Study On Rapid Quantitative Analysis Of Natural Fiber By Using Near Infrared Spectroscopy

Natural fiber is an important material in textile industry. In recent years, facing huge demands of relevant adjustment and optimization of industrial structure, reducing cost for enterprises, and ensure quality of products, fast or online detection of fiber need to addressed. But on application of NIR online analysis of fiber, the influence of light scattering and the change of sample moisture caused by humidity effect predicted performance of quantitative model, thus affecting measurement accuracy and precision. In addition, fiber market still lacks some corresponding on-site and on-line analysis. Therefore, this paper studied the effect on scoured cashmere spectra by moisture and light scatting, spectral processing method, establishment rapid analysis method of cashmere content of scoured cashmere, development of an on-line analysis system of natural cellulose pulp and establishment of on-line analysis method.?1? This paper studied the influence of moisture to spectra. The study indicated the spectra of the samples were badly influenced by moisture. A subtraction H₂O method based on vector-subspace angle criterion algorithm, was proposed to decrease the effect of moisture on the NIR spectra by subtracting spectra of water from the spectra of scoured cashmere samples. Combined cashmere content with raw spectra and pretreatment spectra, two quantitative models were developed by partial least squares ?PLS? method, respectively. The correlation coefficient of predictive model with no subtraction H₂O method is 0.87, and root mean square prediction error ?SEP? is 8.53%. The correlation coefficient of predictive model with subtraction H₂O method is 0.94, and SEP is 5.28%. The results showed subtraction H₂O method can improve the predict ability of model.?2? The characteristic of NIR diffused reflectance spectra of cashmere and guard hair was studied. The results indicated The spectral difference of cashmere and guard hair is tiny, mainly comes from light scattering effect of physical properties such as the diameter. Noise, caused by light scattering effect, is eliminated effective by multiple scattering correction ?MSC?, but some useful information for determining cashmere content is eliminated too. A spectral reconstruction method was designed to take full advantage of chemical and physical information of cashmere in this paper. PCA was applied to the spectral residues resulted from multiplicative scatter correction and then the spectrum of the principal component which was highly correlated to cashmere content was added into the spectral signal. The new spectra and cashmere content are used to build calibration model by PLS method. The correlation coefficient of predictive model is 0.92, and SEP is 6.10.?3? Both the two methods can improve the prediction performance of model, and their combination can achieve the best model with SEP of 5.19%, which can satisfy the accuracy requirement for measuring cashmere content.?4? A new online NIR instrument is developed for analyzing natural cellulose pulp quickly.86 pulp samples were collected. Their NIR diffused reflectance spectra were obtained by new online NIR instrument. The spectra are preprocessed by S-G derivative, mean centralized and MSC method. Combined NIR spectra with a-cellulose content and DP, quantitative models for a-cellulose and DP are developed by partial least squares ?PLS? method, respective. The correlation coefficient of predictive models are 0.89 and 0.98, and SEP are 0.94 and 25.1. The models have good repeatability. The study result indicated online detection of natural cellulose pulp in continuous production process can be come true.