Sulfur wavelength and concentration analysis via preprocessing with laser-induced breakdown spectroscopy measurement data

Abstract

Laser-induced breakdown spectroscopy (LIBS) was used to measure the sulfur concentration inside coal. LIBS has the advantage of not requiring sample pretreatment and being sensitive to low-intensity elements. There are several representative sulfur wavelengths within the wavelength range observed by LIBS; however, there are limitations to sulfur analysis because the intensity of the sulfur emission line in the main measurement wavelength range is weak. To improve the reliability of sulfur concentration analysis, five representative sulfur wavelengths were selected, and the intensity of each wavelength was compared. To create an environment similar to coal used in actual power plants, samples were prepared by mixing coals with known compositions. This was compared with the data obtained through LIBS measurements and expressed as a partial least-squares regression (PLSR) model. Additionally, data preprocessing was performed to better analyze the weak sulfur emission wavelength. Using two preprocessing methods, Savitzky-Golay (SG) smoothing and an SG derivative, two additional PLSR models were developed for the mixed-coal composition relationship obtained via LIBS measurements and industrial analysis. The reliability of the three PLSR models was compared by selecting the R2 and root mean square error as variables.