Llario Rafael, Iñón Fernando A, Garrigues Salvador, de la Guardia Miguel
Departamento de Química Analítica, Universidad de Valencia, Edificio Jerònim Muñoz, C/Dr. Moliner 50, Burjassot, Valencia, Spain.
Talanta. 2006 Apr 15;69(2):469-80. doi: 10.1016/j.talanta.2005.10.016. Epub 2005 Nov 17.
The estimation of important quality parameters of beers, such as original and real extracts and alcohol content, has been evaluated by attenuated total reflectance-Fourier transform infrared spectroscopy (ATR-FTIR) using a partial least square (PLS) calibration approach. Two sample populations, one consisting of 24 samples and other of 21 samples, obtained from the Spanish market and covering different types of beer were used. The first set was used for building and validating the model, whereas the second, measured 6 months after, was used for evaluating its robustness. The spectral range and the size of the calibration set and its suitability for building the PLS model have been evaluated. Considering a calibration set comprised of 12 samples, selected via hierarchical cluster analysis, and a validation data set of 11 samples, the absolute mean difference (d(x-y)) and standard deviation of mean differences (s(x-y)) of the real extract, original extract and alcohol content were 0.009 and 0.069% (w/w), -0.021 and 0.20% (w/w) and -0.003 and 0.130% (v/v), respectively. The maximum error for the prediction of any of these three parameters for a new sample did not exceed 2.5%. These values were practically invariant for both tested data sets. The developed methodology favourably compares with the automatic reference methodology in terms of speed and reagent consumption and waste generation.
采用偏最小二乘（PLS）校准方法，通过衰减全反射傅里叶变换红外光谱（ATR-FTIR）对啤酒的重要质量参数进行了评估，这些参数包括原麦汁浓度、真正浓度和酒精含量。使用了两个样本群体，一个由24个样本组成，另一个由21个样本组成，这些样本均取自西班牙市场，涵盖了不同类型的啤酒。第一组样本用于构建和验证模型，而6个月后测量的第二组样本则用于评估模型的稳健性。对光谱范围、校准集大小及其对构建PLS模型的适用性进行了评估。考虑到通过层次聚类分析选择的由12个样本组成的校准集和11个样本的验证数据集，真正浓度、原麦汁浓度和酒精含量的绝对平均差（d(x-y)）和平均差的标准偏差（s(x-y)）分别为0.009和0.069%（w/w）、-0.021和0.20%（w/w）以及-0.003和0.130%（v/v）。对新样本这三个参数中任何一个的预测最大误差均未超过2.5%。对于两个测试数据集，这些值实际上是不变的。在速度、试剂消耗和废物产生方面，所开发的方法与自动参考方法相比具有优势。
