Chemometrics Methods Study On Near-Infrared Spectra Of Reformed Gasoline And Quantitaive Stucture-Activity Relationship Investigation Of Saffron Components

The first part of this paper was about using chemometics methods on near-infrared spectra as calibration models in predicting the distillation ranges and octane number of reformed gasoline. Robust partial least-squares algorithm (RPLS) and Wavelet transform-artificial neural network (WT-ANN) methods were first offered used on Near-Infrared (NIR) Spectra, which were proved to be better than the methods in common use.RPLS was first carried out on Near-Infrared (NTR) Spectra to determine distillation ranges and octane number of reformed gasoline. The key step was to confirm the initial sensitive factor. Performed RPLS model, and we obtained singularities and calibrated results of good samples at the initial sensitive factor. After adjusting appreciably the sensitive factor several times, the calibrated results of bad samples were also obtained. The standard error of calibration of distillation ranges and octane number were less than 0.0349. RPLS not only eliminated effectively singularities, but also was rapid and robust, and had good accuracy and repeatability. It will be an effective method in calibrating distillation ranges and octane number of reformed gasoline.WT-ANN method was offered using on Near-Infrared (NIR) Spectra to predict distillation ranges and octane number of reformed gasoline. Using the discrete wavelet transform as a preprocessing method in calibration modeling on NIR data, good compression was achieved with almost no loss of information. The calibration method used here was Gaussian potential function network (GPFN). Gaussian recursive least-squares (GRLS) algorithm was used to train the weights and Gaussian function parameters. Predictive results indicated that WT-ANN model had better accuracy than multiple linear regression (MLR) and partial least-squares (PLS) models.The second part of this paper was about quantum-chemical investigation of saffron molecules. Semi-empirical PM3 was carried out to compute structure parameters of large molecules of crocin and crocetin, such parameters as heat of formation, bond level, dipole and the potential energy surface. According the computing results, the comformational was analyzed and the structure-activity relationship of crocin-I and crocetin was investigated. The conclusions were in consistence with experimental results. Ab initio method at 6-31G level was used to optimize conformations and compute parameters of the series of picrocrocin and safranal. The stability and thermolysis mechanism of picrocrocin were discussed which was based on the analysis of the frontier orbital energies, charge distributions and potential energy surfaces. Quantitative structure-activity relationship of picrocrocin and safranal were also studied. It illuminated that the study of quantitative structure-activity relationship of natural medical molecules was valuable, which offered the theoretic foundation for drug design.