食品与发酵工业2025，Vol.51Issue(22)：382-388,7.DOI:10.13995/j.cnki.11-1802/ts.042114
基于低场核磁弛豫特性的驼奶粉掺假识别模型的建立与评价
Establishment and evaluation of a camel milk powder adulteration identification model based on low-field nuclear magnetic relaxation characteristics
摘要
Abstract
To investigate the effects of adulterating camel milk powder with different concentrations of cow milk powder,protein pow-der,and starch,low-field nuclear magnetic resonance technology was employed.The T2 transverse relaxation time spectra revealed distinct differences between pure camel milk powder and adulterated samples,providing valuable data for further analysis.Various machine learning algorithms,including support vector machine,k-nearest neighbors,random forest,multilayer perceptron,and extreme gradient boosting,were applied to classify camel milk powder with different types of adulteration.The RF classification model performed most effectively,achie-ving accuracy and F1 score of 96.35％and 97.53％,respectively.Principal component analysis(PCA)was also utilized to differentiate camel milk samples adulterated with varying concentrations of cow milk powder,protein powder,and starch.PCA results showed clustering trends among samples with the same concentration and clear separation between samples with different concentrations.In quantitative analy-sis,a partial least squares regression model was constructed to predict adulteration concentrations,with results indicating good predictive performance for three types of adulterants.The model's predictive capability was validated using an independent sample set,demonstrating its high generalizability.Furthermore,the precision of the method was evaluated through repeatability experiments,with intra-day precision ranging from 3.0％to 6.8％and inter-day precision ranging from 6.1％to 10.3％,thus confirming the method's stability and reliability.关键词
低场核磁共振/化学计量学/驼奶粉/掺假检测/品质控制Key words
low-field nuclear magnetic resonance/chemometrics/camel milk powder/adulteration detection/quality control引用本文复制引用
张婧怡,周然..基于低场核磁弛豫特性的驼奶粉掺假识别模型的建立与评价[J].食品与发酵工业,2025,51(22):382-388,7.基金项目
国家市场监督管理总局重点实验室(乳及乳制品检测与监控技术)(MDPDMT-2023-01) （乳及乳制品检测与监控技术）
