SUPPORT VECTOR MACHINES; ELEMENTAL ANALYSIS; IRON-ORES; CLASSIFICATION; SPECTROMETRY; SAMPLES; PLASMA; INDUSTRY; ALLOYS; MODEL
The basicity of sintered ore, which is related to the melting point of the sinter, is vital to ore mining and blastfurnace smelting. Laser-induced breakdownspectroscopy (LIBS) with random forest regression (RFR) has been applied for measuring the basicity of sintered ore, which can be defined by the concentrations of oxides: CaO, SiO2, Al2O3 and MgO. In this work, thirty sintered ore samples are used, of which twenty samples are used for the calibration set to construct the random forest regression (RFR) calibration model for the above-mentioned oxides and ten samples are used for the test set. The characteristic lines of the main components in the sintered ore are identified using the National Institute of Standards and Technology (NIST) database. Two model parameters (the number of decision trees -ntree and the number of random variables - m(try)) of the RFR were optimized by out-of-bag (OOB) error estimation for improving the predictive accuracy of the RFR model. The RFR model was applied to sample measurements and the results were compared with partial least squares regression (PLSR) models. The RFR model has shown better predictive capabilities than the PLSR model. In order to verify the stability of the RFR model, fifty measurements were made and the relative standard deviation (RSD) of the data is between 0.27% and 0.59%. Therefore, LIBS combined with RFR could be a promising method for realtime online, rapid analysis in mining and mineral processing industries.