Efficient identification method for adulteration of dairy products based on endogenous nanometer characteristics

Abstract

According to the method for efficiently identifying adulteration of the dairy products through the endogenous nanometer characteristics, natural nanometer particles existing in endogenous sources in the dairy products are detected, multiple nanometer characteristic indexes are integrated to form discrimination parameters, an artificial learning-machine algorithm is utilized to construct a discrimination model, the discrimination model is optimized, the dairy products to be detected are discriminated, and the adulteration of the dairy products to be detected is identified. The dairy product adulteration can be efficiently identified on the premise that adulterants are not known. According to the method, various dairy products including bulk milk and rare camel milk can be identified, a judgment model needs few optimization parameters, the fitting performance is good, the equation fitting degree can reach 0.98 or above, and the prediction effect is good. Experiments prove that the method provided by the invention has a relatively high discrimination rate on common adulterants such as foreign protein, starch and water in dairy products, and the discrimination rate can reach 90% or above.

Description

Technical Field

[0001] The present invention belongs to the field of food detection, and specifically relates to a method for detecting dairy products, and more specifically to a highly efficient method for identifying adulteration of dairy products based on endogenous nano properties. Background Art

[0002] Dairy products not only have a unique taste but are also nutritious and essential for the growth and development of newborn mammals. They are also believed to prevent or improve osteoporosis, botulism, metabolic syndrome, cardiovascular disease, cognitive decline, and digestive system diseases.

[0003] Currently, there are numerous types of dairy products. In addition to the predominantly cow's and goat's milk, buffalo milk, camel's milk, and other dairy products have also recently entered the market. However, to date, no specific evaluation indicators exist for each type of dairy product, making it difficult for manufacturers or consumers to effectively distinguish between dairy products and adulterated dairy products. Existing technologies primarily focus on evaluating specific molecular components in dairy products. However, these well-defined molecular components offer counterfeiters opportunities for fraud and pose significant challenges in identifying adulteration. Dairy adulteration typically involves diluting and / or adding cheap, low-quality, and sometimes even industrial-grade products to increase production, mask inferior quality, and / or replace target substances in dairy products to maximize revenue. This not only results in economic losses for the industry but also poses significant risks to consumer health. Furthermore, as a key ingredient in the food industry, cumulative effects can be observed in the processing of dairy products.

[0004] At present, chemical methods, chromatography methods, etc. are commonly used to detect adulteration and additives in dairy products, and local single-point sampling is used for detection. However, different adulterants require different detection methods, and the detection technology is relatively difficult, the process is cumbersome, and the cycle is long. Chinese patent CN103399091A discloses a method for determining the authenticity of raw milk protein by applying ultra-high pressure liquid chromatography technology combined with data analysis technology, but the determination of this method relies on manual identification and the determination conditions are complex. Therefore, it is urgent to establish an effective, accurate, fast and efficient dairy product adulteration detection technology. Summary of the Invention

[0005] To address the challenges of existing dairy product testing and analysis, which include relatively difficult detection techniques, cumbersome processes, and lengthy processing times, this paper provides a highly efficient method for identifying dairy product adulteration based on endogenous nanoparticle characteristics. This method detects endogenous natural nanoparticles in dairy products and combines them with an artificial learning-machine algorithm and database modeling analysis to rapidly identify adulterated dairy products without requiring knowledge of the adulterant.

[0006] The present invention provides an efficient method for identifying adulteration of dairy products based on endogenous nano-properties, comprising the following steps: S1. Take the test sample and standard dairy product, respectively, and perform membrane treatment or centrifugation, and take the filtrate or supernatant; S2. Characterizing the obtained filtrate or supernatant by nano-characterization to obtain nano-characteristic parameters of the test sample and the nano-characteristic parameters of the standard dairy product, and calculating the wave width coefficient of the test sample and the wave width coefficient of the standard dairy product respectively; S3. Using a machine learning-based algorithm, we fit the obtained nanostructured parameters and wave width coefficients of the test sample and standard dairy products to obtain a discriminant model. S4. Using the obtained discriminant model, compare the nano-characteristic parameters and wave width coefficients of the test sample and the standard dairy product to obtain a comparison result; and determine whether the test sample is an adulterated dairy product based on the comparison result.

[0007] Early studies have shown that there are a large number of natural micro-nano colloidal particles in dairy products, which are mainly composed of proteins. These proteins can form casein micelles through hydrophobic interactions. Therefore, each dairy product has its own characteristic particle size information, and once the dairy product is adulterated, the properties of the nano-colloid particles will inevitably change, and the most significant change is the change in particle size distribution. In addition, the nano-properties of dairy products also include many indicators such as average particle size, distribution coefficient, light scattering intensity, surface charge, and number of particles in addition to particle size distribution. Therefore, the present invention detects the endogenous natural nanoparticles in dairy products, and forms a discrimination parameter by integrating a number of nano-property indicators. The discrimination model is constructed using an artificial learning-machine algorithm and the discrimination model is optimized. The dairy product to be tested is discriminated, and the adulteration of dairy products can be efficiently identified without knowing the adulterant.

[0008] Preferably, the dairy product is selected from any one of cow's milk, buffalo milk, goat's milk, camel's milk, horse's milk and pig's milk, preferably any one of cow's milk, buffalo milk, goat's milk and camel's milk.

[0009] Preferably, the dairy product adulteration comprises adding adulterants to the dairy product; the adulterants include one or more of exogenous proteins, sugars, and water. The exogenous proteins include whey protein isolate and animal protein. The sugars include starch, sucrose, glucose, fructose syrup, and dextrin.

[0010] Preferably, in step S1, the membrane pore size of the membrane treatment is 0.45-1.0 μm; the rotation speed of the centrifugal treatment is 3000-5000 g, and the time of the centrifugal treatment is 10-20 min.

[0011] Preferably, in step S2, the nano characteristic parameters include one or more of particle size peak width, average particle size, hydrated particle size, particle number, light scattering intensity, and surface potential.

[0012] Preferably, in step S2, the calculation formula of the wave width coefficient is: wave width coefficient = (d / 2) 2 / D; where d / is the particle size peak width and D is the average particle size.

[0013] Preferably, in step S3, the machine learning-artificial algorithm includes one or more of the least squares method, gradient boosting method, Fisher algorithm, random forest algorithm, and logistic regression algorithm.

[0014] Preferably, in step S3, the machine learning-artificial algorithm is selected from the Fisher algorithm or the gradient boosting-random forest combined algorithm.

[0015] Preferably, in step S3, the fitting degree of the discriminant model is not less than 0.95.

[0016] Preferably, in step S4, the method for determining whether the sample to be tested is adulterated based on the comparison results includes: determining whether the sample to be tested is adulterated by setting optimal parameters based on the machine learning-artificial algorithm used in step S3, combined with experience in hyperparameter optimization and a cross-validation strategy. If the machine learning-artificial algorithm used in step S3 is the least squares method, the optimal parameters are set as follows: regularization term (L2 regularization / Ridge regression): alpha = 0.1. If the machine learning-artificial algorithm used in step S3 is the gradient boosting method, the optimal parameters are set as follows: learning rate = 0.05; number of trees = 300; maximum tree depth = 6; minimum number of leaf node samples = 25; subsampling rate = 0.7; regularization = 0.5. If the machine learning-artificial algorithm used in step S3 is the Fisher algorithm, the optimal parameter for dimensionality reduction is set to 6. If the machine learning-artificial algorithm used in step S3 is the random forest algorithm, the optimal parameter is set as follows: number of trees = 600. If the machine learning-artificial algorithm used in step S3 is a logistic regression algorithm, the optimal parameter is set as: regularization = 0.99.

[0017] Beneficial effects of the present invention: The present invention provides a highly efficient method for identifying dairy product adulteration using endogenous nano-properties. It can identify a wide range of dairy products, including bulk cow's milk and less common camel milk. This method can identify adulteration without requiring knowledge of the adulterant. The discriminant model requires few parameters to optimize and exhibits excellent fit, with an equation fit exceeding 0.98 and robust prediction results. Experiments have demonstrated that the method has a high discrimination rate of over 90% for common dairy adulterants, including exogenous protein, starch, and water. DETAILED DESCRIPTION

[0018] The following describes the embodiments of the present invention by specific embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. The various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, in the case of no conflict, the features in the following examples and embodiments can be combined with each other. In the embodiments of the present invention, unless otherwise specified, the methods used are all conventional methods, and the reagents used can be obtained from commercial sources.

[0019] An efficient method for identifying adulterated dairy products containing exogenous proteins Example 1: This embodiment provides an efficient method for identifying adulterated cow's milk with added exogenous protein, comprising the following steps: S1. The test samples were milk spiked with 0 (real milk) and 40 mL / L exogenous protein, respectively. The test samples were centrifuged at 4000 g for 15 min, and the supernatant was collected. S2. The sample filtrate was subjected to dynamic light scattering measurement using a Malvern Zetasizer Nano ZS (Malvern Instruments Limited, UK) and Zetasizer software version 8.02 to determine the average particle size (nm) and light scattering intensity (10 6 Kcps) and particle size peak width. The sample was tested at a constant temperature of 25°C with an equilibrium time of 120 seconds. The width coefficient was calculated based on the particle size peak width and average particle size. The calculation formula is: Width coefficient = (d / 2) 2 / D; where d / is the peak width of particle size, and D is the average particle size; see Table 1 for specific parameters; S3. Use machine learning and manual algorithms to fit the obtained nanostructured characteristic parameters and wave width coefficients to obtain a discriminant model; S4. Use the obtained discriminant model to perform machine learning on the nano-property parameters and wave width coefficients of the test sample and authentic dairy products to obtain comparison results; based on the comparison results, predict whether the test sample is adulterated dairy product.

[0020] The results showed that the model fitting rate was 1.0 when modeled using the Fisher algorithm, and the discrimination rate of adulterated milk reached 95.2%; using the gradient enhancement-random forest combined algorithm, the model fitting rate was 1.0, and the discrimination rate of adulterated milk reached 100%.

[0021] Example 2: This example provides a highly efficient method for identifying goat milk adulterated with exogenous protein. Compared to Example 1, this example differs in that the test samples in step S1 are goat milk adulterated with either 0 (authentic goat milk) or 40 mL / L of exogenous protein, respectively. Results showed that using the Fisher algorithm, the model fit was 0.99, with a discrimination rate of 97.3% for adulterated goat milk. Using a combined gradient boosting and random forest algorithm, the model fit was 0.99, with a discrimination rate of 100% for adulterated goat milk.

[0022] Example 3: This example provides an efficient method for identifying camel milk adulterated with exogenous protein. Compared to Example 1, this example differs in that the test samples in step S1 are camel milk adulterated with either 0 (authentic camel milk) or 40 mL / L of exogenous protein, respectively. Results show that using the Fisher algorithm, the model fit was 0.99, with a discrimination rate of 96.4% for adulterated camel milk. Using a combined gradient boosting and random forest algorithm, the model fit was 1.0, with a discrimination rate of 99.3% for adulterated camel milk.

[0023] Example 4: This example provides an efficient method for identifying buffalo milk adulterated with exogenous protein. Compared to Example 1, this example differs in that the test samples in step S1 are buffalo milk adulterated with either 0 (genuine buffalo milk) or 40 mL / L of exogenous protein, respectively. Results show that using the Fisher algorithm, the model fit was 0.99, with a discrimination rate of 97.4% for adulterated buffalo milk. Using a combined gradient boosting and random forest algorithm, the model fit was 1.0, with a discrimination rate of 99.6% for adulterated buffalo milk.

[0024] An efficient method for identifying adulterated dairy products containing added sugars Example 5: This example provides an efficient method for identifying sugar-adulterated milk. Compared to Example 1, this example differs in that the test samples in step S1 are milk adulterated with either 0 (genuine milk) or 5 g / L of starch, respectively. Results show that using the Fisher algorithm, the model fit was 0.98, with a discrimination rate of 93.6% for adulterated milk. Using a combined gradient boosting and random forest algorithm, the model fit was 1.0, with a discrimination rate of 98.6% for adulterated milk.

[0025] Example 6: This example provides an efficient method for identifying sugar-adulterated goat milk. Compared to Example 1, this example differs in that the test samples in step S1 are goat milk adulterated with either 0 (genuine goat milk) or 5 g / L of starch, respectively. The results showed that using the Fisher algorithm, the model fit was 0.98, with a discrimination rate of 97.5% for adulterated goat milk. Using a combined gradient boosting and random forest algorithm, the model fit was 1.0, with a discrimination rate of 100% for adulterated goat milk.

[0026] Example 7: This example provides an efficient method for identifying sugar-adulterated camel milk. Compared to Example 1, this example differs in that the test samples in step S1 are camel milk adulterated with either 0 (genuine) or 5 g / L of starch, respectively. Results show that using the Fisher algorithm, the model fit was 0.99, with a discrimination rate of 98.9% for adulterated camel milk. Using a combined gradient boosting and random forest algorithm, the model fit was 0.99, with a discrimination rate of 100% for adulterated camel milk.

[0027] Example 8: This example provides an efficient method for identifying sugar-adulterated buffalo milk. Compared to Example 1, this example differs in that the test samples in step S1 are buffalo milk adulterated with either 0 (genuine) or 5 g / L of starch, respectively. Results show that using the Fisher algorithm, the model fit was 0.99, with a discrimination rate of 98.9% for adulterated buffalo milk. Using a combined gradient boosting and random forest algorithm, the model fit was 0.99, with a discrimination rate of 100% for adulterated buffalo milk.

[0028] An efficient method for identifying water-adulterated dairy products Example 9: This example provides an efficient method for identifying sugar-adulterated milk. Compared to Example 1, this example differs in that the test samples in step S1 are milk adulterated with either 0 (genuine milk) or 100 mL / L of water, respectively. The results show that using the Fisher algorithm, the model fit rate was 1.0, and the discrimination rate for adulterated milk reached 100%. Using the combined gradient boosting and random forest algorithm, the model fit rate was 1.0, and the discrimination rate for adulterated milk reached 100%.

[0029] Example 10: This example provides an efficient method for identifying sugar-adulterated goat milk. Compared to Example 1, this example differs in that the test samples in step S1 are goat milk adulterated with either 0 (genuine goat milk) or 100 mL / L of water, respectively. The results showed that using the Fisher algorithm, the model fit rate was 1.0, and the discrimination rate for adulterated goat milk reached 96.3%. Using the combined gradient boosting and random forest algorithm, the model fit rate was 1.0, and the discrimination rate for adulterated goat milk reached 99.5%.

[0030] Example 11: This example provides an efficient method for identifying sugar-adulterated camel milk. Compared to Example 1, this example differs in that the test samples in step S1 are camel milk adulterated with either 0 (genuine camel milk) or 100 mL / L of water, respectively. Results show that using the Fisher algorithm, the model fit was 0.99, with a discrimination rate of 99.5% for adulterated camel milk. Using a combined gradient boosting and random forest algorithm, the model fit was 1.0, with a discrimination rate of 100% for adulterated camel milk.

[0031] Example 12: This example provides an efficient method for identifying sugar-adulterated buffalo milk. Compared to Example 1, this example differs in that the test samples in step S1 are buffalo milk adulterated with either 0% (genuine buffalo milk) or 100 mL / L of water, respectively. The results show that using the Fisher algorithm, the model fit rate was 1.0, and the discrimination rate for adulterated buffalo milk reached 100%. Using the combined gradient boosting and random forest algorithm, the model fit rate was 1.0, and the discrimination rate for adulterated buffalo milk reached 100%.

[0032] Table 1 Nano-property parameters of genuine milk and adulterated milk Average particle size (nm) <![CDATA[Light scattering intensity (10 6 Kcps)]]> Wavewidth coefficient real milk 268.9 1.37 0.29 Milk adulterated with protein 296.3 1.25 0.35 Sugar-adulterated milk 222.8 1.57 0.46 water adulterated milk 225.9 1.04 0.36 Table 2 Nano-property parameters of genuine goat milk and adulterated goat milk Average particle size (nm) <![CDATA[Light scattering intensity (10 6 Kcps)]]> Wavewidth coefficient Real goat milk 476.5 1.16 0.27 Protein adulteration of goat milk 519.6 1.29 0.34 Sugar adulteration of goat milk 435.7 0.87 0.39 Water-adulterated goat milk 400.2 0.68 0.28 Table 3 Nano-property parameters of genuine camel milk and adulterated camel milk Average particle size (nm) <![CDATA[Light scattering intensity (10 6 Kcps)]]> Wavewidth coefficient Real camel milk 1292.3 1.48 0.56 Camel milk adulterated with protein 1250.6 2.03 0.61 Sugar adulteration of camel milk 1487.7 1.92 0.59 Water adulteration of camel milk 1025.5 1.35 0.48 Table 4 Nano-property parameters of authentic buffalo milk and adulterated buffalo milk Average particle size (nm) <![CDATA[Light scattering intensity (10 6 Kcps)]]> Wavewidth coefficient Real buffalo milk 415.5 1.51 0.28 Buffalo milk adulterated with protein 412.3 1.55 0.38 Sugar-adulterated buffalo milk 451.7 1.82 0.34 Water adulteration of buffalo milk 339.0 1.21 0.22 Combined with the discrimination results of Examples 1 to 12, it can be seen that the efficient identification method for dairy product adulteration using endogenous nano-properties provided by the present invention can identify various types of dairy products, including bulk milk and rare camel milk, and can be widely used in adulteration detection in the dairy industry. This identification method can identify whether dairy products are adulterated without knowing the adulterants, and the discrimination model requires few parameters to be optimized, has good fitting, and the equation fitting degree can reach above 0.98, with good prediction effect. Experiments have shown that the method provided by the present invention has a high discrimination rate for common adulterants in dairy products, such as exogenous protein, starch, and water, with a discrimination rate of more than 90%.

[0033] The embodiments described above are merely descriptions of preferred implementations of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should fall within the scope of protection of the present invention.

Claims

1. An efficient method for identifying adulteration of dairy products based on endogenous nano-properties, characterized in that: The following steps are involved: S1. Take the test sample and standard dairy product, respectively, and perform membrane treatment or centrifugation, and take the filtrate or supernatant; S2. Characterizing the obtained filtrate or supernatant by nano-characterization to obtain nano-characteristic parameters of the test sample and the nano-characteristic parameters of the standard dairy product, and calculating the wave width coefficient of the test sample and the wave width coefficient of the standard dairy product respectively; S3. Using a machine learning-based algorithm, we fit the obtained nanostructured parameters and wave width coefficients of the test sample and standard dairy products to obtain a discriminant model. S4. Using the obtained discriminant model, compare the nano-characteristic parameters and wave width coefficients of the test sample and the standard dairy product to obtain a comparison result; and determine whether the test sample is an adulterated dairy product based on the comparison result.

2. The method according to claim 1, wherein The dairy product is selected from any one of cow's milk, buffalo's milk, goat's milk, camel's milk, horse's milk and pig's milk.

3. The method according to claim 1, wherein The adulteration of dairy products includes adding adulterants into dairy products; the adulterants include one or more of exogenous proteins, sugars, and water.

4. The method according to claim 1, wherein In step S1, the membrane pore size of the membrane treatment is 0.45-1.0 μm; the rotation speed of the centrifugal treatment is 3000-5000 g, and the time of the centrifugal treatment is 10-20 min.

5. The method according to claim 1, wherein In step S2, the nano characteristic parameters include one or more of particle size peak width, average particle size, hydrated particle size, particle number, light scattering intensity, and surface potential.

6. The method according to claim 1, wherein In step S2, the calculation formula of the wave width coefficient is: wave width coefficient = (d / 2) 2 / D; Where d / is the peak width of particle size, and D is the average particle size.

7. The method according to claim 1, wherein In step S3, the machine learning-artificial algorithm includes one or more of the least squares method, gradient boosting method, Fisher algorithm, random forest algorithm, and logistic regression algorithm.

8. The method according to claim 1, wherein In step S3, the machine learning-artificial algorithm is selected from the Fisher algorithm or the gradient boosting-random forest combination algorithm.

9. The method according to claim 1, wherein In step S3, the fitting degree of the discriminant model is not less than 0.

95.

10. The method according to claim 1, wherein In step S4, the method for determining whether the sample to be tested is an adulterated dairy product based on the comparison results includes: setting optimal parameters based on the machine learning-artificial algorithm used in step S3, combined with hyperparameter optimization experience and cross-validation strategy, to determine whether the sample to be tested is an adulterated dairy product.