论文部分内容阅读
采用短波近红外光谱仪器在线检测保健酒调配液生产线上产品的酒精度。通过使用一阶倒数(First derivative,FD)和平滑处理(Norris derivative filter,ND),对近红外图谱进行预处理,使用偏最小二乘法(Partial least square,PLS)建立了酒精度检测近红外模型。模型的校正集均方根误差(Root mean square error of calibration,RMSEC)为0.737,交互验证相关系数为0.9189;预测集均方根误差(Root mean square error of prediction,RMSEP)为0.788,交互验证相关系数为0.9425。实验数据显示,近红外计算酒精度数值与标准法测量数值相对偏差主要集中在±2%之间,该方法可以满足生产过程中在线检测酒精度的要求。
Using shortwave near infrared spectroscopy instrument online testing health wine blending solution production line alcohol content. Near infrared spectra were preprocessed using First derivative (FD) and Norris derivative filter (ND), and the alcohol detection NIR model was established by partial least square (PLS) . The root mean square error of calibration (RMSEC) was 0.737 and the correlation coefficient of cross validation was 0.9189. The Root mean square error of prediction (RMSEP) of the model was 0.788, The coefficient is 0.9425. The experimental data show that the relative deviation between the calculated values of near-infrared spectra and those measured by the standard method is mainly between ± 2%. This method can meet the requirements of on-line alcohol detection in the production process.