李卫群，朱慧，汪涓涓

（杭州娃哈哈集团有限公司研究院分析中心，浙江 杭州 310018）

摘要：采用基于全数据库奶粉分类模型对样品进行识别，再采用既定类型的掺杂物模型对未知样品进一步定性识别，并对掺伪种类进行判断。建立一种利用近红外光谱法结合Adulterant Screen 算法快速鉴别奶粉单组份和多组分掺假的检测方法。对于单组份掺杂奶粉的识别浓度三聚氰胺和甘氨酸可达到0.05%以下，其它掺杂物如面粉、植脂末、尿素、大豆蛋白和麦芽糊精等的识别浓度达到了1%以下。对于二组分和三组分掺假奶粉在两倍到五倍检出限浓度的准确识别率分别达到了75%和42%，掺杂物含量越高，识别准确率将更高。利用近红外光谱法结合AdulterantScreen 算法可以快速鉴别奶粉单组份和多组分物质的掺假。但随着组分数的增加，识别准确率会有相应的下降。

关键词：近红外光谱；奶粉；掺假；多组分

Study on the Multicomponent Adulteration of Milk Powder by Near Infrared Spectroscopy

LIWei-qun，ZHU Hui，WANG Juan-juan

（R &D Department of Hangzhou Wahaha Group Co. Ltd，Hangzhou 310018，China)

Abstract: Firstly，by means of NIRS，the spectroscopic curves of milk powder samples were obtained. Then based on theAdulterant Screen algorithm，the full database milk powder model and the established adulterated ingredients model，theacquired spectroscopic curves were analyzed and the adulterated ingredient were given. This study was conducted todevelop a method which could detect the single component and multicomponent adulterant in milk powder rapidly by usingnear infrared spectroscopy（NIRS）combined with Adulterant Screen algorithm. To single component adulterated milkpowder，the detection limits of melamine and glycine were 0.05% by this method，the other adulterants such as flour、vegetal Cream Powder、urea、soybean protein and maltodextrin were lower than 1% .To two component adulterated milkpowder and three component adulterated milk powder，the accurate recognition rate of adulterated ingredient were 75%and42% in the concentration range of double detection limit to five times the detection limit . The adulterant could be rapidlydetected in the one-component adulterated milk powder and multicomponent adulterated milk powder by using nearinfrared spectroscopy （NIRS）combined with Adulterant Screen algorithm. but the accurate recognition rate would bedeclined when the component increased.

Key words: near infrared spectroscopy；milk powder；adulteration；multicomponent

(2017年第20卷第1期）