基于激光诱导击穿光谱技术的咖啡豆中咖啡因含量快速检测方法

发布时间：2018-06-25 05:41

本文选题：激光诱导击穿光谱 + 咖啡豆　；参考：《光谱学与光谱分析》2017年07期

【摘要】：应用激光诱导击穿光谱(LIBS)技术研究了快速检测咖啡豆中咖啡因含量的可行性。将咖啡豆磨粉压成片状作为采集LIBS光谱数据的样本,应用原子吸收分光光度计测量每个样本中咖啡因的含量。应用基线校正,小波变换和归一化等数据预处理方法;针对基于全部变量的偏最小二乘(PLS)模型会出现过拟合,分别应用回归系数和主成分分析(PCA)选择特征变量,并建立了基于特征变量的PLS和BP神经网络模型。结果表明:基于回归系数所选特征变量的PLS模型中,建模集相关系数Rc=0.96,预测集Rp=0.91;基于PCA提取特征变量的PLS模型中,Rc=0.94,Rp=0.90;基于PCA所选特征变量的BP神经网络模型中,Rc=0.96,Rp=0.96。两种方法所提取特征变量均对应C,H,O,N,Na,Mn,Mg,Ca和Fe,且基于上述两种方法所选特征变量的PLS模型均对预测集样本有较好的预测结果,说明上述元素与咖啡因含量存在联系,应用回归系数和PCA选择的特征变量是有效的,但是咖啡豆内C,H,O,N,Na,Mn,Mg,Ca,Fe与咖啡因含量的确切关系需要进一步研究。基于PCA所选特征变量的BP神经网络模型有更优的预测结果,说明所选特征变量适用于不同的建模方法。研究表明LIBS技术结合化学计量学方法可以实现咖啡豆中咖啡因含量的快速检测。
[Abstract]:The feasibility of rapid determination of caffeine in coffee beans was studied by laser induced breakdown spectroscopy (LIBS). The coffee bean grinding powder was pressed into sheets as the sample to collect the Libs spectral data. The caffeine content in each sample was measured by atomic absorption spectrophotometer (AAS). The content of caffeine in each sample was measured by atomic absorption spectrophotometer (AAS). Based on baseline correction, wavelet transform and normalization, the partial least squares (PLS) model based on all variables is overfitted, and regression coefficients and principal component analysis (PCA) are used to select feature variables, respectively. The PLS and BP neural network models based on characteristic variables are established. The results show that in the PLS model based on the characteristic variables selected by the regression coefficient, the correlation coefficient of the modeling set is 0.96, the prediction set is Rp0.91; the PLS model based on PCA is used to extract the feature variables, and the Rc0.96Rp0.96 is found in the BP neural network model based on the feature variables selected by PCA. The characteristic variables extracted by the two methods correspond to Ca and Fe. the PLS models based on the above two methods have good prediction results for the predicted set samples, indicating that the above elements are related to the caffeine content. The regression coefficient and the characteristic variables selected by PCA are effective, but the exact relationship between the content of caffeine and the content of caffeine in coffee bean needs further study. The BP neural network model based on PCA selected feature variables has better prediction results, which shows that the selected feature variables are suitable for different modeling methods. The results showed that LIBS combined with chemometrics could be used to detect caffeine in coffee beans.
【作者单位】：浙江大学生物系统工程与食品科学学院;
【基金】：国家科技支撑计划项目(2014BAD10B02) 浙江省自然科学基金项目(LY15C130003)资助
【分类号】：O657.319;TS273

【相似文献】