中红外反射光谱结合偏最小二乘法快速定量分析葡萄酒

利用傅里叶变换中红外光谱(FTIR)和衰减全反射(ATR)技术,结合偏最小二乘算法,建立了一种同时快速测定葡萄酒中总酚含量、滴定酸度、pH值和酒精度的方法。通过光谱预处理方法、异常值诊断和建模参数的优化,提高了模型的预测精度,各指标预测值与实测值相关性良好,总酚含量、滴定酸度、pH值和酒精度模型的相关系数(r2)依次为0.9356,0.9649,0.9935,0.9837;交互验证均方根误差(RMSECV)依次为:0.445,0.378,0.087,0.507。t检验结果表明模型预测结果与标准方法测定结果之间无显著性差异(显著性大于0.05)。

Simultaneous Quantified Analysis of Wine by Mid-Infrared Reflectance Spectrometry and Partial Least Square

A new simultaneous analytical method for the quantitative determination of total phenols,titratable acidity,pH,alcohol of wine was developed using attenuated total reflection （ATR） FTIR spectroscopy combined with the partial least square （PLS） algorithm.Spectral pretreatment,outliers diagnostics and optimization factors for building the models were studied to improve precision and predictive ability of the models.The correlation coefficient （r2） of calibration sets of phenols,titratable acidity,pH,alcohol models was 0.9356,0.9649,0.9935 and 0.9837.The RMSECV （Root Mean Square Error of Cross Validation） was 0.445,0.378,0.087 and 0.507.There was no significant difference at 0.05 level between the predicted values and standardly determined values by statistical significance test.