绿茶茶多酚近红外光谱定量分析模型优化研究

Optimization of Models for Quantitative Analysis of Tea Polyphenols in Green Tea by Near Infrared Spectrophotometry

采用NIR Systems6500和InfraXact Lab型近红外仪分别对158份绿茶未粉碎品和粉碎样品进行光谱扫描,利用正交试验设计,分别采用主成分回归、偏最小二乘、改进偏最小二乘3种校正方法,并对原始光谱分别进行不同的预处理,建立了绿茶茶多酚定标模型,利用目标函数法对模型进行评定,并对评定的最优模型适用性进行验证。试验结果为:利用NIR Systems6500型近红外仪对绿茶粉碎样品扫描的光谱采用改进偏最小二乘法进行定标,在标准正常化+趋势变化散射处理、二阶导数处理(取点间隔为1)、平滑处理取点间隔为4和二次平滑处理取点间隔为8组合的光谱预处理下建立的模型最优,其目标函数值为95.74%,验证相对标准差(RSD)为4.52%,相对分析误差(RPD)为2.52%。结果表明:采用正交试验设计能够综合考察不同的校正方法和预处理方法对近红外定标模型的影响,利用近红外光谱分析法能够实现绿茶中茶多酚含量的定量检测,所建立的最优模型具有很好的预测准确性和适用性。

158 non-comminuted and comminuted green tea samples were scanned by NIR Systems 6500 and InfraXact NIR analyzers. An analysis model for quantitation of polyphenols in green tea was developed through orthogonal test design and validated with principle component regression （PCR）, partial least square （PLS） and modified PLS. The models were evaluated by the objective function method and the applicability of the optimal models was validated. The optimal model was obtained by using the NIR Systems 6500 analyzer with comminuted samples by modified PLS and the optimum analytic conditions for spectral data preprocessing were standard normal variate （SNV）/detrend scatter, second derivative treatment （the interval for one point）, smoothing interval for four points and the secondary-smoothing interval for eight points. This optimal condition leaded to the objective function value of 95.74% and RSD and RPD of validation of 4.52% and 2.52%, respectively. The results showed that orthogonal experiment design could comprehensively evaluate the effect of methods for validation and spectral data preprocessing on the models. Quantitative analysis of polyphenols could be realized by NIRS and this optimal NIRS model had very good precision and applicability.