含水率对烟草多组分近红外预测结果的影响
Effects of moisture content on near-infrared predictions of tobacco chemical components
-
摘要:
为研究含水率对烟草多组分近红外预测结果的影响，选取代表性样品50个，分别制备5个含水率梯度的系列样品，采集其近红外光谱，使用统一的烟草近红外多组分预测模型进行预测。结果表明，随含水率的增加，样品近红外光谱图不同波段的吸收均有不同程度增加，仅少量化学成分含量近红外预测值受样品含水率的影响较小。建立6%~8%和10%~12%两个区间含水率的烟草近红外多组分预测模型，再次对这5个含水率梯度的样品系列进行近红外预测，发现除个别指标外，采用建模样品含水率区间和样品含水率区间一致的烟草近红外多组分预测模型进行预测，有效降低了水分对近红外预测值的影响。 Abstract:In order to investigate the effects of moisture content on near-infrared predictions of tobacco chemical components, 50 representative samples were selected, and sample series of 5 moisture content gradients were prepared. After the spectra acquisition of all samples, a tobacco multicomponent NIR model was used to make predictions. The results showed that the bands in NIR spectra were intensified to different degrees with the increase of moisture content, and only the prediction values of a few chemical components were less affected by moisture content. Two NIR prediction models of tobacco multicomponent NIR models were established for two moisture intervals (6%-8% and 10%-12%) and near-infrared predictions of samples of 5 moisture gradients were made. The results showed that except for a few indexes, making predictions by using one of the two models with moisture interval consistent with that of test samples effectively minimized the effects of moisture content.
下载: