Application Of Partial Least Squares Method In Non-destructed Quantitative Analysis With Near Infrared Spectroscopy

With the rapid development of computation technique and chemometrics, the analytical problems which are aroused by the poor effected information in the near infrared spectroscopy (NIRS), difficulties for regression because of overlapped and conlinear spectra have been solved. Near infrared spectroscopy has become one of the fast (rapid) developments and most mentioned analytical technologies in this century. In this paper, the application of partial least squares regression (PLSR) method for the relationship between the NIR spectra and the content of paracetamol in antivirus drug (compound paracetamol and amantadine tablets) has been studied, the relationship between the NIR spectra and the rudimental shikimic acid in Illicium verum Hook.f. also has been studied. The results demonstrated that this method was good at extracting the effect information from the NIR spectra and parsed the problem of overlapped and conlinear spectra. With its advantages of no pretreatment, no pollution, facility, non-destructibility and online measurement, it has been taken as an appealing tool for solving all sorts of analytical problems, for all sorts of samples, such as liquid, dense, coat, powder and solid. In recent years, with the developing of chemometrics, optics, electronics and computer science, NIRS is being applied to numerous fields, including agriculture, food, chemical, pharmaceuticals, textiles, polymers, cosmetics, tobacco and medical.1. The application of near infrared spectra technique with partial least squares regression in determining the content of paracetamol in compound Paracetamol and Amantadine Hydrochlorde: Applied data pretreatment approaches including Savitzky-Golay smoothing, Fast Fourier Transfer(FFT), first derivative and second derivative to dispose the NIR spectra of compound paracetamol and amantadine tablets, and introduced a novel method using the RSD of the spectra and the correlation coefficient between the absorbance at every wavelength and the contents of the paracetamol in the samples to choose the optimal wavelength regions. The original spectra and every pretreated spectrum at each spectra region were applied to develop the quantitative analysis models for determining the content of the paracetamol in compound paracetamol and amantadine tablets. Suitable numbers of the factor for he developed models were selected depended on the root mean squares error of calibration set by cross-validation method (RMSECV), and then, the optimum models were selected according as RMSECV, root mean squares error of calibration set (RMSEC), the correlation coefficient between the precicted values by cross-validation and actual values (Rv), the correlation coefficient between the precicted values and actual values (Rc)and the root mean square error of predictive set (RMSEP). The optimum model that for determining the content of the paracetamol in compound paracetamol and amantadine tablets applied the second derivative spectra with the region of 1100～2121 nm, its suitable number of factor were 2, its suitable floating window is 50, RMSECV were 0.00577, RMSEC were 0.00506, Rc were 0.98520, Rvwere 0.98089, RMSEP were 0.00416. These results indicate that this method is feasible, rapid, non-pretreated, and non-destructed, it have great application foreground.2. The application of near infrared spectra technique with partial least squares regression in determining the contents of rudimental shikimic acid in Illicium verum Hook.f.: Applied data pretreatment approaches including Savitzky-Golay smoothing, Fast Fourier Transfer(FFT), first derivative and second derivative to dispose the NIR spectra of Illicium verum Hook.f.. The original spectra and every pretreated spectrum at each spectral region were applied to develop the quantitative analysis models for determination the contents of the rudimental shikimic acid in Illicium verum Hook.f.. Suitable numbers of the factor for he developed models were selected depended on the root mean squares error of calibration set by cross-validation method (RMSECV) , and then, the optimum models were selected according as according as RMSECV, root mean squares error of calibration set (RMSEC), the correlation coefficient between the precicted values by cross-validation and actual values (R_v), the correlation coefficient between the precicted values and actual values (R_c)and the root mean square error of predictive set (RMSEP). The optimum model that for determining the content of rudimental shikimic acid in Illicium verum Hook.f. applied the FFT spectra with the region of 1100～2500 nm, its suitable floating window is 10, RMSECV were 0.28488, RMSEC were 0.0.22400, Rc were 0.94895, Rvwere 0.91740, RMSEP were 0.27300. These results indicate that this method is feasible, rapid, non-pretreated, and non-destructed, it have great application foreground.In this paper, I studied on extracting information and pretreatment of NIR spectra, applied the PLS to construct a quantitative determination model to predict the content of the paracetamol in compound paracetamol and the contents of the rudimental shikimic acid in Illicium verum Hook.f.. It's not necessary to pretreat the samples in chemical means. PLS can analyze the overlapped NIR spectra well, it also can reduce interfere of the component. It can be generalized to on-line and real-time quality control for Chinese traditional medicine.