Determination of carbon and sulfur content in coal by laser induced breakdown spectroscopy combined with kernel-based extreme learning machine

Abstract

The carbon and sulfur content is an important index of coal property. In the present work, kernel-based extreme learning machine (K-ELM) model was built and applied to laser induced breakdown spectroscopy (LIBS) to improve the quantitative analysis accuracy of carbon and sulfur in coal. The different preprocessing techniques, input variables and model parameters were optimized by 5-fold cross validation to find which combination can provide an appropriate calibration model. Then the optimized K-ELM model was applied to quantitative analysis of carbon and sulfur content in coal, and a comparison with support vector machine (SVM), least squares support vector machine (LS-SVM) and back propagating artificial neutral net (BP-ANN) was carried out. The three quantitative techniques were evaluated in terms of Root Mean Square Error (RMSE) and correlation coefficients (R2). The results show that K-ELM model has excellent performance compared to the others both in calibration and prediction set, and the optimum results of the K-ELM model were achieved with RMSE = 0.3762%, R2 = 0.9994 for C and RMSE = 0.7704%, R2 = 0.9832 for S in the prediction set. The overall results sufficiently demonstrate that LIBS coupled with K-ELM method has the potential to measure carbon and sulfur content in coal, and is a promising technique for real-time online, rapid analysis in coal industry.