Accuracy improvement of quantitative LIBS analysis of coal properties using a hybrid model based on a wavelet threshold de-noising and feature selection method.

A hybrid model based on a wavelet threshold de-noising (WTD) and recursive feature elimination with cross-validation (RFECV) method was proposed to improve the measurements in quantitative analysis of coal properties using laser-induced breakdown spectroscopy (LIBS). First, a modified threshold of WTD was proposed based on wavelet coefficient theory. Interference of noise in the LIBS spectrum was reduced by using this modified method. Then, the RFECV method was applied to extract effective features from the de-noised LIBS spectrum. Finally, support vector regression (SVR) models of coal properties were established by the selected features. A validation set was used to verify the effectiveness and robustness of the hybrid model. The improvement of the hybrid model on the quantitative analysis of each index of coal properties (heat value, ash, volatile content) was studied and discussed. By using the proposed model, the determination coefficient (), root mean square error of prediction, average relative error, and relative standard deviation were all significantly improved over the original spectra model. The results demonstrated that the proposed model could effectively improve the accuracy and precision of LIBS quantitative analysis for coal properties.