Study On Detection Of Adulteration In Raw Milk By Near-infrared Spectrometers

Milk adulteration is a common phenomenon in many countries and is a big concern for humans due to health hazards that could result in some fatal diseases..Recently,there has been an increase in milk consumption globally,especially in developing countries,and consumption of milk is now forming an important part of the diet for the most proportion of the global population.As a result of the most increased demands,high growth in competition in the dairy markets and increasing complexity of the supply chain,some unscrupulous milk producers are indulging in milk adulteration.This malpractice in milk has become a big common problem in the developing countries.Due to the common practice of milk adulteration and its consequences on human health,therefore it is necessitated to have a suitable method for adulteration detection.So far,many types of techniques have been investigated for the quantitative detection of adulterants in milk.These techniques are challenged by their complexity,time-consuming and high cost.Among the techniques that have been used recently for food quality control is Near-infrared(NIR)spectroscopy.In this study,portable Near-infrared(MicroNIR 1700)spectrometer and bench-top Near-infrared(NIRDS2500,FOSS XDS analyzer)combined with multivariate analysis(Chemometrics in excel)were used to detect and quantify milk adulteration.The data obtained by portable Near-infrared(MicroNIR 1700)were analyzed by Chemometrics in excel as a new and advanced statistical method.Samples of fresh cow milk were collected from different dairy farms were containing different breeds of cows around Beijing and Hebei province,China over a 3-month period in different seasons.Those fresh cow milk samples then were adulterated with water,urea,starch and goat milk as milk adulterants at 10 different concentrations for each farm samples.The data driven soft independent modeling of class analogy(DD-SIMCA)method was employed for qualitative analysis.Partial Least Squares Regression(PLSR)was applied for statistical analysis of the obtained Near-infrared(NIR)spectral data.The results showed that DD-SIMCA approach was showed tight and well-separated clusters allowing discrimination of control samples from adulterated milk,these results of classification for both instruments.The results regarding model Acceptance percentage were presented.The best results for the model of control were obtained with 4PCs and type I error ?=0.01.Acceptance percentage of training sample of control=100% whereas Acceptance percentage of testing sample were 100%,7%,7%,0% and 0% for fresh cow milk,water,urea,starch,and goat milk,respectively.At the same time,Rejection percentage for non target classes were 0%,93%,93%,100% and 100% with type II error ? = 0.998,0.127,0.103.0.001 and 0.081 for fresh cow milk,water,urea,starch,and goat milk,respectively.The best results for the model for(class 2)milk adulterated with water as a target class were obtained with 9PCs and type I error ?=0.01.Acceptance percentage of training samples of water=100% whereas Acceptance percentage of testing samples were 0%,100%,0%,0% and 0% for fresh cow milk,water,urea,starch,and goat milk,respectively.At the same time,Rejection percentage for non target classes were 100%,0%,100%,100% and 100% with type II error ? = 0.120,1.000,0.059.0.000 and 0.003 for fresh cow milk,water,urea,starch,and goat milk,respectively.Results for the model for(class 3)milk adulterated with urea as a target class were obtained with 7 PCs and type I error ?=0.01.Acceptance percentage of training samples of urea = 98% whereas Acceptance percentage of testing samples were 0%,0%,100%,0% and 0% for fresh cow milk,water,urea,starch,and goat milk,respectively.At the same time,Rejection percentage for non target classes were 100%,100%,0%,100% and 100% with type II error ? = 0.073,0.064,0.981.0.000 and 0.123 for fresh cow milk,water,urea,starch,and goat milk,respectively.The good results for the model for(class 4)milk adulterated with starch as a target class were obtained with 4 PCs and type I error ?=0.01.Acceptance percentage of training samples of starch = 95% whereas Acceptance percentage of testing samples were 0%,0%,0%,100% and 0% for fresh cow milk,water,urea,starch,and goat milk,respectively.At the same time,Rejection percentage for non target classes were 100%,100%,100%,0% and 100% with type II error ? = 0.014,0.010,0.003.0.986 and 0.001 for fresh cow milk,water,urea,starch,and goat milk,respectively.The PLS regression model obtained standard error of prediction of 4.35,0.34,4.74 and 5.56 g/L for water,urea,starch,and goat milk,respectively for the portable Near-infrared NIR and 2.75,0.31,4.35 and 8.56g/L for water,urea,starch,and goat milk,respectively for the bench-top Near-infrared(NIR).The new established model for qualitative analysis by the data driven soft independent modeling of class analogy(DD-SIMCA)was validated by a random samples obtained from Inner Mongolia and Hebei province by Yili industrial group company limited and scanned to detect the same adulterants(water,urea,starch and goat milk).A total of 60 samples(30 samples from Inner Mongolia + 30 samples from Hebei)were scanned by portable NIR spectrometer.The results showed that about 79% of the total was not adulterated(47 samples)whereas about 8% was adulterated with high concentration of urea(5 samples)and the remaining was adulterated by unknown adulterants 13%(8 samples).To confirm that results were shown high level of urea,when detect by the model.Another method was used(Urea nitrogen in milk and milk product-Spectrophotometric method)to check the level of urea in these 5 samples.We get the same results that indicate of high level of urea in these 5 milk samples These results demonstrated the feasibility and reliability of near-infrared spectroscopy,combined with the multivariate analysis in the prediction of the total content of investigated adulterants in cow milk.