摘要
提出了一种基于近红外漫反射光谱技术快速测定烟草pH值的方法。采集不同烟草粉末样品的近红外漫反射光谱,并对其原始光谱数据进行一阶微分、二阶微分及平滑等预处理,用偏最小二乘法(PLS)方法建立pH值预测模型(建模样品572个)。从光谱主成分分布和pH值分布方面考察了81个验证集样品,结果表明验证集样品分布范围较大,分布较合理。利用主仪器模型对验证集样品进行预测,结果表明主仪器一阶微分模型和二阶微分模型对验证集样品的pH值预测与实际测量值的平均绝对偏差分别为0.057、0.065,t检验表明预测值和实测值之间没有显著性差异,达到了较好的结果。考察了主仪器pH值一阶微分、二阶微分模型在同一型号不同仪器间的直接转移效果,一阶微分模型转移给了子仪器A^F,二阶微分模型转移给了子仪器G,7台子仪器pH值预测的平均绝对偏差为0.049~0.070,且都通过了F检验。实验表明,该主仪器模型能够快速预测烟叶中的pH值,并成功转移到同类仪器上进行检测。
A rapid method was developed for measurement of pH value based on the combination of near infrared（NIR） diffuse reflection spectroscopic and chemometries methods. The spectral data of tobacco were collected by Thermo Antaris Near-IR Analyzer, and analyzed by first derivative, second derivative and Savitzky - Golay filter smoothing. The pH prediction models were established with prediction 572 calibration samples by partial least squares （PLS） method, and the distribution of principal component scores and pH values of 81 validation samples was investigated. Results indicated that the 81 validation samples have broad and fine distribution. The pH values of validation samples and fine average absolute deviation（AAD） were tested by first and second derivative prediction models of master instrument. The AAD of predicted values and measurement values were 0. 057 and 0. 065, respectively, t-Test results indicated that no significant difference was found between prediction value and test value. In the meantime, the calibration transport effect of pH first derivative and second derivative prediction models of master instrument among 7 NIR devices with the same types was also investigated, respectively. The AAD were between 0. 049 and 0. 070, and all devices were passed Ftest. Therefore, the prediction models of master instrument could be applied to pH fast test, and be successfully transported to other instruments of the same type.
出处
《分析测试学报》
CAS
CSCD
北大核心
2009年第6期742-745,共4页
Journal of Instrumental Analysis
