Model for relation between quantitative water quantity and fermented grain moisture of Luzhou-flavor liquor and establishment method thereof

Abstract

According to the relation model of the weighted water quantity and the fermented grain moisture of the Luzhou-flavor liquor and the establishment method of the relation model, weighted water quantity data and corresponding fermented grain moisture content data serve as a data set, an appropriate model is selected for regression analysis and adjustment optimization, and the qualified relation model of the weighted water quantity and the fermented grain moisture is obtained. Fine guidance is provided for measuring the water dosage in the Baijiu production, production fluctuation caused by human experience operation is avoided, the Baijiu production efficiency and the product stability are improved, and the method is suitable for large-scale popularization and application.

Description

Technical Field

[0001] The invention relates to the technical field of liquor production, and in particular to a model for the relationship between the measured water volume of Luzhou-flavor liquor and the moisture content of grain mash and a method for establishing the model. Background Art

[0002] Liquor has a long history, among which Luzhou-flavor liquor accounts for the largest share in the market. There are many operating processes and parameters in the production process of liquor that need to rely on traditional experience. Among them, measuring water is a very important operating step in the production process of Luzhou-flavor liquor. It means that after the fermented grains are distilled and steamed, their moisture content is not enough to reach the moisture content required for entering the cellar, and the grain starch has not been fully gelatinized. After a certain amount of water is added, it can ensure that the grains absorb enough water, and the grain starch can also absorb water and swell and gelatinize, thereby ensuring the normal fermentation of microorganisms in the later stage. Therefore, reasonable control of the moisture content of grains is of great significance to the production of Luzhou-flavor liquor, and measuring the amount of water is directly related to the moisture content of grains. In the actual production process, workers often rely more on experience to determine the amount of water used. The accuracy and stability of the amount of water used are difficult to guarantee. Too little water will cause insufficient moisture in the grains, and too much water will easily cause waste, and the starch in the grain will be taken away while the water is lost, which will have an adverse effect on fermentation.

[0003] CN103387906A discloses a production process of sesame-flavor liquor, comprising the following steps: mixing sorghum, wheat and bran, adding water to moisten the grains, and obtaining a first mixture; steaming the first mixture, discarding the steaming water, and obtaining a cooking material; cooling the cooking material, adding white koji, bacterial koji, yeast koji and a large koji prepared with a temperature greater than 60°C during the koji-making process, and mixing to obtain grain-fed material; adding mash, white koji, bacterial koji, yeast koji and a large koji prepared with a temperature greater than 60°C during the koji-making process after seven times of extracting sauce-flavor liquor to the grain-fed material, and performing a first fermentation to obtain a second mixture; transferring the second mixture to the bottom of a cellar and performing a second fermentation to obtain a third mixture; distilling the third mixture to obtain raw liquor; blending the raw liquor to obtain sesame-flavor liquor.

[0004] CN110317692A discloses a method for preparing a novel liquor. The method comprises steps 1, 2 and 3; Step 1 comprises the following steps: sorghum and wheat are used as liquor-making grains, wheat koji is selected, and the grains are crushed, moistened, mixed, steamed, taken out of the steamer and watered, cooled, added with koji, piled and fermented, turned over, and fermented in a cellar; Step 2 comprises the following steps: taking out the fermented mash, mixed with ingredients, picking liquor, steaming, taking out of the steamer and watered, cooled, added with koji, piled and fermented, turned over, and fermented in a cellar; each step in Step 2 is repeated three times in sequence; Step 3 comprises the following steps: taking out the mash that has been fermented in the cellar for the third time in Step 2, mixing with ingredients, picking liquor, and steaming.

[0005] CN108441381A discloses a process for brewing strong-flavor liquor, comprising the following steps: a. selecting main ingredients and auxiliary ingredients, and preparing the main ingredients in proportion; b. adding the auxiliary ingredients in step a and the prepared main ingredients in proportion to the fermented mother dregs, and cooking and gelatinizing them in three steamers; c. after cooking and gelatinizing, the material to be fermented is obtained through the steps of measuring water, cooling and spreading koji, and then the material to be fermented is placed in a fermentation tank for fermentation; d. after the fermentation period expires, the fermentation tank is opened, the fermented grain dregs in the fermentation tank are taken out for steaming, and the steamed wine is steamed in a 4-steamer mode, and the fermented mash obtained after steaming is used as the mother tank in step b for secondary fermentation, and the steamed wine is selected by quantity and quality to obtain the original wine; e. the original wine obtained in step d is graded and stored, and after the stored original wine is aged, it is blended and filtered, and then it can be put into storage after storage, filling and light inspection.

[0006] However, the method of controlling the amount of water in the above-mentioned liquor production process still relies on manual experience. There is an urgent need for a model of the relationship between the amount of water and the moisture content of grain mash to scientifically predict the required amount of water and achieve refined control of the production process. Summary of the invention

[0007] In view of the problems existing in the prior art, the present invention provides a relationship model between the amount of water measured for Luzhou-flavor liquor and the moisture content of mash and a method for establishing the model. The required amount of water measured can be scientifically and accurately predicted through the relationship model, thus avoiding production fluctuations caused by human experience and improving the stability of the liquor product.

[0008] To achieve this object, the present invention adopts the following technical solutions:

[0009] In a first aspect, the present invention provides a method for establishing a relationship model between the amount of water in a Luzhou-flavor liquor and the moisture content of grain mash, the method comprising the following steps:

[0010] (1) The water volume data and the corresponding grain mash moisture content data are used as the data set, and 5% to 10% of the water volume data and the corresponding grain mash moisture content data in the data set are randomly selected as the validation set, and the remaining data set is divided into a training set and a test set;

[0011] (2) using the regression model to perform model training and testing on the training set and test set described in step (1), respectively, and adjusting and optimizing the hyperparameters in the regression model to obtain a regression model after the test;

[0012] (3) using a validation set to validate the regression model tested in step (2) to obtain a validated regression model;

[0013] (4) The regression model verified in step (3) is verified in actual production. When the average difference between the predicted result of the established model and the actual moisture content in production is greater than 0.8%, steps (2) to (4) are repeated; otherwise, a model for the relationship between the amount of water in the fermented grains and the moisture content of the fermented grains of the Luzhou-flavor liquor is obtained.

[0014] The method for establishing a model of the relationship between the amount of water measured for Luzhou-flavor liquor and the moisture content of grain mash described in the present invention uses the water amount data and the corresponding grain mash moisture content data as data sets, selects a suitable model for regression analysis and adjustment optimization, and obtains a regression model that passes the test. The actual production data of the verification set is used for double verification to ensure that the final model of the relationship between the amount of water measured for Luzhou-flavor liquor and the moisture content of grain mash can accurately obtain the appropriate amount of water measured, provide refined guidance for the amount of water measured in winemaking production, avoid production fluctuations caused by human experience operations, improve efficiency and product stability, and save resources.

[0015] In step (1) of the present invention, 5% to 10% of the water volume data and the corresponding grain mash moisture content data in the data set are randomly selected as a validation set, for example, it can be 5%, 6%, 7%, 8%, 9% or 10%, etc., but it is not limited to the listed values, and other unlisted values ​​within the numerical range are also applicable.

[0016] In step (4) of the present invention, when the average difference between the predicted result of the established model and the actual moisture content in production is greater than 0.8%, for example, it can be 0.82%, 0.85%, 0.9%, 1%, 2%, 3% or 5%, etc., but is not limited to the listed values, and other unlisted values ​​within the numerical range are also applicable.

[0017] Preferably, the mash in step (1) comprises mash after distillation in the production of Luzhou-flavor liquor.

[0018] Preferably, the water volume measurement in step (1) includes a series of water volume measurement gradient data.

[0019] Preferably, the remaining data in step (1) is split into a training set and a test set using a five-fold cross split.

[0020] Preferably, the regression model in step (2) comprises a Gaussian process regression model.

[0021] Preferably, Bayesian optimization is used to adjust and optimize the hyperparameters in the regression model.

[0022] The present invention does not limit the specific steps of Bayesian optimization in detail, and conventional Bayesian optimization steps in the art can be used to perform hyperparameter adjustment and optimization.

[0023] Preferably, the hyperparameters in step (2) include width parameters, basis functions, kernel functions and kernel scales.

[0024] Preferably, the search range of the width parameter is 0.0001 to 0.30689, for example, it can be 0.0001, 0.001, 0.01, 0.1, 0.2, 0.25 or 0.30689, but is not limited to the listed values, and other unlisted values ​​within the range are also applicable.

[0025] Preferably, the basis function includes any one of a constant, zero or a linear function.

[0026] Preferably, the kernel function includes any one of a non-isotropic exponential function, a non-isotropic Mattern first differentiable function, a non-isotropic Mattern second differentiable function, a non-isotropic quadratic rational function, a non-isotropic square exponential function, an isotropic exponential function, an isotropic Mattern first differentiable function, an isotropic Mattern second differentiable function, an isotropic quadratic rational function or an isotropic square exponential function.

[0027] Preferably, the search range of the kernel scale is 0.7 to 700, for example, 0.7, 1, 10, 50, 100, 500 or 700, etc., but is not limited to the listed values, and other unlisted values ​​within the numerical range are also applicable.

[0028] Preferably, the evaluation index for adjusting and optimizing the hyperparameters in the regression model in step (2) includes: root mean square error <0.01, for example, it can be 0.009, 0.008, 0.005, 0.003, 0.001 or 0.0001, etc., but is not limited to the listed values, and other values ​​not listed in the numerical range are also applicable;

[0029] The mean square error is less than 0.0001, for example, it can be 0.00009, 0.00008, 0.00005, 0.00003, or 0.000001, but it is not limited to the listed values, and other values ​​not listed in the numerical range are also applicable;

[0030] R 2 >0.9, for example, it may be 0.91, 0.93, 0.95, 0.97, 0.98 or 0.99, etc., but it is not limited to the listed values, and other values ​​not listed in the numerical range are also applicable;

[0031] The mean absolute error is <0.008, for example, it can be 0.0079, 0.0077, 0.0075, 0.0073, 0.0071 or 0.007, etc., but it is not limited to the listed values, and other unlisted values ​​within the numerical range are also applicable.

[0032] Preferably, the validation set used in step (3) is used to validate the regression model after the test in step (2), and the root mean square error is less than 0.01, for example, it can be 0.009, 0.008, 0.005, 0.003, 0.001 or 0.0001, etc., but it is not limited to the listed values, and other values ​​not listed in the numerical range are also applicable;

[0033] The mean square error is less than 0.0001, for example, it can be 0.00009, 0.00008, 0.00005, 0.00003, or 0.000001, but it is not limited to the listed values, and other values ​​not listed in the numerical range are also applicable;

[0034] R 2 >0.9, for example, it may be 0.91, 0.93, 0.95, 0.97, 0.98 or 0.99, etc., but it is not limited to the listed values, and other values ​​not listed in the numerical range are also applicable;

[0035] and the mean absolute error < 0.008, for example, it can be 0.0079, 0.0077, 0.0075, 0.0073, 0.0071 or 0.007, etc., but it is not limited to the listed values. Other values ​​not listed in the numerical range are also applicable to obtain the verified regression model.

[0036] As a preferred technical solution of the present invention, the establishment method comprises the following steps:

[0037] (1) A series of water volume gradient data and the corresponding grain mash moisture content data are used as the data set. 5% to 10% of the water volume gradient data and the corresponding grain mash moisture content data are randomly selected as the validation set, and the remaining data set is split into a training set and a test set using a five-fold cross split;

[0038] The grain mash includes the grain mash after distillation in the production of Luzhou-flavor liquor;

[0039] (2) The training set and test set described in step (1) are trained and tested using a Gaussian process regression model, and the hyperparameters in the Gaussian process regression model are adjusted and optimized using Bayesian optimization, with the root mean square error <0.01, mean square error <0.0001, and R 2 >0.9 and mean absolute error <0.008, resulting in a tested regression model;

[0040] The hyper parameters include width parameter, basis function, kernel function and kernel scale;

[0041] The search range of the width parameter is 0.0001 to 0.30689; the basis function includes any one of a constant, zero or a linear function; the kernel function includes any one of a non-isotropic exponential function, a non-isotropic Mattern first-differentiable function, a non-isotropic Mattern second-differentiable function, a non-isotropic second-rational function, a non-isotropic square exponential function, an isotropic exponential function, an isotropic Mattern first-differentiable function, an isotropic Mattern second-differentiable function, an isotropic second-rational function or an isotropic square exponential function; the search range of the kernel scale is 0.7 to 700;

[0042] (3) Use the validation set to validate the regression model after the test in step (2), and meet the requirements of root mean square error < 0.01, mean square error < 0.0001, R 2 >0.9 and mean absolute error <0.008, resulting in a validated regression model;

[0043] (4) The regression model verified in step (3) is verified in actual production. When the average difference between the predicted result of the established model and the actual moisture content in production is greater than 0.8%, steps (2) to (4) are repeated; otherwise, a model for the relationship between the amount of water in the fermented grains and the moisture content of the fermented grains of the Luzhou-flavor liquor is obtained.

[0044] In a second aspect, the present invention further provides a model for the relationship between the measured water volume of Luzhou-flavor liquor and the moisture content of grain mash obtained by adopting the method for establishing the model for the relationship between the measured water volume of Luzhou-flavor liquor and the moisture content of grain mash described in the first aspect.

[0045] The model of the relationship between the measured water volume and the moisture content of the mash of the Luzhou-flavor liquor described in the present invention can realize the refined control of the measured water volume, ensure that the mash fully absorbs moisture, and prevent starch loss due to excessive moisture, thereby reducing resource waste and saving liquor preparation costs.

[0046] Compared with the prior art, the present invention has at least the following beneficial effects:

[0047] The method for establishing a model of the relationship between the measured water volume and the moisture content of mash of Luzhou-flavor liquor provided by the present invention uses the measured water volume data and the corresponding mash moisture content data as data sets, and selects a suitable model for regression analysis and adjustment optimization, thereby obtaining a qualified model of the relationship between the measured water volume and the moisture content of mash, providing refined guidance for the measured water dosage in liquor production, avoiding production fluctuations caused by human experience, improving liquor production efficiency and product stability, and having a prospect for large-scale promotion and application. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a flow chart of a method for establishing a model of the relationship between the amount of water measured for Luzhou-flavor liquor and the moisture content of grain mash provided in a specific embodiment of the present invention.

[0049] Figure 2 It is an iteration diagram of the modeling process in a specific implementation manner of the present invention.

[0050] Figure 3 It is the optimal model residual graph in the specific implementation manner of the present invention.

[0051] Figure 4 It is a comparison chart between the actual value and the predicted value in a specific implementation manner of the present invention. DETAILED DESCRIPTION

[0052] The technical solution of the present invention is further described below with reference to the accompanying drawings and through specific implementation methods.

[0053] The present invention is further described in detail below. However, the following examples are only simplified examples of the present invention and do not represent or limit the scope of protection of the present invention. The scope of protection of the present invention shall be subject to the claims.

[0054] As a specific embodiment of the present invention, a method for establishing a relationship model between the amount of water used in making Luzhou-flavor liquor and the moisture content of grain mash is provided, and its flow chart is as follows: Figure 1 shown.

[0055] The establishment method comprises the following steps:

[0056] (1) 548 water volume gradient data and corresponding grain mash moisture content data were used as a data set, 48 water volume gradient data and corresponding grain mash moisture content data were randomly selected from the data set as a validation set, and the remaining data set was divided into a training set of 250 and a test set of 250 by a five-fold cross split;

[0057] The grain mash includes the grain mash after distillation in the production of Luzhou-flavor liquor;

[0058] (2) using a Gaussian process regression model to perform model training and testing on the training set and test set described in step (1), respectively, and using Bayesian optimization to adjust and optimize the hyperparameters in the Gaussian process regression model to obtain a regression model after the test;

[0059] The modeling process iteration diagram in this specific implementation method is as follows Figure 2 As shown, from Figure 2 It can be seen that the minimum MSE hyperparameter is obtained at 23 iterations, and the optimal hyperparameter appears in the second iteration of 30.

[0060] Through model training and testing, the Gaussian process regression model was finally obtained. The width parameter is 0.0168, the basis function is a linear function, the kernel function is a non-isotropic Mattern first-order differentiable function, and the kernel scale is 0.93451. The root mean square error of the regression model test set is 0.005912, and the mean square error is 3.4951×e -5 , R 2 It is 0.96, and the mean absolute error is 0.0044331, which meets the requirements;

[0061] (3) Use the validation set to validate the regression model after the test in step (2), and meet the requirements of root mean square error < 0.01, mean square error < 0.0001, R 2 >0.9 and mean absolute error <0.008, resulting in a validated regression model;

[0062] In this specific implementation, the optimal model residual diagram is as follows: Figure 3 As shown, from Figure 3 It can be seen that the residual value is between ±0.02 and concentrated between ±0.005, which shows a good effect.

[0063] (4) The regression model verified in step (3) is actually verified in production. The comparison diagram between the actual value and the predicted value in this specific implementation is as follows: Figure 4 As shown, from Figure 4 It can be seen that the average difference between the prediction results of the established model and the actual moisture content in production is 0.4%, which is less than 0.8%. The relationship model between the amount of water in Luzhou-flavor liquor and the moisture content of grain mash is obtained, which is a Gaussian process regression model, in which the width parameter is 0.0168, the basis function is a linear function, the kernel function is a non-isotropic Mattern first-order differentiable function, and the kernel scale is 0.93451.

[0064] The amount of water obtained in this specific implementation method is reduced by an average of 150kg compared to the amount of water given by traditional manual experience. The moisture content of the grain mash meets the production requirements, saving production water.

[0065] In summary, the model of the relationship between the amount of water measured for Luzhou-aroma liquor and the moisture content of grain mash provided by the present invention can scientifically and accurately predict the amount of water measured for liquor brewing, realize refined control of the production process, avoid production fluctuations caused by human experience operations, and improve the stability of liquor products.

[0066] The applicant declares that the above is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Those skilled in the art should understand that any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention shall fall within the protection scope and disclosure scope of the present invention.

Claims

1. A method for establishing a relationship model between the amount of water in Luzhou-flavor liquor and the moisture content of grain mash, characterized in that: The establishment method comprises the following steps: (1) The water volume data and the corresponding grain mash moisture content data are used as the data set, and 5% to 10% of the water volume data and the corresponding grain mash moisture content data in the data set are randomly selected as the validation set, and the remaining data set is divided into a training set and a test set; (2) using the regression model to perform model training and testing on the training set and test set described in step (1), respectively, and adjusting and optimizing the hyperparameters in the regression model to obtain a regression model after the test; (3) using a validation set to validate the regression model tested in step (2) to obtain a validated regression model; (4) The regression model verified in step (3) is verified in actual production. When the average difference between the predicted result of the established model and the actual moisture content in production is greater than 0.8%, steps (2) to (4) are repeated; otherwise, a model for the relationship between the amount of water in the fermented grains and the moisture content of the fermented grains of the Luzhou-flavor liquor is obtained.

2. The establishment method according to claim 1, characterized in that: The mash in step (1) includes the mash after distillation in the production of Luzhou-flavor liquor.

3. The establishment method according to claim 1 or 2, characterized in that: The water volume measurement in step (1) includes a series of water volume gradient data.

4. The establishment method according to any one of claims 1 to 3, characterized in that: The remaining data in step (1) is split into a training set and a test set using a five-fold cross split.

5. The establishment method according to any one of claims 1 to 4, characterized in that: The regression model in step (2) includes a Gaussian process regression model; Preferably, Bayesian optimization is used to adjust and optimize the hyperparameters in the regression model.

6. The establishment method according to any one of claims 1 to 5, characterized in that: The hyper parameters in step (2) include width parameter, basis function, kernel function and kernel scale; Preferably, the search range of the width parameter is 0.0001 to 0.30689; Preferably, the basis function includes any one of a constant, zero or a linear function; Preferably, the kernel function includes any one of a non-isotropic exponential function, a non-isotropic Mattern first differentiable function, a non-isotropic Mattern second differentiable function, a non-isotropic second rational function, a non-isotropic square exponential function, an isotropic exponential function, an isotropic Mattern first differentiable function, an isotropic Mattern second differentiable function, an isotropic second rational function or an isotropic square exponential function; Preferably, the search range of the kernel scale is 0.7-700.

7. The establishment method according to any one of claims 1 to 6, characterized in that: The evaluation indicators for adjusting and optimizing the hyperparameters in the regression model in step (2) include: root mean square error < 0.01, mean square error < 0.0001, R 2 >0.9 and mean absolute error <0.

008.

8. The establishment method according to any one of claims 1 to 7, characterized in that: In step (3), the validation set is used to validate the regression model after the test in step (2), and the root mean square error < 0.01, the mean square error < 0.0001, and R 2 >0.9 and mean absolute error <0.008, resulting in a validated regression model.

9. The establishment method according to any one of claims 1 to 8, characterized in that: The establishment method comprises the following steps: (1) A series of water volume gradient data and the corresponding grain mash moisture content data are used as the data set. 5% to 10% of the water volume gradient data and the corresponding grain mash moisture content data are randomly selected as the validation set, and the remaining data set is split into a training set and a test set using a five-fold cross split; The grain mash includes the grain mash after distillation in the production of Luzhou-flavor liquor; (2) using a Gaussian process regression model to perform model training and testing on the training set and test set described in step (1), and using Bayesian optimization to adjust and optimize the hyperparameters in the Gaussian process regression model, with a root mean square error of <0.01, a mean square error of <0.0001, an R2 of >0.9, and a mean absolute error of <0.008, to obtain a regression model after testing; The hyper parameters include width parameter, basis function, kernel function and kernel scale; The search range of the width parameter is 0.0001 to 0.30689; the basis function includes any one of a constant, zero or a linear function; the kernel function includes any one of a non-isotropic exponential function, a non-isotropic Mattern first-differentiable function, a non-isotropic Mattern second-differentiable function, a non-isotropic second-rational function, a non-isotropic square exponential function, an isotropic exponential function, an isotropic Mattern first-differentiable function, an isotropic Mattern second-differentiable function, an isotropic second-rational function or an isotropic square exponential function; the search range of the kernel scale is 0.7 to 700; (3) Use the validation set to validate the regression model after the test in step (2), and meet the requirements of root mean square error < 0.01, mean square error < 0.0001, R 2 >0.9 and mean absolute error <0.008, resulting in a validated regression model; (4) The regression model verified in step (3) is verified in actual production. When the average difference between the predicted result of the established model and the actual moisture content in production is greater than 0.8%, steps (2) to (4) are repeated; otherwise, a model for the relationship between the amount of water in the fermented grains and the moisture content of the fermented grains of the Luzhou-flavor liquor is obtained.

10. A model for the relationship between the measured water volume of Luzhou-flavor liquor and the moisture content of grain mash obtained by using the method for establishing a model for the relationship between the measured water volume of Luzhou-flavor liquor and the moisture content of grain mash as described in any one of claims 1 to 9.

Images