Automatic selection of low-permeability sandstone acoustic emission feature parameters and its application in moisture identification

Abstract

Moisture is a vital factor in the structural stability of sandstone, which is the main component of low-permeability reservoir rocks. Hence, studies into moisture identification are crucial. Diverse information about rock, such as its structural and mechanical parameters, can be obtained from the acoustic emission (AE) signal. However, the types of AE parameters are varied, and the rock information that is represented by them is different. Traditional methods of parameter selection are mostly based on the correlation between parameters and the experience of researchers, which are not accurate when the correlation between parameters is fuzzy and does not meet automation requirements. In this study, a method of signal feature selection based on a data fluctuation rule and clustering analysis is proposed. This method takes the fluctuation law of the signal itself and the correlation degree of cluster labels as the basis, and the selection step is divided into two steps. An experimental platform is established, and uniaxial compression on sandstones with different moisture contents is carried out to verify the efficiency of this method. The selected feature parameters are used for moisture classification combined with a support vector machine (SVM) classifier, and the identification results verify the efficiency of energy security monitoring in low-permeability rocks.