Study Of The Spectral Data Processing In Laser Induced Breakdown Spectroscopy Analysis And Its Application In Elemental Analysis Of Coal

Since the early 1960s, Laser-Induced Breakdown Spectroscopy (LIBS), as a rising atomic emission spectroscopy, has been researched as an element analysis technique in many fields because of its advantages of remote distance, in-situ and multi-element analytical capability. Recently, the data processing with the purpose of improving the measure accuracy and stability of LIBS analysis has become an important research subject and attracted lots of interests and attention. This work chooses the spectral data processing in LIBS elemental analysis and its application in coal analysis as the topic investigated, and aims at developing some new methods and models for the optimization of the spectral data processing, spectral analysis and elementary analysis of coal. The following investigations were made in this dissertation.Some data preprocessing approaches and models were introduced and verified for their validity by the corresponding example test. The "jump degree" method was used to discern and remove the abnormal spectral data in series of repeated measurement spectroscopic data. The analyzed spectral line profile was fitted by Lorentzian function for determining the spectral background, which was proved to be an effective method for the baseline correction of the spectrum. For the resolution of adjacent overlapped spectral lines, a model based on Fourier self-deconvolution combined with wavelet transfer was established and tested. The automatic peak-seeking algorithm as well as the qualitative analysis rules was founded for automatic spectral line identification. Also the wavelet analysis was used for the data compression of LIBS, and a high compression ratio and nice recovery coefficient was obtained by this method.The special attention was paid to the matrix effect and self-absorption effect, which were two common and negative phenomenons in the LIBS analysis. The existing correction methods proposed by the other researchers were presented and analyzed by some example experiments. And then, a kind of inter-element correction method was studied to reduce the influence of matrix effects on the calibration curves, and the experimental result showed that this method was more effective for correcting the matrix effect than internal standard method. As well, a correction model for emission line self-absorption, based on the information of line broadening in the analytical spectrum and the Stark broadening parameter, was introduced and used to calculate the degree of line self-absorption and then correct the line intensity. The results showed that this correction model was very effective to improve the LIBS calibrating measurement when the analytical spectral line was obviously self-absorbed in the optical thick plasma.Two existing LIBS quantitative analysis methods, the calibration curve analysis and free-calibration analysis, were introduced in detail. Also some experiments were carried out to analyze the feature and applications of these methods. Then, a new quantitative analysis method, based on the observed dependence of the linear correlation coefficient between spectra of sample analyzed and the spectrum of reference certified samples, was introduced and analyzed for its application. Beside, another new spectral analysis method based on neural networks was developed and used for the LIBS analysis, and its effectiveness was examined and certified by the elemental analysis of aluminum alloy standard samples.The experiment investigation about coal analysis by LIBS was executed at the typical LIBS setup. Some important experimental parameters, such as laser energy, the distance of lens to sample, delay time for spectrum acquisition and particle size of coal samples, were put to test for finding an optimal measurement condition. On this basic, the elemental components of coal were analyzed by LIBS. In this experimental investigation, the quantitative results of metallic elements and silicon were at a high confidence level with relative error between 0.7% and 9.2%, and the detection limits between 0.0013% and 0.0743% were obtained foe these elements. The fractionation effect of volatile matter and self-absorption of thick cool flame were considered to have negative effect on the calibration quantitative analysis of nonmetallic elements. Thus, the neural network method integrated the spectral line intensity and volatile matter content was used in the LIBS to analyze the nonmetallic elements of coal, and an acceptable result was gained. Beside, the carbon content in ash was analyzed by LIBS based on internal standard method, and the relative error was under 5%.Finally, the results of the whole research works were summarized and the directions for the further studies were suggested.