近红外光谱测定小麦粉常规营养成分的模型优化  

Model Optimization for Determination of Common Nutritional Components in Wheat Flour Using Near Infrared Spectroscopy

作　　者：高慧宇 王国栋 赵佳 王竹 GAO Hui-yu;WANG Guo-dong;ZHAO Jia;WANG Zhu(Natiomnal Instiute for Nutrition and Health,Chinese Center for Disease Control and Preention/Key Laboratory of Trace Elenent Nutrition of National Health Cnmnission,Bejing 100050,China)

机构地区：[1]中国疾病预防控制中心营养与健康所/国家卫生健康委员会微量元素与营养重点实验室,北京100050

出　　处：《中国食物与营养》2021年第5期30-34,共5页Food and Nutrition in China

基　　金：国家自然科学基金“全谷物对过激炎症的预防作用及适宜标志物筛选”(项目编号:81673169)

摘　　要：目的:应用近红外光谱(NIR)结合偏最小二乘法(PLS)建立小麦粉常规营养成分蛋白质、水分和脂肪的含量预测模型,并选择最佳模型。方法:收集117份小麦粉样品的近红外光谱,化学法测定蛋白质、水分和脂肪的含量,利用主成分分析(PCA)随机分组,81份样品用于构建模型、36份样品用作验证模型的预测能力。探讨波长范围和光谱预处理方法对所建模型预测能力的影响。结果:3个营养成分预测能力最好的模型分别是:对于蛋白质,预处理采用矢量归一化(SNV),波长选取7505.9~5446.2 cm^(-1)和4605.4~4242.8 cm^(-1),预测模型的RPD值是7.02;对于水分,无预处理,波长选择全谱12800~3960 cm^(-1),模型的RPD值是6.83;对于脂肪,无预处理,波长在9000~4000 cm^(-1),模型的RPD值是5.06。结论:近红外光谱法可以实现对小麦粉常规营养成分的快速预测,通过选择波长范围和光谱预处理方法可以显著提高模型的预测能力。Objective Near-infrared(NIR)spectroscopy and partial least squares(PLS)were applied to determine the content of the common nutritional components in wheat flour samples.The rapid prediction models for protein,moisture and fat were established,and the optimal model was selected.Method NIR spectra and analytical measurements of protein,moisture and fat were collected from 117 wheat flour samples.Partial least-squares(PLS)regression models were developed with principal components analysis(PCA)to assign 81 wheat flour samples to the calibration data set and 36 accessions to the validation set.The effects of wavelength range and spectral pretreatment methods on the predictive ability of the model were discussed.Result The best predictive models for protein,moisture and fat in wheat flour samples were achieved.For protein,the best predictive model possessed 7.02 RPD value in validation sets at 7505.9~5446.2 cm^(-1) and 4605.4~4242.8 cm^(-1) using standard normal variate(SNV).For moisture and fat,the best predictive NIR models were obtained without pretreatment,at full spectra 12800~3960 cm^(-1) and 9000~4000 cm^(-1),with the PRD value 6.83 and 5.06,respectively.ConclusionNIR spectroscopy can realized the rapid prediction of the common nutritional components of wheat flour samples.By selecting wavelength range and spectral pretreatment method,the predictive model can be significantly improved.

关 键 词：近红外光谱 小麦 蛋白质 脂肪 营养成分 

分 类 号：O657.33[理学—分析化学] TS211.7[理学—化学]