In this paper, a double artificial neural network （DANN） algorithm was used to parse near infrared （NIR） reflectance spectrum of Cofrel medicines. The contents of benproperine phosphate, which is the effective ing...In this paper, a double artificial neural network （DANN） algorithm was used to parse near infrared （NIR） reflectance spectrum of Cofrel medicines. The contents of benproperine phosphate, which is the effective ingredient in Cofrel medicines, were accurately nondestructive quantitatively predicted. Compared the results with those of HPLC, the relative errors （RE %） were less than 0.18%. The analytical results could be applied to qualitative control of Cofrel medicines.展开更多
Liquid state methanol and ethanol under different temperatures have been investigated by FT-NIR(Fourier transform nearinfrared) spectroscopy,generalized two-dimensional(2D) correlation spectroscopy,and PCA(principal c...Liquid state methanol and ethanol under different temperatures have been investigated by FT-NIR(Fourier transform nearinfrared) spectroscopy,generalized two-dimensional(2D) correlation spectroscopy,and PCA(principal component analysis) . First,the FT-NIR spectra were measured over a temperature range of 30-64(or 30-71) °C,and then the 2D correlation spectra were computed.Combining near-infrared spectroscopy,generalized 2D correlation spectroscopy,and references,we analyzed the molecular structures(especially the hydrogen bond) of methanol and ethanol,and performed the NIR band assignments. The PCA method was employed to verify the results of the 2D analysis.This study will be helpful to the understanding of these reagents.展开更多
研究了偏最小二乘(partial least squares,PLS)与广义回归神经网络(generalized regression neuralnetworks,GRNN)联用在近红外光谱多组分定量分析中的应用。以饲料样品为实验材料,采用PLS-GRNN法建立了饲料中水溶性氯化物、粗纤维、脂...研究了偏最小二乘(partial least squares,PLS)与广义回归神经网络(generalized regression neuralnetworks,GRNN)联用在近红外光谱多组分定量分析中的应用。以饲料样品为实验材料,采用PLS-GRNN法建立了饲料中水溶性氯化物、粗纤维、脂肪三项组分含量近红外光谱定量分析模型。马氏距离法剔除强影响点和奇异点,用PLS法将原始数据压缩为主成分,取8个主成分吸收峰与4个原始图谱特征峰值输入GRNN网络,网络光滑因子σi为0.1。PLS-GRNN模型对样品3个组分含量的预测决定系数(r2)分别为:0.9840,0.9870,0.9830;样品平行扫描光谱预测值的标准偏差分别为:0.00326,0.0655,0.0314。结果表明所建PLS-GRNN模型通过近红外光谱能够准确预测饲料中水溶性氯化物、粗纤维、脂肪三项组分含量,为近红外光谱进行多组分定量分析提供了新思路,同时为解决近红外快速检测技术在预测组分含量较低的样品时误差相对较大的问题提供了可靠的方法。展开更多
利用近红外和拉曼光谱法定量分析了甲醇汽油中甲醇的含量,采用偏最小二乘法(partial least squares,PLS)建立甲醇的定量模型.近红外光谱法测定甲醇定量模型的预测集相关系数RP为0.998,预测均方根误差(RMSEP)为0.289%;拉曼光谱法测定甲...利用近红外和拉曼光谱法定量分析了甲醇汽油中甲醇的含量,采用偏最小二乘法(partial least squares,PLS)建立甲醇的定量模型.近红外光谱法测定甲醇定量模型的预测集相关系数RP为0.998,预测均方根误差(RMSEP)为0.289%;拉曼光谱法测定甲醇定量模型的预测集相关系数RP为0.982,预测均方根误差(RMSEP)为1.141%.实验表明,近红外与拉曼光谱技术均可用于甲醇汽油中甲醇含量的快速检测.展开更多
文摘In this paper, a double artificial neural network （DANN） algorithm was used to parse near infrared （NIR） reflectance spectrum of Cofrel medicines. The contents of benproperine phosphate, which is the effective ingredient in Cofrel medicines, were accurately nondestructive quantitatively predicted. Compared the results with those of HPLC, the relative errors （RE %） were less than 0.18%. The analytical results could be applied to qualitative control of Cofrel medicines.
基金supported by the Medical Scientific Research Foundation of Guangdong Province,China(B2009043)
文摘Liquid state methanol and ethanol under different temperatures have been investigated by FT-NIR(Fourier transform nearinfrared) spectroscopy,generalized two-dimensional(2D) correlation spectroscopy,and PCA(principal component analysis) . First,the FT-NIR spectra were measured over a temperature range of 30-64(or 30-71) °C,and then the 2D correlation spectra were computed.Combining near-infrared spectroscopy,generalized 2D correlation spectroscopy,and references,we analyzed the molecular structures(especially the hydrogen bond) of methanol and ethanol,and performed the NIR band assignments. The PCA method was employed to verify the results of the 2D analysis.This study will be helpful to the understanding of these reagents.
文摘研究了偏最小二乘(partial least squares,PLS)与广义回归神经网络(generalized regression neuralnetworks,GRNN)联用在近红外光谱多组分定量分析中的应用。以饲料样品为实验材料,采用PLS-GRNN法建立了饲料中水溶性氯化物、粗纤维、脂肪三项组分含量近红外光谱定量分析模型。马氏距离法剔除强影响点和奇异点,用PLS法将原始数据压缩为主成分,取8个主成分吸收峰与4个原始图谱特征峰值输入GRNN网络,网络光滑因子σi为0.1。PLS-GRNN模型对样品3个组分含量的预测决定系数(r2)分别为:0.9840,0.9870,0.9830;样品平行扫描光谱预测值的标准偏差分别为:0.00326,0.0655,0.0314。结果表明所建PLS-GRNN模型通过近红外光谱能够准确预测饲料中水溶性氯化物、粗纤维、脂肪三项组分含量,为近红外光谱进行多组分定量分析提供了新思路,同时为解决近红外快速检测技术在预测组分含量较低的样品时误差相对较大的问题提供了可靠的方法。
文摘利用近红外和拉曼光谱法定量分析了甲醇汽油中甲醇的含量,采用偏最小二乘法(partial least squares,PLS)建立甲醇的定量模型.近红外光谱法测定甲醇定量模型的预测集相关系数RP为0.998,预测均方根误差(RMSEP)为0.289%;拉曼光谱法测定甲醇定量模型的预测集相关系数RP为0.982,预测均方根误差(RMSEP)为1.141%.实验表明,近红外与拉曼光谱技术均可用于甲醇汽油中甲醇含量的快速检测.
文摘探索了近红外光谱(near infrared spectra,NIRS)结合支持向量机(support vector machine,SVM)检测甘薯粉丝掺假(掺杂木薯淀粉和玉米淀粉)的可行性。以掺假甘薯粉丝为研究对象,建立了基于NIRS及SVM的甘薯粉丝掺假定性判别及定量分析模型,并通过光谱预处理及光谱变量筛选对模型进行了优化。结果显示,采用标准正态变量变换和一阶导数对全光谱预处理后,甘薯粉丝掺假SVM定性判别模型的识别准确率可达100%,优于马氏距离判别模型;用标准正态变量变换和一阶导数对光谱预处理,并通过前向区间支持向量机(forward interval support vector machine,fi-SVM)筛选光谱变量后,木薯淀粉含量SVM预测模型的相关系数( r )和预测均方差(RMSEP)可达到0.92和11.20,玉米淀粉含量SVM预测模型的 r 和RMSEP可达到0.96和7.49。结果表明,基于NIRS和SVM的甘薯粉丝掺假定性判别及定量分析检测模型具有较高的识别率和预测精度,用于检测甘薯粉丝的掺假是可行的。
