High-accuracy quantification of soil elements by laser-induced breakdown spectroscopy based on PCA-GS-ELM

Abstract

Laser-induced breakdown spectroscopy (LIBS) quantitative analysis is susceptible to matrix effects, especially in samples with significant differences in texture, such as soil and coal. Adding additional information such as the physicochemical properties of the sample and plasma images based on the original spectrum is an effective measure to reduce substrate effects. In this study, a new strategy to mitigate the impact of matrix effects and a high-accuracy quantification method for elements in soil by LIBS called PCA-GS-ELM are proposed. No additional equipment is required to obtain auxiliary information. Principal component analysis (PCA) is employed to extract spectral differences between different samples, and the differential spectrum is combined with the original spectrum to form the generalized spectra (GS), which is then input into the extreme learning machine (ELM) model. The model is trained to simultaneously focus on the element characteristic spectral lines and matrix differences between samples. In the experiment, a self-developed portable high-frequency LIBS is used. In the quantitative analysis of six major elements in 13 soil samples, the PCA-GS-ELM method has significantly improved accuracy. The RMSEP for Si, Al, Ca, Fe, Mg, and Ti is 0.946, 0.278, 0.394, 0.08, 0.169, and 0.034 wt%, respectively. The results demonstrate that the proposed generalized spectral method can mitigate matrix effects and enhance the performance of multivariate analysis methods.