摘要
Abstract
This study selected 160 commercially available and laboratory-prepared E-liquid samples for spectral scanning using a Fourier transform near-infrared(NIR)spectrometer(Thermo Fisher,USA).By comparing with chemical reference values,the near-infrared spectroscopy(NIRS)tech-nique was validated for accurate prediction of six components in E-liquids:benzoic acid,nicotine,WS-23,WS-3,propylene glycol,and glycerol.Utilizing TQ Analyst software,partial least squares(PLS)regression was employed to establish quantitative models for these additives.Spectral preprocessing involved first-derivative transformation combined with Savitzky-Golay smoothing,and optimal wavelength ranges were determined based on multivariate correlation spectroscopy.Results demonstrated that the calibration correlation coefficient(R2C)and prediction correlation coefficient(R2p)for nicotine,benzoic acid,WS-23,propylene glycol,and glycerol all exceeded 0.98,indi-cating high accuracy in predicting these five additives.The cross-validation correlation coefficient(R2CV)for the WS-3 model was 0.95,suggesting potential for further optimization.The established quantitative models provide an effective tool for rapid quality assessment of E-liquids.关键词
近红外光谱/电子烟油/添加剂/定量模型/相关系数Key words
near-infrared spectroscopy/E-liquids/additives/quantitative model/correlation co-efficient分类
化学
