Quantitative Analysis Of LIBS Liquid Steel Components Based On Multivariate Regression

The steel industry is an important pillar industry of China's national economy. The traditional method of steel liquid detection has been unable to meet the needs of the current stage of the metallurgical level, so it is necessary to use the laser induced breakdown spectroscopy(LIBS) to detect the composition of steel.Using the LIBS technique in the steel liquid component quantitative analysis, a LIBS test system is set up in the laboratory environment, according to the experimental requirements, the selected experimental apparatus includes a light detecting system, a laser, an intermediate frequency furnace, etc. The interaction between the laser and the steel liquid sample is studied. A variety of multivariate quantitative analysis were studied,including multivariate linear quantitative analysis, multivariate nonlinear quantitative analysis,improved multivariate nonlinear quantitative analysis and the particle swarm optimization support vector machine(PSO-SVM) analysis method.Applying the multivariate regression to the LIBS system, the R2 increased from0.843 to 0.893, the RSD decreased from 23.5% to 15.1%, and the RMSE was decreased from 0.197% to 0.126%, by using multivariate linear quantitative analysis. After using the multivariate nonlinear quantitative analysis, compared with multivariate linear quantitative analysis, the R2 is increased from 0.893 to 0.910, RSD is decreased from15.1% to 10.3%, RMSE is changed from 0.126% to 0.091%. Finally, it makes a further improvement on the model of multivariate nonlinear quantitative analysis, the R2, RSD,RMSE are 0.918, 9.3% and 0.085% respectively. The forecast results obtained by PSOSVM quantitative analysis show that the algorithm has good generalization ability, the R2,RSD, RMSE are 0.937, 8.16% and 0.051% respectively. The results of the analysis show that the multivariate quantitative analysis can be well applied to the quantitative analysis of LIBS steel.