The Research On Identification Technology Of Carbohydrates Adulteration In Raw Milk

Carbohydrate is one of important nutrients in raw milk and dairy products, and plays an important role in human health. However, the adulteration phenomenon is widespread and the types are varied. This paper studied the chemical method (nephelometry and polarimetry), HPLC with refractive index dectetor, and mid-infrared spectroscopy combined with chemometrics methods of MLP neural network and PLS for carbohydrates adulteration in raw milk, and established a relatively perfect identification system. The main results were as follo wings:We used trichloroacetic acid as the precipitant and anhydrous ethanol as the chromogenic agent, under the condition of600nm, to establish a colorimetric method to measure the dextrin adulteration in raw milk. The results showed that the R2were0.9982and0.9987, the average recovery rates were98.8%and98.0%, and the relative standard deviations were0.99%and0.90%. At the same time, we used trichloroacetic acid as precipitant to establish a differential polarimetry to measure the lactose (whey powder form) adulteration in raw milk under the589.3-589.44nm of polarimeter. The results of polarimetry for the determination of lactose adulteration in raw milk were that the R2was0.9991, the average recovery rate was98.1%, and the relative standard deviation was0.99%. The results demonstrated that this two chemical methods are repeatable, accurate and reliable.We used Hypersil amino column as separation column, CH3CN/H2O(80:20, v/v) as mobile phase, flow rate of1.0mL/min, the method of high performance liquid chromatography with refractive index dectetor was established for differentiation of raw milk adulteration of fructose, glucose, sucrose, and lactose. The results showed that the linear range of such method ranged from0.5mg/mL to20mg/mL, the linear equations of fructose, glucose, sucrose, and lactose were y=552642x-15597, y=r287299x+3292, y=299217x+29805, and y=283607x+28975, respectively. The R2ranged from0.9990to0.9994, the recovery rate ranged from98.5%to101.0%, the minimum detection limit ranged from0.20to0.25mg/mL and the relative standard deviation ranged from1.500%to2.975%. Results were accurate and reliable, and it was suggested that the method obtained can be used for study of carbohydrates in raw milk adulteration. We used mid-infrared spectroscopy combined with Multilayer Perceptron neural network and partial least squares to identify6kinds of single carbohydrate adulteration in raw milk(378samples) and established the corresponding qualitative and quantitative models. The average judgment accuracy of the MLP neural network correction model was94.574%, the average judgment accuracy of the validation set model was91.667%. Based on these results to established the PLS quantitative models, the R2ranged from0.9298to0.9544, and RMSECV ranged from1.020%to1.632%. The quantitative calibration models were tested, the R2ranged from0.9274to0.9502, and RMSEP ranged from1.271%to1.530%; At the same time, mid-infrared spectroscopy combined with Multilayer Perceptron neural network and partial least squares were also used to identify8kinds of multiple carbohydrates adulteration in raw milk(756samples), qualitative and quantitative analysis models were established. The average diagnosis rate of the MLP neural network correction model was94.747%, the average correct diagnosis rate of the validation set was95.060%. Based on this, PLS quantitative models were established. The R2ranged from0.8470to0.9209, and RMSECV ranged from0.569%to1.607%. The quantitative calibration models were tested, the R2ranged from0.8241to0.9216, and RMSEP ranged from0.681%to1.537%. The results showed that the discriminant models possess many advantages such as high accuracy and operability, based on mid-infrared spectroscopy combined with stoichiometry method to identify the adulteration of carbohydrates in raw milk.