Method For Modeling Near-infrared Spectra And Temperature Effect

Near-infrared (NIR) spectroscopy has been widely used due to fastness, accuracy and green. However, the technique needs to combine the chemometric methods for quantitative and qualitative analysis. Moreover, it is necessary to establish effective methods to extract informative variables from the high dimensional data of complex samples. NIR spectra are sensitive to external conditions, especially temperature. This limits the application of the technique. Therefore, in this dissertation, discrimination analysis of Chinese patent medicines was performed by using chemical pattern recognition. Variable selection method was proposed to build a robust quantitative model. The effect of temperature on the spectra was studied using multilevel analysis. Moreover, the effect of temperature on the structure of N-methylacetamide (NMA) in aqueous solution was investigated by multilevel simultaneous component analysis (MSCA). The main contexts are as follows:1. A method for the discrimination of Chinese patent medicines is developed. Discrimination of pharmaceutical products has been an important task in pharmaceutical industry and pharmaceutical safety. In this study, principal component accumulation (PCAcc) method combined with near-infrared spectroscopy is investigated for discrimination of Chinese patent medicines. In PCAcc method, an accumulation strategy is utilized to combine the classification information contained in multiple PC subspaces by using a rotation, a projection and a summation operation. To improve the performance of classification, continuous wavelet transform (CWT) is applied as the pretreatment method to eliminate the background. The results show that, among the12classes of Chinese patent medicines,8classes are correctly classified, and a total of ten samples are misclassified for the other four classes. Compared with the results obtained by principal component analysis (PCA), radial basis function artificial neural network (RBF-ANN) and partial least squares discriminant analysis (PLSDA), PCAcc produces the best classification. 2. A new method based on locally linear embedding (LLE) mapping for variable selection is proposed. LLE is a nonlinear dimensionality reduction method that can preserve the relationship between samples in the mapping space. The neighbors in high dimensional space will keep their relative position in LLE space. A method based on the effect of the variables on the relative position of the samples in LLE space is proposed for variable selection in NIR spectral analysis. In the method, the spectra are mapped into LLE space with all variables at first, and then the mapping is repeated by removing a variable from the spectra. Therefore, the movement of the samples in LLE space caused by a variable can be used to evaluate the effect of the variable on the spectra. The variables that cause a large movement will be the important ones to affect the relationship of the spectra. For further selection of the informative variables specific to the target component, a forward stepwise selection is applied to the variables selected by LLE method. To validate the performance of the proposed method, it is applied to the partial least squares (PLS) modeling of three NIR spectral datasets of corn, pharmaceutical tablets and tobacco lamina samples. Results show that the proposed method can effectively select the informative variables from the NIR spectra, and build a parsimonious model by using several tens of selected variables.3. Temperature dependent near-infrared spectra are studied using multilevel analysis. Quantitative spectra-temperature relationship (QSTR) between near-infrared (NIR) spectra and temperature has been used for quantitative determination of the compositions in mixtures. In this work, QSTR is studied using multilevel simultaneous component analysis (MSCA) and the spectral data of the samples with different concentrations measured at different temperatures. MSCA model contains a between-individual model describing the differences between the individuals and a within-individual model capturing the differences within the data of all the individuals. NIR spectra of five different compositions (water-ethanol-isopropanol) measured at seven temperatures are analyzed. A between-temperature model describing the effect of temperature and a within-temperature model describing the variation of concentration are obtained, from which QSTR model is established and quantitative analysis is achieved. Furthermore, the difference between the between-temperature or within-temperature models of different mixtures is used to study the composition of the solvent.4. Molecular interactions of N-methylacetamide in aqueous solution are analyzed using temperature dependent near-infrared spectroscopy. For investigating the effect of temperature on the interactions of N-methylacetamide (NMA) in aqueous solution, near-infrared (NIR) spectra of NMA aqueous solutions with different concentrations over a temperature range of30-80℃are measured, and MSCA is adopted to analyze the variation in the data induced by temperature. A between-temperature model is built. The loading and scores of the between-temperature model in MSCA describe the effect of temperature on hydrogen bonding in both water and NMA. The results demonstrated that an increase in temperature reduces the number of hydrogen bonds, making the longer-chain oligomers dissociate into monomer or smaller size oligomers. Compared with the results obtained by two-dimensional correlation spectroscopy (2DCOS) analysis, MSCA can capture the structural changes information of both water and NMA. However,2DCOS can only provide the dominant spectral changes of water.