Lin Tao, Yu Hai-yan, Xu Hui-rong, Ying Yi-bin
College of Biosystems Engineering and Food Science, Zhejiang University, Hangzhou 310029, China.
Guang Pu Xue Yu Guang Pu Fen Xi. 2009 Apr;29(4):950-5.
Near infrared (NIR) spectroscopy and chemometrics were applied to determine the effect of pathlength variation on the spectra of the Chinese rice wine and the consequences of the prediction precision of calibration models developed for measuring alcoholic degree, sugar content, and pH was investigated in the present research Samples were scanned in transmission mode using a commercial FT-NIR spectrometer and a demountable liquid cell for versatile path length liquid sampling. By comparing the results of performance between models based on different optical pathlength (0.5, 1, 1.5, 2, 2.5, and 3 mm), the best indicators of optical pathlength were identified. Based on the optimum pathlength, the models for alcoholic degree, sugar content and pH were established. The best optical pathlength for the alcoholic degree was 2 mm, using partial least squares regression (PLSR) model with the original spectra, correlation coefficient (r) was 0.942, root mean standard error of calibration (RMSEC) and root mean standard error of cross-validation (RMESCV) were 0.256 (%, (phi)) and 0.292 (%, (phi)), respectively; the best optical pathlength for the sugar content was 1 mm, using PLSR model with the original spectra, r was 0.945, and RMSEC and RMESCV were 0.125% and 0.149%, respectively; the best optical pathlength for the pH was 2 mm, using PLSR model with the original spectra, r was 0.947, and RMSEC and RMESCV were 0.018 and 0.039, respectively. This study showed that pathlength variation had influence on the performance of calibration models for Chinese rice wine, and a suitable pathlength could effectively improve detection accuracy.
本研究应用近红外（NIR）光谱法和化学计量学方法，测定光程变化对中国黄酒光谱的影响，并研究了所建立的用于测量酒精度、糖含量和pH值的校准模型的预测精度。使用商用傅里叶变换近红外光谱仪和可拆卸液体池以透射模式对样品进行扫描，以便进行多光程液体采样。通过比较基于不同光程（0.5、1、1.5、2、2.5和3mm）的模型之间的性能结果，确定了最佳光程指标。基于最佳光程，建立了酒精度、糖含量和pH值的模型。酒精度的最佳光程为2mm，使用原始光谱的偏最小二乘回归（PLSR）模型，相关系数（r）为0.942，校准均方根误差（RMSEC）和交叉验证均方根误差（RMESCV）分别为0.256（%，（phi））和0.292（%，（phi））；糖含量的最佳光程为1mm，使用原始光谱的PLSR模型，r为0.945，RMSEC和RMESCV分别为0.125%和0.149%；pH值的最佳光程为2mm，使用原始光谱的PLSR模型，r为0.947，RMSEC和RMESCV分别为0.018和0.039。本研究表明，光程变化对中国黄酒校准模型的性能有影响，合适的光程可以有效提高检测精度。
