Method for controlling moisture content at tobacco shred drying inlet under different temperature and humidity conditions based on PSO-ELM

Abstract

This invention discloses a method for controlling the inlet moisture content of leaf filament drying under different temperature and humidity conditions based on PSO-ELM. The method includes steps such as S1 dividing the ambient temperature and humidity range, S2 classifying the quality of inlet moisture content under different temperature and humidity conditions, S3 establishing a classification model for inlet moisture content under different temperature and humidity conditions, and S4 recommending optimal process parameters. This method uses machine learning and modeling for different temperatures, humidity levels, and moisture qualities. It selects the key process parameters corresponding to the highest probability that the inlet moisture content classified as high quality in medium-temperature low-humidity, high-temperature high-humidity, medium-temperature medium-humidity, and low-temperature medium-humidity zones actually belongs to the high-quality category. These parameters are then used as optimal process parameters for filament production, providing a reference for production process control and improving the stability of subsequent drying process quality.

Description

Technical Field

[0001] This invention relates to the field of tobacco processing moisture control, specifically to a method for controlling the moisture content at the tobacco leaf drying inlet under different temperature and humidity conditions based on PSO-ELM. Background Technology

[0002] In the cigarette processing, the moisture content of tobacco products is a crucial factor in ensuring the quality of tobacco processing and the internal and external quality of cigarette products. It is also a decisive indicator for evaluating the quality stability of the tobacco processing process. Tobacco raw materials belong to the category of biomass, which is a porous medium with hygroscopic and desiccant properties. Its moisture content is easily affected by changes in environmental conditions, thus impacting the quality of tobacco processing and cigarette rolling.

[0003] The leaf drying process is one of the key steps in cigarette manufacturing. The inlet moisture content is a crucial control point in cigarette manufacturing; its stability reflects the ability to control moisture in the early stages of processing and directly affects the moisture control effect on the finished tobacco in the later stages. Therefore, steady-state control of the inlet moisture content in the leaf drying process under different temperature and humidity conditions is essential.

[0004] Therefore, this application takes the external temperature and humidity of Zhangjiakou City as an example to divide different temperature and humidity ranges, and classifies the quality of the inlet moisture of the leaf drying process under different temperature and humidity ranges, so as to conduct research on the control of the inlet moisture of this process. The aim is to provide a theoretical basis for the reasonable optimization design of the cigarette production under different environmental conditions, to construct the inlet moisture control strategy of leaf drying under different environmental conditions, to optimize the process parameters of key processes, and to effectively improve the quality of cigarette processing and the stability of product quality. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a method for controlling the inlet moisture content of leaf filament drying under different temperature and humidity conditions based on PSO-ELM. This invention utilizes statistical analysis to differentiate the quality of the inlet moisture content of leaf filament drying under different temperature and humidity ranges. The optimal process parameters for filament production under different temperature and humidity conditions selected based on PSO-ELM can control the inlet moisture content of leaf filament drying within a high-quality range, effectively improving the stability of the leaf filament drying process.

[0006] The technical solution adopted by this invention to solve its technical problem is as follows:

[0007] A method for controlling inlet moisture for leaf drying under different temperature and humidity conditions based on PSO-ELM, including

[0008] S1 Environmental Temperature and Humidity Range Division

[0009] Temperature and humidity ranges are divided according to the annual ambient temperature and humidity of the production area;

[0010] Based on the monthly average temperature and humidity, K-means cluster analysis was performed on the annual environmental temperature and humidity of the city to divide it into different temperature and humidity ranges:

[0011] For example, there are four temperature and humidity ranges: medium temperature and low humidity, high temperature and high humidity, medium temperature and medium humidity, and low temperature and medium humidity.

[0012] S2 Leaf filament drying inlet moisture quality classification under different temperature and humidity conditions

[0013] Production data on leaf drying inlet moisture and key parameters were selected based on the months corresponding to different temperature and humidity ranges.

[0014] Key parameters include the moisture content at the loosening and rehydration inlet, the opening degree of the loosening and rehydration steam valve, the cumulative amount of water dispensed during loosening and rehydration, the moisture content at the loosening and rehydration outlet, the moisture content at the rehydration feeder inlet, the output opening degree of the blade feeding compensation steam valve, the moisture content at the rehydration feeder outlet, the moisture content at the hot air leaf humidifier inlet, the output opening degree of the hot air leaf humidifier compensation steam valve, and the moisture content at the hot air leaf humidifier outlet.

[0015] Excel was used to preprocess the raw production data, interpolation was used to fill in the missing values ​​in the data, the data was filtered for validity and outliers were removed according to the 3σ principle, and the relevant production data of tobacco raw materials were matched and associated according to the same batch and the same date.

[0016] The moisture content at the inlet of leaf drying under different temperature and humidity ranges was discretized by statistical analysis, and a discretization standard for the moisture content at the inlet of leaf drying was established. The moisture content at the inlet of leaf drying under different temperature and humidity ranges was classified into categories.

[0017]

[0018] Where μ is the average moisture content at the inlet of leaf drying for different temperature and humidity ranges; σ is the standard deviation of the moisture content at the inlet of leaf drying for different temperature and humidity ranges.

[0019] Establishment of a classification model for inlet moisture of leaf filaments under different temperature and humidity conditions (S3)

[0020] S3.1 Data Processing under Different Temperature and Humidity Conditions

[0021] The training and test sets are divided according to the method of taking one and then three for the association data of different categories under different temperature and humidity conditions, so as to ensure that the training and test sets have representative sample distributions under different temperature and humidity conditions.

[0022] The target variable, leaf filament drying inlet moisture content, was transformed using one-hot encoding, and the input variables were standardized.

[0023]

[0024] S3.2 Establishment of a PSO-ELM leaf filament drying inlet moisture classification model under different temperature and humidity conditions

[0025] The standardized loose rehydration inlet moisture, loose rehydration steam valve opening, loose rehydration water accumulation, loose rehydration outlet moisture, rehydration feeder inlet moisture, blade feeding compensation steam valve output opening, rehydration feeder outlet moisture, hot air leaf humidification inlet moisture, hot air leaf humidification machine compensation steam valve output opening, and hot air leaf humidification outlet moisture were used as the input layer nodes of the PSO-ELM classification model, and the leaf filament drying inlet moisture after unique thermal encoding was used as the output layer node, thus constructing a 3-layer network structure model with 10 inputs and 4 outputs.

[0026] Sigmoid and Linear were chosen as activation functions for ELM to minimize negative accuracy. PSO was used to optimize the number of neurons in the ELM algorithm. The population size in PSO was set to 20, the maximum number of iterations was set to 50, and a 5-fold crossover method was used for validation.

[0027] Recommended optimal process parameters for S4

[0028] Based on the PSO-ELM classification model, the inlet moisture of leaf filament drying is classified under different temperature and humidity conditions;

[0029] For leaf filament drying inlet moisture that is classified as high quality under different temperature and humidity conditions, the probability that it actually belongs to the high quality category is calculated, and the key process parameters corresponding to the leaf filament drying inlet moisture with the highest probability are selected as the optimal process parameters for filament production. These optimal process parameters can maximize the possibility of the leaf filament drying inlet moisture reaching high quality under the current temperature and humidity conditions.

[0030] As an improvement to the above technical solution, the method also includes S5 model self-learning and optimization.

[0031] The PSO-ELM classification model is continuously self-learned and optimized based on actual production data to improve the model's accuracy and its matching with actual production.

[0032] As an improvement to the above technical solution, S3 further includes

[0033] Evaluation of the S3.3 classification model

[0034] The predictive performance of the classification model is evaluated using accuracy (the proportion of correct predictions out of all predictions), precision (the proportion of true positives out of the total number of predictions that the model predicts as positive), recall (the proportion of true positives out of the actual positive samples), and F1 score (the harmonic mean of precision and recall). The prediction results of the classification model on different class attributes are visualized using a confusion matrix.

[0035] The present invention also aims to provide a leaf filament drying inlet moisture control system based on PSO-ELM under different temperature and humidity conditions, including...

[0036] The environmental temperature and humidity classification module is used to divide the temperature and humidity ranges based on the annual environmental temperature and humidity of the production area. For example, based on the monthly average temperature and humidity, it performs cluster analysis on the annual environmental temperature and humidity of the city and divides it into different temperature and humidity ranges.

[0037] The data filtering and processing module is used to select leaf drying inlet moisture data according to the months corresponding to different temperature and humidity ranges and to filter key parameters of the processes before leaf drying. At the same time, it associates and matches relevant pre-processed production data according to the same batch and the same date, and then uses statistical analysis to classify the leaf drying inlet moisture content under different temperature and humidity ranges into categories. Among them, key parameters include loose rehydration inlet moisture, loose rehydration steam valve opening, loose rehydration water accumulation, loose rehydration outlet moisture, rehydration feeder inlet moisture, leaf feeding compensation steam valve output opening, rehydration feeder outlet moisture, hot air leaf humidification inlet moisture, hot air leaf humidification machine compensation steam valve output opening, and hot air leaf humidification outlet moisture.

[0038] The classification model building module first divides the associated data of different categories under different temperature and humidity conditions into training and testing sets by taking one and then three samples respectively. It then transforms the target variable, leaf drying inlet moisture, using unique thermal encoding and standardizes the input variables. Subsequently, it uses the standardized loose rehydration inlet moisture, loose rehydration steam valve opening, loose rehydration water accumulation, loose rehydration outlet moisture, rehydration feeder inlet moisture, leaf feeding compensation steam valve output opening, rehydration feeder outlet moisture, hot air leaf moistening inlet moisture, and hot air leaf moistening machine compensation steam... The valve output opening degree and the moisture content at the hot air leaf-moistening outlet are used as input layer nodes of the PSO-ELM classification model, and the moisture content at the leaf filament drying inlet after unique thermal encoding is used as the output layer node. PSO-ELM classification models are constructed under different temperature and humidity conditions. The PSO-ELM classification model uses Sigmoid and Linear as activation functions for ELM, with minimizing negative accuracy as the optimal objective. The number of neurons in the PSO-ELM algorithm is optimized using PSO. The population size in PSO is set to 20, the maximum number of iterations is set to 50, and a 5-fold crossover method is used for validation.

[0039] The optimal process parameter recommendation module, based on the PSO-ELM classification model, classifies the inlet moisture content of the filament drying process under different temperature and humidity conditions. For the inlet moisture content of the filament drying process classified as high quality under different temperature and humidity conditions, it calculates the probability that it actually belongs to the high quality category, and selects the key process parameters corresponding to the inlet moisture content of the filament drying process with the highest probability as the optimal process parameters for filament production. The optimal process parameters can maximize the possibility of the inlet moisture content of the filament drying process reaching high quality under the current temperature and humidity conditions.

[0040] As an improvement to the above technical solution, the system also includes a model self-learning and optimization module, which continuously learns and optimizes the PSO-ELM classification model based on actual production data to improve the model's accuracy and its matching with actual production.

[0041] The beneficial effects of this invention are as follows:

[0042] This invention classifies the moisture content at the inlet of tobacco leaf drying under different temperature and humidity conditions, comprehensively considering the impact of environmental conditions on the production process, and effectively improving the stability of the tobacco leaf drying process. Machine learning and modeling are used for different temperatures, humidity levels, and moisture qualities. The key process parameters corresponding to the maximum probability that the inlet moisture content of tobacco leaf drying classified as high-quality actually belongs to the high-quality category under different temperature and humidity conditions are selected as the optimal process parameters for tobacco processing. By controlling the inlet moisture content of tobacco leaf drying to reach the optimal range, the stability of the tobacco leaf drying process is ensured. This aims to provide a reasonable optimization design for tobacco processing under different environmental conditions, construct a control strategy for the inlet moisture content of tobacco leaf drying under different environmental conditions, optimize key process parameters, and effectively improve the quality and stability of cigarette processing.

[0043] This invention enhances the controllability and predictability of the moisture content at the inlet of tobacco leaves during the tobacco processing by introducing the PSO-ELM classification model, thereby improving the stability of the quality of subsequent drying processes and providing an effective way for refined control of the tobacco processing process and optimization of the quality of finished tobacco products. Attached Figure Description

[0044] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0045] Figure 1 A schematic diagram of the process for controlling the inlet moisture of leaf filament drying under different temperature and humidity conditions;

[0046] Figure 2 System block diagram of the leaf filament drying inlet moisture control system under different temperature and humidity conditions;

[0047] Figures 3-6The following are confusion matrices representing the classification results of the PSO-ELM classification model in Example 6 in the medium-temperature and low-humidity zone, the high-temperature and high-humidity zone, the medium-temperature and medium-humidity zone, and the low-temperature and medium-humidity zone. Detailed Implementation

[0048] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments and accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0049] Example 1

[0050] according to Figure 1 A method for controlling inlet moisture of leaf filament drying under different temperature and humidity conditions based on PSO-ELM, including

[0051] S1 Environmental Temperature and Humidity Range Division

[0052] Temperature and humidity ranges are divided according to the annual ambient temperature and humidity of the production area;

[0053] Based on the monthly average temperature and humidity, K-means cluster analysis was performed on the annual environmental temperature and humidity of the city to divide it into different temperature and humidity ranges:

[0054] For example, it can be divided into four temperature and humidity ranges: medium temperature and low humidity, high temperature and high humidity, medium temperature and medium humidity, and low temperature and medium humidity.

[0055] S2 Leaf filament drying inlet moisture quality classification under different temperature and humidity conditions

[0056] Production data on leaf drying inlet moisture and key parameters were selected based on the months corresponding to different temperature and humidity ranges.

[0057] Key parameters include the moisture content at the loosening and rehydration inlet, the opening degree of the loosening and rehydration steam valve, the cumulative amount of water dispensed during loosening and rehydration, the moisture content at the loosening and rehydration outlet, the moisture content at the rehydration feeder inlet, the output opening degree of the blade feeding compensation steam valve, the moisture content at the rehydration feeder outlet, the moisture content at the hot air leaf humidifier inlet, the output opening degree of the hot air leaf humidifier compensation steam valve, and the moisture content at the hot air leaf humidifier outlet.

[0058] Excel was used to preprocess the raw production data, interpolation was used to fill in the missing values ​​in the data, the data was filtered for validity and outliers were removed according to the 3σ principle, and the relevant production data of tobacco raw materials were matched and associated according to the same batch and the same date.

[0059] The moisture content at the inlet of leaf drying under different temperature and humidity ranges was discretized by statistical analysis, and a discretization standard for the moisture content at the inlet of leaf drying was established. The moisture content at the inlet of leaf drying under different temperature and humidity ranges was classified into categories.

[0060] Table 1 Discretization and Classification Criteria of Leaf Filament Drying Inlet Moisture Content Attributes

[0061]

[0062]

[0063] Where μ is the average moisture content at the inlet of leaf drying for different temperature and humidity ranges; σ is the standard deviation of the moisture content at the inlet of leaf drying for different temperature and humidity ranges.

[0064] Establishment of a classification model for inlet moisture of leaf filaments under different temperature and humidity conditions (S3)

[0065] S3.1 Data Processing under Different Temperature and Humidity Conditions

[0066] The training and test sets are divided according to the method of taking one and then three for the association data of different categories under different temperature and humidity conditions, so as to ensure that the training and test sets have representative sample distributions under different temperature and humidity conditions.

[0067] The target variable, leaf filament drying inlet moisture content, was transformed using one-hot encoding, and the input variables were standardized.

[0068] First, calculate the average of the input variables, where x i Here, n is the value of the i-th data point, and n is the total number of data points.

[0069]

[0070] Next, calculate the standard deviation of the input variables:

[0071] Perform Z-score standardization:

[0072] S3.2 Establishment of a PSO-ELM leaf filament drying inlet moisture classification model under different temperature and humidity conditions

[0073] The standardized loose rehydration inlet moisture, loose rehydration steam valve opening, loose rehydration water accumulation, loose rehydration outlet moisture, rehydration feeder inlet moisture, blade feeding compensation steam valve output opening, rehydration feeder outlet moisture, hot air leaf humidification inlet moisture, hot air leaf humidification machine compensation steam valve output opening, and hot air leaf humidification outlet moisture were used as the input layer nodes of the PSO-ELM classification model, and the leaf filament drying inlet moisture after unique thermal encoding was used as the output layer node, thus constructing a 3-layer network structure model with 10 inputs and 4 outputs.

[0074] Sigmoid and Linear were chosen as activation functions for ELM to minimize negative accuracy. PSO was used to optimize the number of neurons in the ELM algorithm. The population size in PSO was set to 20, the maximum number of iterations was set to 50, and a 5-fold crossover method was used for validation.

[0075] Recommended optimal process parameters for S4

[0076] Based on the PSO-ELM classification model, the inlet moisture of leaf filament drying is classified under different temperature and humidity conditions;

[0077] For the leaf filament drying inlet moisture that is classified as high quality under different temperature and humidity conditions, calculate the probability that it actually belongs to the high quality category, and select the key process parameters corresponding to the leaf filament drying inlet moisture with the highest probability as the optimal process parameters for filament production. The optimal process parameters can maximize the possibility of the leaf filament drying inlet moisture reaching high quality under the current temperature and humidity conditions.

[0078] S5 Model Self-Learning and Optimization

[0079] The PSO-ELM classification model is continuously self-learned and optimized based on actual production data to improve the model's accuracy and its matching with actual production.

[0080] Example 2

[0081] A method for controlling inlet moisture for leaf drying under different temperature and humidity conditions based on PSO-ELM, including

[0082] S1 Environmental Temperature and Humidity Range Division

[0083] Zhangjiakou City has a temperate continental semi-arid monsoon climate with four distinct seasons, significant seasonal variations in precipitation, large diurnal temperature range, long and cold winters, dry and windy springs, hot and short summers with concentrated rainfall, and sunny and mild autumns.

[0084] Based on the monthly average temperature and humidity, K-means cluster analysis was conducted on the annual environmental temperature and humidity of Zhangjiakou City. According to the cluster centers, the whole year was clustered into 4 categories, which can be divided into 4 different temperature and humidity ranges: the first category is the medium temperature and low humidity zone, which is from April to May; the second category is the low temperature and medium humidity zone, which is from November to March; the third category is the high temperature and high humidity zone, which is from June to August; and the fourth category is the medium temperature and medium humidity zone, which is from September to October.

[0085] Table 2. Environmental conditions in Zhangjiakou City under different temperature and humidity ranges in 2022.

[0086]

[0087] S2 Leaf filament drying inlet moisture quality classification under different temperature and humidity conditions

[0088] Production data on leaf drying inlet moisture and key parameters were selected based on the months corresponding to different temperature and humidity ranges.

[0089] The following data were collected from the Manufacturing Information Management System (MES) of Zhangjiakou Cigarette Factory Co., Ltd. for the 2023 Diamond (Hard Yingbin) brand: Moisture content at the inlet of loose rehydration (A), opening of the loose rehydration steam valve (B), cumulative amount of water applied during loose rehydration (C), moisture content at the outlet of loose rehydration (D), moisture content at the inlet of leaf feeding (E), opening of the leaf feeding compensation steam valve (F), moisture content at the outlet of leaf feeding (G), moisture content at the inlet of hot air leaf humidification (H), opening of the hot air leaf humidification machine compensation steam valve (I), moisture content at the outlet of hot air leaf humidification (J), and moisture content at the inlet of leaf drying (all calculated as the average value for that batch). Environmental temperature and humidity data for Zhangjiakou City in 2022 were also collected from the China Meteorological Administration website.

[0090] The collected data was initially processed using Excel. Interpolation was used to fill in missing values ​​in the data. The data was then filtered for validity and outliers were removed according to the 3σ principle. The relevant data of tobacco raw materials were then matched and associated according to the same batch and the same date.

[0091] The moisture content at the inlet of leaf drying under different temperature and humidity ranges was discretized using statistical analysis. A discretization standard for the moisture content at the inlet of leaf drying was established. Referring to Table 1, the moisture content at the inlet of leaf drying under different temperature and humidity ranges was classified into categories. The discretized moisture content at the inlet of leaf drying under each category in different temperature and humidity ranges is shown in Table 3.

[0092] Table 3. Moisture content at the inlet of leaf filament drying under different temperature and humidity ranges.

[0093]

[0094] Establishment of a classification model for inlet moisture of leaf filaments under different temperature and humidity conditions (S3)

[0095] S3.1 Data Processing under Different Temperature and Humidity Conditions

[0096] The association data of different categories under different temperature and humidity conditions were divided into training and test sets by taking one and then three samples at a time to ensure that both the training and test sets have representative sample distributions under different temperature and humidity conditions. The number of samples of each category in each temperature and humidity interval after the division is shown in Table 4.

[0097] Table 4. Number of samples of each category under different temperature and humidity conditions

[0098]

[0099]

[0100] To meet the needs of ELM models in handling multi-class classification problems, one-hot encoding is used to transform the target variable, leaf filament drying inlet moisture, and the input variables are standardized.

[0101] S3.2 Establishment of a PSO-ELM leaf filament drying inlet moisture classification model under different temperature and humidity conditions

[0102] The standardized loose rehydration inlet moisture, loose rehydration steam valve opening, loose rehydration water accumulation, loose rehydration outlet moisture, rehydration feeder inlet moisture, blade feeding compensation steam valve output opening, rehydration feeder outlet moisture, hot air leaf humidification inlet moisture, hot air leaf humidification machine compensation steam valve output opening, and hot air leaf humidification outlet moisture were used as the input layer nodes of the PSO-ELM classification model, and the leaf filament drying inlet moisture after unique thermal encoding was used as the output layer node, thus constructing a 3-layer network structure model with 10 inputs and 4 outputs.

[0103] Sigmoid and Linear were chosen as activation functions for ELM to minimize negative accuracy. PSO was used to optimize the number of neurons in the ELM algorithm. The population size in PSO was set to 20, the maximum number of iterations was set to 50, and a 5-fold crossover method was used for validation.

[0104] The optimal classification results of the PSO-ELM model are shown in Table 5.

[0105] Table 5. Classification results of PSO-ELM model under different temperature and humidity conditions.

[0106]

[0107] As shown in Table 5, the PSO-ELM model achieves high accuracy, precision, recall, and F1 score in the high temperature and high humidity and medium temperature and medium humidity ranges.

[0108] Recommended optimal process parameters for S4

[0109] Based on the PSO-ELM classification model, the inlet moisture of leaf filament drying is classified under different temperature and humidity conditions;

[0110] For leaf filament drying inlet moisture that is classified as high quality under different temperature and humidity conditions, the probability that it actually belongs to the high quality category is calculated, and the key process parameters corresponding to the leaf filament drying inlet moisture with the highest probability are selected as the optimal process parameters for filament production. These optimal process parameters can maximize the possibility of the leaf filament drying inlet moisture reaching high quality under the current temperature and humidity conditions.

[0111] Based on the PSO-ELM classification model, the maximum probability that the inlet moisture of leaf filament drying, which is classified as high quality, actually belongs to the high quality category under different temperature and humidity conditions is shown in Table 6.

[0112] Table 6. Maximum Probability of Inlet Moisture Classification in High-Quality Leaf Filament Drying

[0113]

[0114] Table 7 shows the key process parameters corresponding to the highest probability that the inlet moisture content of the leaf filaments classified as high-quality actually belongs to the high-quality category, and their corresponding inlet moisture content of the leaf filaments in the original data.

[0115] Table 7 Optimal process parameters for filament production and their corresponding inlet moisture content for filament drying in the raw data.

[0116]

[0117]

[0118] To verify the correctness and effectiveness of the optimal process parameters, the obtained optimal process parameters for yarn production under different temperature and humidity conditions were compared with the original data. The quality of the inlet moisture content of the selected samples corresponding to the yarn drying process in the original dataset was observed to be consistent with the classification results. Table 7 shows that the inlet moisture content of the yarn drying process corresponding to the optimal process parameters selected by the PSO-ELM model under different temperature and humidity conditions is of high quality in the original data, consistent with the prediction results.

[0119] This embodiment demonstrates that the selected optimal process parameters for silk production can effectively control the inlet moisture content of the silk drying process within a high-quality range under different temperature and humidity conditions. The PSO-ELM model can accurately control the inlet moisture content of the silk drying process under different temperature and humidity conditions.

[0120] Example 3

[0121] according to Figure 2A leaf filament drying inlet moisture control system based on PSO-ELM under different temperature and humidity conditions, including

[0122] The environmental temperature and humidity classification module is used to divide the temperature and humidity ranges based on the annual environmental temperature and humidity of the production area. For example, based on the monthly average temperature and humidity, it performs cluster analysis on the annual environmental temperature and humidity of the city and divides it into different temperature and humidity ranges.

[0123] The data filtering and processing module is used to select leaf drying inlet moisture data according to the months corresponding to different temperature and humidity ranges and to filter key parameters of the processes before leaf drying. At the same time, it associates and matches relevant pre-processed production data according to the same batch and the same date, and then uses statistical analysis to classify the leaf drying inlet moisture content under different temperature and humidity ranges into categories. Among them, key parameters include loose rehydration inlet moisture, loose rehydration steam valve opening, loose rehydration water accumulation, loose rehydration outlet moisture, rehydration feeder inlet moisture, leaf feeding compensation steam valve output opening, rehydration feeder outlet moisture, hot air leaf humidification inlet moisture, hot air leaf humidification machine compensation steam valve output opening, and hot air leaf humidification outlet moisture.

[0124] The classification model building module first divides the associated data of different categories under different temperature and humidity conditions into training and testing sets by taking one and then three samples respectively. It then transforms the target variable, leaf drying inlet moisture, using unique thermal encoding and standardizes the input variables. Subsequently, it uses the standardized loose rehydration inlet moisture, loose rehydration steam valve opening, loose rehydration water accumulation, loose rehydration outlet moisture, rehydration feeder inlet moisture, leaf feeding compensation steam valve output opening, rehydration feeder outlet moisture, hot air leaf moistening inlet moisture, and hot air leaf moistening machine compensation steam... The valve output opening degree and the moisture content at the hot air leaf-moistening outlet are used as input layer nodes of the PSO-ELM classification model, and the moisture content at the leaf filament drying inlet after unique thermal encoding is used as the output layer node. PSO-ELM classification models are constructed under different temperature and humidity conditions. The PSO-ELM classification model uses Sigmoid and Linear as activation functions for ELM, with minimizing negative accuracy as the optimal objective. The number of neurons in the PSO-ELM algorithm is optimized using PSO. The population size in PSO is set to 20, the maximum number of iterations is set to 50, and a 5-fold crossover method is used for validation.

[0125] The optimal process parameter recommendation module, based on the PSO-ELM classification model, classifies the inlet moisture content of the filament drying process under different temperature and humidity conditions. For the inlet moisture content of the filament drying process classified as high quality under different temperature and humidity conditions, it calculates the probability that it actually belongs to the high quality category, and selects the key process parameters corresponding to the inlet moisture content of the filament drying process with the highest probability as the optimal process parameters for filament production. The optimal process parameters can maximize the possibility of the inlet moisture content of the filament drying process reaching high quality under the current temperature and humidity conditions.

[0126] The system also includes a model self-learning and optimization module, which continuously learns and optimizes the PSO-ELM classification model based on actual production data to improve the model's accuracy and its matching with actual production.

[0127] Example 4

[0128] Establishment of a GS-SVM inlet moisture classification model for leaf filament drying under different temperature and humidity conditions

[0129] The standardized inlet moisture content of loose rehydration, the opening degree of the loose rehydration steam valve, the cumulative amount of water sprayed during loose rehydration, the outlet moisture content of loose rehydration, the inlet moisture content of the blade feeder, the output opening degree of the blade feeder compensation steam valve, the outlet moisture content of the blade feeder, the inlet moisture content of the hot air humidifying leaf, the output opening degree of the hot air humidifying leaf compensation steam valve, and the outlet moisture content of the hot air humidifying leaf are used as inputs, and the inlet moisture content of the leaf filament drying after unique thermal coding is used as the output to construct a GS-SVM classification model for the inlet moisture content of leaf filament drying.

[0130] In this embodiment, RBF is used as the kernel function of SVM. GS is used to screen and optimize the penalty factor C in SVM and the parameter gamma in the kernel function. The 5-fold cross-validation method is used for verification. The optimal classification results of the GS-SVM model are shown in Table 8.

[0131] Table 8. Classification results of GS-SVM model under different temperature and humidity conditions.

[0132]

[0133] As shown in Table 8, the performance of the GS-SVM model fluctuates under different temperature and humidity conditions. It performs poorly under medium temperature and humidity conditions, but performs well under low temperature and medium humidity and medium temperature and low humidity conditions.

[0134] Example 5

[0135] Establishment of a GS-RF leaf filament drying inlet moisture classification model under different temperature and humidity conditions

[0136] The following parameters are used as inputs: loose rehydration inlet moisture, loose rehydration steam valve opening, loose rehydration water accumulation, loose rehydration outlet moisture, rehydration feeder inlet moisture, blade feeding compensation steam valve output opening, rehydration feeder outlet moisture, hot air leaf humidification inlet moisture, hot air leaf humidification machine compensation steam valve output opening, and hot air leaf humidification outlet moisture. The following parameter is used as output: leaf filament drying inlet moisture.

[0137] In this embodiment, GS is used to screen and optimize the depth of the tree and the number of classifiers in RF, and the 5-fold cross-validation method is used for verification. The optimal classification results of the GS-RF model are shown in Table 9.

[0138] Table 9. Classification results of the GS-RF model under different temperature and humidity conditions.

[0139]

[0140] As shown in Table 9, the GS-RF model exhibits a certain classification ability under various temperature and humidity conditions, with the best classification performance under low temperature and medium humidity conditions.

[0141] Example 6

[0142] This embodiment comprehensively compares the accuracy, precision, recall, and F1 score of the leaf filament drying inlet moisture classification models PSO-ELM, GS-SVM, and GS-RF established in four temperature and humidity ranges.

[0143] The results showed that in the medium-temperature and low-humidity, high-temperature and high-humidity, and medium-temperature and medium-humidity zones, the PSO-ELM model significantly outperformed the GS-SVM and GS-RF models in classifying leaf filament drying inlet moisture. In the low-temperature and medium-humidity zone, the PSO-ELM model outperformed the GS-SVM and GS-RF models in accuracy, recall, and F1 score, while the GS-RF model had slightly higher accuracy than the PSO-ELM and GS-RF models.

[0144] The confusion matrix of the PSO-ELM leaf filament drying inlet moisture classification model under different temperature and humidity conditions is as follows: Figures 3-6 As shown in the figure, in the confusion matrix, the diagonal elements represent the number of correct classifications, and the off-diagonal elements represent the number of incorrect classifications. As can be seen from the figure, the PSO-ELM classification model is relatively accurate in classifying the leaf filament drying inlet moisture of each category under different temperature and humidity conditions. The number of misclassified samples for each category is less than 1, and the misclassified leaf filament drying inlet moisture is mainly classified as adjacent quality.

[0145] The PSO-ELM model exhibits good classification performance under different temperature and humidity conditions, making it suitable for classifying inlet moisture in leaf drying under varying temperature and humidity conditions.

[0146] It should be noted that the above description 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 foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for controlling inlet moisture of leaf filament drying under different temperature and humidity conditions based on PSO-ELM, characterized in that: include S1 Environmental Temperature and Humidity Range Division Temperature and humidity ranges are divided according to the annual ambient temperature and humidity of the production area; Based on the monthly average temperature and humidity, cluster analysis was performed on the annual environmental temperature and humidity of the city to divide it into different temperature and humidity ranges. S2 Leaf filament drying inlet moisture quality classification under different temperature and humidity conditions Production data on leaf drying inlet moisture and key parameters were selected based on the months corresponding to different temperature and humidity ranges. Key parameters include the moisture content at the loosening and rehydration inlet, the opening degree of the loosening and rehydration steam valve, the cumulative amount of water dispensed during loosening and rehydration, the moisture content at the loosening and rehydration outlet, the moisture content at the rehydration feeder inlet, the output opening degree of the blade feeding compensation steam valve, the moisture content at the rehydration feeder outlet, the moisture content at the hot air leaf humidifier inlet, the output opening degree of the hot air leaf humidifier compensation steam valve, and the moisture content at the hot air leaf humidifier outlet. The raw production data is preprocessed, and the relevant production data of tobacco raw materials are matched and associated according to the same batch and the same date. The moisture content at the inlet of leaf drying under different temperature and humidity ranges was discretized by statistical analysis, and a discretization standard for the moisture content at the inlet of leaf drying was established. The moisture content at the inlet of leaf drying under different temperature and humidity ranges was classified into categories. The moisture content category of the leaf drying inlet includes poor quality, medium quality - low moisture, high quality, and medium quality - high moisture. Establishment of a classification model for inlet moisture of leaf filaments under different temperature and humidity conditions (S3) S3.1 Data Processing under Different Temperature and Humidity Conditions The association data of different categories of attributes under different temperature and humidity conditions are divided into training and test sets; The target variable, leaf filament drying inlet moisture content, was converted using unique thermal coding, and the input variables were standardized. S3.2 Establishment of a PSO-ELM leaf filament drying inlet moisture classification model under different temperature and humidity conditions The standardized loose rehydration inlet moisture, loose rehydration steam valve opening, loose rehydration water accumulation, loose rehydration outlet moisture, rehydration feeder inlet moisture, blade feeding compensation steam valve output opening, rehydration feeder outlet moisture, hot air leaf humidification inlet moisture, hot air leaf humidification machine compensation steam valve output opening, and hot air leaf humidification outlet moisture were used as the input layer nodes of the PSO-ELM classification model, and the leaf filament drying inlet moisture after unique thermal encoding was used as the output layer node, thus constructing a 3-layer network structure of PSO-ELM classification model with 10 inputs and 4 outputs. Recommended optimal process parameters for S4 Based on the PSO-ELM classification model, the inlet moisture of leaf filament drying is classified under different temperature and humidity conditions; For leaf filament drying inlet moisture that is classified as high quality under different temperature and humidity conditions, the probability that it actually belongs to the high quality category is calculated, and the key process parameters corresponding to the leaf filament drying inlet moisture with the highest probability are selected as the optimal process parameters for filament production. These optimal process parameters can maximize the possibility of the leaf filament drying inlet moisture reaching high quality under the current temperature and humidity conditions.

2. The method for controlling the inlet moisture content of leaf filament drying according to claim 1, characterized in that: The method also includes S5 model self-learning and optimization. The PSO-ELM classification model is continuously self-learned and optimized based on actual production data to improve the model's accuracy and its matching with actual production.

3. The method for controlling the inlet moisture content of leaf filament drying according to claim 1, characterized in that: The classification criteria corresponding to the different leaf filament drying inlet moisture content categories in S2 are as follows: Inferior quality: Other; Medium quality - low moisture content: μ-1.5σ~μ-0.5σ; High quality: μ-0.5σ~μ+0.5σ; Medium quality - slightly high moisture content: μ+0.5σ~μ+1.5σ; Where μ is the average moisture content at the inlet of leaf drying for different temperature and humidity ranges; σ is the standard deviation of the moisture content at the inlet of leaf drying for different temperature and humidity ranges.

4. The method for controlling the inlet moisture content of leaf filament drying according to claim 1, characterized in that: The PSO-ELM leaf filament drying inlet moisture classification model in S3.2 uses Sigmoid and Linear as activation functions for ELM to minimize negative accuracy as the optimal objective. PSO is used to optimize the number of neurons in the ELM algorithm. The population size in PSO is set to 20, the maximum number of iterations is set to 50, and the 5-fold crossover method is used for verification.

5. The method for controlling the inlet moisture content of leaf filament drying according to claim 1, characterized in that: S3 also includes the evaluation of the S3.3 classification model. The predictive performance of the classification model is evaluated using the harmonic mean score of precision, accuracy, recall, and accuracy, and the prediction results of the classification model on different class attributes are visualized using a confusion matrix.

Images