Quantitative Analysis of Phosphorus in Compound Fertilizer by Support Vector Machine

Abstract

In the production of compound fertilizer, real-time detection of components was required to achieve the purpose of guiding production. In this paper, laser-induced breakdown spectroscopy technique combined with support vector machine (SVM)was used to quantitatively analyze phosphorus (P)in compound fertilizer. 58 compound fertilizer samples were provided by Anhui Huilong Group. Three spectral lines of P I: 213.5 nm, 214.9 nm, and 215.4 nm were selected as analytical lines. 58 samples were divided into a training set (43 samples)and a test set (15 samples)using the K-S method. By constructing the SVM of P element in compound fertilizer analysis model, the grid search method was used to optimize the parameters of quantitative analysis model. The results showed that the correlation coefficient R2 of the established training set calibration model was 0.981. It indicated the correlation between the reference value and the predicted value of the training set was relatively high. In the test set, the correlation coefficient R2 between the reference value and the predicted value of the sample phosphorus(P)element was 0.992, and the mean squared error CVmse was 4.951*10-3. It indicated that the constructed SVM model had strong applicability. The average absolute error of the training set was 0.059% and the relative error was 0.399%. The average absolute error and relative error of the test set were 0.056% and 0.328%, respectively. It meant that the accuracy of the LIBS measurement was greatly improved. This shows that combining the SVM algorithm with LIBS technology can achieve rapid detection of phosphorus in compound fertilizers.