Research On The Method Of Component Analysis By Laser-induced Breakdown Spectroscopy

Laser-Induced Breakdown Spectroscopy(LIBS) is a recently developed form of analysis used identify elemental constituents of virtually any type of sample. LIBS has great potential in many applications because of its unique advantages, such as needless sample preparation, real-time on-line analysis, multiple elements detection, and so on. But the analysis methods of LIBS can not meet actual applied demand at present, especially its measurement accuracy is much lower than traditional spectroscopy technique.Research aimed at developing appropriate analytic methods to promote LIBS applications. Some key problems on both qualitative and quantitative analysis were studied, and the emphasis is the research of quantitative analysis methods. The following works had been finished in this dissertation.1)A LIBS experimental setups,which can obtain the spectrum and image of plasma simultaneously, were designed and built. The influence of several key experimental parameters was studied and the parameters were optimized.2)A unsupervised LIBS classification method based on principal component analysis and self-organization mapping neural network was proposed, and applied in fast lithological identification successfully.3)The continuum emission from laser-induced plasma results in unexpected background in LIBS. A dynamic background correction method was proposed to reduce the effect of background to signal. This method was examinated by a serial experiments. The experimental results show that the determination coefficients of calibration curves can be improved by using this background correction method.4)A new spectral normalization method, which takes the intensity of the plasma image as reference signal, was proposed to reduce the fluctuation of spectral line intensity. This method was compared with conventional normalization method that use the whole spectrum intensity as reference signal, the experimental results show that this new method do better in improving signal stability.5)Most existing quantitative models are based on the hypothesis of local thermal equilibrium, which is not satisfied in most cases actually. A new multivariable nonlinear model was proposed, which is insensitive to the plasma temperature. So it does not require the LTE assumption was valid. The analysis results of alloy steel show that the accuracy and repeatability of the measurement results are both better than those of conventional internal calibration method.6) The basic partial least square(PLS) model was improved to reduce interelements effects. The improved PLS quantitative model considers the interference between various elements. Two PLS models were used in slag quantitative analysis respectively. The results show that measurement errors of four components in slag using improved PLS model are all less that those using in basic PLS model.The works in this dissertation have developed various qualitative and quantitative analysis methods, this promote LIBS application development.