A Method for Establishing an Intelligent Blending Model of Chinese Baijiu

Abstract

The present invention discloses a method for establishing an intelligent blending model of liquor, which relates to the technical field of liquor brewing and blending. The present invention includes collecting the data of the flavor component content and the sensory evaluation score data of liquor as the basic data, then respectively constructing a relationship model between the flavor component content and the sensory evaluation score, and further predicting the content of each flavor component; then, based on the data of the sensory evaluation score of liquor, the flavor component content data predicted by the relationship model between the flavor component content and the sensory evaluation score, the measured flavor substance content data of the base liquor, and the price coefficient of the base liquor, the quality requirements of the target liquor are processed as constraint conditions, and the dosage of the base liquor is processed as a variable, and an improved genetic algorithm-based optimization goal programming is used to establish a blending formula model. This method enables the content ratio of various flavor components in the blended target liquor to be moderate, thereby obtaining a more excellent sensory performance, making the liquor products of different production batches more stable in quality, and having a relatively low production cost.

Description

Technical Field

[0001] The invention relates to the technical field of liquor brewing and blending, and in particular to a method for establishing a liquor intelligent blending model. Background Art

[0002] There are still some misunderstandings in many people's cognition about liquor blending, that is, the final liquor is blended through edible alcohol and food additives. However, in the actual production process, blending is one of the more critical processes among many, and the production of each bottle of commercial liquor will go through the blending process. The real liquor blending emphasizes the adjustment of wines, so that the blended wine presents the effect of 1+1 greater than 2. In order to ensure the quality of the blended target wine while improving the efficiency of the enterprise, in addition to the cost factor, the characteristics of the base wine cannot be ignored. The selection of raw base wine must first have quality requirements, because the finished liquor blended from the base wine of lower grade is also not very good in trace components and sensory manifestations.

[0003] So far, traditional mathematical statistics methods such as linear programming and goal programming are an important part of the theoretical basis of the formulation model. The current formulation models of these methods have the following shortcomings: (1) Linear programming has certain limitations. Since its constraints are system constraints and lack flexibility, the constraints are prone to conflict with each other, which can easily lead to linear programming having no solution. Moreover, if the goal is only a minimum cost formulation, its optimization goal is single and it is easy to fall into local no solution. (2) The goal programming algorithm can well solve the shortcomings of linear programming's single optimization goal and easy local no solution, and has the advantage of being able to obtain high-quality formulations with multi-objective requirements according to the setting of priority factors, but the "priority factor" in the algorithm is difficult to determine. Summary of the invention

[0004] The purpose of the present invention is to propose a method for establishing an intelligent blending model for liquor, establish a relationship model based on the liquor sensory tasting score data and the flavor component content measurement data by a joint support vector regression algorithm, then use the data predicted by the relationship model together with the measured base liquor flavor component content data and the base liquor price coefficient as the data basis, and use an improved genetic algorithm to optimize the target programming to establish a blending formula model, so that the liquor products of different production batches are more stable in quality and the production cost is relatively low.

[0005] The technical solution adopted by the present invention is as follows:

[0006] The present invention is a method for establishing a liquor intelligent blending model, comprising the following steps:

[0007] Obtain sensory tasting score data of liquor;

[0008] Collect and process the flavor component content data of Baijiu and base liquor respectively, and classify the flavor component content data into four major categories: acid compounds, ester compounds, alcohol compounds, and hydroxy and furan compounds;

[0009] Adopt the support vector regression algorithm, and based on the sensory evaluation scores and flavor component content of Baijiu, establish a relationship model to predict the content values of acid compounds, ester compounds, alcohol compounds, hydroxy and furan compounds in the target liquor respectively;

[0010] Optimize the goal programming algorithm through the improved genetic algorithm. Based on the flavor component content data predicted by the relationship model of Baijiu sensory evaluation scores and flavor component content, the measured flavor component content data of the base liquor, and the price coefficient of the base liquor, process the quality requirements of the target liquor as constraint conditions, and process the dosage of the base liquor as a variable to establish a blending formula model.

[0011] Furthermore, the sensory evaluation score is the result obtained by professional wine tasters scoring in a Baijiu blending laboratory without vibration and noise, clean and tidy without peculiar smell, fresh air, sufficient light, and appropriate temperature and humidity according to the G·R Baijiu sensory evaluation score criterion.

[0012] Furthermore, the specific steps for collecting the flavor component content data of Baijiu measured by gas chromatography-mass spectrometry are as follows:

[0013] Prepare the liquor sample, and conduct preparatory treatment on the liquor sample. The preparatory treatment includes quantitative packaging and label setting of the liquor sample, and adding internal standard compounds to the liquor sample;

[0014] Set the parameters of the gas chromatography-mass spectrometry, including that before the injector of the gas chromatography-mass spectrometry performs the injection operation, the injection needle must be rinsed with the rinsing solution, and the injection flow rate and injection port temperature of the sample are set;

[0015] Conduct qualitative analysis, and adopt the method of combining artificial experience with library retrieval to determine and analyze the types of flavor components contained in the Baijiu sample;

[0016] Conduct quantitative analysis, and adopt the internal standard method for quantitative analysis to determine and analyze the specific content of the flavor components contained in the Baijiu.

[0017] Furthermore, for the data processing of the flavor component content data of Baijiu and base liquor, select 24 flavor components that are commonly contained in the liquor and have a high importance level, and classify these flavor components into four major categories: acid compounds, ester compounds, alcohol compounds, and hydroxy and furan compounds as the flavor component content data.

[0018] Further, the flavor component content data of the Chinese liquor is composed of four major categories: acid compounds, ester compounds, alcohol compounds, and hydroxy and furan compounds, forming four different basic data sets. The sensory evaluation score data is divided into four categories: color, aroma, taste, and style. The entire sensory evaluation score data set is merged with the data sets of the four major categories of acid compounds, ester compounds, alcohol compounds, and hydroxy and furan compounds to form four new different basic data sets. A support vector regression algorithm based on grid search for parameter optimization is used to establish a relationship model between the sensory evaluation score and flavor component content of the Chinese liquor.

[0019] Further, the specific steps for establishing a relationship model between the sensory evaluation score and flavor component content of the Chinese liquor using a support vector regression algorithm based on grid search for parameter optimization are as follows:

[0020] Import the basic data, perform correlation analysis on the data, and perform normalization processing;

[0021] Then divide the data set into a training set and a test set in a ratio of 4:1;

[0022] Select the R2 evaluation index to judge different kernel functions. After determining the kernel function, use the grid search method to optimize the C and γ hyperparameters and use R2 as the evaluation index as the basis for determining the hyperparameters;

[0023] Finally, establish a relationship model by training the training set, and the test set is used to test the relationship model.

[0024] Further, the specific steps for constructing a blending formula model using the goal programming algorithm are as follows: Using the flavor component content of acid compounds, ester compounds, alcohol compounds, hydroxy and furan compounds in the base liquor, as well as the content of acid compounds, ester compounds, alcohol compounds, hydroxy and furan compounds in the target blended liquor obtained from the relationship model between the sensory and flavor components of the Chinese liquor and the price coefficient of the base liquor as the data basis, and using the quality requirements of the target liquor as the constraint conditions and the dosage of the base liquor as the variable, while meeting the quality requirements, pursue the lowest cost, and attach weak constraint conditions that the ratio of ethyl lactate to ethyl hexanoate is less than 1 and the ratio of ethyl butyrate to ethyl hexanoate is equal to 0.1 to establish a goal programming blending formula model.

[0025] Further, the improved genetic algorithm is used to optimize the goal programming algorithm. The blending cost of the Chinese liquor is processed as the objective function, and the lowest cost is used as the optimization goal to perform optimal solution on the ratio of the raw liquor during the blending process and establish an intelligent blending formula model.

[0026] In summary, due to the adoption of the above technical solutions, the beneficial effects of the present invention are:

[0027] The present invention relates to a method for establishing an intelligent blending model of Chinese liquor. By using the support vector regression algorithm, a relationship model between the sensory properties and flavor components of Chinese liquor is established. Then, based on the content data of flavor components predicted by the relationship model between the sensory evaluation scores and flavor component contents of Chinese liquor, the measured content data of flavor components in base liquor, and the price coefficient of base liquor, the quality requirements of the target liquor are processed as constraint conditions, and the dosage of base liquor is processed as a variable. The improved genetic algorithm-based optimization goal programming is used to optimize and solve the dosage ratio of each raw liquor during the blending process, thereby realizing the establishment of a blending formula model. The content ratios of various flavor components such as acids, esters, alcohols, hydroxyl groups, and furan compounds blended by this method are moderate, resulting in more excellent sensory performance. It not only stabilizes the quality of Chinese liquor products in different production batches but also reduces production costs, which is of great significance for guiding Chinese liquor production and promoting the digital and intelligent development of Chinese liquor. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings, where:

[0029] Figure 1 It is the total spectrum diagram of a single sample measured by a gas chromatography-mass spectrometry instrument;

[0030] Figure 2 It is the G·R Chinese liquor sensory evaluation score standard adopted by the present invention;

[0031] Figure 3 It is a partial result of the Chinese liquor sensory evaluation score;

[0032] Figure 4 It is the SVR regression prediction distribution diagram of acid substances;

[0033] Figure 5 It is the analysis table diagram of the prediction results of acid compounds;

[0034] Figure 6 It is the SVR regression prediction distribution diagram of ester substances;

[0035] Figure 7 It is the analysis table diagram of the prediction results of ester compounds;

[0036] Figure 8 It is the SVR regression prediction distribution diagram of alcohol substances;

[0037] Figure 9 It is the analysis table diagram of the prediction results of alcohol compounds;

[0038] Figure 10 Regression prediction distribution chart of hydroxyl and furan substances by SVR

[0039] Figure 11 Analysis table chart of prediction results of hydroxyl and furan compounds

[0040] Figure 12 Comparison chart of flavor component contents based on improved genetic algorithm optimized goal programming for blending

[0041] Figure 13 Details table chart of blending cost based on improved genetic algorithm optimized goal programming

[0042] Figure 14 Optimization iteration process chart of improved genetic algorithm Specific implementation manners

[0043] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention, that is, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Generally, the components of the embodiments of the present invention described and illustrated herein can be arranged and designed in various different configurations.

[0044] It should be noted that the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0045] The features and performance of the present invention will be further described in detail below in conjunction with the embodiments.

[0046] Embodiment 1

[0047] The present invention is a method for establishing a liquor intelligent blending model. The data collection of the present invention is mainly divided into collecting the data of the flavor component contents of liquor measured by a gas chromatography-mass spectrometry (GC-MS) instrument and collecting the sensory evaluation score data given by wine tasters;

[0048] The specific steps for collecting the data of the flavor component contents of liquor measured by the gas chromatography-mass spectrometry instrument are as follows:

[0049] First, prepare the samples. Considering the large variety and quantity of samples for each measurement and the small sample volume required for GC-MS, it is necessary to prepare the liquor samples before the measurement experiment. The preparation process mainly includes two parts: quantitative aliquoting and label setting. For subsequent quantitative calculations, use a micropipette to take 1 mL of each sample and deposit it into a sample vial for measurement. Use the internal standard method as the analysis method for GC-MS measurement. Adding internal standard compounds to the liquor samples can effectively assist in the qualitative and quantitative analysis of liquor. The internal standard compounds should be selected as substances that are relatively stable, have weak volatility, do not react with other compounds in the liquor, and do not exist in the liquor. Considering the basic characteristics of the liquor to be measured and the peak time distribution of the compounds, finally, two compounds, 2-ethylbutyric acid and amyl acetate, are selected as the internal standards for the samples.

[0050] Secondly, set the parameters for GC-MS. The injector of GC-MS is a multi-functional automatic injector. Before performing the injection operation, it is necessary to rinse the injection needle with a rinsing solution. During the process, methanol solution is selected as the rinsing solution, and the injection needle is rinsed about 10 times in total. To further reduce the errors in the experimental operation process, after the first rinsing of the injection needle, it is also necessary to rinse the injection needle with the sample liquor to be measured, about 3 - 5 times. After the two rinsings are completed, the sample injection is carried out. The sample injection method is split injection, the split ratio is set to 40:1, the rated injection volume per time is 1 μL, and the temperature of the equipment injection port is set to 250 °C. For the GC-MS measurement of liquor samples in this embodiment, a DB-Wax capillary column is selected, with a column length of 60 m and an inner diameter of 0.25 mm. To improve the accuracy of the measurement, the temperature change during the whole process is controlled by using a programmed temperature rise method during the measurement. The details of the entire temperature control program are as follows: the initial column temperature is 30 °C, maintained for 6 min; then, at a heating rate of 2.5 °C / min, the temperature is raised to 40 °C; then, at a rate of 5 °C / min, the temperature is raised to 100 °C; then, at a heating rate of 10 °C / min, the temperature is controlled to 200 °C; finally, at a rate of 20 °C / min, the temperature is raised to 220 °C and maintained for 10 min. From the injection of the sample to the end of the test, the total time-consuming is about 40 min.

[0051] Then, qualitative analysis is carried out, that is, the types of flavor components contained in the liquor samples are determined and analyzed; GC-MS is used for determination, and the main qualitative analysis method is the combination of artificial experience and library search; in the database corresponding to GC-MS, there are standard spectra of nearly 300,000 compounds stored, which can fully meet the determination and analysis of liquor; during the experiment, the mass spectrum corresponding to each component can be obtained from the total ion chromatogram; select the corresponding mass spectrum and automatically compare it with the spectra in the standard library; the results of the comparison mainly include chemical molecular formula, CAS number, molecular weight, base peak, and similarity ratio, etc. Then, considering the physical and chemical properties of the unknown substance, chromatographic retention value, infrared, nuclear magnetic spectrum, etc. comprehensively, combined with personal experience and similarity, the compound is determined, and the compound name is obtained by querying the CAS number.

[0052] Finally, quantitative analysis is carried out, that is, the specific content of the flavor components contained in the liquor is determined and analyzed; in the GC-MS quantitative analysis method, due to the addition of the internal standard in the internal standard method, when the injection volume and the instrument response value change, the content ratio of the compound to be measured and the internal standard compound in the same sample remains unchanged, and the ratio of their response values also remains unchanged; therefore, the internal standard method can eliminate the systematic errors caused by fluctuations in injection volume, instrument response value, etc.; it is best to add the same amount of internal standard to all standard samples and samples to be measured to avoid the deviation caused by the non-linear change of the response value due to the change in the concentration of the internal standard substance; to overcome the inevitable measurement errors, during the experiment, the internal standard method is selected for quantitative analysis; when using the internal standard method for quantitative operation, first prepare a mixed standard sample solution, that is, select the internal standard compound and the chromatographic standard products (purity greater than 99%) of the compounds obtained from qualitative analysis to quantitatively prepare the mixed standard solution; select a total of 24 component substances such as acetaldehyde, ethyl formate, ethyl acetate, propanol, butyric acid, etc., and two internal standard substances, 2-ethylbutyric acid and amyl acetate, as the mixed standard sample; the preparation of the mixed standard solution should first accurately prepare a mother liquor, in which the content of each component needs to be accurately added and recorded, and its content should be higher than that in the sample to be measured; then gradually dilute the prepared mother liquor, and the content after the last dilution should be lower than that in the sample to be measured; finally, take 1 ml of the diluted solution and measure it by GC-MS to obtain the spectral data of the mixed standard solution; obtain the total ion chromatogram of the liquor sample by GC-MS determination; the peak area of each absorption peak in the chromatogram is proportional to the corresponding component content; select the reference substance (internal standard compound) and add it to the standard sample and the sample to be measured for determination, calculate the ratio of the response values of the compound to be measured and the internal standard compound (referred to as the relative correction factor), and perform quantitative analysis based on the relative response factor and the amount of the internal standard compound added. The total spectrum of a single sample measured by gas chromatography-mass spectrometry is shown in the figure.

[0053] The specific quantitative calculation of the quantitative analysis is as follows: GC-MS measurement obtains the total ion chromatogram of the liquor sample, the peak area of each absorption peak in the chromatogram is proportional to the corresponding component content, selects an appropriate reference substance (internal standard compound) and adds it to the standard sample and the sample to be tested for measurement, calculates the ratio of the response value of the test compound and the internal standard compound (called the relative correction factor), and quantifies based on the relative response factor and the amount of the added internal standard compound; wherein the relative correction factor is calculated as follows: (ω i Indicates the content of the standard sample of the compound to be tested, A i represents the peak area of the standard sample of the test compound), (ω s Indicates the content of internal standard compound in the standard sample; A s represents the peak area of the internal standard compound in the standard sample), ( is the relative correction factor); the calculation method of the content of the compound to be tested is: (P represents the content of the compound to be tested; f represents the relative response factor; A represents the peak area of the compound to be tested, ω n Indicates the content of internal standard in the wine sample to be tested, A n represents the peak area of the internal standard in the wine sample to be tested).

[0054] The specific steps of obtaining the sensory tasting score data are as follows: select a liquor blending laboratory with no vibration and noise, clean and tidy, no odor, fresh air, sufficient light, and appropriate temperature and humidity as the location of the tasting experiment, organize professional wine tasters to conduct the tasting, and use the following method: Figure 2 The liquor is evaluated and scored according to the G·R liquor sensory tasting score criteria shown in the figure; the corresponding score is given according to the degree of conformity with the standard to complete the liquor sensory tasting score. The liquor sensory tasting score result is as follows Figure 3 shown.

[0055] Taking the sensory evaluation score data and flavor component content data obtained from experiments as basic data, 24 flavor components that are commonly contained in the liquor and have a high importance level are classified into four major categories: acids, esters, alcohols, hydroxyls, and furans; the sensory data of each liquor is corresponded to the acid, ester, alcohol, hydroxyl, and furan compounds it contains respectively to form four different basic data sets; using the liquor sensory data as the input data of the model and the flavor component data as the output data of the model, then establish a relationship model between liquor sensory and flavor components; for acids, esters, alcohols, hydroxyls, and furan compounds, use the support vector regression algorithm with parameter optimization based on the grid search method to establish a relationship model between liquor sensory evaluation scores and flavor component contents; the prediction of acids, esters, alcohols, hydroxyls, and furan compounds uses the support vector regression algorithm that specifically targets the case of limited samples, realizes the minimization of structural risk, and seeks a compromise between the accuracy of approximating the given data and the complexity of the approximation function in order to obtain the best generalization ability.

[0056] From Figure 5 , Figure 7 , Figure 9 , Figure 11 it can be seen that the support vector regression algorithm is optimal among various compounds.

[0057] The specific steps of establishing a relationship model between liquor sensory evaluation scores and flavor component contents using the support vector regression algorithm with parameter optimization based on the grid search method are as follows: import the basic data, conduct a correlation analysis on the data and perform normalization processing; then divide the data set into a training set and a test set according to a ratio of 4:1, select the R2 evaluation index to judge different kernel functions, after determining the kernel function, use the grid search method to optimize the C and γ hyperparameters and use R2 as the evaluation index as the basis for determining the hyperparameters; finally, train the training set to establish a model and make predictions; the analysis table of the prediction results of acid compounds is as Figure 5 shown, the analysis table of the prediction results of ester compounds is as Figure 7 shown, the analysis table of the prediction results of alcohol compounds is as Figure 9 shown, and the analysis table of the prediction results of hydroxyl and furan compounds is as Figure 11 shown.

[0058] Taking the contents of acid, ester, alcohol, hydroxyl, and furan flavor components in the blended base liquor collected from experiments, as well as the contents of acid, ester, alcohol, hydroxyl, and furan compounds in the blended target liquor obtained from the relationship model between liquor sensory and flavor component contents and the price coefficient of the base liquor as basic data, use the goal programming algorithm optimized by the improved genetic algorithm to establish a blending formula model.

[0059] The establishment of the blending formula model uses a goal programming algorithm that does not consider minimizing or maximizing each goal, but rather hopes that under the constraints, each goal can approach the given value as closely as possible.

[0060] The specific steps for constructing the blending formula model based on the goal programming algorithm are as follows: Using the contents of acid, ester, alcohol, hydroxyl, and furan flavor components in the base liquor, as well as the relationship model between the sensory and flavor components of the liquor, to obtain the contents of acid, ester, alcohol, hydroxyl, and furan compounds in the target blended liquor and the price coefficient of the base liquor as basic data, and treating the quality requirements of the target liquor as constraints and the usage of the base liquor as variables. While meeting the quality requirements, pursue the lowest cost, and attach weak constraints such as the ratio of ethyl lactate to ethyl caproate being less than 1 is better, and the ratio of ethyl butyrate to ethyl caproate being equal to 0.1 is better to establish a goal programming formula model.

[0061] The parameter optimization of goal programming has a natural advantage in solving multi-objective problems and can optimize multiple objective functions simultaneously, thus effectively solving multi-objective problems using a genetic algorithm.

[0062] The specific steps for optimizing the goal programming algorithm using an improved genetic algorithm (GA) are as follows: Treat the blending cost of the liquor as the objective function, with the lowest cost as the optimization goal, add a penalty term, and punish individuals that do not meet the constraints by setting weights. Perform the selection operation of the genetic algorithm according to the roulette wheel selection method. At the same time, copy the individual with the highest fitness in the current population intact to the next generation population, and then perform the crossover operation, which is achieved by randomly selecting using the two-point crossover method in combination with the crossover probability. After the crossover operation, bring it into the constraints for judgment, and only leave the individuals that meet the constraints. Use the uniform mutation method to implement the mutation operation in GA. After mutation, bring the mutated individuals into the constraint terms for judgment, and leave the individuals that meet the constraints for iteration, finally achieving the optimal solution for the ratio of the raw liquor in the blending process; From Figure 12 、 13 、14, it can be seen that the goal programming optimized by the improved GA shows the best performance in terms of cost and quality.

[0063] The present invention establishes a relationship model between the sensory tasting score of liquor and the content of flavor components through a support vector regression algorithm, then takes the flavor component content data predicted by the relationship model of the liquor sensory tasting score and the content of flavor components, the measured base liquor flavor component content data and the price coefficient of the base liquor as data basis, treats the quality requirements of the target liquor as constraints, treats the amount of the base liquor as a variable, and uses an improved genetic algorithm-based optimization target planning to accurately calculate the proportion and amount of each raw material liquor in the blending process, thereby realizing the establishment of a blending formula model. The content ratios of various flavor components of acid, ester, alcohol, hydroxyl and furan compound blended by the method are moderate, which not only stabilizes the product quality of liquor in different production batches, but also reduces the production cost, which is of great significance for guiding liquor production and promoting the digitalization and intelligent development of liquor.

[0064] The above is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be thought of by a person skilled in the art within the technical scope disclosed by the present invention without creative work should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope defined in the claims.

Claims

1. A method for establishing an intelligent blending model of Chinese liquor, characterized in that, It includes the following steps: Obtain the sensory evaluation score data of white liquor; Collect and process the flavor component content data of white liquor and base liquor respectively, and divide the flavor component content data into four categories: acid compounds, ester compounds, alcohol compounds, hydroxyl and furan compounds; Adopt the support vector regression algorithm, based on the sensory evaluation score of white liquor and the flavor component content as basic data, establish a relationship model, and predict the content values of acid compounds, ester compounds, alcohol compounds, hydroxyl and furan compounds in the target liquor respectively; among them, adopt the support vector regression algorithm with parameter optimization based on the grid search method to establish the relationship model between the sensory evaluation score of white liquor and the flavor component content; The specific steps to establish the relationship model between the sensory evaluation score of white liquor and the flavor component content by using the support vector regression algorithm with parameter optimization based on the grid search method are as follows: Import the basic data, conduct correlation analysis on the data and perform normalization processing; Then divide the data set into a training set and a test set according to a ratio of 4:1; Select the R2 evaluation index to judge different kernel functions. After determining the kernel function, then use the grid search method to optimize the C and γ hyperparameters and use R2 as the evaluation index as the basis for determining the hyperparameters; Finally, establish a relationship model by training the training set, and the test set is used to test the relationship model; Optimize the goal programming algorithm through the improved genetic algorithm. Based on the flavor component content data predicted by the relationship model between the sensory evaluation score of white liquor and the flavor component content, the measured flavor component content data of the base liquor and the price coefficient of the base liquor, process the target liquor quality requirements into constraint conditions, and process the dosage of the base liquor into variables to establish a blending formula model; Among them, the specific steps to construct the blending formula model based on the goal programming algorithm are: taking the flavor component content of acid compounds, ester compounds, alcohol compounds, hydroxyl and furan compounds in the base liquor, and the content of acid compounds, ester compounds, alcohol compounds, hydroxyl and furan compounds in the blended target liquor obtained from the relationship model between the sensory and flavor components of white liquor and the price coefficient of the base liquor as basic data, and taking the target liquor quality requirements as constraint conditions and the dosage of the base liquor as variables, pursue the lowest cost while meeting the quality requirements, and attach the weak constraint conditions that the ratio of ethyl lactate to ethyl hexanoate is less than 1 and the ratio of ethyl butyrate to ethyl hexanoate is equal to 0.1 to establish a goal programming blending formula model; Among them, the specific steps of optimizing the goal programming algorithm based on the improved genetic algorithm are as follows: Treat the blending cost of white liquor as the objective function, take the lowest cost as the optimization goal, add a penalty term, and punish the individuals that do not meet the constraint conditions by setting weights. Perform the selection operation of the genetic algorithm according to the roulette wheel selection method. At the same time, copy the individual structure with the highest fitness in the current population intact to the next-generation population, and then perform the crossover operation. After the crossover operation, bring it into the constraint conditions for judgment, and only leave the individuals that meet the constraints for the mutation operation. After mutation, bring the mutated individuals into the constraint terms for determination, and leave the individuals that meet the constraints for iteration to optimize the ratio of raw liquor in the blending process and establish an intelligent blending formula model.

2. The method for establishing an intelligent blending model of Chinese liquor according to claim 1, wherein: The sensory evaluation score is the result obtained by professional wine tasters scoring in a white liquor blending laboratory without vibration and noise, clean and tidy without peculiar smell, fresh air, sufficient light, and appropriate temperature and humidity according to the G·R white liquor sensory evaluation score criterion.

3. The method for establishing an intelligent blending model of Chinese liquor according to claim 1, wherein: The specific process of separately collecting the flavor component content data of white liquor and base liquor is as follows: Use a gas chromatography-mass spectrometry (GC-MS) instrument to measure the flavor component content data in the liquor sample, and use the internal standard method to perform qualitative and quantitative analysis on the liquor sample to obtain the results.

4. The method for establishing an intelligent blending model of Chinese liquor according to claim 3, characterized in that, The specific steps of collecting the flavor component content data of white liquor measured by the gas chromatography-mass spectrometry instrument are as follows: Prepare the liquor sample, and perform preparatory treatment on the liquor sample. The preparatory treatment includes quantitative packaging and label setting of the liquor sample, and adding an internal standard compound to the liquor sample. Set the parameters of the gas chromatography-mass spectrometry instrument, including that before the injector of the gas chromatography-mass spectrometry instrument performs the injection operation, the injection needle must be rinsed with the rinsing solution, and the injection flow rate and injection port temperature of the sample are set. Perform qualitative analysis, and use the method of combining artificial experience and library retrieval to determine and analyze the types of flavor components contained in the white liquor sample. Perform quantitative analysis, and use the internal standard method for quantitative analysis to determine and analyze the specific content of the flavor components contained in the white liquor.

5. The method for establishing an intelligent blending model of Chinese liquor according to claim 4, characterized in that: Perform data processing on the flavor component content data of white liquor and base liquor, select 24 flavor components that are commonly contained in the liquor and have a high importance level, and classify these flavor components into four major categories: acid compounds, ester compounds, alcohol compounds, and hydroxyl and furan compounds as the flavor component content data.

6. The method for establishing an intelligent blending model of Chinese liquor according to claim 5, characterized in that: The flavor component content data of the white liquor is composed of four major categories: acid compounds, ester compounds, alcohol compounds, and hydroxyl and furan compounds, forming four different basic data sets. The sensory evaluation score data is divided into four categories: color, aroma, taste, and style. And the entire sensory evaluation score data set is combined with the data sets of the four major categories of acid compounds, ester compounds, alcohol compounds, and hydroxyl and furan compounds to form four new different basic data sets.

Images