【作者】 王雁飞；

【导师】 张桂荣；

【作者基本信息】 吉林大学 ， 生物工程， 2010， 硕士

【摘要】 随着现代分析仪器和计算机科学技术的快速发展,多形式化的近红外光谱采集模式开发,结合日益发展的化学计量学,近红外光谱技术越来越受到人们的关注。然而我国近红外光谱技术发展起步较晚,发展空间还很大,目前尚未成为国家药典规定的标准化方法,因此需要加大力度,对近红外外光谱分析技术在各领域的应用研究,推进近红外光谱技术的快速发展,发挥其强大的应用潜力。本文研究了采用偏最小二乘法(PLS)结合近红外漫反射光谱建立了测定八角茴香中八角茴香油含量的定量分析模型。对建模过程进行了深入研究,分别采用了蒙特卡罗偏最小二乘法(MCPLS)法识别异常样本,分别采用Savitzky-Golay平滑法、傅里叶变换(FFT)、一阶导数光谱法、二阶导数光谱法和标准正态转换法(SNV)对所有样品的近红外漫反射光谱进行预处理,采用可移动窗口偏最小二乘法(MWPLS)筛选波长变量和选择模型的最适主因子数等步骤对模型进行了优化,最后建立八角茴香中八角茴香油含量的最优偏最小二乘法模型,采用最优的模型对校正集样本和预测集样本中的八角茴香油含量进行预测,模型的校正均方根误差(RMSEC)为0.215,而预测均方根误差(RMSEP)为0.350,校正集样本八角茴香油含量的模型预测值与参考值间的相关系数(Rc)为0.9955,而预测集样本八角茴香油含量的模型预测值与参考值间的相关系数(Rp)为0.9890,结果表明该方法准确可靠,重现性、稳定性均良好,有望成为药物常规检测方法之一。更多还原

【Abstract】 The near infrared spectroscopy (NIRS) analysis technology is so popular in chemical analysis since the rapid development of computer technology, chemometrics and electronics. As the near infrared spectroscopy analysis technology is based on the frequency of organic media’s chemical bond, the absorbance is weak and the spectroscopy is overlapped and complex. A powerful calibration method is needed to parse the near infrared spectra. Chemometrics is much suitable. However, the calibration of chemometrics is complex. It must be assisted with the computer technology. As near infrared spectroscopy analysis technology is nondestructive, rapid, low cost and muti-component analysis, it is greatly interested by researchers. However, near infrared spectroscopy technology is a indirect technology. The application of this technology is based on developing a credible calibration models. Therefore, the development of calibration model is a key process and this process is much complex. In this paper, the process of developing the calibration model for determining of the content of anise Oil in Illicium verum Hook. f. is investigated. In this paper, the background of Illicium verum Hook. f., anise oil, near infrared spectroscopy and chemometrics has been reviewed in the first chapter. In the second chapter, partial least square (PLS) was applied in modeling the relationship between near infrared spectra and the contents of anise oil in Illicium verum Hook. f.. The modeling processes has been Further studied. First, the initial partial least square model has been developed using original near infrared spectra and leave-one-cross-validation method has been employed to select the initial number of factors.During this process, the root mean square error of cross-validation (RMSECV) was employed as criterion. Second, Monte Carlo partial least square (MCPLS) was employed to recognize the outliers. The samples were divided into two sample sets depends their contents of anise oil and the messages of the spectra. Average smoothing method, Savitzky-Golay smoothing method, fast fourier transform (FFT), first order derivative, second order derivative and standard normalize transfer (SNV) were used for preprocessing the spectra respectively. The original spectra and the preprocessed spectra were used for modeling respectively. Each model has been optimized by selecting the effective wavelength variables using moving window partial least square method (MWPLS) and selecting the most suitable number of principal factors depending on RMSECV. And then, the optimum model has been developed. The contents of anise oil in Illicium verum Hook. f. has been predicted by the optimum model. The root mean square error of calibration set (RMSEC) was 0.215, and the root mean square error of prediction set (RMSEP) was 0.350. The coefficient between the predictive contents of anise oil and the reference contents in calibration set (Rc) was 0.99955, and coefficient between the predictive contents of anise oil and the reference contents in prediction set (Rp) was 0.9890. These results demonstrated that this method was precise, stabilized and reproducible. It will be popular in the pharmacy. A comprehensive summarization of this research has been presented in the last chapter.更多还原