Source Type Classification Based On the Support Vector Machine Method

Abstract

Summary Attaining information of the source mechanism involved in micro-seismic events will greatly help us understand the reservoir fracturing and the stress evolved. The components of moment tensor can tell us the information involving magnitudes, modes, and orientations of fractures. Meanwhile, its singular value decomposition (SVD) exposes the difference between three main kinds of source types that may present itself in a moment tensor solution. We propose to use support vector machine (SVM), which is a type of machine learning approach, to classify the source type of a micro-seismic event by using the normalized eigenvalues of moment tensor matrix as classification principal components. The tests on moment tensor matrices based on typical source type and real cases yield reliable classification results.