Study On Rapid Characterization Of Compound Fertilizers And Textile Fibers Using Near Infrared Spectroscopy

Near infrared spectroscopy has been applied in agriculture, petrochemical industry and other areas, with its rapid analysis, easy operation, non-destruction and suitability for online analysis. This research focuses on the applications of NIR in compound fertilizers and textile fibers areas, for the purpose of solving technical problems with its rapidity and easiness.First, this thesis proposed a new approach to rapid determination of contents of total nitrogen, P2O5 and K2O and stone powder in compound fertilizers, and built PLS models of N, P and stone powder content with the SEP values of 0.8,0.8 and 1.4 respectively. The information on which stone powder content model is built is the spectrum of crystal water existed in stone powder. K content was calculated using other ingredient contents by normalization principle with a SEP value of 1.5. Then a new approach to rapid determination of compound fertilizer composition was introduced to improve the accuracy of analysis of inorganic constituents, namely first preparing aqueous solution of solid fertilizer, then predicting the compound composition using the near infrared transmission spectra of the solution. By extracting spectral information of components in the aqueous solution, models were built by means of PLS regression, and the SEP values of N, P, K and stone powder content are 0.5,0.7,0.8, and 2.0, respectively. This method solved the problem of determination of K2O content, achieved rapid analysis of all compositions of compound fertilizers. The results obtained by the new method are agreed with the national standards.In this thesis, a lot of cotton/spandex blended fabrics has been collected, and the PLS model of content of spandex using near infrared reflectance spectroscopy has been built which has a SEP value of 0.8 and prediction residual of 2%. The results show that the determination of spandex of cotton/spandex blended fabrics using near infrared spectroscopy can be realized in the laboratory. The new method is fast, accurate and efficient. It can meet the needs of actual work, and may greatly improve efficiency of routine composition analysis of cotton/spandex blended fabrics.The NIR reflectance spectra of four types of viscose fibers made from bamboo, flax, cotton and wood have been collected, and the identification of different type of viscose fibers was carried out using chemometric methods of PCA, SIMCA and LS-SVM. The results show that PCA has a clustering effect, but a clear distinction is still not reached; the results from SIMCA based on PCA are similar with those from PCA; LS-SVM classification method is satisfactory for discriminating bamboo fiber from other types of viscose fibers and its recognition rate is up to 90%. It is feasible to make rapid identification of types of viscose fibers using near infrared reflectance spectroscopy. This method has broad application prospects, which is very simple and less time-consuming.