Accuracy improvement for multimetal analysis by laser-induced breakdown spectroscopy with least squares support vector machine

An efficient method has been developed to analyze metal species by using laser-induced breakdown spectroscopy (LIBS). A least squares support vector machine (LSSVM) was applied to quantitative analysis of multimetal samples with different matrix content, as LSSVM can select the input variables and convert the mapping to the feature space to obtain the linear performance. In this work, three kinds of metal species (alloy steel, ferrochromium alloy, and ferromanganese alloy) were used to construct the LSSVM model. The correlation coefficients (R-2), the root-mean-square error of prediction (RMSEP), and the average relative error (ARE) were calculated according to the estimated concentrations of chromium (Cr), manganese (Mn), titanium (Ti), and molubdenum (Mo) using the LSSVM model. The results showed that RMSEP decreased from 1.7047%, 1.3453%, 1.9253%, and 2.9431% to 0.0747%, 0.0942%, 0.0192%, and 0.0809%. The ARE of Cr, Mn, Ti, and Mo decreased from 10.2370, 9.8276%, 13.2460%, and 16.2386% to 1.0179%, 1.3937%, 2.1228%, and 1.6257%, respectively. This study demonstrated that LSSVM is an effective algorithm to improve both the accuracy and stability for metal analysis using LIBS.