Identification of Coals Using Terahertz Spectroscopy Combined with Manifold Learning and SVM Discriminant Analysis

LIANG LIANG, YU-XIN DU, ZHONG-WEN WU, ZI-JIAN YANG, HONG-YAN ZHANG

Abstract


The coal quality links to the efficiency and toxic gas emission of coal combustion, which has led to the increasing demand for quick and non-destructive detection method to identify various coal. In this study, terahertz spectroscopy combined with manifold learning algorithm and support vector machine (SVM) discriminant analysis was applied to analysis six types of coal materials. To evaluate the effectiveness of the proposed method, interval PLS (iPLS) and genetic algorithm combined PLS (GA-PLS) were used for spectral variable selection, principal component analysis (PCA) and stochastic neighbor embedding (SNE) algorithm were applied for spectral dimensional reduction. The experimental result showed that the SNE algorithm combined with SVM (SNE-SVM) has a higher correlation coefficient of prediction set (0.9842), a lower root mean squared error of prediction (0.2144), and the prediction accuracy of different coal materials reached 100%. This study indicates that the terahertz spectrum analysis combined with manifold learning algorithm is a promising method for the classification of different coals.

Keywords


THz-TDS, Pulverized coal, Manifold learning, SVM, Discriminant analysisText


DOI
10.12783/dteees/peems2019/33941

Full Text:

PDF

Refbacks

  • There are currently no refbacks.