frequency Domain Technique for the Filtering of Acoustic Emission/Microseismic Data

The performance of acoustic emission/microseismic (AE/MS) monitoring systems in mines is greatly impacted by background noise interference. Mine trucks/vehicles, fans, blasting and electrical spikes generate noise which negatively affect AE/MS signal quality. Using such datasets without filtering may lead to inaccurate determination of first arrivals and hence wrong source location. Also, data contaminated with noise makes manual identification and picking of the first arrivals impossible or difficult to achieve. This paper explored the use of stationary discrete wavelet transform as a filter to augment noise reduction on mine AE/MS data. The output of a test run on both noisy synthetic and field data showed cleaner AE/MS signal quality and clear emergence of first arrivals. As a result, manual picking of the first arrivals were much easier and faster. The use of the technique in this study presents an opportunity for picking accurately the first arrivals manually or automatically. Reliable first arrival picks lead to accurate computation of event source location and therefore, better safety monitoring and emergency responses in the event of an underground failure.