近红外光谱快速测定电子烟油中烟碱含量的波段优选研究

Rapid Determination of Nicotine in E-Liquid by Near Infrared Spectroscopy Based on Wavelength Optimization

围绕电子烟油中烟碱含量的近红外测量模型,采用63个不同烟碱含量和不同口味的电子烟油样品进行光谱数据的采集,利用蒙特卡罗交互验证法剔除异常样本后,采用区间偏最小二乘法(iPLS)、组合区间偏最小二乘法(SiPLS)选择最佳波长并建立校正模型。结果表明,iPLS算法识别出的近红外测量电子烟油中烟碱含量的特征波段分布在1090~1228 nm和1370~1508 nm这两个波段附近,SiPLS算法通过对不同波段进行组合,进一步确定了最佳波段为1126~1240 nm,1358~1414 nm,1474~1530 nm这一波段组合,SiPLS-PLS模型与全谱-PLS模型相比,所采用的变量数降低了2/3,预测均方误差值从1.188降低到了0.963,模型预测的准确性得到了提高。

To establish the near infrared measurement model of nicotine content in e-liquid,63 samples with different nicotine contents and tastes were collected.After eliminating the abnormal samples via Monte Carlo cross validation,the interval partial least squares(iPLS),and synergy interval partial least squares(SiPLS) methods were used to select the optimal wavelength,and subsequently,the correction model was established.The results show that iPLS identified the characteristic wavelength bands of 1090-1228 nm and 1370-1508 nm.By combining the different bands,SiPLS determined the optimal bands as 1126-1240 nm,1358-1414 nm,and 1474-1530 nm.When compared with the full-spectrum PLS model,the variables used in the SiPLS-PLS model reduced by two-thirds and the mean square error of prediction value reduced from 1.188 to 0.963;thus,improving the accuracy of the model.