Method for identifying flavor of finished milk based on flavor of raw milk

Abstract

The invention discloses a method for establishing an algorithm model for identifying finished milk based on flavor of raw milk, which comprises the following steps: (1) establishing a flavor data set, (2) establishing a model algorithm: taking each element in sensory evaluation data and electronic nose data as an independent dependent variable, and measuring each dependent variable to obtain a flavor data set; the maximum value and the minimum value of different dependent variables are obtained, the interval between the maximum value and the minimum value serves as the evaluation criterion of one dependent variable, and the similarity conditions are calculated through discussing multiple potential conditions in the calculation process. According to the method, UHT milk flavor judgment in the actual production process is effectively simplified, the sensory flavor quality of subsequent finished products is predicted in advance, pre-screening and pre-judgment are achieved, the quality stability and the production efficiency are greatly improved, the satisfaction degree of consumers is improved, and meanwhile the production cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of dairy product detection, and particularly relates to a method for establishing an algorithm model for identifying the flavor of finished milk based on the flavor of raw milk and its application. Background Art

[0002] At present, the production of liquid milk generally goes through processes such as pre-inspection, standardization, heat treatment (such as ultra-high temperature instantaneous sterilization (UHT)), homogenization, filling, and packaging. Ultra-high temperature instantaneous sterilization (UHT) milk is the largest category of liquid milk products on the market at present. However, during the production process, differences in raw milk can cause UHT milk to possibly have different flavor characteristics, and even form off-flavors, which will have an irreversible impact on the quality of subsequent finished products. UHT milk with different flavors entering the subsequent production process often results in uneven product quality, seriously affecting the enterprise's benefits and further exacerbating the potential threat to consumers' health. At present, for the comparison of the flavor changes of raw milk and the flavor array of UHT finished milk during the production and processing process, both at home and abroad, only simple sensory evaluation and differential analysis of volatile flavor compounds based on GC-MS (such as VIP value, P value, and FC value analysis) are used. This single differential analysis cannot comprehensively and accurately evaluate the true difference degree between the flavor of raw milk and the finished product during the production process. Therefore, analyzing and predicting the flavor quality of raw milk after heat processing can provide timely and accurate guiding information for the production and processing of UHT milk, and timely take control measures to avoid economic losses.

[0003] Currently, the convenient and efficient identification and differentiation of UHT milk with different flavor characteristics have not attracted attention. More previous technical solutions focused on differentiating and identifying raw milk, pasteurized milk, and UHT milk based on their physical and chemical property characteristics. For example, Lu Jiaping (Lu Jiaping, Zhang Shuwen, Liu Lu, etc. Method for differentiating raw milk, pasteurized milk, and UHT milk [P]. Beijing: CN103245602B, November 5, 2014.) et al. used trisodium citrate to effectively chelate colloidal calcium in milk and adopted a turbidimetry method to differentiate raw milk, pasteurized milk, and UHT milk.

[0004] At present, the differentiation of the different flavor characteristics between raw milk and its processed UHT finished milk during the processing process is still based on sensory evaluation and basic difference analysis of volatile compounds. For example, Wen Rong et al. (Wen Rong, Zhang Dongjie, Gao Fei, etc. A method for evaluating the oxidative flavor of fermented milk [P]. Inner Mongolia Autonomous Region: CN202210503128.4, 2023-11-17.) used comprehensive two-dimensional gas chromatography-olfactometry-mass spectrometry (GC×GC-O-MS) technology to qualitatively and quantitatively analyze the flavor substances enriched in fermented milk, and combined sensory tasting, dilution analysis and olfactometry to determine the oxidative markers and the contribution degree of each marker to the oxidative flavor. However, this sensory and machine evaluation method is not only single, but also cannot comprehensively judge the flavor quality stability of processed raw milk, and may even cause large deviations, seriously reducing production efficiency.

[0005] In addition, heat treatment will affect the flavor precursor substances in milk, such as peptides and amino acids produced by protein hydrolysis, and compounds generated through the Maillard reaction. The changes of these precursor substances during heat treatment are the key factors affecting the different flavors of the final products. Therefore, it is necessary to standardize the evaluation of its flavor during the processing of liquid milk to ensure the sensory flavor and quality stability of the subsequent finished products during the circulation process. Summary of the Invention

[0006] Aiming at the single method for judging the difference degree between the flavor of raw milk and the standardized finished product during the heat processing process, which makes the judgment of product quality lag, may affect the product quality stability, and seriously affects the enterprise reputation. The inventor of the present invention developed an accurate and efficient algorithm model for evaluating the similarity of milk flavor based on electronic nose discrimination and human sensory analysis database. Using the algorithm model, the group closest to the flavor of UHT milk can be screened to realize the rapid pre-identification of the flavor of raw milk after heat processing. The present invention effectively simplifies the flavor determination of UHT milk in the actual production process, predicts the sensory flavor quality of the subsequent finished products in advance, realizes pre-screening and pre-judgment, greatly improves the quality stability and production efficiency, enhances the consumer satisfaction, and reduces the production cost at the same time.

[0007] Therefore, to solve the above technical problems, the technical solution provided by the present invention is: a method for establishing an algorithm model for identifying finished milk based on the flavor of raw milk, comprising the following steps: (1) Construction of flavor data set s1. Sensory evaluation The sensory analysis adopts the quantitative descriptive analysis method, that is: QDA. The temperature of the sensory analysis laboratory is controlled at 25 °C ± 2 °C, and the online Appsense software is used to collect test data; s2. Electronic nose analysis Take raw milk samples, add water for incubation, perform electronic nose analysis, and analyze the volatile substances in the samples; based on the analysis results of the MXT-5-FID1 and MXT-1701-FID2 chromatographic columns in the electronic nose and comparison with the electronic nose compound database, evaluate and determine the types and intensities of the detected compounds; (2) Establishment of model algorithm Regard each element in the sensory evaluation data and electronic nose data as a separate dependent variable. By measuring each dependent variable, obtain the maximum and minimum values of different dependent variables. Take the interval between the maximum and minimum values as the evaluation standard for a dependent variable. During the calculation process, explore multiple potential situations and calculate the similarity situations respectively. The specific operation method is as follows: s1. When neither the dependent variable of the finished product nor the dependent variable at a certain temperature exists, the dependent variable is strongly similar, and the similarity is 100%; s2. When the dependent variable of the finished product does not exist while the dependent variable at a certain temperature exists, the dependent variable is strongly dissimilar, and the similarity is 0%; s3. When the maximum value of the dependent variable of the finished product A i is less than the maximum value at a certain temperature B i If the minimum value of the finished product a i is less than or equal to the minimum value at a certain temperature b i it is weakly similar, and the differentiation is calculated through the following range interval formula l i where d is for ensuring the effectiveness of the operation and adjusting the order of magnitude. For the sensory evaluation data and electronic nose data, the d brought in are different; ; For the sensory evaluation data and electronic nose data, the d brought in are different; s4. When the maximum value of the dependent variable of the finished product A i is less than the maximum value at a certain temperature B i If the minimum value of the finished product a i is greater than the minimum value at a certain temperature b i it is weakly similar, and the differentiation is calculated through the following range interval formula l i where the relevant letters are the same as above; ; s5. When the maximum value of the dependent variable of the finished product Ai The maximum value greater than a certain temperature B i When, if the minimum value of the finished product a i Less than or equal to the minimum value of a certain temperature b i , the dependent variable is strongly similar, and the similarity is 100%; S6. When the maximum value of the dependent variable of the finished product A i Greater than the maximum value of a certain temperature B i When, if the minimum value of the finished product a i Greater than the minimum value of a certain temperature b i , calculate the differentiation through the following range interval formula l i , where the relevant letters are the same as above; ; S7. For all differentiations l i , convert it into similarity through the following formula s i ; ; S8. Finally, average the similarities of all dependent variables as the final similarity result S ; ; This algorithm is measured with multiple batches of data, and finally adjusts the parameters to obtain a relatively stable effect.

[0008] Furthermore, in step (1), the Appsense software in s1 collects sensory test data.

[0009] Preferably, for the electronic nose analysis in step (1) of s2, take 5 g of raw milk sample and place it in a 20 ml headspace bottle, add 3 - 5 ml of water, incubate at 40 °C for 15 - 40 min, the inlet temperature is 200 - 220 °C, desorb for 10 - 30 s, and conduct electronic nose analysis to desorb the volatile substances in the sample.

[0010] Meanwhile, the present invention also provides an application method of the model, and this method includes the following steps: (1) Raw material preparation: Select fresh and high-quality raw milk as the test sample, and conduct preliminary detection and filtration; (2) Pretreatment: Pretreat the raw milk, including removing impurities, uniformly stirring, and sterilization treatment to ensure the smooth progress of subsequent treatment processes; (3)Thermal processing: Feed the pre-treated raw milk into a UHT processing device for sterilization to obtain UHT milk for standby use, and subject the remaining raw milk to low-temperature treatment under pressure to obtain treated milk for standby use; (4)Collect the treated raw milk and UHT milk, conduct sensory evaluation and electronic nose analysis respectively, and incorporate the sensory evaluation and electronic nose data into the calculation of the algorithm model to evaluate the flavor characteristics of the samples.

[0011] Further, in step (3), feed the pre-treated raw milk into a UHT processing device, continuously heat it by quickly raising the temperature to an ultra-high temperature (usually exceeding 135 °C), and then quickly cool it in an extremely short time.

[0012] Further, in step (3), subject the raw milk to heat pressing treatment for 5, 10, and 18 minutes respectively.

[0013] The present invention provides an identification system based on an algorithm model for identifying finished milk from raw milk flavor, which includes a data entry module, a data processing module, and a result output module, wherein the data processing module executes the algorithm model for calculation.

[0014] The present invention also provides a device for identifying finished milk, which encapsulates the system in a computer, such as a portable computer. Specifically, it includes the system as claimed in claim 8, a display, a mouse, and a keyboard; the display serves as an output terminal for displaying the identification result; the mouse and the keyboard serve as input terminals for realizing human-computer interaction, such as outputting or importing data.

[0015] Due to the quality differences of raw milk itself, the raw milk after UHT treatment often presents different flavor states, and even forms bad flavors, which will have an irreversible impact on the subsequent finished product quality. Therefore, it is very necessary to pre-screen the milk according to the flavor characteristics of UHT milk. However, in the actual production process, only analyzing the basic flavor differences of the heat-treated milk is not sufficient to comprehensively reflect the product quality stability. The present invention establishes a calculation model of similarity with UHT milk based on the sensory evaluation and electronic nose analysis data sets, and quantifies its system similarity from the overall flavor perspective. Based on this method, the raw milk can be processed and its flavor can be judged efficiently and quickly, and the flavor analysis result of UHT milk can be approximated to the greatest extent.

[0016] The present invention analyzes the similarity between the raw milk treated under different thermal processing conditions and UHT milk. In terms of data selection, in order to comprehensively consider from both subjective and objective perspectives, the sensory evaluation data and electronic nose data of the treated raw milk are respectively selected. Since this algorithm calculates by measuring the ranges of different dependent variables, the following points are mainly considered for the treatment of the dependent variables: 1. After imposing certain restrictions on temperature and milk batches, there is still a certain degree of randomness in the obtained data. For the existing data, this randomness cannot be explored for its regularity by computer means. Therefore, by setting the possible range it may exist in, the impact of a certain random data on the whole can be reduced. 2. The interaction within the dependent variable will also affect the result. That is, in the sensory evaluation data, there will be a certain correlation within different tastes and flavors; however, this correlation only shows a weak difference in the sensory evaluation results. In the electronic nose data, a general electronic nose only reflects the substance intensity with signal values, which is very vague. Using a mass spectrometry-based electronic nose with substance analysis can help to specify the intensity changes between different compounds and further quickly clarify the changes in its flavor compounds. However, due to the large variety, different contents, and interactive intensity of milk source flavor compounds, it is difficult to directly evaluate the flavor differences of milk through the intensity of several substances; in the sensory evaluation and electronic nose data, these factors are interrelated and there is also an interaction relationship, that is, the difference groups that cannot be distinguished by human sensory evaluation may have differences in compound composition and intensity. Correspondingly, the compound differences reflected in the electronic nose data do not necessarily correspond to the sensory evaluation caused by the corresponding compounds. Therefore, by setting the possible range it may exist in and calculating by separating each variable as independently as possible, it helps to reduce this interaction. 3. In the existing similarity calculation methods, generally, the similarity is calculated through the difference of a single value or the distance of an overall function, but both of these methods are greatly affected by unstable factors. Due to the consideration of the first two factors, this algorithm measures by using the range calculation method.

[0017] Based on the electronic nose discrimination and the human sensory analysis database, the present invention develops an accurate and efficient algorithm model for evaluating the similarity of milk flavors. Using the algorithm model, the group closest to the flavor of UHT milk can be screened to achieve rapid pre-discrimination of the flavor of raw milk after heat processing, quantify the overall flavor proximity between different milk samples, systematically evaluate the flavor difference between heat-processed raw milk and standardized UHT milk, and provide a reference for realizing the flavor simulation prediction of UHT milk. The present invention effectively simplifies the determination of UHT milk flavor in the actual production process, predicts the sensory flavor quality of the subsequent finished product in advance, realizes pre-screening and pre-determination, greatly improves the quality stability and production efficiency, enhances the satisfaction of consumers, and at the same time reduces the production cost. Brief Description of the Drawings

[0018] Figure 1 This is the technical roadmap of the present invention.

[0019] Figure 2 This is the illustration of the milk flavor similarity algorithm model of the present invention.

[0020] Figure 3Sensory evaluation (A), electronic nose data (B), flavor substance classification (C), and flavor proximity (D) for raw milk treated for 5, 10, 15, and 18 minutes.

[0021] Figure 4 Sensory evaluation (A), electronic nose data (B), flavor substance classification (C), and flavor proximity (D) for raw milk treated for 11 - 17 minutes. Detailed implementation manner

[0022] Based on the sensory evaluation and electronic nose data sets, the present invention establishes an algorithm model for evaluating the flavor similarity between raw milk and UHT milk, quantifies the overall flavor proximity between different milk samples, and can systematically evaluate the flavor difference between heat - processed raw milk and standardized UHT milk. For the specific technical route, please refer to Figure 1 .

[0023] The following further describes the present invention in detail with reference to the drawings and embodiments. It should be noted that the following embodiments are intended to facilitate the understanding of the present invention and do not limit it in any way.

[0024] Embodiment 1. The method for establishing the algorithm model of the present invention is as follows: (1) Construction of flavor data set s1. Sensory evaluation The sensory analysis adopts the quantitative descriptive analysis method (QDA). The sensory evaluation is carried out by 11 evaluators in the sensory panel. All members have received strict professional training in accordance with ISO8586 and GB / T16291 standards. The temperature in the sensory analysis laboratory is controlled at about 25 °C ± 2 °C. The online Appsense software is used to collect test data; s2. Electronic nose analysis Take 5 g of raw milk sample and place it in a 20 - ml headspace vial, add 3 - 5 ml of water, incubate at 40 °C for 15 - 40 min, the inlet temperature is 200 - 220 °C, and desorb for 10 - 30 s for electronic nose analysis to analyze the volatile substances in the sample. Based on the analysis results of the MXT - 5 - FID1 and MXT - 1701 - FID2 chromatographic columns in the electronic nose and the comparison with the electronic nose compound database, the types and intensities of the detected compounds are evaluated and determined.

[0025] (2) Establishment of model algorithm Regard each element in the sensory evaluation data and the electronic nose data as a separate dependent variable. By measuring each dependent variable, obtain the maximum and minimum values of different dependent variables, and take the interval between the maximum and minimum values as the evaluation standard for a dependent variable. During the calculation process, explore various potential situations and calculate the similarity situations respectively. The specific operation method is as follows: s1. When the dependent variable of the finished product and the dependent variable at a certain temperature do not exist, the dependent variable is strongly similar, and the similarity is 100%; s2. When the dependent variable of the finished product does not exist while the dependent variable at a certain temperature exists, the dependent variable is strongly dissimilar, and the similarity is 0%; s3. When the maximum value of the dependent variable of the finished product A i is less than the maximum value at a certain temperature B i and if the minimum value of the finished product a i is less than or equal to the minimum value at a certain temperature b i it is weakly similar, and the differentiation is calculated through the following range interval formula l i where d is for ensuring the effectiveness of the operation and adjusting the order of magnitude. For sensory evaluation data and electronic nose data, the d brought in is different; ; For sensory evaluation data and electronic nose data, the d brought in is different; s4. When the maximum value of the dependent variable of the finished product A i is less than the maximum value at a certain temperature B i and if the minimum value of the finished product a i is greater than the minimum value at a certain temperature b i it is weakly similar, and the differentiation is calculated through the following range interval formula l i where the relevant letters are the same as above; ; s5. When the maximum value of the dependent variable of the finished product A i is greater than the maximum value at a certain temperature B i and if the minimum value of the finished product a i is less than or equal to the minimum value at a certain temperature b i the dependent variable is strongly similar, and the similarity is 100%; s6. When the maximum value of the dependent variable of the finished product A i is greater than the maximum value at a certain temperature B i and if the minimum value of the finished product a i is greater than the minimum value at a certain temperatureb i , calculate the differentiation through the following range formula l i , where the relevant letters are the same as above; ; S7. For all differentiations, convert them into similarity degrees through the following formula s i ; ; S8. Finally, average the similarity degrees of all dependent variables as the final similarity degree result S ; ; This algorithm is measured with multiple batches of data, and finally adjusts the parameters to obtain a relatively stable effect.

[0026] Embodiment 2. An application embodiment of the algorithm model of the present invention is as follows: 1. Raw material preparation: Select fresh and high-quality raw milk as the test sample, and conduct preliminary detection and filtration to ensure that the quality of the raw milk (protein, fat content, etc.) meets the requirements.

[0027] 2. Pretreatment: Pretreat the raw milk, including removing impurities, uniformly stirring, and sterilization treatment to ensure the smooth progress of subsequent treatment processes.

[0028] 3. Thermal processing: Feed the pretreated raw milk into the UHT treatment equipment, continuously heat it by quickly raising the temperature to an ultra-high temperature (usually exceeding 135 °C), and then quickly cool it in an extremely short time to achieve an effective sterilization effect and obtain UHT milk for later use. The remaining raw milk is treated by low-pressure heat treatment (the raw milk is respectively subjected to heat pressing treatment for 5, 10, and 18 minutes) to obtain the treated group milk for later use.

[0029] 4. Collect the treated group raw milk and UHT milk, conduct sensory evaluation and electronic nose analysis respectively, and incorporate the sensory evaluation and electronic nose data into the calculation of the above-mentioned algorithm model to evaluate the flavor characteristics of the samples.

[0030] According to the analysis results ( Figure 2 in A), in the sensory evaluation, the score gap between UHT milk and the treated group milk in each specific evaluation index is relatively small (within 1-2 points), and it is difficult to distinguish the evaluation. And in the electronic nose data (such as Figure 3 in B, Figure 3 in C), compared with different treated groups and UHT milk, it shows an interactive state in terms of the types and intensities of compounds such as aldehydes, alcohols, amines, esters, acids, and ketones, and it is difficult to directly distinguish the evaluation. Therefore, a model algorithm is established to calculate the flavor proximity between the treated group and UHT milk. The results are asFigure 3 As shown in D, it was found that as time increased, the flavor proximity between the treatment group and UHT milk showed a trend of first increasing and then decreasing. At 5 minutes, the flavor proximity was only 67%, while at around 10 and 15 minutes, the flavor proximity exceeded 80%. After 18 minutes, the flavor proximity dropped to 72.9% again. Therefore, there may be an optimal treatment time point within the range of 10 - 18 minutes. Thus, the raw milk was further treated for 11 - 17 minutes (each minute as a treatment group) and flavor analysis was carried out.

[0031] Further analysis of the electronic nose results of the 11 - 17 min treatment group found that as the treatment time increased, the intensities of aldehydes (acetaldehyde, octanal) in the raw milk increased to varying degrees, the proportion of alkanes (decane, pentane, butane) decreased, and the intensities of alcohols (3 - heptanol, ethanol, 2 - octanol, 2 - hexanol, cyclohexanol, methanol) also increased.

[0032] Please refer to Figure 4 , through model calculation, it was determined that the treatment time for the raw milk treated by this method to have the best flavor proximity to UHT milk was between 11 - 17 minutes, and the final overall flavor proximity could reach over 90% (the highest was 91.4%), achieving a relatively high flavor approximation level with UHT milk. From the perspective of the intensities of specific compounds, compared with real UHT milk, there were varying degrees of differences in the contents of aldehydes, acids, esters, and ketones in the treated raw milk. However, through the systematic analysis of this algorithm, this degree of difference was quantitatively determined in the form of a similarity index.

Claims

1. A method for establishing an algorithm model for identifying finished milk based on the flavor of raw milk, characterized in that It includes the following steps: (1) Construction of flavor dataset s1. Sensory evaluation The sensory analysis adopts the quantitative descriptive analysis method, i.e., QDA. The temperature of the sensory analysis laboratory is controlled at 25 °C ± 2 °C; s2. Electronic nose analysis Take raw milk samples, add water for incubation, and conduct electronic nose analysis to analyze the volatile substances in the samples; Based on the comparison of the analysis results of the MXT-5-FID1 and MXT-1701-FID2 chromatographic columns in the electronic nose with the electronic nose compound database, evaluate and determine the types and intensities of the detected compounds; (2)Establishment of model algorithm Regard each element in the sensory evaluation data and electronic nose data as a separate dependent variable. By measuring each dependent variable, obtain the maximum and minimum values of different dependent variables, and take the interval between the maximum and minimum values as the evaluation criterion for a dependent variable. During the calculation process, explore various potential situations and calculate the similarity situations respectively. The specific operation method is as follows: s1. When neither the dependent variable of the finished product nor the dependent variable at a certain temperature exists, the dependent variable is strongly similar, and the similarity is 100%; s2. When the dependent variable of the finished product does not exist while the dependent variable at a certain temperature exists, the dependent variable is strongly dissimilar, and the similarity is 0%; S3. When the maximum value of the dependent variable of the finished product A i is less than the maximum value of a certain temperature B i and the minimum value of the finished product a i is less than or equal to the minimum value of a certain temperature b i , it is weakly similar, and the differentiation is calculated through the following range interval formula l i , where d For the purpose of ensuring the effectiveness of the operation and adjusting the order of magnitude, the d brought in for sensory evaluation data and electronic nose data is different; ； For the sensory evaluation data and the electronic nose data, the d differ; S4. When the maximum value of the dependent variable of the finished product A i is less than the maximum value of a certain temperature B i and the minimum value of the finished product a i is greater than the minimum value of a certain temperature b i , it is weakly similar, and the differentiation is calculated through the following range interval formula l i , where the relevant letters are the same as above; ； S5. When the maximum value of the dependent variable of the finished product A i is greater than the maximum value of a certain temperature B i and if the minimum value of the finished product a i is less than or equal to the minimum value of a certain temperature b i then the dependent variable is strongly similar with a similarity of 100%; S6. When the maximum value of the dependent variable of the finished product A i is greater than the maximum value of a certain temperature B i and if the minimum value of the finished product a i is greater than the minimum value of a certain temperature b i calculate the differentiation through the following range interval formula l i where the relevant letters are the same as above; ； s7. For all differentiations l i , convert them into similarity through the following formula s i ; ； s8. Finally, average the similarities of all dependent variables as the final similarity result S ; ； This algorithm is calculated with multiple batches of data, and finally adjusts the parameters to obtain a relatively stable effect.

2. The establishment method according to claim 1, wherein: The Appsense software in s1 of step (1) collects sensory test data.

3. The establishment method according to claim 1, characterized in that: For the electronic nose analysis in s2 of step (1), take 5 g of raw milk samples and place them in a 20 ml headspace vial, add 3 - 5 ml of water, incubate at 40 °C for 15 - 40 min, the inlet temperature is 200 - 220 °C, and desorb for 10 - 30 s, then conduct electronic nose analysis to analyze the volatile substances in the samples.

4. An algorithm model for identifying finished milk based on the flavor of raw milk obtained by the establishment method according to any one of claims 1 to 3.

5. A method for identifying the flavor of finished milk based on the flavor of raw milk, the method includes the following steps: (1)Raw material preparation: Select fresh and high-quality raw milk as the test sample, and conduct preliminary detection and filtration; (2)Pretreatment: Pretreat the raw milk, including removing impurities, uniformly stirring, and sterilization treatment to ensure the smooth progress of subsequent treatment processes; (3)Thermal processing treatment: Send the pretreated raw milk into a UHT treatment device for sterilization to obtain UHT milk for standby, and subject the remaining raw milk to low-temperature treatment under pressure to obtain treated group milk for standby; (4)Collect the treated group raw milk and UHT milk, conduct sensory evaluation and electronic nose analysis respectively, and jointly input the sensory evaluation and electronic nose data into the algorithm model described in claim 4 for calculation to evaluate the flavor characteristics of the samples.

6. The method according to claim 5, wherein: In step (3), send the pretreated raw milk into the UHT treatment device, continuously heat it by quickly raising the temperature to an ultra-high temperature, and then quickly cool it in a very short time.

7. The method according to claim 5, wherein: In step (3), the raw milk is respectively subjected to heat pressing treatment for 5, 10, and 18 minutes.

8. An identification system for an algorithm model that identifies finished milk based on the flavor of raw milk, characterized in that, It includes a data entry module, a data processing module, and a result output module, wherein the data processing module performs calculations according to the algorithm model described in claim 4.

9. An apparatus for identifying finished milk, characterized in that, Enclose the system as claimed in claim 8 in a computer, such as a portable computer.

10. The device according to claim 9, characterized in that, Comprise the system as claimed in claim 8, a display, a mouse and a keyboard; the display serves as an output end for displaying the recognition result; the mouse and the keyboard serve as input ends for realizing human-computer interaction, such as outputting or importing data.

Images