近红外透射法预测再造烟叶中的5种主要化学成分

Prediction of Five Major Chemical Components in Reconstituted Tobacco Sheet with NIR Transmission Spectroscopy

为了解近红外(NIR)透射法预测再造烟叶中烟碱、总糖、还原糖、总氮和钾含量的可行性,以100个有代表性的再造烟叶样品作为模型校正集,15个样品为验证集,通过偏最小二乘法(PLS)建立了这5种成分的NIR透射模型,并对模型的预测效果和重复性进行评价。结果表明:①烟碱、总糖、还原糖、总氮和钾NIR透射模型的相关系数分别为0.9268、0.9575、0.9252、0.8024和0.9665;除烟碱外,总糖、还原糖和总氮模型平均相对预测偏差均低于5%,钾的模型平均绝对预测偏差低于0.20;5种化学成分模型的预测相对标准偏差都低于5%;②NIR透射模型的预测精确度比漫反射法略低。NIR透射法适合于批量再造烟叶样品中烟碱、总糖、还原糖、总氮和钾含量的快速分析。

In order to investigate the feasibility of predicting nicotine, total sugar, reducing sugar, total nitrogen and potassium in reconstituted tobacco sheet with NIR transmission spectroscopy, the prediction models for the five components were established with partial least square （PLS） by using 100 representative samples of reconstituted tobacco sheet as the model calibration set and other 15 samples as the validation set. The predictive results and repeatability of the models were evaluated as well. The results showed that the correlation coefficients of NIR transmission spectrum prediction models for nicotine, total sugar, reducing sugar, total nitrogen and potassium were 0. 9268, 0. 9575, 0. 9252, 0. 8024 and 0. 9665, respectively. Aside from nicotine, the average relative predictive deviations of total sugar, reducing sugar and total nitrogen models were below 5% , and the average absolute predictive deviation of potassium was lower than 0. 201 the predictive RSDs of these five components were below 5%. The predictive accuracy of NIR transmission models was slightly lower than that of NIR diffuse reflection spectrum models. NIR transmission spectrometry is suitable for rapid analysis of these five components in batches of reconstituted tobacco sheet samples.