Study On Spectra Data Analysis Methods For Laser-induced Breakdown Spectroscopy

Laser-induced breakdown spectroscopy?LIBS?has great potential for material analyses with its unique properties.The unique features of LIBS such as requiring little or no sample preparation,allowing remote detection,and rapid online multi-element analyses have contributed to its wide applications in many fields including industrial production,environmental monitoring,food safety monitoring,and space exploration.However,the laser-induced breakdown spectroscopy formation is a complicated process,subject to many factors such as noise,continuum background,spectral broadening,self-absorption and matrix effects.As a result,the analytical accuracy and precision of LIBS is limited.To further improve the analytical accuracy of LIBS,we studied three spectra data analysis methods,and the detailed contents are as follows:?1?Based on the theory of plasma emission spectroscopy,a peak recognition approach was proposed to recognize the spectral peak subject to severe spectral overlapping interference.The proposed approach recognizes the interfered spectral peak by using the correlation matrix.The proposed approach was applied in the analysis of LIBS spectrum of iron ore.The results showed that the the interfered spectral peak can be recognized effectively by the proposed approach.?2?A wavelet-based interference correction method was developed for processing LIBS spectra of steel samples with strong spectral interference.With the method developed in this study,the simultaneous correction of continuum backgrounds and spectral interferences have been realized without requiring any prior knowledge of the interference elements.The RMSECV values of Cr,Si,Ti,and Mn decreased from 0.0948,0.0199,0.0285,and 0.1979 wt%to 0.0295,0.0140,0.0183%,and 0.0558 wt%,respectively,using the method developed.Quantitative analyses of the referred four elements demonstrated that the interference from major or trace elements could be effectively corrected using the method developed with optimized wavelet function,decomposition level,and scaling factor?.?3?A hybrid sparse partial least squares?SPLS?and least-squares support vector machine?LS-SVM?model was proposed to improve the accuracy of iron ore analysis using LIBS in this study.SPLS was used for variables selection and establishing the multi-linear regression model between spectral data and concentrations.To deal with nonlinear self-absorption and matrix effects,LS-SVM was used to fit the residual errors of the SPLS regression model.Variables selection can be achieved with the hybrid model.Furthermore,the linear relationships and nonlinear effects in LIBS analysis were both taken into account by the hybrid model.With the hybrid model,the RMSEP values of TFe,SiO₂,Al₂O₃,CaO,and MgO were 0.6242,0.3569,0.0456,0.0962,and 0.2157 wt%,respectively.The results showed that the proposed spectra data analysis methods can contribute to the further improvement of analytical accuracy of LIBS.Data analysis is one of the key solution to the further improvement of analytical performance of LIBS.This study on spectra data analysis methods may contribute to the fundamental research and industrial application of LIBS.