Quantitative analysis of Cr in soil using laser induced breakdown spectroscopy combined with relevance vector machine method

Abstract

The laser induced breakdown spectroscopy (LIBS) technology coupled with a principal component analysis and relevance vector machine (PCA-RVM) was used for rapidly detect the concentration of heavy metal Cr in soil. The PCA-RVM was obtained by optimizing the relevance vector machine (RVM) with PCA.LIBS spectrum of 14 soil samples with different concentration of elements were collected and used for the model training and prediction. 10 of those were selected as training sample sets to build the PCA-RVM model, and the others as test sample sets for model evaluation. Comparing with the prediction results of PCA-RVM model with that obtained using RVM and support vector machines (SVM) model, the analytical accuracy improved by 84.17% to RVM and 92.62% to SVM at the maximum, respectively. Indicating that the PCA-RVM model can effectively improve the detection accuracy and repeatability of LIBS analysis of elemental concentration in soil.