A method for rapid identification and evaluation of different oxidized flavors of milk based on electronic nose

Abstract

The invention belongs to the technical field of food quality monitoring, and discloses a method for quickly identifying and evaluating different oxidation smells of milk based on an electronic nose, which specifically includes the following steps: S1, establishing reference samples of three oxidation smells of milk; S2, electronic nose detection, respectively using metal oxide sensors on the electronic nose to detect the gases emitted by the established three oxidation smell reference samples; S3, electronic nose effective sensor screening, based on the detection results of the three oxidation smell reference samples, analyzing and identifying sensors useful for sample differentiation, and establishing an electronic nose sensor response graph; S4, based on the electronic nose sensor response graph, establishing a discrimination model and a sensory evaluation model; S5, sample testing. The present invention fundamentally distinguishes three oxidation smells that are difficult to distinguish by manual sensory evaluation by making reference samples of different raw materials and different oxidation degrees, and establishes a sensory evaluation model through electronic nose technology to achieve sensory score prediction of the sample to be tested.

Description

Technical Field

[0001] The invention relates to the technical field of food quality monitoring, and in particular to a method for quickly identifying and evaluating different oxidation flavors of milk based on an electronic nose. Background Art

[0002] During the storage and sales process, milk is easily affected by factors such as light and temperature, which can cause oxidation and produce complex odors, thereby reducing its quality. The reason why milk has a complex odor after oxidation is that the substances that cause milk oxidation are protein and fat. The odors produced by the oxidation of the two are different, and the conditions and time of oxidation of the two are different, so the odors presented under different oxidation conditions are also different. In addition, as consumers' demand for milk nutrition and flavor continues to increase, high-protein, low-fat, thick milk and other milk products are emerging in an endless stream. Due to the different protein and fat content of these products, the oxidized taste after oxidation will also be different.

[0003] When conducting sensory evaluation of milk odor, evaluators often describe the various odors of milk after oxidation as metallic, plastic, paint, cardboard, stale, rancid, and sunny. Since the taste of milk after oxidation is different and it is easy to be confused with the odor caused by packaging materials, microorganisms, etc., it is difficult to establish a unified sensory evaluation standard for oxidation taste.

[0004] The existing methods for detecting milk odor mainly include artificial sensory evaluation technology, protein and fat oxidation detection technology and electronic nose technology. The current status of the above three methods is as follows:

[0005] Artificial sensory evaluation technology is mature, but it has not differentiated and evaluated the different oxidation flavors in milk. For the sensory evaluation of milk odor, the usual practice is to form a sensory evaluation team and then let the evaluation team describe and analyze the odors of different milk products and evaluate their intensity. However, artificial sensory technology has the disadvantages of poor reproducibility, low accuracy, and long time. At present, artificial sensory technology can hardly classify the oxidation taste of milk into protein oxidation taste, fat oxidation taste and comprehensive oxidation taste for evaluation.

[0006] The protein and fat oxidation detection technology is complex to operate and has poor correlation with artificial sensory organs. Protein oxidation usually causes protein degradation, conformational changes, etc. Fat oxidation is mainly caused by the easy oxidation of unsaturated fatty acids to produce rancidity, and when metal ions are present, the fat oxidation rate will be greatly increased. Protein detection methods are mainly through the detection of protein content (such as biuret method, Kjeldahl method) and structural changes (such as infrared method, electrophoresis method). These methods require the use of chemical reagents, are time-consuming, and have a low degree of integration with artificial sensory evaluation; while fat detection methods are mainly to quantify the loss of unsaturated fatty acids and the generation of peroxides and conjugated dienes, such as the peroxide value (POV) method and thiobarbituric acid reactants (TBARs) value. Experiments have found that the determination results of TBARs values ​​are poorly correlated with sensory evaluation. Studies have shown that sensory evaluation is more sensitive to the analysis of oxidation properties in dairy products than traditional instrumental methods (such as POV, TBARs value and gas chromatography). Sensory evaluation technology is simple and intuitive to operate, and can most directly reflect the acceptability of food.

[0007] Research status of electronic nose technology. Electronic nose is an intelligent sensory analysis instrument. It is more and more widely used in many industries because of its objective, accurate, fast, easy to carry and non-damaging characteristics. The depth and breadth of the application of electronic nose in the dairy industry are also expanding. At present, electronic nose is mainly used in the following aspects in the dairy industry: prediction of the maturity and shelf life of dairy products, analysis of volatile substances in dairy products, analysis of microorganisms in dairy products, identification of the origin of dairy products, etc. There are many studies on the application of electronic nose to establish models, and there are many data statistical methods for establishing models, including neural networks, support vector machines, popular algorithms, cluster analysis, principal component analysis, fuzzy comprehensive evaluation, discriminant analysis, etc. For example, the patent with publication number CN105181761A discloses a method for quickly identifying the absorbed dose of tea irradiation using an electronic nose. In this patent, the response diagram of different sensors of the electronic nose to tea with different irradiation doses is established, PCA and LDA are used to extract characteristic signals, and then the Mahalanobis distance, Euclidean distance, and Fisher method are used to establish a qualitative discriminant model. For example, the patent with publication number CN106501470A discloses a method for evaluating the flavor grade of mustard sauce by jointly using the taste system and the electronic nose. The patent combines the traditional sensory system, electronic nose, and electronic tongue with chemical measured values ​​to score and judge the mustard sauce products, and uses a random forest algorithm for nonlinear fitting, thereby achieving grade evaluation of the mustard sauce products.

[0008] Although the above three methods can detect the degree of milk deterioration to a certain extent, they cannot accurately classify the milk oxidation taste into protein oxidation taste, fat oxidation taste and comprehensive oxidation taste for evaluation. Summary of the invention

[0009] The purpose of the present invention is to propose a method for quickly identifying and evaluating different oxidation smells of milk based on an electronic nose, mainly to solve the problem that the prior art cannot accurately classify the oxidation smell of milk into protein oxidation smell, fat oxidation smell and comprehensive oxidation smell for evaluation.

[0010] The technical solution used to achieve the purpose of the present invention is: a method for quickly identifying and evaluating different oxidation flavors of milk based on an electronic nose, which specifically includes the following steps:

[0011] Step S1, establishing reference samples of three oxidation smells of milk, the three oxidation smells being protein oxidation smell, fat oxidation smell, and comprehensive oxidation smell of protein oxidation and fat oxidation;

[0012] Step S2, electronic nose detection, using the metal oxide sensor on the electronic nose to detect the gases emitted by the three established oxidative odor reference samples;

[0013] Step S3, screening of effective electronic nose sensors, based on the detection results of three oxidative taste reference samples, analyzing and identifying sensors useful for sample differentiation, and establishing an electronic nose sensor response graph;

[0014] Step S4, establishing a discrimination model and a sensory evaluation model based on the electronic nose sensor response graph.

[0015] Furthermore, the method further includes the following steps: Step S5, sample testing, taking whole milk treated with different light exposure times as the sample to be tested for sample testing.

[0016] Furthermore, in step S1, the preparation methods of the three reference samples of oxidation taste are as follows:

[0017] The method for preparing a protein oxidation taste reference sample is as follows: sterile fresh skim milk is irradiated with strong light to obtain oxidized milk, and then the fresh skim milk and the oxidized skim milk are compounded to obtain oxidation taste reference samples with different oxidation degrees;

[0018] The method for preparing a fat oxidation taste reference sample is as follows: diluting sterile fresh cream to obtain a fresh cream solution, and irradiating the fresh cream solution with strong light to obtain an oxidized cream solution; then compounding the fresh cream solution and the oxidized cream solution to obtain oxidation taste reference samples with different oxidation degrees;

[0019] The preparation method of the comprehensive oxidation taste reference sample is as follows: sterile fresh whole milk is irradiated with strong light to obtain oxidized milk, and then the fresh whole milk and the oxidized whole milk are compounded to obtain oxidation taste reference samples with different oxidation degrees.

[0020] Furthermore, in step S2, the electronic nose detection method is specifically as follows:

[0021] Step S201, placing the sample to be tested into a stoppered sample bottle and leaving it at room temperature for 3 to 4 hours;

[0022] Step S202, inserting the air supply needle and the sample absorption needle of the electronic nose into the stoppered sample bottle, and the 10 metal sensors of the electronic nose detect the gas emitted by the sample.

[0023] Furthermore, in step S3, a loading analysis method is used to analyze and identify sensors useful for sample differentiation.

[0024] Further, in step S4, the regression equations of different oxidation flavors of the discriminant model are established as follows:

[0025] Y1=a1X1+b1X2+c1X3+d1X4+e1X5……n1X n +k1

[0026] Y2=a2X1+b2X2+c2X3+d2X4+e2X5……n2X n +k2

[0027] Y3=a3X1+b3X2+c3X3+d3X4+e3X5……n3X n +k3

[0028] In the formula, Y1, Y2, and Y3 represent the discrimination results of protein oxidation smell, fat oxidation smell, and comprehensive oxidation smell, respectively;

[0029] X1, X2, X3, X4, X5...X n They respectively represent the values ​​of the first sensor, the second sensor, the third sensor, the fourth sensor, the fifth sensor, ... the nth sensor useful for sample differentiation;

[0030] a1, a2, and a3 represent the discrimination coefficients of the first sensor useful for sample discrimination in the protein oxidation taste discrimination results, fat oxidation taste discrimination results, and comprehensive oxidation taste discrimination results, respectively;

[0031] b1, b2, and b3 represent the discrimination coefficients of the second sensor useful for sample discrimination in the protein oxidation taste discrimination results, fat oxidation taste discrimination results, and comprehensive oxidation taste discrimination results, respectively;

[0032] c1, c2, and c3 represent the discrimination coefficients of the third sensor useful for sample discrimination in the protein oxidation taste discrimination results, fat oxidation taste discrimination results, and comprehensive oxidation taste discrimination results, respectively;

[0033] d1, d2, and d3 represent the discrimination coefficients of the fourth sensor useful for sample discrimination in the protein oxidation taste discrimination results, fat oxidation taste discrimination results, and comprehensive oxidation taste discrimination results, respectively;

[0034] e1, e2, and e3 represent the discrimination coefficients of the fifth sensor useful for sample discrimination in the protein oxidation taste discrimination results, fat oxidation taste discrimination results, and comprehensive oxidation taste discrimination results, respectively;

[0035] n1, n2, and n3 represent the discrimination coefficients of the nth sensor useful for sample discrimination in the protein oxidation taste discrimination results, fat oxidation taste discrimination results, and comprehensive oxidation taste discrimination results, respectively;

[0036] k1, k2, and k3 are the constants of the odor discrimination models of protein oxidation smell, fat oxidation smell, and comprehensive oxidation smell, respectively.

[0037] Further, in step S4, the regression equations of different oxidation flavors of the sensory evaluation model are established as follows:

[0038] Y i =α i +β i1 X1+β i2 X2+β i3 X3+β i4 X4+β i5 X5……+β in X n

[0039] Y j =α j +β j1 X1+β j2 X2+β j3 X3+β j4 X4+β j5 X5……+β jn X n

[0040] Y k =α k +β k1 X1+β k2 X2+β k3 X3+β k4 X4+β k5 X5……+β kn X n

[0041] Where Y i , Y j , Y k They represent the sensory evaluation prediction results of protein oxidation taste, fat oxidation taste and comprehensive oxidation taste respectively;

[0042] α i , α j , α kare the constants of the sensory evaluation models for protein oxidation taste, fat oxidation taste and comprehensive oxidation taste, respectively;

[0043] X1, X2, X3, X4, X5...X n They respectively represent the values ​​of the first sensor, the second sensor, the third sensor, the fourth sensor, the fifth sensor, ... the nth sensor useful for sample differentiation;

[0044] β i1 , β i2 , β i3 , β i4 , β i5 ...β in are the regression coefficients of the first sensor, the second sensor, the third sensor, the fourth sensor, the fifth sensor, and so on, the nth sensor in the sensory evaluation model of protein oxidation taste;

[0045] Β j1 , β j2 , β j3 , β j4 , β j5 ...β jn are the regression coefficients of the first sensor, the second sensor, the third sensor, the fourth sensor, the fifth sensor, and so on, the nth sensor in the sensory evaluation model of fat oxidation taste;

[0046] β k1 , β k2 , β k3 , β k4 , β k5 ...β kn It is the regression coefficient of the first sensor, the second sensor, the third sensor, the fourth sensor, the fifth sensor…the nth sensor in the comprehensive oxidative taste sensory evaluation model.

[0047] In the above step S4, a discrimination model is established with the electronic nose sensor parameters as independent variables and the oxidation taste type as the dependent variable; and a sensory evaluation model is established with the electronic nose sensor parameters as independent variables and the sensory score as the dependent variable.

[0048] The principle adopted by the present invention is: first, three oxidation taste reference samples with different oxidation degrees are established, and the odor intensity of the three reference samples is preliminarily evaluated through manual sensory evaluation; then, the detection data of the three reference samples are collected through an electronic nose, and an electronic nose sensor response diagram is established; finally, a discrimination model and a sensory evaluation model are established to achieve discrimination and evaluation of milk with different oxidation tastes, solving the problem that the current oxidation taste is difficult to accurately evaluate. In the present invention, the electronic nose is used to achieve rapid data collection, and the classification and identification of different oxidation tastes are achieved by establishing a discrimination model, and on the basis of classification, the sensory score prediction of different oxidation tastes is achieved by respectively establishing three sensory evaluation models.

[0049] The beneficial effects of the present invention are as follows: the present invention fundamentally distinguishes three kinds of oxidation flavors that are difficult to distinguish by manual sensory evaluation by preparing reference samples with different raw materials and different oxidation degrees, and establishes a sensory evaluation model through electronic nose technology to realize the prediction of sensory scores of samples to be tested. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is an overall flow chart of the method of the present invention.

[0051] Figure 2 This is the electronic nose sensor response diagram of the protein oxidation taste reference sample of the present invention.

[0052] Figure 3 This is the electronic nose sensor response diagram of the fat oxidation taste reference sample of the present invention.

[0053] Figure 4 It is the electronic nose sensor response diagram of the comprehensive oxidation taste reference sample of the present invention. DETAILED DESCRIPTION

[0054] like Figure 1 As shown, a method for quickly identifying and evaluating different oxidized flavors of milk based on an electronic nose specifically comprises the following steps:

[0055] Step S1, establishing reference samples of three oxidation smells of milk, wherein the three oxidation smells are protein oxidation smell, fat oxidation smell and comprehensive oxidation smell of protein oxidation and fat oxidation.

[0056] Photo-oxidation is the most common and most likely cause of milk protein and fat oxidation. Therefore, reference samples with different oxidation flavors are established by photo-oxidation method to simulate different oxidation conditions. The three oxidation flavors established by the present invention are protein oxidation flavor dominated by protein oxidation, fat oxidation flavor dominated by fat oxidation, and comprehensive oxidation flavor of protein oxidation and fat oxidation. The specific preparation methods of the above three oxidation flavor reference samples are as follows:

[0057] Preparation method of protein oxidation taste reference sample: fresh whole milk is sterilized at high temperature, and then skimmed to obtain fresh skimmed milk (the protein content of skimmed milk is the same as that of whole milk). Fresh skimmed milk is irradiated with strong light to prepare oxidized milk, and other methods can also be used for oxidation, as long as the purpose of oxidation can be achieved. Then, fresh skimmed milk and oxidized skimmed milk are compounded to prepare 20 to 30 oxidation taste reference samples of different strengths (i.e., different oxidation degrees), wherein the addition amount of oxidized milk is 0% to 100%. In the actual test process, the more the number of compounded samples, the better. Since the addition amount of oxidized milk can be adjusted at will, the present invention sets 20 to 30 oxidation taste reference samples of different strengths between 0% and 100%. Since the fat content of skimmed milk can be ignored, the reference sample prepared after oxidation has a protein oxidation taste.

[0058] Preparation method of fat oxidation taste reference sample: fresh whole milk is sterilized at high temperature, and then prepared through the steps of concentration and the like to obtain fresh cream, and the fresh cream is diluted to obtain a fresh cream solution (the fat content of the cream solution is the same as that of whole milk). The fresh cream solution is irradiated with strong light to prepare an oxidized cream solution, and other methods can also be used for oxidation, as long as the purpose of oxidation can be achieved. Then, the fresh cream solution and the oxidized cream solution are compounded to prepare 20 to 30 oxidation taste reference samples of different strengths (i.e., different oxidation degrees), wherein the addition amount of the oxidized cream solution is 0% to 100%. In the actual test process, the more the number of compounded samples, the better. Since the addition amount of the oxidized cream solution can be adjusted at will, the present invention can set 20 to 30 oxidation taste reference samples of different strengths (i.e., different oxidation degrees) between 0% and 100%. Since the protein content of the cream solution can be ignored, the reference sample prepared after oxidation has a fat oxidation taste.

[0059] Preparation method of comprehensive oxidation taste reference sample: fresh whole milk is sterilized at high temperature, placed in a beaker and placed in a 4°C refrigerator to eliminate the influence of packaging materials and microorganisms. Fresh whole milk is irradiated with strong light to prepare oxidized milk. It can also be oxidized in other ways as long as the purpose of oxidation is achieved. Then, fresh and oxidized milk are compounded to prepare 20 to 30 oxidation taste reference samples of different strengths (i.e., different oxidation degrees), wherein the addition amount of oxidized milk is 0% to 100%. In the actual test process, the more the number of compounded samples, the better. Since the addition amount of oxidized milk can be adjusted at will, the present invention sets 20 to 30 oxidation taste reference samples of different strengths between 0% and 100%. Since fresh milk contains protein and fat, the reference sample prepared after oxidation is a comprehensive oxidation taste.

[0060] Step S2, electronic nose detection, using the metal oxide sensor on the electronic nose to detect the gases emitted by the three established oxidative odor reference samples.

[0061] Electronic nose model: PEN3 electronic nose, selected from AIRSENSE company of Germany.

[0062] Electronic nose detection parameters: sensor cleaning time 60s, sample preparation time 5s, sample test time 60s, injection flow rate 300mL / L.

[0063] Electronic nose detection method: First, place the sample to be tested in a stoppered sample bottle and let it stand at room temperature for 3 to 4 hours; then insert the air supply needle and sample absorption needle of the electronic nose into the stoppered sample bottle, and the 10 metal oxide sensors of the electronic nose detect the gas emitted by the sample. The 10 metal sensors of the electronic nose are W1C (sensitive to aromatic substances), W5S (very sensitive to nitrogen oxides), W3C (sensitive to ammonia and aromatic substances), W6S (selective to hydrogen compounds), W5C (sensitive to short-chain alkanes and aromatic substances), W1S (sensitive to methane, similar to W2S), W1W (sensitive to inorganic sulfides), W2S (sensitive to alcohols, aldehydes and ketones), W2W (sensitive to aromatic substances and organic sulfides), and W3S (sensitive to long-chain alkanes).

[0064] Step S3, screening of effective electronic nose sensors, based on the results of the three oxidative taste reference samples, analyze and identify sensors useful for sample differentiation, and establish an electronic nose sensor response graph.

[0065] Loading analysis is helpful to identify sensors that are useful for sample differentiation and to establish an electronic nose sensor response graph (also known as an electronic nose sensor load analysis graph), as shown in the following figure: Figures 2 to 4 The direction of the sensor in the Loading analysis is consistent with the direction of the principal component analysis (PCA), so the PCA thinking can be used to analyze the sensor's ability to distinguish. Sensors with larger values ​​on the horizontal and vertical axes are effective in distinguishing, and sensors located near the origin (0, 0) have little effect on distinguishing samples and can be ignored in subsequent analysis.

[0066] Protein oxidation taste reference sample results: Figure 2 As shown in the figure, the sensors with distinguishing ability include W1W, W2W, W5S and W1S, while W2S is slightly stronger than the remaining overlapping sensors located near the origin, so it can be taken into consideration. The remaining sensors can be ignored because of their small effects. Among them, the contribution rate of the first principal component is 81.74%, and the contribution rate of the second principal component is 10.91%. The sensors with greater distinguishing ability are W2W>W1W>W1S>W5S.

[0067] Fat oxidation taste reference sample results: Figure 3 As shown in the figure, the sensors with distinguishing ability include W1W, W2W, W5S, W1S and W2S, and the remaining sensors overlap and are all located near the origin, indicating that the distinguishing effect is small and can be ignored. The contribution rate of the first principal component is 98.95%, which is far more than the contribution rate of the second principal component of 0.77%. Therefore, the sensor distinguishing ability can be directly judged from the first principal component, and the sensors with better distinguishing ability are W1S>W5S>W2S.

[0068] Comprehensive oxidation taste reference sample results: Figure 4 As shown in the figure, the sensors with distinguishing ability include W1W, W2W, W5S, W1S and W2S, and the remaining sensors overlap and are all located near the origin, indicating that the distinguishing effect is small and can be ignored. Among them, the horizontal axis represents the first principal component, whose contribution rate is 98.56%, and the vertical axis represents the first principal component, whose contribution rate is only 0.78%. It can be inferred that the effect of each sensor is: W1W>W2W>W5S>W1S>W2S.

[0069] In summary, under the three oxidation conditions, the sensors with distinguishing ability include W1W, W2W, W5S, W1S and W2S, so the discrimination model of different oxidation tastes can be established through these five sensors. In addition, because each sensor plays a different role in identifying different oxidation tastes, different oxidation tastes can screen out sensors with better distinguishing ability to establish sensory evaluation models. The specific types of screened sensors are shown in Table 1. According to Table 1, it can be seen that the effective sensor types for protein oxidation taste are W2W, W1W, W1S, and W5S; the effective sensor types for fat oxidation taste are W1S, W5S, and W2S; the effective sensor types for comprehensive oxidation taste are W1W, W2W, W5S, W1S, and W2S.

[0070] Table 1 Effective sensor types for three oxidative odors

[0071]

[0072] Step S4, establishing an odor discrimination model and a sensory evaluation model based on the electronic nose sensor response graph.

[0073] (I) Establishing a model for distinguishing odors

[0074] Taking the sensor parameters of the electronic nose as independent variables and the three types of oxidation smell as dependent variables, a discriminant model was constructed. The regression equations of different oxidation smells of the discriminant model were established as follows:

[0075] Y1=a1X1+b1X2+c1X3+d1X4+e1X5+k1

[0076] Y2=a2X1+b2X2+c2X3+d2X4+e2X5+k2

[0077] Y3=a3X1+b3X2+c3X3+d3X4+e3X5+k3

[0078] In the formula, Y1, Y2, and Y3 represent the discrimination results of protein oxidation taste, fat oxidation taste, and comprehensive oxidation taste, respectively. Substitute the new case into the three functions, and the case with the larger value will be classified into that category;

[0079] X1, X2, X3, X4, and X5 represent the values ​​of sensors W2W, W2S, W1W, W1S, and W5S, respectively;

[0080] a1, a2, and a3 represent the discrimination coefficients of sensor W2W in the protein oxidation taste discrimination results, fat oxidation taste discrimination results, and comprehensive oxidation taste discrimination results, respectively;

[0081] b1, b2, and b3 represent the discrimination coefficients of sensor W2S in the protein oxidation taste discrimination results, fat oxidation taste discrimination results, and comprehensive oxidation taste discrimination results, respectively;

[0082] c1, c2, and c3 represent the discrimination coefficients of sensor W1W in the protein oxidation taste discrimination results, fat oxidation taste discrimination results, and comprehensive oxidation taste discrimination results, respectively;

[0083] d1, d2, and d3 represent the discrimination coefficients of sensor W1S in the protein oxidation taste discrimination results, fat oxidation taste discrimination results, and comprehensive oxidation taste discrimination results, respectively;

[0084] e1, e2, and e3 represent the discrimination coefficients of sensor W5S in the protein oxidation taste discrimination results, fat oxidation taste discrimination results, and comprehensive oxidation taste discrimination results, respectively;

[0085] k1, k2, and k3 are the constants of the discrimination models of protein oxidation smell, fat oxidation smell, and comprehensive oxidation smell, respectively.

[0086] 54 groups of data detected by the electronic nose were randomly selected as training sets and 36 groups as test sets. The discriminant model established is as follows:

[0087] Y1=6669.132X1+55889.535X2-5838.971X3-18140.816X4+3253.757X5-21258.336

[0088] Y2=6780.279X1+55840.517X2-5910.443X3-17954.134X4+3061.0528X5-21225.227

[0089] Y3=6088.850X1+55919.235X2-5303.621X3-18196.258X4+3277.400X5-21213.309

[0090] In the above formula, the specific sources of the discrimination coefficient and constant are as follows: the electronic nose data of three oxidative taste reference samples are brought into the SPSS software to fit the function model, as well as the accuracy results of the training set and the test set.

[0091] Among the 54 samples in the training set, 52 samples were accurately classified, and the discrimination accuracy reached 98.1%; among the 36 samples in the prediction set, 33 samples were accurately classified, and the discrimination accuracy reached 94.4%. It can be seen that this model has good classification ability for different oxidation flavors (generally the accuracy is required to be greater than 80%). The test classification results of the training set and the test set are shown in Table 2 below.

[0092] Table 2. Test classification results of training set and test set

[0093]

[0094] 2. Establishing a sensory evaluation model

[0095] Collection of sensory evaluation data: A sensory evaluation team consisting of 11 people was formed to evaluate the odor intensity of the reference samples through manual sensory evaluation, and a sensory scoring database was constructed using a 15-point scale, with 0 representing no odor and 15 representing very strong odor.

[0096] Taking the sensor parameters of the electronic nose (each sensor parameter needs to subtract the unoxidized parameter, that is, the increment relative to the unoxidized) as the independent variable and the sensory score as the dependent variable, three sensory evaluation models were constructed as follows:

[0097] Y i =α i +β i1 X1+β i3 X3+β i4 X4+β i5 X5

[0098] Y j =α j +β j2 X2+β j4 X4+β j5 X5

[0099] Y k =α k +β k1 X1+β k2 X2+β k3 X3+βk4 X4+β k5 X5

[0100] Where Y i , Y j , Y k They represent the sensory evaluation prediction results of protein oxidation taste, fat oxidation taste and comprehensive oxidation taste respectively. Based on the discrimination results, the new case is substituted into one of the corresponding functions to obtain the sensory evaluation prediction value.

[0101] α i , α j , α k are the constants of the sensory evaluation models for protein oxidation taste, fat oxidation taste and comprehensive oxidation taste, respectively;

[0102] X1, X2, X3, X4, and X5 correspond to the values ​​of sensors W2W, W2S, W1W, W1S, and W5S respectively;

[0103] β i1 , β i3 , β i4 , β i5 They are the regression coefficients corresponding to sensors W2W, W1W, W1S, and W5S in the sensory evaluation model of protein oxidation taste;

[0104] β j2 , β j4 , β j5 They are the regression coefficients corresponding to sensors W2S, W1S, and W5S in the sensory evaluation model of fat oxidation taste;

[0105] β k1 , β k2 , β k3 , β k4 , β k5 These are the regression coefficients corresponding to sensors W2W, W2S, W1W, W1S, and W5S in the comprehensive sensory evaluation model of oxidative taste.

[0106] After processing the electronic nose detection data, three sensory evaluation models and correlation coefficients (R 2 ) are as follows:

[0107] Y i =1.8855+0.1699X1+0.6556X3-0.0907X4-0.8229X5, R 2 =0.9268

[0108] Y j =1.7369+1.019X2+1.5605X4+1.7132X5, R 2=0.9012

[0109] Y k =2.2901+0.3737X1+0.2065X2+0.4066X3-0.1473X4+0.2743X5, R 2 =0.8945

[0110] In the formula, the correlation coefficients of the three oxidative taste sensory evaluation models (R 2 ) are all controlled above 0.8, indicating that the model has good predictive ability.

[0111] In the above formula, the specific sources of regression coefficients and constants are as follows: The electronic nose data of three oxidative taste reference samples were brought into the SIMCA software to fit the function model and the correlation coefficient R 2 The numerical value of .

[0112] Step S5, sample testing, taking the whole milk treated with different light exposure times as the test samples for sample testing.

[0113] Firstly, the sample electronic nose detection data is substituted into the three functions of the discriminant model to judge the type of oxidation taste; then, based on the discrimination, one of the corresponding functions is selected in the sensory evaluation model for calculation to obtain the sensory score prediction result; finally, the sensory score prediction result is compared with the manual sensory evaluation result to judge the prediction ability of the model.

[0114] The sample test results are shown in Table 3 below. It can be seen from the table that the oxidation taste discrimination model can well distinguish the protein oxidation taste and the comprehensive oxidation taste, indicating that the oxidation taste of whole milk treated with different light exposure times is different. In addition, the sensory prediction value obtained by the sensory evaluation model is close to the sensory evaluation panel score, and the change pattern is similar, so the sensory evaluation model has a good prediction ability.

[0115] Table 3 Sample oxidation taste identification and evaluation results

[0116]

[0117]

[0118] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the scope of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for quickly identifying and evaluating different oxidized flavors of milk based on an electronic nose, characterized in that: The specific steps include: Step S1, establishing reference samples of three oxidation smells of milk, the three oxidation smells being protein oxidation smell, fat oxidation smell, and comprehensive oxidation smell of protein oxidation and fat oxidation; Step S2, electronic nose detection, using the metal oxide sensor on the electronic nose to detect the gases emitted by the three established oxidative odor reference samples; Step S3, screening of effective electronic nose sensors, based on the detection results of three oxidative taste reference samples, analyzing and identifying sensors useful for sample differentiation, and establishing an electronic nose sensor response graph; Step S4, establishing a discrimination model and a sensory evaluation model based on the electronic nose sensor response graph; In step S1, the preparation methods of the three reference samples of oxidation taste are as follows: The method for preparing a protein oxidation taste reference sample is as follows: sterile fresh skim milk is irradiated with strong light to obtain oxidized milk, and then the fresh skim milk and the oxidized skim milk are compounded to obtain oxidation taste reference samples with different oxidation degrees; The method for preparing a fat oxidation taste reference sample is as follows: diluting sterile fresh cream to obtain a fresh cream solution, and irradiating the fresh cream solution with strong light to obtain an oxidized cream solution; then compounding the fresh cream solution and the oxidized cream solution to obtain oxidation taste reference samples with different oxidation degrees; The preparation method of the comprehensive oxidation taste reference sample is as follows: sterile fresh whole milk is irradiated with strong light to obtain oxidized milk, and then the fresh whole milk and the oxidized whole milk are compounded to obtain oxidation taste reference samples with different oxidation degrees.

2. The method for quickly identifying and evaluating different oxidation flavors of milk based on an electronic nose according to claim 1, characterized in that: The following steps are also included: Step S5, sample testing, taking the whole milk treated with different light exposure times as the test samples for sample testing.

3. The method for quickly identifying and evaluating different oxidized flavors of milk based on an electronic nose according to claim 1 or 2, characterized in that: In step S2, the electronic nose detection method is specifically as follows: Step S201, placing the sample to be tested into a stoppered sample bottle and leaving it at room temperature for 3 to 4 hours; Step S202, inserting the air supply needle and the sample absorption needle of the electronic nose into the stoppered sample bottle, and the 10 metal sensors of the electronic nose detect the gas emitted by the sample.

4. The method for quickly identifying and evaluating different oxidized flavors of milk based on an electronic nose according to claim 1 or 2, characterized in that: In step S3, a loading analysis method is used to analyze and identify sensors useful for sample differentiation.

5. The method for rapidly identifying and evaluating different oxidation flavors of milk based on an electronic nose according to claim 3, characterized in that: In step S3, a loading analysis method is used to analyze and identify sensors useful for sample differentiation.

6. The method for quickly identifying and evaluating different oxidized flavors of milk based on an electronic nose according to claim 1, 2 or 5, characterized in that: In step S4, the regression equations of different oxidation flavors of the discriminant model are established as follows: Y1=a1X1+b1X2+c1X3+d1X4+e1X5……n1X n +k1 Y2=a2X1+b2X2+c2X3+d2X4+e2X5……n2X n +k2 <h2 style=";text-align:left;direction:ltr">Y3 = a3X1 + b3X2 + c3X3 + d3X4 + e3X5……n3X<h2 style=";text-align:left;direction:ltr"> n <h2 style=";text-align:left;direction:ltr"> +k3 In the formula, Y1, Y2, and Y3 represent the discrimination results of protein oxidation smell, fat oxidation smell, and comprehensive oxidation smell, respectively; X1, X2, X3, X4, X5...X n They respectively represent the values ​​of the first sensor, the second sensor, the third sensor, the fourth sensor, the fifth sensor, ... the nth sensor useful for sample differentiation; a1, a2, and a3 represent the discrimination coefficients of the first sensor useful for sample discrimination in the protein oxidation taste discrimination results, fat oxidation taste discrimination results, and comprehensive oxidation taste discrimination results, respectively; b1, b2, and b3 represent the discrimination coefficients of the second sensor useful for sample discrimination in the protein oxidation taste discrimination results, fat oxidation taste discrimination results, and comprehensive oxidation taste discrimination results, respectively; c1, c2, and c3 represent the discrimination coefficients of the third sensor useful for sample discrimination in the protein oxidation taste discrimination results, fat oxidation taste discrimination results, and comprehensive oxidation taste discrimination results, respectively; d1, d2, and d3 represent the discrimination coefficients of the fourth sensor useful for sample discrimination in the protein oxidation taste discrimination results, fat oxidation taste discrimination results, and comprehensive oxidation taste discrimination results, respectively; e1, e2, and e3 represent the discrimination coefficients of the fifth sensor useful for sample discrimination in the protein oxidation taste discrimination results, fat oxidation taste discrimination results, and comprehensive oxidation taste discrimination results, respectively; n1, n, and n3 represent the discrimination coefficients of the nth sensor useful for sample discrimination in the protein oxidation taste discrimination results, fat oxidation taste discrimination results, and comprehensive oxidation taste discrimination results, respectively; k1, k2, and k3 are the constants of the odor discrimination models of protein oxidation smell, fat oxidation smell, and comprehensive oxidation smell, respectively.

7. The method for quickly identifying and evaluating different oxidized flavors of milk based on an electronic nose according to claim 1, 2 or 5, characterized in that: In step S4, the regression equations of different oxidation flavors of the sensory evaluation model are established as follows: Y i =a i +b i1 X1+β i2 X2+β i3 X3+β i4 X4+b i5 X5……+β in X n Y j =a j +b j1 X1+β j2 X2+β j3 X3+β j4 X4+b j5 X5……+β jn X n Y k =a k +b k1 X1+β k2 X2+β k3 X3+β k4 X4+b k5 X5……+β kn X n Where Y i , Y j , Y k They represent the sensory evaluation prediction results of protein oxidation taste, fat oxidation taste and comprehensive oxidation taste respectively; α i , α j , α k are the constants of the sensory evaluation models for protein oxidation taste, fat oxidation taste and comprehensive oxidation taste; X1, X2, X3, X4, X5...X n They respectively represent the values ​​of the first sensor, the second sensor, the third sensor, the fourth sensor, the fifth sensor, ... the nth sensor useful for sample differentiation; β i1 , β i2 , β i3 , β i4 , β i5 ...β in They represent the regression coefficients of the first sensor, the second sensor, the third sensor, the fourth sensor, the fifth sensor, and so on, the nth sensor in the sensory evaluation model of protein oxidation taste; β j1 , β j2 , β j3 , β j4 , β j5 ...β jn They represent the regression coefficients of the first sensor, the second sensor, the third sensor, the fourth sensor, the fifth sensor, ... the nth sensor in the sensory evaluation model of fat oxidation taste; β k1 , β k2 , β k3 , β k4 , β k5 ...β kn They respectively represent the regression coefficients of the first sensor, the second sensor, the third sensor, the fourth sensor, the fifth sensor...the nth sensor in the comprehensive oxidative taste sensory evaluation model.

Images