摘要
采集标准香精样品和不同生产批次香精样品的傅里叶变换多次衰减全反射红外指纹图谱,通过主成分分析(PCA)法,剔除偏离主成分区域的光谱,对PCA校正后的光谱进行Karl Norris滤波(段长为5,段间距为2)和二阶微分预处理,结合Gram-Schmidt正交化方法,建立香精品质鉴定模型。首先,以落在主成分区域内的某个样品为基准建立模型,计算主成分区域内其余样品的相似度,作为检验标准;其次,以落在主成分区域内的全部样品为基准建立模型,分别计算30个生产批次香精样品的相似度,作为各批次香精品质鉴定的依据。所建立的模型能够准确鉴别不合格品,且与常规物理指标检测方法相比,该方法准确度更高,操作方便快速,可直接应用于香精生产过程中的品质在线检测和监控。
The multi-bounce attenuated total reflectance-fourier transform Infrared spectroscopy fingerprints of the standard flavors and flavors final products were collected. The spectrums which were not in the main region were eliminated through PCA analysis. The Karl Norris wave filter （segment length 5, segment width2） and the second differential were used as pre-processing. Combined with the method of Gram-Schmidt, the quality control model was established. The model was based on one sample spectrum characteristics in the main region, and used to calculate the similarities of the rest samples in the main region. Based on the whole samples spectrums in the main region, the other model was established and the similarity of different production batches was calculated as the quality indexes. In this way, the unqualified products were detected. The method is highly accurate and easier to operate. It could be used in the online quality control.
出处
《中国食品添加剂》
CAS
北大核心
2015年第4期111-115,共5页
China Food Additives
基金
云南省科技厅重点新产品开发计划资助[2014]01号
云南中烟工业有限责任公司项目基金资助(2014JC07)
