Automatic P-wave Picking of Microseismic Events in Underground Mines

Loading...
Thumbnail Image

Authors

Johnson, Stephanie

Date

2014-05-01

Type

thesis

Language

eng

Keyword

Mining , Seismology

Research Projects

Organizational Units

Journal Issue

Alternative Title

Abstract

This thesis investigates microseismic P-wave arrival time detection performance of automatic picking algorithms, as well as the handpicking performance of human experts. The data set used in this project was collected from Malmberget mine (LKAB, Sweden) and the handpicked P-wave arrivals were prepared by multiple expert analysts from the Institute of Mine Seismology (IMS). Characterization of the event records in the data set was completed including the magnitude distribution of the events, noise content of the traces, and frequency spectrum of the traces. Three promising automatic P-wave picking algorithms from previous seismological research were investigated: the short-term average to long-term average ratio detector (STA/LTA), the characteristic function detector (CF), and the autoregressive modelling detector (ARfpe). Several versions of each algorithm were implemented, and the most promising versions were tested on the full dataset of microseismic events. The STA/LTA algorithm and CF algorithm were superior to the ARfpe algorithm in terms of accuracy and percentage of false negatives (missed P-wave arrival time picks). The analyst P-wave arrival times were compared and statistical distributions of the analyst P-wave arrival time differences were studied. The analyst P-wave arrival time difference and algorithm P-wave arrival time difference were defined as the mean analyst P-wave arrival time minus the specific analyst P-wave arrival time pick or the specific algorithm P-wave arrival time pick. The analyst and algorithm P-wave arrival time differences were combined into separate statistical distributions and compared. The analyst P-wave arrival time distribution lengths varied by a factor of 5, and the percentage of outliers in the distribution varied between 12% and 32%. The STA/LTA algorithm had comparable distribution statistics to the worst analyst P-wave arrival in terms of median value, distribution length, and percentage of outliers. However when the number of traces with automatic P-wave picks within the analyst handpicking range was calculated the STA/LTA algorithm had only 32.0% of picks and the CF algorithm had only 11.6% of picks.

Description

Thesis (Master, Mining Engineering) -- Queen's University, 2014-04-30 21:45:13.741

Citation

Publisher

License

This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.

Journal

Volume

Issue

PubMed ID

External DOI

ISSN

EISSN