辊压法烟草薄片在线检测模型建立与验证

Establishment and Verification of On-line Detection Model of Rolled Tobacco Sheet

为建立辊压法烟草薄片化学成分的在线预测模型,采用近红外光谱检测技术(NIRS)对辊压法烟草薄片进行化学成分检测以获得实验室数据,采用标准正态变量变换(SNV)进行数据预处理,采用马氏距离(MD)剔除异常数据,采用留一验证(LOOCV)为内部交叉验证方法,采用偏最小二乘法(PLS)进行建模,得出烟碱预测模型的决定系数(R^(2))达0.930,交叉验证标准偏差(RMSECV)为0.02,总糖预测模型的决定系数(R^(2))达0.977,交叉验证标准偏差(RMSECV)为0.16。分析表明,所建立的预测模型可应用于辊压法薄片关键化学成分的检测,其检测数据可为烟叶薄片的化学成分波动、质量分析等提供可靠的数据支撑。

In order to establish the on-line detection model of rolled tobacco sheet,its chemical composition was detected by near infrared spectroscopy(NIRS)to obtain laboratory data,the data were preprocessed by standard normal variable transformation(SNVT),the abnormal data were eliminated by mahalanobis's distance(MD),the leave one out cross validation(loocv)was used as the internal cross verification method,the partial least squares(PLS)is used for modeling,the determination coefficient(R^(2))of nicotine prediction model is 0.930,and the Root Mean Square Error of Cross Validation(RMSECV)is 0.02,the determination coefficient(R^(2))of the total sugar prediction model was 0.977 and the Root Mean Square Error of Cross Validation(RMSECV)was 0.16.The analysis shows that the established prediction model can be applied to the detection of key chemical components of rolled tobacco flakes,and the detection data can provide reliable data support for chemical component fluctuation and quality analysis of rolled tobacco sheet.