Abstract There is a growing need of measurement technologies that can be used close to the sample source and optical spectroscopy is an excellent example of this genre of technology: from the lab to the field. This study investigates the possibility to quantify the major components and to detect the presence or absence of lactose in commercial milks with ATR-MIR spectroscopy. We explored the possibility to use a portable and economical ATR-MIR instrument, comparing the results with a benchtop system. Commercial milk samples from Italy, Switzerland and Spain were chosen covering the maximum range of variation for protein, carbohydrate and fat content. The analytical protocol was optimized to make it as fast and useable as possible for both instruments, from the sample pretreatment to the instrumental parameters. Multivariate calibration was used to correlate the recorded spectra to the content of the major milk components, while a classification was done in order to classify samples with or without lactose. A comparison was performed between the predictive capabilities of the models built with different data pretreatments, different variable selection methods and different validation systems to obtain the best results and to assure robust models.

中文翻译：

---

用于预测商业牛奶主要成分的 ATR-MIR 光谱：手持式和台式仪器之间的比较

摘要 对可在样品源附近使用的测量技术的需求不断增长，而光谱学是此类技术的一个很好的例子：从实验室到现场。本研究调查了使用 ATR-MIR 光谱量化主要成分和检测商业牛奶中乳糖存在与否的可能性。我们探索了使用便携式和经济型 ATR-MIR 仪器的可能性，并将结果与​​台式系统进行了比较。选择了来自意大利、瑞士和西班牙的商业牛奶样品，涵盖了蛋白质、碳水化合物和脂肪含量的最大变化范围。对分析方案进行了优化，使其尽可能快速且适用于两种仪器，从样品预处理到仪器参数。多变量校准用于将记录的光谱与主要牛奶成分的含量相关联，同时进行分类以对含有或不含乳糖的样品进行分类。对使用不同数据预处理、不同变量选择方法和不同验证系统构建的模型的预测能力进行了比较，以获得最佳结果并确保模型稳健。