| Laser-induced breakdown spectroscopy (LIBS) is a newly developed method of atomic emission spectroscopy (AES). It has several advantages, such as in situ, online, fast and remote sensing without any preparation and so on. Qualitative and quantitative analysis are carried out by monitoring the wavelength and intensity of emission lines. In order to quickly select the suitable analytical lines from the emission lines and improve the precision of quantitative analysis of LIBS, it is essential to propose an automatic and efficient line selection method.The related theory of LIBS, including the system structure, the basic theory of plasma, quantitative and qualitative analysis method is summarized. Then we analyze the main factors affecting the spectral quality (spectral overlap, self-absorption, electromagnetic interference, etc.) and the corresponding impact on results. An automatic line selection method is prompted. Two parameters, Wavelength Difference of Detected and Theory (WDDT) and the Relative Detected-to-Theory Intensity Ratio (RDTIR), are defined. By setting reasonable thresholds of RDTIR and WDDT, we can automatically exclude spectral lines that seriously disturbed by self-absorption and spectral interference.Five kinds of high-alloy steels and six kinds of the standard stainless are examined to prove the efficiency and versatility of the method. The analytical lines of iron (Fe), chromium (Cr), nickel (Ni), manganese (Mn), copper (Cu) for each sample are reserved by automatic selection process, which coincide with the results acquired by traditional method. Cr (I) 524.76nm and Fe (I) 324.72nm from the results are used for quantitative analysis of high-alloy steels, and the linear fitting coefficient R2 between the detected intensities and concentrations are 0.97 and 0.89 respectively, while the relative errors of analysis are 6% and 17% respectively. For Standard stainless steel sample, Fe(I)295.69nm, Mn(I)482.35nm and Cr(I)470.8nm from the selected lines are used for analysis, and the linear fitting coefficient R2 are 0.93, 0.98 and 0.89 respectively, while the relative errors of analysis are 7%, 13% and 9% respectively. Finally, using analytical line Mn(I)482.352nm selected by our algorithm, we analyzed the test samples quantitatively by means of external standard method. The results show that the root mean square error (RMSE) is 0.118%. Compared with the results obtained by manual selection line (Mn(I)403.076nm), the RMSE decrease about 4% for the same samples. Therefore, the application of automatic line selection algorithm can help to improve the accuracy of quantitative analysis. Automatic selection of analytical lines proposed by this study has resolved the problem of complex and low efficiency in process of line selecting, and also improved the precision of LIBS online analysis. |
PDF Full Text Request