A LIBS quantitative analysis method for samples with changing temperature via functional data analysis

Abstract

The accuracy of quantitative analysis of laser-induced breakdown spectroscopy (LIBS) will decrease if the temperatures of the testing samples are different from the temperature at which the calibration model is established. For the samples with changing temperature, it is not practical to build a calibration model for each temperature. Therefore, a novel LIBS quantitative analysis method for alloy steel samples with changing temperature via functional data analysis (FDA) is proposed, in which a conversion model for finding the non-linear relationship of LIBS spectra at varying sample temperatures is built by functional data analysis, and a calibration model with spectra at room temperature as the input and elemental concentrations as the output is established based on least squares support vector machines (LSSVMs). In the testing, the measured spectra at any temperature in the model range can be easily converted into their mappings at room temperature by the conversion model and used as the input of the calibration model to obtain the concentration results. Experiments on certified alloy steel standard samples were conducted, in which 5 samples at room temperature and at 6 sets of high temperature were used to build the calibration model and conversion model. Another 4 samples at 3 sets of high temperature were used as the testing samples. The experimental results of the Cr concentration show that, with the functional data analysis method, the relative errors are all below 10%, some even within 5%. With the proposed method, each testing sample does not have to be kept at the same temperature as that of the calibration model, which provides a feasible and effective way for LIBS analysis of samples at varying temperatures in iron and steel smelting production processes.