Classification of Coal Origin of Fulvic Acid According to Spectroscopic and Chemometric analyses Coupled with Discriminant Analysis and Machine Learning Methods

Abstract

We investigated amodel to classify the coal origins of fulvic acid (FA) according to their chemical difference coupled with discriminant analysis(DA) and machine learning methods. Ten parameters involved Ash analysis, Ultraviolet-visible(UV-Vis), and fluorescence(FL) spectrometry of 16 FA samples derived from peat, lignite and weathered coal were record. By using Pearson correlation analysis, 8 parameters were filtered from the 10 parameters to classify the type of FA. A simple linear discriminant function employed the 8 parameters was established and verified with Leave-one-out Cross Validation(LOOCV). For further validation of FA classification model,we used traditional machine learning algorithm Support Vector Machine (SVM) and deep learning algorithm Convolutional Neural Networks (CNN) respectively to train the samples for testing the accuracy of classification. The results show all the samples with the 8 parameters were classified into three groups by SVM and CNN with near 100% accuracy. In all, based on the chemical differences among FA, the potential of parameters obtained from Ash analysis, UV-Vis, and FL coupled with DFhas been proven to be able to classify the coal origin of FA.