Moisture Prediction and Control Method in Tobacco Making Process Based on Moisture Content of Cigarette

Abstract

The present invention discloses a method for predicting and controlling moisture content in the tobacco-making process based on cigarette moisture content, comprising the steps of S1: principal component parameter screening, S2: neural network prediction model establishment, S3: model prediction and control, and S4: feedback control, adjustment, and optimization. The present invention uses the moisture content of finished cigarettes as a target and reversely predicts the optimal moisture content at the tobacco-mixing and flavoring outlet. This in turn provides a control basis for the amount of water added during the loosening and leaf-feeding steps and the amount of dehydration during the tobacco-making process. The moisture prediction and control system for the tobacco-making process based on cigarette moisture content has adaptive and self-learning capabilities, and can continuously adjust the adaptability of the system to the production process based on the current production status. This effectively shortens the moisture feedback time between the rolling and packaging workshops and the tobacco-making workshops, improving the compatibility and applicability of tobacco-making and the tobacco-making process.

Description

Technical Field

[0001] The present invention relates to the field of tobacco shred processing, and in particular to a moisture prediction and control method for a shred processing process based on the moisture content of cigarettes. Background Art

[0002] Moisture content is an important factor affecting the sensory quality of cigarette products. The moisture content of finished cigarettes not only affects the physical indicators of the cigarettes, but also affects the smoking quality and mainstream smoke of the cigarettes. Therefore, stable moisture content in cigarettes plays a vital role in ensuring the quality of cigarettes.

[0003] The moisture content of cigarettes is affected by parameters such as the tobacco-making and rolling processes. Currently, tobacco moisture testing is performed in ovens or offline, where the tobacco is removed from the cigarette. This method is time-consuming and has a lag in controlling moisture during the tobacco-making process, resulting in unstable moisture control on the tobacco production line. Summary of the Invention

[0004] In order to overcome the above problems, the present invention provides a moisture prediction and control method for the tobacco-making process based on the moisture content of the cigarettes.

[0005] The present invention takes the moisture content of finished cigarettes as the target, reversely predicts the most suitable moisture content at the outlet of mixed and flavored tobacco, and further provides a control basis for the amount of water added in the loosening and moisture regaining process of the tobacco leaves, and the amount of dehydration in the tobacco drying process, thereby continuously adjusting the adaptability of the system and the production process according to the current production situation.

[0006] The technical solution adopted by the present invention to solve the technical problem is:

[0007] The moisture prediction and control method of tobacco-making process based on tobacco moisture includes:

[0008] S1 principal component parameter screening

[0009] Conduct principal component analysis on the tobacco storage conditions and locomotive parameters of the same batch of historical production data to determine the principal component influencing parameters that affect the moisture content of cigarettes;

[0010] The main component influencing parameters include the storage time of wire, storage humidity, amount of wire waste, vehicle speed, soldering iron temperature, VE suction chamber negative pressure, and target cigarette moisture;

[0011] S2 prediction model establishment

[0012] With the influencing parameters selected by principal component analysis as the input layer and the outlet moisture of mixed and flavored cigarettes as the output layer, a prediction model for the outlet moisture of mixed and flavored cigarettes based on the moisture content of cigarettes was established.

[0013] Train a model for predicting outlet moisture content of blended and flavored cigarettes based on historical production data sets;

[0014] Model training parameters: set the training target to 0.05, the training speed to 0.01, and the maximum number of steps to 100;

[0015] The accuracy of the model is tested by the root mean square difference between the predicted value and the actual value;

[0016] S3 Model Prediction and Control

[0017] S3.1 Combine the moisture process standards of each brand of cigarettes and the current moisture test data of cigarettes to determine a reasonable moisture content as the target value of cigarette moisture:

[0018] Typically, the process standard is the standard value ± tolerance, and the test data fluctuates within the range of [standard value - 0.5% to standard value + 0.5%], occasionally exceeding the standard requirement. Therefore, the method for determining the target moisture value of cigarettes by combining the process standard and the test data is to compare the test data minus the standard value with the tolerance. If the difference between the test data minus the standard value and the tolerance is large, the target moisture value of the cigarettes is adjusted to bring it closer to the standard moisture value of the cigarettes.

[0019] S3.2 obtains the target moisture value (i.e., predicted value) of the mixed yarn and fragrance outlet based on the mixed yarn and fragrance outlet moisture prediction model;

[0020] S3.3 Based on the leaf water absorption and dehydration logic during the silk production process, a linear regression equation for loosening and regaining, tobacco drying, and blending and flavoring is established to form an expert database for silk production process control recommendations:

[0021] The linear regression equation established with the moisture at the entrance of loose rehydration, the amount of water added during loose rehydration and the amount of dehydration of tobacco cut as variables and the moisture at the outlet of mixed tobacco cut and flavoring as the dependent variable is: 混 =M 入 +a*I 松 +b*R 脱 ;

[0022] Based on the target moisture value of the mixed and flavored tobacco outlet and the linear regression equation, the amount of water added for loosening and regaining during the tobacco making process and the amount of dehydration during tobacco drying are obtained. The actual moisture value of the mixed and flavored tobacco outlet and the actual moisture value of the cigarette are also obtained under the control of the target moisture value of the mixed and flavored tobacco outlet.

[0023] S3.4 Secondary accuracy check:

[0024] Conduct deviation analysis based on the target moisture value at the mixed and flavored outlet and the actual moisture value at the mixed and flavored outlet: If the deviation between the target moisture value at the mixed and flavored outlet and the actual moisture value at the mixed and flavored outlet is ≤0.5%, continue to conduct deviation analysis on the moisture content of the cigarettes; if the deviation exceeds 0.5%, feedback is required for adjustment and optimization S4.1;

[0025] Deviation analysis is performed based on the target and actual cigarette moisture values. If the deviation between the target and actual cigarette moisture values ​​is ≤0.5%, the model-predicted moisture content at the blended and flavored outlet meets the current cigarette moisture requirements and the model control accuracy is good. If the deviation exceeds 0.5%, feedback is required for adjustment and optimization (S4.2).

[0026] S4 feedback control adjustment and optimization

[0027] S4.1 Based on the linear regression equation and the loose rehydration inlet moisture content, adjust the loose rehydration water injection amount and the tobacco drying and dehydration amount in subsequent batches to obtain the optimized target moisture content at the mixed tobacco flavoring outlet;

[0028] S4.2 Execute S3.1 to obtain the optimized target moisture value of the cigarette.

[0029] As an improvement of the above technical solution, the moisture prediction and control method of the tobacco-making process based on the moisture content of the cigarette also includes:

[0030] S5 model adaptation

[0031] Continuously improve the adaptability of the prediction model to the actual production process based on historical control results and measured cigarette moisture.

[0032] As an improvement of the above technical solution, the moisture prediction and control method of the tobacco-making process based on the moisture content of the cigarette also includes:

[0033] S6 model self-learning

[0034] The model is learned and optimized by manually detecting the moisture content of cigarettes and the moisture content of mixed and flavored cigarettes.

[0035] As an improvement to the above technical solution, the method for screening the principal component parameters of S1 is:

[0036] Constructing historical production data into a matrix Among them, d nm Represents the mth data of the nth parameter;

[0037] Normalize the matrix D to get the matrix X: in, (n=1,2,...; m=1,2,...);

[0038] Establish the correlation coefficient matrix R between parameters: in,

[0039] Calculation of principal component contribution rate:

[0040] Cumulative contribution rate of principal components:

[0041] The first p parameters corresponding to the cumulative contribution rate of 85% to 95% are taken as the principal components for comprehensive analysis, and the first p principal components corresponding to the contribution rate of 85% to 95% are taken as the principal component parameters.

[0042] As an improvement to the above technical solution, the principal component parameters screened by S1 include:

[0043] Wire storage time, wire storage temperature, wire storage humidity, wire return amount, soldering iron temperature, VE suction chamber negative pressure, vehicle speed and cigarette moisture content.

[0044] As an improvement to the above technical solution, S1 principal component parameter screening also includes

[0045] Perform dimensionality reduction analysis on the selected variables to obtain new unrelated variables:

[0046] Get the correlation matrix of each principal component parameter;

[0047] Get the variance and contribution rate of each principal component;

[0048] Select factors with eigenvalues ​​greater than 1 to replace the original variables;

[0049] The orthogonal rotation method with Kaiser standardization was used to obtain the factor loadings of each environmental factor on different main factors and obtain the component matrix;

[0050] Select the variables that contribute most to each component as the final principal component influencing parameters:

[0051] Wire storage time, wire storage humidity, wire return amount, soldering iron temperature, VE suction chamber negative pressure, vehicle speed and cigarette moisture content.

[0052] The beneficial effects brought by the present invention are:

[0053] The present invention performs principal component analysis on the factors affecting the moisture content of cigarettes, reduces the dimensionality and complexity of the influencing parameters, and forms several new unrelated parameters to construct a moisture prediction and control model and system for the tobacco-making process based on the moisture content of cigarettes. At the same time, an expert database for moisture control in the tobacco-making process based on the moisture content of cigarettes is established. The moisture content of target cigarettes is used as input to predict the moisture content at the mixed tobacco and flavoring outlet, and then the amount of water added for loosening and regaining moisture in the tobacco-making process, the amount of dehydration in the tobacco drying process, and the adjustment range are fed back, providing new ideas for controlling the moisture content of finished cigarette products.

[0054] The present invention can provide a control basis for loosening and regaining moisture in the silk production process, the amount of water added to the leaf material, and the amount of dehydration in the silk drying process, shorten the moisture feedback time between the rolling workshop and the silk-making workshop, and improve the matching and applicability of the silk-making tobacco and the rolling process. At the same time, the moisture prediction control system for the silk-making process based on the moisture of the cigarettes also has a self-learning function. It performs self-learning according to the data recorded in the production process and the process control standards, continuously optimizes the model, ensures the accuracy of the feedback system and the high adaptability of the production process, and improves the coordinated linkage between the silk production and the rolling process. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The present invention will be further described below with reference to the accompanying drawings and specific embodiments:

[0056] Figure 1 This is a prediction control flow chart of the present invention;

[0057] Figure 2 is a system block diagram of the present invention;

[0058] Figure 3 Schematic diagram of the neural network prediction model structure. DETAILED DESCRIPTION

[0059] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0060] Example 1

[0061] The moisture prediction and control method of tobacco-making process based on tobacco moisture includes:

[0062] S1 principal component parameter screening

[0063] Conduct principal component analysis on the tobacco storage conditions and locomotive parameters of the same batch of historical production data to determine the principal component influencing parameters that affect the moisture content of cigarettes;

[0064] The main component influencing parameters include the storage time of wire, storage humidity, amount of wire waste, vehicle speed, soldering iron temperature, VE suction chamber negative pressure, and target cigarette moisture;

[0065] S2 prediction model establishment

[0066] With the influencing parameters selected by principal component analysis as the input layer and the outlet moisture of mixed and flavored cigarettes as the output layer, a prediction model for the outlet moisture of mixed and flavored cigarettes based on the moisture content of cigarettes was established.

[0067] Train a model for predicting outlet moisture content of blended and flavored cigarettes based on historical production data sets;

[0068] Model training parameters: set the training target to 0.05, the training speed to 0.01, and the maximum number of steps to 100;

[0069] The model accuracy test is tested by the root mean square difference between the predicted value and the actual value (see Example 3);

[0070] S3 Model Prediction and Control

[0071] by Figure 1 For example, including

[0072] S3.1 Combine the moisture process standards of each brand of cigarettes and the current moisture test data of cigarettes to determine a reasonable moisture content as the target value of cigarette moisture:

[0073] Typically, the process standard is the standard value ± tolerance, and the test data fluctuates within the range of [standard value - 0.5% to standard value + 0.5%], occasionally exceeding the standard requirement. Therefore, the method for determining the target moisture value of cigarettes by combining the process standard and the test data is to compare the test data minus the standard value with the tolerance. If the difference between the test data minus the standard value and the tolerance is large, the target moisture value of the cigarettes is adjusted to bring it closer to the standard moisture value of the cigarettes.

[0074] S3.2 obtains the target moisture value (i.e., predicted value) of the mixed yarn and fragrance outlet based on the mixed yarn and fragrance outlet moisture prediction model;

[0075] S3.3 Based on the leaf water absorption and dehydration logic during the silk production process, a linear regression equation for loosening and regaining, tobacco drying, and blending and flavoring is established to form an expert database for silk production process control recommendations:

[0076] The linear regression equation established with the moisture at the entrance of loose rehydration, the amount of water added during loose rehydration and the amount of dehydration of tobacco cut as variables and the moisture at the outlet of mixed tobacco cut and flavoring as the dependent variable is: 混 =M 入 +a*I 松 +b*R 脱 ;

[0077] Based on the target moisture value of the mixed and flavored tobacco outlet and the linear regression equation, the amount of water added for loosening and regaining during the tobacco making process and the amount of dehydration during tobacco drying are obtained. The actual moisture value of the mixed and flavored tobacco outlet and the actual moisture value of the cigarette are also obtained under the control of the target moisture value of the mixed and flavored tobacco outlet.

[0078] S3.4 Secondary accuracy check:

[0079] Conduct deviation analysis based on the target moisture value at the mixed and flavored outlet and the actual moisture value at the mixed and flavored outlet: If the deviation between the target moisture value at the mixed and flavored outlet and the actual moisture value at the mixed and flavored outlet is ≤0.5%, continue to conduct deviation analysis on the moisture content of the cigarettes; if the deviation exceeds 0.5%, feedback is required for adjustment and optimization S4.1;

[0080] Deviation analysis is performed based on the target and actual cigarette moisture values. If the deviation between the target and actual cigarette moisture values ​​is ≤0.5%, the model-predicted moisture content at the blended and flavored outlet meets the current cigarette moisture requirements and the model control accuracy is good. If the deviation exceeds 0.5%, feedback is required for adjustment and optimization (S4.2).

[0081] S4 feedback control adjustment and optimization

[0082] S4.1 Based on the linear regression equation and the loose rehydration inlet moisture content, adjust the loose rehydration water injection amount and the tobacco drying and dehydration amount in subsequent batches to obtain the optimized target moisture content at the mixed tobacco flavoring outlet;

[0083] S4.2 Execute S3.1 to obtain the optimized target moisture value of the cigarette.

[0084] As an improvement of the above technical solution, the moisture prediction and control method of the tobacco-making process based on the moisture content of the cigarette also includes:

[0085] S5 model adaptation

[0086] Continuously improve the adaptability of the prediction model to the actual production process based on historical control results and measured cigarette moisture.

[0087] As an improvement of the above technical solution, the moisture prediction and control method of the tobacco-making process based on the moisture content of the cigarette also includes:

[0088] S6 model self-learning

[0089] The model is learned and optimized by manually detecting the moisture content of cigarettes and the moisture content of mixed and flavored cigarettes.

[0090] As an improvement to the above technical solution, the method for screening the principal component parameters of S1 is:

[0091] Constructing historical production data into a matrix Among them, d nm Represents the mth data of the nth parameter;

[0092] Normalize the matrix D to get the matrix X: in, (n=1,2,...; m=1,2,...);

[0093] Establish the correlation coefficient matrix R between parameters: in,

[0094] Calculation of principal component contribution rate:

[0095] Cumulative contribution rate of principal components:

[0096] The first p parameters corresponding to the cumulative contribution rate of 85% to 95% are taken as the principal components for comprehensive analysis, and the first p principal components corresponding to the contribution rate of 85% to 95% are taken as the principal component parameters.

[0097] As an improvement to the above technical solution, the principal component parameters screened by S1 include:

[0098] Wire storage time, wire storage temperature, wire storage humidity, wire return amount, soldering iron temperature, VE suction chamber negative pressure, vehicle speed and cigarette moisture content.

[0099] As an improvement to the above technical solution, S1 principal component parameter screening also includes

[0100] Perform dimensionality reduction analysis on the selected variables to obtain new unrelated variables:

[0101] Get the correlation matrix of each principal component parameter;

[0102] Get the variance and contribution rate of each principal component;

[0103] Select factors with eigenvalues ​​greater than 1 to replace the original variables;

[0104] The orthogonal rotation method with Kai ser standardization was used to obtain the factor loadings of each environmental factor on different main factors and obtain the component matrix;

[0105] Select the variables that contribute most to each component as the final principal component influencing parameters:

[0106] Wire storage time, wire storage humidity, wire return amount, soldering iron temperature, VE suction chamber negative pressure, vehicle speed and cigarette moisture content.

[0107] Example 2

[0108] The moisture prediction and control method of tobacco-making process based on tobacco moisture includes:

[0109] S1 principal component parameter screening

[0110] Principal component analysis was conducted on the tobacco storage conditions and equipment parameters from the blending and flavoring outlet to the cigarette outlet to screen out the main component parameters that affect the moisture content of tobacco.

[0111] Principal component analysis is to explore the correlation between multiple variables by using dimensionality reduction thinking. It conducts principal component analysis on the historical moisture content of mixed silk flavoring outlet, storage time, storage temperature, storage humidity, amount of recycled silk, soldering iron temperature, VE suction chamber negative pressure, vehicle speed and cigarette moisture content. The steps of principal component analysis are as follows:

[0112] Constructing historical production data into a matrix Among them, d nm Represents the mth data of the nth parameter;

[0113] Normalize the matrix D to get the matrix X: in, (n=1,2,...; m=1,2,...);

[0114] Establish the correlation coefficient matrix R between parameters: in,

[0115] Calculation of principal component contribution rate:

[0116] Cumulative contribution rate of principal components:

[0117] The first p parameters corresponding to the cumulative contribution rate of 85% to 95% are taken as the principal components for comprehensive analysis, and the first p principal components corresponding to the contribution rate of 85% to 95% are taken as the principal component parameters.

[0118] Through experimental analysis:

[0119] There is a certain correlation between the selected wire storage conditions and locomotive parameters, as shown in Table 1. If these are directly used as input to the neural network prediction model, the number of models would be large. Therefore, it is necessary to use principal component analysis to reduce the dimension of the wire storage conditions and locomotive parameter data to obtain new uncorrelated variables. The results of the principal component analysis are shown in Table 2.

[0120] Table 1 Correlation matrix

[0121]

[0122] Table 2 Variance and principal component contribution rate

[0123]

[0124] As shown in Table 2, the eigenvalues ​​of the first three principal component factors are all greater than 1, so the first three principal component factors are selected to replace the original variables.

[0125] The orthogonal rotation method with Kaiser standardization was used to obtain the factor loadings of various environmental factors on different principal component factors. The resulting component matrix is ​​shown in Table 3: the parameters that contribute most to component 1 are wire storage time, wire waste amount, VE suction chamber negative pressure, vehicle speed, and cigarette moisture; the parameters that contribute most to component 2 are soldering iron temperature, and the parameter that contributes most to component 3 is wire storage humidity.

[0126] Table 3 Component matrix

[0127] Element Ingredient 1 Ingredient 2 Ingredient 3 Silk storage time 0.890 -0.159 0.207 Wire storage temperature -0.521 0.186 -0.042 Silk storage humidity -0.048 0.246 0.912 Amount of waste yarn 0.812 0.190 -0.212 Soldering iron temperature -0.116 0.862 0.060 VE suction chamber negative pressure -0.606 -0.517 0.340 Speed 0.902 -0.108 0.151 Cigarette moisture -0.920 0.022 -0.095

[0128] Therefore, seven key influencing factors, including wire storage time, wire storage humidity, wire waste amount, soldering iron temperature, VE suction chamber negative pressure, vehicle speed and cigarette moisture, were finally selected as the input layer of the neural network model.

[0129] S2 prediction model establishment

[0130] The model consists of an input layer, a hidden layer, and an output layer. It takes the storage time, storage humidity, waste volume, soldering iron temperature, VE suction chamber negative pressure, vehicle speed, and cigarette moisture as inputs, and the moisture content of the mixed and flavored cigarette outlet as output. The training target is set to 0.05, the training speed is 0.01, and the maximum number of steps is 100. The structure of the artificial neural network prediction model is as follows: Figure 3 shown.

[0131] S3 prediction model application

[0132] The above-mentioned mixed silk fragrance export moisture prediction model was put into operation online.

[0133] Based on the model and the target cigarette moisture content, the moisture content analysis of the mixed and flavored cigarette outlet was fed back. Ten batches were randomly selected, and the comparison between the actual and predicted moisture content of the mixed and flavored cigarette outlet is shown in Table 4.

[0134] Table 4 Application results of the export moisture prediction model for blended and flavored silk

[0135]

[0136]

[0137] As shown in the table above, the model accuracy is 0.034 < 0.05. This shows that this method can achieve moisture control in the tobacco production process based on the moisture content of the cigarette.

[0138] Example 3

[0139] Model accuracy testing methods include

[0140] The neural network was used to predict the moisture content of the blended and flavored silk at the outlet, and the accuracy of the neural network prediction was evaluated.

[0141] The evaluation method was performed by root mean square error (RMSE);

[0142] Root mean square deviation calculation formula: Where: X obs，i ——Actual value of moisture, %; X model,i ——Model predicted value, %; n——number of experiments.

[0143] Table 5 Accuracy evaluation of the export moisture prediction model for mixed silk with fragrance

[0144]

[0145] The three groups of accuracy calculations in Table 5 were all based on the comparison between the model predicted values ​​and the actual values. Each group was repeated 10 times, and the calculated accuracy was all <0.05, and the accuracy test met the requirements.

[0146] Example 4

[0147] Reference Figure 2 , based on the moisture content of cigarettes, the moisture prediction control system for the tobacco-making process includes

[0148] Parameter library unit, including equipment parameter module, wire storage condition module and prediction model module;

[0149] It is used to store equipment parameters for each batch in historical production data, tobacco storage conditions such as storage time, ambient temperature and humidity, etc. Based on this, principal component analysis is performed to determine the main component influencing parameters that affect the moisture content of cigarettes;

[0150] The principal component analysis method can refer to other embodiments. The main component influencing parameters finally determined include the storage time of the wire, the storage humidity of the wire, the amount of wire waste, the speed of the machine, the temperature of the soldering iron, the negative pressure of the VE suction chamber, and the moisture content of the target cigarette.

[0151] It is also used to establish a neural network prediction model based on the selected principal component influencing parameters:

[0152] With the influencing parameters selected by principal component analysis as the input layer and the outlet moisture of mixed and flavored cigarettes as the output layer, a prediction model for the outlet moisture of mixed and flavored cigarettes based on the moisture content of cigarettes was established.

[0153] Train a model for predicting outlet moisture content of blended and flavored cigarettes based on historical production data sets;

[0154] Expert database unit, which is used to obtain the target mixed silk fragrance outlet moisture and form the expert database recommended silk making control mode;

[0155] To obtain the target blended silk fragrance outlet moisture——

[0156] First, combine the moisture process standards of each brand of cigarettes and the current cigarette moisture test data to determine a reasonable cigarette moisture as the cigarette moisture target value:

[0157] Secondly, the target moisture value of the mixed silk flavoring outlet is obtained based on the moisture prediction model of the mixed silk flavoring outlet;

[0158] Recommended silk making control methods to form an expert database——

[0159] First, based on the leaf water absorption and dehydration logic during the silk production process, a linear regression equation for loosening and regaining, tobacco drying, and blending and flavoring was established, forming an expert database for silk production process control recommendations:

[0160] The linear regression equation established with the moisture at the entrance of loose rehydration, the amount of water added during loose rehydration and the amount of dehydration of tobacco cut as variables and the moisture at the outlet of mixed tobacco cut and flavoring as the dependent variable is: 混 =M 入 +a*I 松 +b*R 脱 ;

[0161] Secondly, based on the target moisture value of the mixed tobacco flavoring outlet and the linear regression equation, the loosening moisture recovery and watering amount of tobacco cuttings in the tobacco making process and the drying and dehydration amount of tobacco cuttings were obtained;

[0162] Data collection unit, including moisture self-collection module, cigarette moisture process standard module and mixed tobacco flavoring outlet moisture process standard module;

[0163] It performs secondary accuracy verification by obtaining the actual moisture value of the mixed silk flavoring outlet and the actual moisture value of the cigarette under the silk making control method recommended by the expert database:

[0164] Conduct deviation analysis based on the target moisture value at the mixed and flavored outlet and the actual moisture value at the mixed and flavored outlet: If the deviation between the target moisture value at the mixed and flavored outlet and the actual moisture value at the mixed and flavored outlet is ≤0.5%, continue to conduct deviation analysis on the moisture content of the cigarettes; if the deviation exceeds 0.5%, feedback needs to be sent to the expert database unit for adjustment and optimization;

[0165] Deviation analysis is performed based on the target and actual moisture values ​​of cigarettes. If the deviation between the target and actual moisture values ​​is ≤0.5%, it indicates that the moisture content of the blended and flavored outlet predicted by the model meets the current moisture requirements of the cigarettes and the model control accuracy is good. If the deviation exceeds 0.5%, feedback needs to be sent to the expert library unit for adjustment and optimization.

[0166] The method of feedback adjustment and optimization of the expert library unit is:

[0167] According to the linear regression equation and the moisture content at the inlet of loose rehydration, the amount of water added during the loose rehydration process and the amount of drying and dehydration of cut tobacco were adjusted accordingly in subsequent batches, thereby obtaining the optimized target moisture value at the outlet of mixed and flavored tobacco.

[0168] Or optimize the target moisture value of the cigarette.

[0169] In this embodiment, the moisture prediction control system for the tobacco-making process based on the moisture content of the cigarettes further includes a system maintenance unit, which includes

[0170] Adaptive unit, used to continuously improve the adaptability of the prediction model to the actual production process based on historical control results and measured cigarette moisture content;

[0171] The self-learning unit optimizes the model learning by manually detecting the moisture content of the cigarettes and the moisture content of the mixed and flavored outlet.

[0172] It should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent replacements for some of the technical features therein. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A moisture prediction and control method for the tobacco-making process based on tobacco moisture, characterized by: include S1 principal component parameter screening Conduct principal component analysis on the tobacco storage conditions and locomotive parameters of the same batch in historical production data to determine the principal component influencing parameters that affect the moisture content of cigarettes; The main component influencing parameters include the storage time of wire, storage humidity, amount of wire waste, vehicle speed, soldering iron temperature, VE suction chamber negative pressure, and target cigarette moisture; S2 prediction model establishment With the influencing parameters selected by principal component analysis as the input layer and the outlet moisture of mixed and flavored cigarettes as the output layer, a prediction model for the outlet moisture of mixed and flavored cigarettes based on the moisture content of cigarettes was established. Train a model for predicting outlet moisture content of blended and flavored cigarettes based on historical production data sets; S3 Model Prediction and Control S3.1 Combine the moisture process standards of each brand of cigarettes and the current moisture test data of cigarettes to determine a reasonable moisture content as the target value of cigarette moisture: S3.2 obtains the target moisture value of the mixed yarn flavoring outlet based on the mixed yarn flavoring outlet moisture prediction model; S3.3 Based on the leaf water absorption and dehydration logic during the silk production process, a linear regression equation for loosening and regaining, tobacco drying, and blending and flavoring is established to form an expert database for silk production process control recommendations: The linear regression equation established with the moisture at the entrance of loose rehydration, the amount of water added during loose rehydration and the amount of dehydration of tobacco cut as variables and the moisture at the outlet of mixed tobacco cut and flavoring as the dependent variable is: 混 =M 入 +a*I 松 +b*R 脱 ; Based on the target moisture value of the mixed and flavored tobacco outlet and the linear regression equation, the amount of water added for loosening and regaining during the tobacco making process and the amount of dehydration during tobacco drying are obtained. The actual moisture value of the mixed and flavored tobacco outlet and the actual moisture value of the cigarette are also obtained under the control of the target moisture value of the mixed and flavored tobacco outlet. S3.4 Secondary accuracy check: Conduct deviation analysis based on the target moisture value at the mixed and flavored outlet and the actual moisture value at the mixed and flavored outlet: If the deviation between the target moisture value at the mixed and flavored outlet and the actual moisture value at the mixed and flavored outlet is ≤0.5%, continue to conduct deviation analysis on the moisture content of the cigarettes; if the deviation exceeds 0.5%, feedback is required for adjustment and optimization S4.1; Deviation analysis is performed based on the target and actual moisture values ​​of cigarettes. If the deviation between the target and actual moisture values ​​is ≤0.5%, it indicates that the moisture content of the blended and flavored outlet predicted by the model meets the current moisture requirements of the cigarettes, and the model control accuracy is good. If the deviation exceeds 0.5%, feedback is required for adjustment and optimization S4.2; S4 feedback control adjustment and optimization S4.1 Based on the linear regression equation and the loose rehydration inlet moisture content, adjust the loose rehydration water injection amount and the tobacco drying and dehydration amount in subsequent batches to obtain the optimized target moisture content at the mixed tobacco flavoring outlet; S4.2 Execute S3.1 to obtain the optimized target moisture value of the cigarette.

2. The method for predicting and controlling moisture content in a silk-making process according to claim 1, wherein: The control method also includes S5 model adaptation: Continuously improve the adaptability of the prediction model to the actual production process based on historical control results and measured cigarette moisture.

3. The method for predicting and controlling moisture content in a silk-making process according to claim 1, wherein: The control method also includes S6 model self-learning: The model is learned and optimized by manually detecting the moisture content of cigarettes and the moisture content of mixed and flavored cigarettes.

4. The method for predicting and controlling moisture content in a silk-making process according to claim 1, wherein: The principal component parameters screened by S1 include: Wire storage time, wire storage temperature, wire storage humidity, wire return amount, soldering iron temperature, VE suction chamber negative pressure, vehicle speed and target cigarette moisture content.

5. The method for predicting and controlling moisture content in a silk-making process according to claim 4, wherein: S1 principal component parameter screening also includes Perform dimensionality reduction analysis on the selected variables to obtain new unrelated variables: Get the correlation matrix of each principal component parameter; Get the variance and contribution rate of each principal component; Select factors with eigenvalues ​​greater than 1 to replace the original variables; The orthogonal rotation method with Kaiser standardization was used to obtain the factor loadings of each environmental factor on different main factors and obtain the component matrix; Select the variables that contribute most to each component as the final principal component influencing parameters: Wire storage time, wire storage humidity, wire return amount, soldering iron temperature, VE suction chamber negative pressure, vehicle speed and target cigarette moisture content.

6. The method for predicting and controlling moisture content in a silk-making process according to claim 1, wherein: The S2 prediction model also includes Model training parameters: set the training target to 0.05, the training speed to 0.01, and the maximum number of steps to 100; The accuracy of the model is tested by the root mean square difference between the predicted value and the actual value.

Images