A liquor fermentation prediction and feedback intervention system based on digital twin

Abstract

The present invention discloses a liquor fermentation prediction and feedback intervention system based on digital twins, which relates to the field of liquor fermentation technology; it comprises: a modeling data selection unit for screening out data types with strong correlation with liquor quality from historical data and empirical data, and generating five twin data modules in order of correlation; a fusion modeling unit for fusion modeling of the five twin data modules with each stage of the entire life cycle of liquor, and then establishing an ester composition prediction function model; an early warning unit for completing real-time prediction of the composition of ester organic matter according to the ester composition prediction function model; a feedback unit for receiving information sent by the early warning unit and taking corresponding processing measures; with strong data support, the present invention realizes the prediction of the quality results of liquor after fermentation in the cellar is completed, and this prediction can be performed at any time during the fermentation process, so that intervention can be made during the fermentation process to ensure the quality of liquor.

Description

Technical Field

[0001] The present invention relates to the field of liquor fermentation technology, and specifically to a liquor fermentation prediction and feedback intervention system based on digital twins. Background Art

[0002] Traditional brewing methods rely primarily on the brewer's experience and manual labor. This approach is plagued by issues such as high manual intervention, low production efficiency, and high costs. Digital twin technology is widely used in industrial manufacturing, logistics, healthcare, and other fields. By building simulation models, it can simulate and predict the physical world, improving manufacturing efficiency and product quality. Therefore, digital twin technology can be applied to the baijiu brewing process, enabling predictive and feedback intervention, thereby improving both quality and production efficiency.

[0003] Numerous factors influence baijiu (white liquor) quality, including fermentation temperature, fermentation time, bacterial population and distribution, geographical temperature variations, and the moisture content of the mash. These factors are even influenced by the production staff's operational experience and techniques, as well as the composition of the raw materials used, such as the water and fermentation grain sources. Therefore, digital twin technology is necessary. This involves testing and integrating data from incoming brewing ingredients. By integrating this extensive data with historical data, we can conduct multi-level monitoring and early warning optimization based on this vast amount of testing data, encompassing regions, work sections, and units. This allows us to minimize factors that can cause product quality fluctuations throughout the production process, providing timely early warnings and feedback interventions to ensure product quality. To this end, we provide a baijiu fermentation prediction and feedback intervention system based on digital twins. Summary of the Invention

[0004] The purpose of the present invention is to provide a liquor fermentation prediction and feedback intervention system based on digital twin.

[0005] The technical problems solved by the present invention are:

[0006] (1) How to select data with strong correlation from the massive amount of data that may affect the quality of liquor through the modeling data selection unit, so as to more accurately determine the parameters that affect the quality of liquor and monitor them, thereby solving the problem in the existing technology that it is difficult to conduct targeted quantitative analysis of the parameters that affect the quality of liquor;

[0007] (2) How to fuse the twin data modules with strong correlation through the fusion modeling unit, and model the static data and dynamic data separately, and finally obtain the ester organic matter composition ratio prediction function model. The early warning unit uses this model to make predictions, which solves the problem in the existing technology that it is difficult to complete the real-time prediction of the quality of liquor during the production and fermentation process.

[0008] (3) How to set up an early warning unit and a feedback unit to compare the predicted ester organic matter composition ratio with the target value, and use the deviation value to adjust the parameters, so as to solve the problem that the existing technology can only perform feedback adjustment based on a single parameter of each node in the fermentation sequence, but cannot perform adaptive adjustment based on the quality of the wine.

[0009] The present invention can be implemented through the following technical solutions: A liquor fermentation prediction and feedback intervention system based on digital twins, comprising: a modeling data selection unit for screening out data types with strong correlation with liquor quality from historical data and empirical data, and generating five twin data modules in order of correlation;

[0010] The fusion modeling unit is used to integrate the five twin data modules with each stage of the entire life cycle of liquor to build a prediction function model for ester composition;

[0011] The early warning unit draws a real-time curve based on the real-time data stream of fermentation and matches it with a fitting standard curve. The composition ratio of ester organic matter is mapped according to the function of the fitting standard curve to complete the prediction. At any time during fermentation, the twin data at the current moment is encapsulated as a twin data module and imported into the ester composition prediction function to complete the real-time prediction of the ester organic matter composition.

[0012] The feedback unit is used to receive the information sent by the early warning unit and take corresponding processing measures.

[0013] A further technical improvement of the present invention is that the specific steps of the modeling data selection unit performing relevance screening on data that may be relevant to liquor fermentation include:

[0014] S1: Classify potentially related data into categories and group the same type of data, and then standardize, normalize, and discretize them;

[0015] S2: Pearson product-moment correlation coefficient is used to measure the correlation between the corresponding data and the quality of liquor;

[0016] S3: Rank according to the correlation coefficient in S2, and classify and integrate similar data to obtain the corresponding twin data module.

[0017] A further technical improvement of the present invention is that the fusion modeling unit divides the twin data module into static data and dynamic data, and adopts different data processing methods accordingly:

[0018] For static data, combined with historical data, a matrix relationship function between the twin data module belonging to static data and the composition ratio of ester organic matter is established separately;

[0019] For dynamic data, multiple groups of data are extracted from historical data. The fermentation time variance of the data in the group does not exceed the set value. A curve is established in the coordinate system according to the fermentation time sequence and the corresponding ester organic matter composition ratio is marked. The deviation between it and the target ester organic matter composition ratio is calculated, and the weight value is assigned according to the deviation. A curve is fitted to obtain the fitting standard curve.

[0020] A further technical improvement of the present invention is to integrate the data of the same fermentation time into one group, thereby obtaining a large number of data groups, obtaining a fitting standard curve corresponding to the fermentation time and a curve driving function of the curve, and establishing a mapping relationship between the curve driving function and the composition ratio of ester organic matter, that is, f i (t)→P i , P i Indicates the composition ratio of the mapped ester organic matter.

[0021] A further technical improvement of the present invention is that the fusion modeling unit fuses the matrix relationship function of static data with the curve driving function of dynamic data to obtain an ester composition prediction function model:

[0022]

[0023] Among them, a+b+c+∑d i =1, a, b, c, d are the distribution coefficients ranked according to correlation, and the higher the ranking, the larger the distribution coefficient; W(x, y, z, t) represents the composition ratio of ester organic matter.

[0024] A further technical improvement of the present invention is that when the early warning unit predicts the composition ratio of ester organic matter, if there is a deviation from the target value, a feedback adjustment signal is generated and sent to the feedback unit.

[0025] A further technical improvement of the present invention is that the early warning unit also uses the fitted standard curve to monitor data and issue early warnings during production, performs vertical and horizontal comparisons based on the real-time data obtained at the corresponding fermentation moment, and determines whether the data is incorrect or abnormal based on the comparison results.

[0026] A further technical improvement of the present invention is that: during the vertical comparison, the total fermentation time is considered to be consistent with the selected time of the historical data, the data deviation coefficient is calculated and it is determined whether it is within an acceptable range; if it exceeds the acceptable range, a horizontal comparison is performed. At this time, it is first assumed that the acquired data is normal, and the fitted standard curve is translated left and right along the fermentation time sequence, each time with a step length of thirty minutes, and it is analyzed whether the real-time data can be matched with the curve data. When the match can be completed, the total translation step length at this time is obtained and sent to the feedback unit. When the match cannot be completed, the data is considered to be abnormal, and an abnormal alarm signal is generated and sent to the feedback unit.

[0027] A further technical improvement of the present invention is that the feedback unit receives a feedback adjustment signal and adjusts the parameter values ​​in the subsequent fermentation process according to the fermentation stage of the corresponding cellar and expert experience data. The parameter values ​​include temperature and humidity, pressure and fermentation time. The adjustment of the parameter values ​​is carried out within a set range in the direction of compensating for data anomalies.

[0028] A further technical improvement of the present invention is that after the feedback unit obtains the total translation step length, it will translate the opening time of the corresponding cellar, the translation step length is consistent with the obtained total translation compensation, and the translated fitting standard curve is used for comparison in subsequent early warning feedback.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] 1. By using the modeling data selection unit to select data with strong correlation from the massive data that may affect the quality of liquor, and encapsulating similar twin data, the parameters affecting the quality of liquor can be determined more systematically and accurately for monitoring, making it easier to predict the composition of ester organic matter in liquor based on changes in parameters, and to use strong data support to guide production.

[0031] 2. Through the fusion modeling unit, the twin data modules with strong correlation are distinguished, and the static data and dynamic data therein are modeled separately and fused together to finally obtain the ester composition ratio prediction function model. The early warning unit uses this model to make predictions, thereby realizing the prediction of the quality results of the liquor after the fermentation in the cellar is completed. This prediction can be made at any time during the fermentation process, so that intervention can be made during the fermentation process to ensure the quality of the liquor.

[0032] 3. By setting up an early warning unit and a feedback unit, the predicted ester organic matter composition ratio is compared with the target value, and the deviation value is used to adjust the parameters, thereby avoiding the existing technology that can only perform feedback adjustment based on a single parameter of each node in the fermentation time sequence. In addition, this adjustment is empirical and has no data model for data support, so it is not persuasive. In the present invention, the parameters can be adjusted at various stages of production by utilizing the strongly correlated data association relationship, which is highly persuasive and stable. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] To facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0034] Figure 1 It is a schematic diagram of the workflow of the present invention;

[0035] Figure 2 This is a system block diagram of the present invention. DETAILED DESCRIPTION

[0036] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0037] See also Figure 1-2 As shown, a liquor fermentation prediction and feedback intervention system based on digital twins includes a modeling data selection unit, a fusion modeling unit, an early warning unit and a feedback unit.

[0038] The modeling data selection unit selects data types that are highly correlated with liquor quality from a large amount of historical and empirical data. Data that may be relevant to the quality of brewed liquor include geographic location, climate, origin of grain raw materials, water source, fermentation temperature, fermentation time, water content, production staff proficiency, health status, bacterial species, quantity, and hierarchical arrangement. Among them, liquor quality is generally closely related to the ester organic matter in the brewed liquor, of which ethyl lactate, ethyl acetate, and ethyl hexanoate are the main ones, accounting for more than 90% of the total esters. Therefore, the quality of liquor can be tested by focusing on the proportion and composition of these three esters, thereby obtaining a data-supported quality assessment.

[0039] The specific steps of screening the collected data for relevance include:

[0040] S1: Classify the above possible related data into types, group the data of the same type, and convert the data of different types into the set type and perform standardization, normalization and discretization processing;

[0041] S2: The Pearson product-moment correlation coefficient is used to measure the correlation between the corresponding data and the quality of the liquor. The calculation method is as follows:

[0042]

[0043] Where cov(X,Y) is the covariance of variables X and Y; σ X is the standard deviation of variable X; σ Y is the standard deviation of the variable Y; the result r(corr) ranges from [–1, 1].

[0044] The absolute value of r reflects the magnitude of the correlation. Generally speaking, when r is between 0.00 and ±0.30, the two variables are slightly correlated; when r is between ±0.30 and ±0.50, the two variables are actually correlated; when r is between ±0.50 and ±0.80, the two variables are significantly correlated; when r is between ±0.80 and ±1.00, the two variables are highly correlated.

[0045] S3: Rank the correlation coefficients between the above possible related data and the liquor quality obtained from the correlation analysis in step S2, and classify the data of the same type. For example, if the fermentation temperature ranks first in the correlation coefficient, and the fermentation temperature is the liquor fermentation condition, then the fermentation temperature, fermentation time and water content are classified as a type of twin data and marked as the fermentation condition twin data module. Similarly, five twin data modules are generated in the order of correlation size, namely, the fermentation condition twin data module, the material source twin data module, the climate environment twin data module, the microbial distribution twin data module and the production personnel twin data module;

[0046] The fusion modeling unit associates the five twin data modules generated above with each stage of the entire life cycle of liquor fermentation and performs fusion modeling to obtain a prediction feedback model throughout the entire life cycle. When constructing the prediction feedback model, the five twin data are divided into static data with fixed values ​​and dynamic data that changes in real time as the fermentation sequence progresses;

[0047] The material source twin data module, colony distribution twin data module, and production personnel twin data module are defined as static data and marked as A2, A4, and A5 respectively. The fermentation condition twin data module is defined as dynamic data and marked as B1 and B3 respectively.

[0048] Specifically, the material source twin data module includes the material type and the corresponding material origin and storage time; the bacterial colony distribution twin data module includes the bacterial colony type, the corresponding bacterial colony number, and the bacterial colony arrangement method (position and hierarchical order); the production personnel twin data module includes the production personnel's work number, job length and health status; the climate environment twin data module includes the geographical location, ambient temperature and humidity, and atmospheric pressure;

[0049] With the support of a large amount of historical data, the fusion modeling unit first separately establishes the matrix relationship functions between A2, A4 and A5 in each of the above twin data modules and the proportion of ester organic matter in liquor, which are G(x), G(y) and G(z) respectively, where x, y and z are group variables, x = [x1, x2, ..., xn] T , y=[y1,y2,...,yn] T , z=[z1,z2,...,zn] T ,n is a positive integer;

[0050] Since B1 and B3 are real-time data streams, there are problems such as data instability, data duplication, and data similarity. In addition, the two real-time data streams are correlated in the time dimension and the previous and next data. We use the non-local mean algorithm to denoise the real-time data streams and randomly extract multiple groups of data from different fermentation batches from the historical data. It should be noted that the variance value of the fermentation time of the corresponding fermentation batch does not exceed the set value. In addition, different data points are marked in the corresponding virtual plane rectangular coordinate system according to the fermentation time sequence, and adjacent points are connected with smooth curves. In each virtual plane rectangular coordinate system, multiple curves are generated (such as the temperature-time curve in the cellar), and each curve corresponds to the composition ratio of ester organic matter.

[0051] Set the target ester organic compound composition ratio, compare the marked ester organic compound composition ratio with the target ester organic compound composition ratio and calculate the deviation, assign a weight value to the data of each curve according to the deviation, and fit a new curve based on the curve data after the weight value is assigned. Mark it as the fitting standard curve, and obtain the curve driving function f that drives the curve i (t), where i is used to distinguish different data types and t represents the fermentation time;

[0052] Repeat the above operation and integrate the data of the same fermentation time into one group, thereby obtaining a large number of data groups, obtaining the fitting standard curve corresponding to the fermentation time and the curve driving function of the curve, and establishing a mapping relationship between the curve driving function and the composition ratio of ester organic matter, that is, f i (t)→P i , P i Indicates the composition ratio of the mapped ester organic matter;

[0053] The matrix relationship function of static data and the curve driving function of dynamic data are integrated into the model to obtain the ester composition prediction function model:

[0054]

[0055] Among them, a+b+c+∑d i =1, a, b, c, d are the distribution coefficients obtained by ranking according to the correlation, and the higher the ranking, the larger the distribution coefficient; W(x, y, z, t) represents the composition ratio of ester organic matter;

[0056] The early warning unit draws a real-time curve with the acquired real-time data stream, and matches the closest fitting standard curve in the process of establishing the ester composition prediction function model according to the trend of the real-time curve, so as to map the corresponding ester organic matter composition ratio according to the fitting standard curve; during the entire fermentation process, at any time, the current twin data is substituted into the ester composition prediction function model, and the ester organic matter composition after the completion of the liquor fermentation can be predicted in real time; when the predicted ester organic matter composition ratio deviates from the target value, a feedback adjustment signal is generated and sent to the feedback unit;

[0057] The early warning unit can also use the fitted standard curve to perform data monitoring and early warning during production, and perform vertical and horizontal comparisons based on the real-time data at the corresponding fermentation time: vertical comparison means that the total fermentation time is considered to be consistent with the selected time of the historical data, and the data deviation coefficient is calculated. When the data deviation coefficient is within the acceptable range, it is determined that the current fermentation state is normal; when the data deviation coefficient exceeds the acceptable range, the next five groups of data are continuously compared. If the next five groups of data are all beyond the acceptable range, a horizontal comparison is performed. At this time, it is first assumed that the acquired data is normal, and the fitted standard curve is translated left and right along the fermentation time sequence, each time with a step length of thirty minutes, and the real-time data is analyzed to see whether it can be matched with the curve data. If it can be matched, the total translation step length at this time is obtained and sent to the feedback unit. If it cannot be matched, the data is considered abnormal, and an abnormal alarm signal is generated and sent to the feedback unit.

[0058] The feedback unit receives the feedback adjustment signal and adjusts the parameters of the subsequent fermentation process based on the fermentation stage of the corresponding cellar and expert experience data. The parameters include temperature, humidity, pressure, and fermentation time. It should be noted that these parameters are adjusted within the set range to compensate for data anomalies.

[0059] After obtaining the total translation step length, the feedback unit will appropriately shift the opening time of the corresponding cellar. The translation step length will be consistent with the total translation compensation obtained, and the shifted fitting standard curve will be used for comparison in subsequent early warning feedback. At the same time, after receiving the abnormal alarm signal, the feedback unit will indicate that the data abnormality in the cellar is not caused by time series shift, but by problems with the fermentation conditions of the cellar. Through wireless communication, the oldest person in the production staff of the corresponding cellar will be designated to go and handle the matter, ensuring efficient and timely abnormal handling.

[0060] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as a preferred embodiment as above, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A liquor fermentation prediction and feedback intervention system based on digital twin, characterized by: include: The modeling data selection unit is used to screen out data types that are highly correlated with liquor quality from historical data and empirical data, and generate five twin data modules in order of correlation: the fermentation condition twin data module, the material source twin data module, the climate environment twin data module, the microbial population distribution twin data module, and the production personnel twin data module; The fusion modeling unit is used to integrate the five twin data modules with each stage of the entire life cycle of liquor to build a prediction function model for ester composition; The fusion modeling unit divides the twin data module into static data and dynamic data, and adopts different data processing methods accordingly: For static data, combined with historical data, a matrix relationship function between the twin data module belonging to static data and the composition ratio of ester organic matter is established separately; For dynamic data, multiple groups of data are extracted from historical data. The fermentation time variance of the data in the group does not exceed the set value. A curve is established in the coordinate system according to the fermentation time sequence and the corresponding ester organic matter composition ratio is marked. The deviation between the curve and the target ester organic matter composition ratio is calculated. The weight value is assigned according to the deviation degree, and a curve is obtained by fitting, which is the fitting standard curve. The data of the same fermentation time are integrated into one group to obtain a large number of data groups, and the fitting standard curve corresponding to the fermentation time and the curve driving function of the curve are obtained, and a mapping relationship between the curve driving function and the composition ratio of ester organic matter is established, that is, , Indicates the composition ratio of the mapped ester organic matter; The fusion modeling unit fuses the matrix relationship function of static data with the curve driving function of dynamic data to obtain the ester composition prediction function model: in, , a, b, c, d are the distribution coefficients obtained by ranking according to the correlation, and the distribution coefficient value of the top ranking is larger; Indicates the composition ratio of ester organic matter; The early warning unit draws a real-time curve based on the real-time data stream of fermentation and matches it with a fitting standard curve. The composition ratio of ester organic matter is mapped according to the function of the fitting standard curve to complete the prediction. At any time during fermentation, the twin data at the current moment is encapsulated as a twin data module and imported into the ester composition prediction function to complete the real-time prediction of the ester organic matter composition. The feedback unit is used to receive the information sent by the early warning unit and take corresponding processing measures.

2. A liquor fermentation prediction and feedback intervention system based on digital twin according to claim 1, characterized in that: The specific steps of the modeling data selection unit performing relevance screening on data that may be relevant to liquor fermentation include: S1: Classify potentially related data into categories and group the same type of data, and then standardize, normalize, and discretize them; S2: Pearson product-moment correlation coefficient is used to measure the correlation between the corresponding data and the quality of liquor; S3: Rank according to the correlation coefficient in S2, and classify and integrate similar data to obtain the corresponding twin data module.

3. The digital twin-based liquor fermentation prediction and feedback intervention system according to claim 1 is characterized in that: When the early warning unit predicts the composition ratio of ester organic matter, if there is a deviation from the target value, a feedback adjustment signal is generated and sent to the feedback unit.

4. The digital twin-based liquor fermentation prediction and feedback intervention system according to claim 3 is characterized in that: The early warning unit also uses the fitted standard curve to perform data monitoring and early warning during production, performs vertical and horizontal comparisons based on the real-time data obtained at the corresponding fermentation time, and determines whether the data is incorrect or abnormal based on the comparison results.

5. The digital twin-based liquor fermentation prediction and feedback intervention system according to claim 4 is characterized in that: During the vertical comparison, the total fermentation time is considered to be consistent with the selected time of the historical data, and the data deviation coefficient is calculated to determine whether it is within the acceptable range; if it exceeds the acceptable range, a horizontal comparison is performed. At this time, it is first assumed that the acquired data is normal, and the fitted standard curve is translated left and right along the fermentation time sequence, each time with a step length of thirty minutes, and it is analyzed whether the real-time data can be matched with the curve data. When the match can be completed, the total translation step length at this time is obtained and sent to the feedback unit. When the match cannot be completed, the data is considered to be abnormal, and an abnormal alarm signal is generated and sent to the feedback unit.

6. The digital twin-based liquor fermentation prediction and feedback intervention system according to claim 5 is characterized in that: The feedback unit receives the feedback adjustment signal and adjusts the parameter values ​​in the subsequent fermentation process according to the fermentation stage of the corresponding cellar and expert experience data. The parameter values ​​include temperature and humidity, pressure and fermentation time. The adjustment of the parameter values ​​is carried out within the set range in the direction of compensating for data anomalies.

7. The digital twin-based liquor fermentation prediction and feedback intervention system according to claim 5 is characterized in that: After obtaining the total translation step length, the feedback unit translates the opening time of the corresponding cellar, and the translation step length is consistent with the obtained total translation compensation, and the translated fitting standard curve is used for comparison in subsequent early warning feedback.

Images