Application Of Laser Induced Breakdown Spectroscopy Technology In Coal Industry

As one of the largest industrial waste production in our country, coal ash has become one of the major factors of the formation of fog haze. It seriously affected the ecological environment. The composition of coal ash determines the fouling tendency of the boiler heating surface and affects the safe operation of the boiler. Therefore, fast and accurate measurement of ash content can judge the slagging tendency of boiling coal, and is helpful for the recovery and reuse of coal ash. Laser induced breakdown spectroscopy, based on atomic emission technology, is a new kind of quantitative analysis method. it has the advantages such as pretreatment, fast, accurate and multi element analysis, it has been widely used in the coal industry and other fields.In the process of LIBS measurement, due to the influence of matrix effect and testing environment, large interferent information could exist in spectral signals inevitably, which seriously affected the qualitative and quantitative analysis of the sample. It can be effectively used to extract characteristic information and optimize spectral data by means of the method of stoichiometry, which is one of the effective ways to improve the accuracy of qualitative and quantitative analysis of LIBS. To solve the practical problems of recovery and reuse of coal ash in this paper, LIBS technique combined with chemometric methods were applied to construct different models for quantitative analysis and discriminant analysis of coal ash. This study enriches the application of LIBS technology, and provides a new method and technical support for the analysis of coal industry. The full text is divided into four chapters, the main research contents are as follows:At first, the linear regression model-(Partial least squares, PLS) was studied. In paper, before constructing the PLS model, the calibration set and prediction set were selected by the K-S, as well as the out samples were eliminated based on the markov distance combined with principle component analysis. The constructed PLS model is used to analysis the major components of coal ash, obtained better forecasting performance (correlation coefficient R>0.9000). Compared to traditional PLS model, the model which eliminated abnormal spectra is more accurate.Then, a kind of nonlinear model--wavelet neural network(WNN) was studied. In the paper, different methods were used select the input variable, and the various model parameters were optimized. The main composition of coal ash was analyzed by the optimal WNN model and the artifical neural network (ANN) model. Based on the evaluation of root mean square error and the correlation coefficient, it can be concluded that the wavelet neural network shows better prediction performance than that of ANN.Finally, Different classification methods for a variety of coal ash were researched. The spectral feature information was extracted by the method of independent component analysis, and WNN and BPNN models were constructed, respectively. Results show that the two kinds of samples can be classified (classification accuracy reached more than 95%), but the classification accuracy of WNN based on independent component is better.