A smart control system and method for controlling moisture and temperature during leaf re-drying and moistening.

Abstract

This invention discloses an intelligent control system and method for moisture and temperature during the first humidification stage of leaf re-drying. The system includes: a main PLC, a data acquisition PLC, a server and a client (including a data acquisition unit and a database). The main PLC control module reads real-time process parameters of the first humidification stage and controls the moisture and temperature at the outlet of the first humidification stage based at least on the predicted results. The data acquisition PLC communicates with the main PLC. The data acquisition unit reads historical and real-time process parameters and writes the predicted values ​​of outlet moisture and temperature. The database stores historical and real-time process parameters of the first humidification stage and the predicted values ​​of outlet moisture and temperature. The client reads real-time process parameters and predicts outlet moisture and temperature using a machine learning model. The intelligent control system and method for moisture and temperature during the first humidification stage of leaf re-drying provided by this invention predicts the moisture and temperature of the tobacco leaves at the outlet in real time through a machine learning-trained model, enabling proactive intervention in the amount of steam and water applied to the humidification cylinder.

Description

Technical Field

[0001] This invention relates to the field of leaf re-drying technology, and in particular to an intelligent control system and method for leaf re-drying, moisture control, and temperature control. Background Technology

[0002] The primary humidification process (referred to as primary humidification) in the tobacco leaf re-drying production line uses a drum-type hot air humidifier to heat and humidify the cut raw tobacco leaves to meet the requirements of subsequent processes. The performance indicators for this process are: outlet moisture content must reach ±0.5% of the set value; outlet temperature must reach ±2.0℃ of the set value. Whether the outlet moisture content and temperature can be stably controlled within the process indicators will seriously affect the intrinsic quality, sensory properties, and smoking experience of the tobacco leaves.

[0003] Currently, the control of moisture and temperature at the humidification drum mostly adopts the traditional result feedback control method. This method uses real-time detection values ​​of the outlet tobacco leaf temperature and moisture to feed back to a PID controller, which then adjusts the amount of steam and water applied to the humidification drum to ensure that the outlet temperature and moisture reach the target values. Result feedback control, which uses real-time detection values ​​of the outlet tobacco leaf moisture and temperature as input, has a significant time lag and is prone to causing periodic oscillations in outlet moisture and temperature within a certain range, resulting in uneven quality of the processed tobacco leaves.

[0004] Therefore, there is an urgent need for an intelligent control system and method for controlling the moisture and temperature of leaf drying, re-drying, and moisturizing. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent control system and method for the moisture and temperature of tobacco leaves during the re-drying and moistening process, in order to solve the problems in the prior art. By training a mathematical model through machine learning, the moisture and temperature of the tobacco leaves at the outlet can be predicted in real time, and the amount of steam and water applied to the moistening cylinder can be intervened in advance.

[0006] This invention provides an intelligent control system for leaf re-drying, moistening, and temperature, comprising:

[0007] The system comprises, in sequence, a main PLC control module, a data acquisition PLC control module, a server, and a client. The server includes, in sequence, a data acquisition unit and a database. The data acquisition unit is connected to the data acquisition PLC control module, and the database is connected to the client.

[0008] The main PLC control module is used to read the real-time process parameters of the first lubricant and control the outlet moisture and / or temperature of the first lubricant at least according to the predicted results of the outlet moisture and / or temperature of the first lubricant.

[0009] The data acquisition PLC control module is used to communicate with the main PLC control module to realize the data exchange of the real-time process parameters of the first runner and to preprocess the data.

[0010] The data acquisition unit is used to read the historical process parameters and real-time process parameters of the first runner in the data acquisition PLC control module, and to write the predicted values ​​of the outlet moisture and / or temperature of the first runner.

[0011] The database is used to store the historical process parameters and real-time process parameters of the first lubricant collected by the data acquisition unit, and to store the predicted values ​​of the outlet moisture and / or temperature of the first lubricant predicted by the client.

[0012] The client is used to read the historical process parameters and real-time process parameters of the first-stage lubricant from the database of the server, and based on the read real-time process parameters of the first-stage lubricant, to predict the outlet moisture and / or temperature of the first-stage lubricant using a machine learning moisture model and / or a machine learning temperature model constructed based on the historical process parameters of the first-stage lubricant, and to store the prediction results of the outlet moisture and / or temperature of the first-stage lubricant in the database.

[0013] In the above-described intelligent control system for the moisture and temperature of the leaf re-drying and drying process, preferably, the main PLC control module uses a PID controller to control the moisture and / or temperature at the drying outlet.

[0014] The PID controller is specifically used to: take the outlet moisture setpoint and the predicted value of the outlet moisture obtained by the machine learning moisture model as input, perform PID control on the outlet moisture of the first humidifier, calculate the corrected water addition amount, and adjust the opening of the water valve according to the calculation result of the corrected water addition amount to adjust the amount of water entering the humidifier tube.

[0015] The PID controller is also used to take the outlet temperature setpoint and the predicted value of the first-stage outlet temperature obtained by the machine learning temperature model as inputs, perform PID control on the first-stage outlet temperature, calculate the corrected steam quantity, and adjust the opening of the steam valve according to the calculation result of the corrected steam quantity to regulate the amount of steam entering the humidifier.

[0016] In the above-described intelligent control system for leaf re-drying and moistening moisture and temperature, preferably, the main PLC control module further includes a data optimization unit, used to smooth the collected real-time process parameters of the moistening process, and / or to adjust the speed of the integral constant of the PID controller, and / or to optimize the calculation results of the corrected water addition, and / or to optimize the calculation results of the corrected steam amount.

[0017] In the intelligent control system for leaf re-drying and moisture and temperature as described above, preferably, the data acquisition PLC control module uses a queue method for data stacking to perform spatiotemporal alignment preprocessing on the data to obtain the historical process parameters of the first drying stage.

[0018] In the intelligent control system for leaf re-drying, moisture and temperature as described above, preferably, the data acquisition unit is further used to clean the acquired historical process parameters and / or real-time process parameters of the first moistening process to obtain valid data, and to smooth the valid data in the historical process parameters and / or real-time process parameters of the first moistening process.

[0019] In the intelligent control system for leaf re-drying, moisture, and temperature as described above, preferably, the database includes a configuration table, a data prediction table, a historical data table, and a real-time data table. Specifically: the configuration table is used to configure PLC variable address settings and communication settings; the data prediction table is used to enable interaction between the client and the data acquisition PLC control module; the historical data table is used to collect historical process parameters for the re-drying process, providing a training dataset for model training of the machine learning moisture model and / or the machine learning temperature model; and the real-time data table is used to collect real-time data of the re-drying process parameters, providing real-time data for model decision-making of the machine learning moisture model and / or the machine learning temperature model.

[0020] The intelligent control system for leaf drying, re-drying, and moisturizing, as described above, preferably includes, in the client, a machine learning moisture model and / or a machine learning temperature model, respectively comprising a feature selection unit, a model selection unit, a model training unit, a model evaluation and optimization unit, and a model decision-making unit, wherein:

[0021] The feature selection unit is used to select the feature variables that rank highest in terms of contribution to outlet moisture and / or temperature from multiple process parameters and equipment parameters of the first-run process. The selected feature variables for the machine learning moisture model are inlet tobacco leaf flow rate, inlet tobacco leaf moisture, pre-injection steam flow rate, heating humid steam flow rate, pre-water supply flow rate, return air temperature, hot air temperature, mixed air temperature, exhaust damper opening, ambient temperature, and ambient humidity. The selected feature variables for the machine learning temperature model are inlet tobacco leaf flow rate, pre-injection steam flow rate, heating humid steam flow rate, pre-water supply flow rate, hot air temperature, mixed air temperature, exhaust damper opening, and ambient temperature.

[0022] The model selection unit is used to select the best model from multiple regression models, including linear regression model, multinomial regression model, ridge regression model, decision tree regression model, neural network model and ensemble learning model. The best model selected is the bagging ensemble algorithm with ridge regression as the basic model.

[0023] The model training unit is used to import the historical process parameters of Yirun into the regression model selected by the model selection unit corresponding to each tobacco leaf type for training according to the tobacco leaf type.

[0024] The model evaluation and optimization unit is used to evaluate the regression models corresponding to each tobacco leaf type according to the prediction effect of each trained regression model on the test dataset, and optimize the model parameters of each regression model by adjusting the evaluation results of different model parameters, and save the optimization results of the regression models corresponding to each tobacco leaf type in the model library.

[0025] The model decision unit is used to make decisions based on the real-time process parameters of the first-run tobacco leaf, using an optimized regression model matched with the tobacco leaf type corresponding to the real-time process parameters of the first-run tobacco leaf, and to obtain the predicted results of the moisture content and / or temperature at the outlet of the first-run tobacco leaf.

[0026] The intelligent control system for leaf threshing, re-drying, and moistening moisture and temperature, as described above, preferably includes a control mode determination unit in the main PLC control module, and a client unit including a current tobacco leaf type determination unit, a result prediction unit, and a prediction timer, wherein:

[0027] The control mode determination unit is used to determine the control mode based on the material presence status and / or production status of the inlet. The control mode includes result feedback mode and feedforward intelligent control mode. If there is no material at the inlet or the production is in the head or tail state, the control mode is determined to be result feedback control mode. If there is material at the inlet and the production is in progress, or if the machine learning moisture model and / or the machine learning temperature model in the client is not operating normally or the network is disconnected, the control mode is determined to be feedforward intelligent control mode.

[0028] The PID controller is specifically used to perform PID analysis on the moisture and temperature at the first humidifier outlet based on the prediction of the first humidifier outlet moisture and temperature and the corresponding set value, so as to adjust the amount of water and / or steam entering the humidifier tube.

[0029] The current tobacco leaf type determination unit is used to query the latest real-time process parameter record from the real-time data table of the database in the feedforward intelligent control mode, and determine the current tobacco leaf type based on the query result.

[0030] The result prediction unit is used to import the last updated machine learning moisture model and / or machine learning temperature model into the model library according to the current tobacco leaf type, and use the imported machine learning moisture model and / or machine learning temperature model to predict the moisture and / or temperature at the outlet of a nutrient solution.

[0031] The prediction timer is used to time the prediction time of the machine learning moisture model and / or the machine learning temperature model to repeat the cyclic control process.

[0032] The model training unit uses an online self-learning method to train the regression model selected by the model selection unit. Specifically, the model training unit is used to: query the corresponding data from the historical data table according to the tobacco leaf type to form a modeling dataset, perform dynamic modeling, and obtain a dynamic model.

[0033] The model evaluation and optimization unit is specifically used to: evaluate the dynamic model, and if the evaluation criteria are met, update the newly created model to the model library;

[0034] The model evaluation and optimization unit is also equipped with a dynamic modeling timer to keep track of the dynamic modeling time and to cycle through the dynamic modeling process.

[0035] In the intelligent control system for leaf re-drying, moistening, and temperature as described above, preferably, the main PLC control module includes a programmable logic controller of model S7-1500, the data acquisition PLC control module includes a programmable logic controller of model S7-400, and the database uses a MySQL database.

[0036] The present invention also provides a method for intelligent control of moisture and temperature during leaf drying and re-drying using the above system, comprising the following steps:

[0037] Read the historical and real-time process parameters of Yirun;

[0038] The moisture content and / or temperature at the outlet of the first-run machine are predicted by using a machine learning moisture model and / or a machine learning temperature model constructed based on the historical process parameters of the first-run machine in real time.

[0039] Based on the real-time process parameters of the first lubricant and the predicted results of the lubricant outlet moisture and / or temperature, the lubricant outlet moisture and / or temperature are controlled.

[0040] The present invention relates to an intelligent control system and method for moisture and temperature control during the first rinsing and re-drying of tobacco leaves. This system uses a machine learning-trained mathematical model to predict the moisture and temperature of the tobacco leaves at the outlet in real time. It can proactively intervene in the amount of steam and water applied to the rinsing cylinder, completely solving the lag and instability of existing control methods. This ensures that the outlet moisture and temperature are strictly controlled within the process parameters, improving product processing quality. The use of real-time dynamic modeling eliminates differences in control effects caused by variations in tobacco leaf quality and environment. Automatic control of the first rinsing moisture and temperature requires no human intervention, reducing the workload of operators and avoiding the impact of individual experience differences on quality. Attached Figure Description

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described below with reference to the accompanying drawings, wherein:

[0042] Figure 1 This is a schematic diagram of the current control strategy for moisture and temperature at the outlet of a lubricant.

[0043] Figure 2 A schematic diagram illustrating the principle behind the current control strategy for outlet moisture and temperature that causes a lag in outlet moisture and temperature control.

[0044] Figure 3 This is a structural block diagram of an embodiment of the intelligent control system for leaf drying, re-drying, and moistening moisture and temperature provided by the present invention.

[0045] Figure 4 This is a schematic diagram of the main PLC control module of the present invention controlling the moisture at the outlet of a lubricant.

[0046] Figure 5 This is a schematic diagram of the main PLC control module of the present invention controlling the outlet temperature of a lubricant.

[0047] Figure 6 A block diagram showing the structural examples of the machine learning moisture model and the machine learning temperature model in the client.

[0048] Figure 7 A schematic diagram of the online self-learning modeling and prediction process;

[0049] Figure 8 This is a schematic diagram showing the trends of moisture and temperature at the outlet of a lubricant obtained using traditional control methods.

[0050] Figure 9 This is a schematic diagram showing the trends of moisture content and temperature at the outlet of a lubricant obtained using the intelligent control method of the present invention.

[0051] Figure 10 This is a flowchart illustrating an embodiment of the intelligent control method for moisture and temperature during leaf re-drying and moistening provided by the present invention. Detailed Implementation

[0052] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. The descriptions of the exemplary embodiments are merely illustrative and are in no way intended to limit the present disclosure or its application or use. The present disclosure may be implemented in many different forms and is not limited to the embodiments described herein. These embodiments are provided so that the present disclosure will be thorough and complete, and will fully express the scope of the disclosure to those skilled in the art. It should be noted that, unless specifically stated otherwise, the relative arrangement of components and steps, the composition of materials, numerical expressions, and values ​​set forth in these embodiments should be interpreted as exemplary only and not as limiting.

[0053] The terms “first,” “second,” and similar terms used in this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different parts. Terms such as “including” or “contains” mean that the element preceding the term encompasses the element listed after it, and do not exclude the possibility of encompassing other elements as well. Terms such as “above” and “below” are used only to indicate relative positional relationships; when the absolute position of the described object changes, this relative positional relationship may also change accordingly.

[0054] In this disclosure, when a specific component is described as being located between a first component and a second component, an intermediary component may or may not be present between the specific component and the first or second component. When a specific component is described as connecting to other components, the specific component may be directly connected to the other components without having an intermediary component, or it may not be directly connected to the other components but may have an intermediary component.

[0055] All terms used in this disclosure (including technical or scientific terms) have the same meaning as understood by one of ordinary skill in the art to which this disclosure pertains, unless otherwise specifically defined. It should also be understood that terms defined in a general dictionary, such as a dictionary, should be interpreted as having a meaning consistent with their meaning in the context of the relevant art, and not as having an idealized or highly formalized meaning, unless expressly defined herein.

[0056] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, they should be considered part of the specification.

[0057] Currently, the control of moisture and temperature at the outlet of the primary heat exchanger adopts a result feedback control method. The control strategy for moisture at the outlet of the primary heat exchanger is as follows: Figure 1 As shown, the outlet moisture control involves feeding back the real-time detected value of the outlet moisture to the PID controller, which then adjusts the amount of water entering the humidifier cylinder to ensure the outlet moisture reaches the process target value. Temperature control works similarly, with the controlled object being the steam flow rate and the temperature measuring instrument being a near-infrared thermometer.

[0058] Current technology uses real-time values ​​of exported tobacco leaves as control input, which has a significant lag and easily causes periodic oscillations in exported moisture and temperature within a certain range. Figure 2 As shown, after the tobacco leaves enter the drum, they take time T1 to reach the drum outlet, and then another time T2 to reach the outlet moisture and temperature detection points. The water and steam application points are located at the drum inlet. When the outlet moisture or temperature deviates from the process target value, the water and steam flow rates at the drum inlet will change. The affected tobacco leaves will then take time T1+T2 to reach the moisture and temperature detection points, resulting in a control lag of T1+T2. During this period, the PLC system will assume that the moisture or temperature is still deviating from the target value and will continuously change the inlet water and steam flow rates, inevitably causing overshoot in the control system. This will lead to periodic oscillations in the outlet moisture and temperature within a certain range, making it difficult to reach a steady state.

[0059] Furthermore, changes in water volume affect not only the outlet moisture content but also the outlet temperature; similarly, changes in steam volume affect not only the outlet temperature but also the outlet moisture content. The moisture control program cannot sense the impact of steam volume changes on moisture, and similarly, the temperature control program cannot sense the impact of water volume changes on temperature. Moisture control and temperature control are interdependent, which can easily lead to instability in both control processes.

[0060] like Figure 3 As shown, the intelligent control system for leaf re-drying, moistening, and temperature provided in this embodiment includes: a main PLC control module 1, a data acquisition PLC control module 2, a server 3, and a client 4 arranged sequentially. The server 3 includes a data acquisition unit 31 and a database 32 arranged sequentially. The data acquisition unit 31 is connected to the data acquisition PLC control module 2, and the database 32 is connected to the client 4.

[0061] The main PLC control module 1 is used to read the real-time process parameters of the first lubricant and control the outlet moisture and / or temperature of the first lubricant at least according to the predicted results of the outlet moisture and / or temperature of the first lubricant.

[0062] The data acquisition PLC control module 2 is used to communicate with the main PLC control module 1 to realize the data exchange of the real-time process parameters of the first runner and to preprocess the data.

[0063] The data acquisition unit 31 is used to read the historical process parameters and real-time process parameters of the first runner in the data acquisition PLC control module 2, and to write the predicted values ​​of the outlet moisture and / or temperature of the first runner.

[0064] The database 32 is used to store the historical process parameters and real-time process parameters of the first lubricant collected by the data acquisition unit 31, and to store the predicted values ​​of the outlet moisture and / or temperature of the first lubricant predicted by the client 4.

[0065] The client 4 is used to read the historical process parameters and real-time process parameters of the first-stage lubricant from the database 32 of the server 3, and based on the read real-time process parameters of the first-stage lubricant, to predict the outlet moisture and / or temperature of the first-stage lubricant through a machine learning moisture model and / or a machine learning temperature model constructed based on the historical process parameters of the first-stage lubricant, and to store the prediction results of the outlet moisture and / or temperature of the first-stage lubricant in the database 32.

[0066] The main PLC control module 1 is a programmable logic controller (PLC) of model S7-1500, the data acquisition PLC control module 2 is a programmable logic controller (PLC) of model S7-400, and the database 32 uses a MySQL database.

[0067] Furthermore, the main PLC control module 1 uses a PID controller to control the moisture content and / or temperature at the outlet of the lubricant. Figure 4 and Figure 5 The diagrams show the principle of the main PLC control module 1 controlling the moisture content and temperature at the outlet of the first humidifier.

[0068] like Figure 4 As shown, the PID controller is specifically used to: take the outlet moisture setpoint and the predicted value of the outlet moisture obtained by the machine learning moisture model as input, perform PID control on the outlet moisture of the first humidifier, calculate the corrected water addition amount, and adjust the opening of the water valve according to the calculation result of the corrected water addition amount to adjust the amount of water entering the humidifier tube.

[0069] like Figure 5 As shown, the PID controller is also used to take the outlet temperature setpoint and the predicted value of the first-stage outlet temperature obtained by the machine learning temperature model as inputs, perform PID control on the first-stage outlet temperature, calculate the corrected steam quantity, and adjust the opening of the steam valve according to the calculation result of the corrected steam quantity to adjust the amount of steam entering the humidifier.

[0070] As can be seen, in the PID control process of moisture and temperature, the main PLC control module 1 uses the outlet moisture and temperature values ​​predicted by the moisture model and temperature model based on machine learning as inputs to achieve advanced control of water and steam volume.

[0071] Furthermore, the main PLC control module 1 also includes a data optimization unit, used to smooth the collected real-time process parameters of the lubrication system, and / or to change the speed of the integral constant of the PID controller, and / or to optimize the calculation results of the corrected water addition, and / or to optimize the calculation results of the corrected steam addition.

[0072] The data acquisition PLC control module 2 enables data exchange and preprocessing of process parameters. Furthermore, the data acquisition PLC control module 2 utilizes a queue method for data stacking to perform spatiotemporal alignment preprocessing of the data, thereby obtaining the historical process parameters. Since the selected characteristic variables (described later) do not correspond in real-time with the outlet moisture and temperature, there is a temporal correspondence between them: for example, if the inlet tobacco flow rate changes, the tobacco leaves with the changed flow rate will take some time to reach the water inlet, and only then will the water volume at the water inlet change. Simultaneously, the tobacco leaves with the applied water volume change will take another 2-3 minutes to reach the outlet, ultimately resulting in a specific moisture value. This invention analyzes the historical trend charts of each characteristic variable and control quantity in WINCC and summarizes their temporal correspondence through experiments. When acquiring data, the data acquisition PLC control module 2 uses a queue program for data stacking to achieve data alignment.

[0073] Furthermore, the data acquisition unit 31 is also used to perform data cleaning on the acquired historical process parameters and / or real-time process parameters to obtain valid data, and to smooth the valid data (partial or all) in the historical process parameters and / or real-time process parameters to reduce noise interference with machine learning modeling. The data cleaning process may, for example, involve removing zero values, negative values, null values, outliers, etc., from the acquired data to obtain valid data. In some embodiments of the present invention, a threshold is applied to each feature during the data acquisition process to avoid the collection of a large amount of invalid data and reduce the amount of subsequent data cleaning work.

[0074] Furthermore, the database 32 includes a configuration table, a data prediction table, a historical data table, and a real-time data table. The configuration table is used to configure PLC variable address settings and communication settings. The data prediction table is used to enable interaction between the client and the data acquisition PLC control module 2. The historical data table is used to collect historical process parameters of the first-stage process to provide a training dataset for model training of the machine learning moisture model and / or the machine learning temperature model. The real-time data table is used to collect real-time data of the first-stage process parameters to provide real-time data for model decision-making of the machine learning moisture model and / or the machine learning temperature model.

[0075] like Figure 6 As shown, the machine learning water model and / or the machine learning temperature model in client 4 respectively include a feature selection unit, a model selection unit, a model training unit, a model evaluation and optimization unit, and a model decision unit. In specific implementation, this invention uses Python language for modeling and prediction, wherein:

[0076] The feature selection unit is used to select the feature variables that rank highest in terms of contribution to outlet moisture and / or temperature from multiple process parameters and equipment parameters of the first-run process. The selected feature variables for the machine learning moisture model are inlet tobacco leaf flow rate, inlet tobacco leaf moisture, pre-injection steam flow rate, heating humid steam flow rate, pre-water supply flow rate, return air temperature, hot air temperature, mixed air temperature, exhaust damper opening, ambient temperature, and ambient humidity. The selected feature variables for the machine learning temperature model are inlet tobacco leaf flow rate, pre-injection steam flow rate, heating humid steam flow rate, pre-water supply flow rate, hot air temperature, mixed air temperature, exhaust damper opening, and ambient temperature.

[0077] The model selection unit is used to select the best model from multiple regression models, including linear regression, multinomial regression, ridge regression, decision tree regression, neural network, and ensemble learning models. The best model selected is the bagging ensemble algorithm with ridge regression as the basic model. Considering that the outlet moisture and temperature of the first run are continuous values, a regression model is required. The model selection unit of this invention uses multiple methods such as linear regression, multinomial regression, ridge regression, decision tree regression, neural network, and ensemble learning to find the most suitable model for machine learning in this process.

[0078] The model training unit is used to import the historical process parameters of the first-run process into the regression model selected by the model selection unit corresponding to each tobacco leaf type for training according to the tobacco leaf type. In specific implementation, the model training unit imports the collected and processed effective dataset into the selected regression model for training.

[0079] The model evaluation and optimization unit is used to evaluate the regression models corresponding to each tobacco leaf type according to the prediction effect of each trained regression model on the test dataset, and optimize the model parameters of each regression model by adjusting the evaluation results of different model parameters, and save the optimization results of the regression models corresponding to each tobacco leaf type in the model library.

[0080] The model decision unit is used to make decisions based on the real-time process parameters of the first-run tobacco leaf, using an optimized regression model matched with the tobacco leaf type corresponding to the real-time process parameters of the first-run tobacco leaf, and to obtain the predicted results of the moisture content and / or temperature at the outlet of the first-run tobacco leaf.

[0081] Furthermore, the machine learning moisture model and / or the machine learning temperature model also include a recursive feature elimination unit for determining feature variables using the recursive feature elimination method (RFE). In a specific implementation, the feature selection unit performs preliminary manual screening and automatic screening to identify feature variables with greater contributions. The manual screening initially identifies 19 feature variables based on operational experience and correlation analysis of the data using statistical tools such as SPSS or Python. Based on the manual screening, the recursive feature elimination unit uses the recursive feature elimination method (RFE) in the Wrapper tool to determine the feature variables. The final determined feature variables for the moisture model are: inlet tobacco leaf flow rate, inlet tobacco leaf moisture content, pre-injection steam flow rate, heating humidified steam flow rate, pre-injection water flow rate, return air temperature, hot air temperature, mixed air temperature, exhaust damper opening, ambient temperature, and ambient humidity. The feature variables for the temperature model are: inlet tobacco leaf flow rate, pre-injection steam flow rate, heating humidified steam flow rate, pre-injection water flow rate, hot air temperature, mixed air temperature, exhaust damper opening, and ambient temperature.

[0082] The main PLC control module 1 further includes a control mode determination unit, and the client 4 includes a current tobacco leaf type determination unit, a result prediction unit, and a prediction timer, wherein:

[0083] The control mode determination unit is used to determine the control mode based on the presence status of the material at the inlet and / or the production status. The control mode includes a result feedback mode and a feedforward intelligent control mode. If there is no material at the inlet or the production is in the head or tail state, the control mode is determined to be the result feedback control mode. If there is material at the inlet and the production is in progress, or if the machine learning moisture model and / or the machine learning temperature model in the client 4 are not operating normally or the network is disconnected, the control mode is determined to be the feedforward intelligent control mode.

[0084] The PID controller is specifically used to perform PID analysis on the moisture and / or temperature at the humidifier outlet based on the prediction of the result prediction unit and the corresponding set value, so as to adjust the amount of water and / or steam entering the humidifier tube.

[0085] The current tobacco leaf type determination unit is used to query the latest real-time process parameter record from the real-time data table of the database 32 in the feedforward intelligent control mode, and determine the current tobacco leaf type based on the query result.

[0086] The result prediction unit is used to import the last updated machine learning moisture model and / or machine learning temperature model into the model library according to the current tobacco leaf type, and use the imported machine learning moisture model and / or machine learning temperature model to predict the moisture and / or temperature at the outlet of a nutrient solution.

[0087] The prediction timer is used to time the prediction time of the machine learning moisture model and / or the machine learning temperature model to repeat the cyclic control process.

[0088] Considering the wide variety of grades, varieties, and origins of raw tobacco leaves, and the differences in their hygroscopicity and heat absorption to water and steam, the model training unit employs an online self-learning method to train the regression model selected by the model selection unit. The specific online self-learning modeling and prediction process is as follows: Figure 7 As shown. The model training unit is specifically used to: query relevant data from the historical data table according to the tobacco leaf type to form a modeling dataset, perform dynamic modeling, and obtain a dynamic model;

[0089] The model evaluation and optimization unit is specifically used to: evaluate the dynamic model, and if the evaluation criteria are met, update the newly created model to the model library;

[0090] The model evaluation and optimization unit is also equipped with a dynamic modeling timer to keep track of the dynamic modeling time and to cycle through the dynamic modeling process.

[0091] This invention employs real-time dynamic modeling, eliminating differences in control effects caused by variations in tobacco leaf quality and environmental conditions. For example... Figure 7 As shown, when in feedforward intelligent control mode, the presence status of material at the inlet and the production status are judged. If there is no material at the inlet or the production is in the head or tail state, the prediction stops and the system adopts the result feedback control method. If there is material at the inlet and the production is in the state, the machine learning moisture model and machine learning temperature model in client 4 start running and execute the following steps in sequence.

[0092] (1) Query the latest real-time process parameter record from the real-time data table to determine the current tobacco leaf type;

[0093] (2) Based on the tobacco leaf type, import the model from the last update into the model library to predict moisture and temperature, and start the prediction timer.

[0094] (3) At the same time, based on the tobacco leaf type, relevant data is queried from the historical data table to form a modeling dataset, dynamic modeling is performed, and the model is evaluated. If the evaluation criteria are met, the newly created model is updated to the model library, and the dynamic modeling timer starts counting down.

[0095] (4) The predicted values ​​of moisture and temperature are transmitted to the main PLC control module 1 and the water volume and steam volume are adjusted accordingly by the PID controller.

[0096] (5) When the prediction timer reaches the set time, repeat the above processes (1), (2), and (4) to achieve cyclic control. When the dynamic modeling timer reaches the set time, repeat the above process (3) to achieve cyclic dynamic modeling.

[0097] If the machine learning moisture model and machine learning temperature model malfunction or the network is disconnected during operation, the main PLC control module 1 will automatically switch to the result feedback control mode.

[0098] This invention utilizes machine learning models for moisture and temperature to predict the outlet moisture and temperature of a primary outlet in real time. Based on this prediction, it controls the water and steam flow rates, completely resolving the lag inherent in existing traditional control methods and achieving uniform and stable control of outlet indicators. Specifically, the standard deviation of outlet moisture was reduced by 45.6%, and the coefficient of variation by 47.0%; the standard deviation of outlet temperature was reduced by 66.0%, and the coefficient of variation by 68.6%. Specific improvements are shown in Table 1.

[0099] Table 1. Improvement effect of the intelligent control method of the present invention compared with the prior art.

[0100]

[0101]

[0102] Figure 8 and Figure 9 The historical trends of outlet temperature and outlet moisture content over a certain period of time are compared between the traditional control method and the intelligent control method of the present invention. Compared with the traditional control method, the intelligent control method of the present invention reduces the fluctuation range of moisture content from ±1.5% to ±0.5% of the set value and the fluctuation range of temperature from ±2.5℃ to ±1℃, fully demonstrating the superiority of the intelligent control method of the present invention.

[0103] The intelligent control system for moisture and temperature during the first wetting and re-drying stage provided in this invention uses machine learning to train a mathematical model to predict the moisture and temperature of the tobacco leaves at the outlet in real time. It can proactively intervene in the amount of steam and water applied to the wetting cylinder, completely solving the lag and instability of existing control methods. This ensures that the outlet moisture and temperature are strictly controlled within the process parameters, improving the processing quality of the product. By adopting real-time dynamic modeling, it eliminates the differences in control effect caused by different tobacco leaf quality and environmental conditions. The automatic control of the moisture and temperature during the first wetting stage requires no human intervention, which reduces the labor intensity of operators and avoids the impact of individual experience differences on quality.

[0104] Correspondingly, such as Figure 10As shown, the present invention also provides an intelligent control method for the moisture and temperature of leaf removal, re-drying, and moistening using the above-described system. In actual implementation, the intelligent control method for the moisture and temperature of leaf removal, re-drying, and moistening provided in this embodiment specifically includes:

[0105] Step S1: Obtain the historical process parameters and real-time process parameters of Yirun.

[0106] Step S2: Predict the outlet moisture and / or temperature of the first lubricant using the real-time process parameters of the first lubricant and the machine learning moisture model and / or machine learning temperature model constructed based on the historical process parameters of the first lubricant.

[0107] Step S3: Control the moisture and / or temperature at the outlet of the lubricant based on the real-time process parameters of the lubricant and the predicted results of the lubricant outlet moisture and / or temperature.

[0108] The intelligent control method for moisture and temperature during the first wetting and re-drying stage provided in this invention uses a mathematical model trained by machine learning to predict the moisture and temperature of the tobacco leaves at the outlet in real time. This allows for proactive intervention in the amount of steam and water applied to the wetting cylinder, completely resolving the lag and instability of existing control methods. It ensures that the outlet moisture and temperature are strictly controlled within the process parameters, improving product processing quality. The real-time dynamic modeling eliminates differences in control effects caused by variations in tobacco leaf quality and environment. Automatic control of the first wetting moisture and temperature requires no human intervention, reducing the workload of operators and avoiding the impact of individual experience differences on quality.

[0109] The embodiments of this disclosure have now been described in detail. To avoid obscuring the concept of this disclosure, some details known in the art have not been described. Those skilled in the art can fully understand how to implement the technical solutions disclosed herein based on the above description.

[0110] While specific embodiments of this disclosure have been described in detail by way of examples, those skilled in the art should understand that the examples are for illustrative purposes only and not intended to limit the scope of this disclosure. Those skilled in the art should understand that modifications can be made to the above embodiments or equivalent substitutions can be made to some technical features without departing from the scope and spirit of this disclosure. The scope of this disclosure is defined by the appended claims.

Claims

1. A smart control system for the moisture and temperature of leaves after drying and re-drying, characterized in that, include: The system comprises, in sequence, a main PLC control module, a data acquisition PLC control module, a server, and a client. The server includes, in sequence, a data acquisition unit and a database. The data acquisition unit is connected to the data acquisition PLC control module, and the database is connected to the client. The main PLC control module is used to read the real-time process parameters of the first lubricant and control the outlet moisture and / or temperature of the first lubricant at least according to the predicted results of the outlet moisture and / or temperature of the first lubricant. The main PLC control module uses a PID controller to control the outlet moisture and / or temperature of the first lubricant. The data acquisition PLC control module is used to communicate with the main PLC control module to realize the data exchange of the real-time process parameters of the first runner and to preprocess the data. The data acquisition unit is used to read the historical process parameters and real-time process parameters of the first runner in the data acquisition PLC control module, and to write the predicted values ​​of the outlet moisture and / or temperature of the first runner. The database is used to store the historical and real-time process parameters of the first-stage lubricant collected by the data acquisition unit, and to store the predicted values ​​of the outlet moisture and / or temperature of the first-stage lubricant predicted by the client. The database includes a configuration table, a data prediction table, a historical data table, and a real-time data table. Specifically: the configuration table is used to configure PLC variable address settings and communication settings; the data prediction table is used to enable interaction between the client and the data acquisition PLC control module; the historical data table is used to collect the historical process parameters of the first-stage lubricant to provide training datasets for model training of the machine learning moisture model and / or the machine learning temperature model; and the real-time data table is used to collect real-time data of the real-time process parameters of the first-stage lubricant to provide real-time data for model decision-making of the machine learning moisture model and / or the machine learning temperature model. The client is used to read the historical process parameters and real-time process parameters of the first-stage lubricant from the database of the server, and based on the read real-time process parameters, to predict the outlet moisture and / or temperature of the first-stage lubricant using a machine learning moisture model and / or a machine learning temperature model constructed based on the historical process parameters, and to store the prediction results of the outlet moisture and / or temperature of the first-stage lubricant in the database. The machine learning water model and / or the machine learning temperature model in the client respectively include a feature selection unit, a model selection unit, a model training unit, a model evaluation and optimization unit, and a model decision unit, wherein: The feature selection unit is used to select the feature variables that rank highest in terms of contribution to outlet moisture and / or temperature from multiple process parameters and equipment parameters of the first-run process. The selected feature variables for the machine learning moisture model are inlet tobacco leaf flow rate, inlet tobacco leaf moisture, pre-injection steam flow rate, heating humid steam flow rate, pre-water supply flow rate, return air temperature, hot air temperature, mixed air temperature, exhaust damper opening, ambient temperature, and ambient humidity. The selected feature variables for the machine learning temperature model are inlet tobacco leaf flow rate, pre-injection steam flow rate, heating humid steam flow rate, pre-water supply flow rate, hot air temperature, mixed air temperature, exhaust damper opening, and ambient temperature. The model selection unit is used to select the best model from multiple regression models, including linear regression model, multinomial regression model, ridge regression model, decision tree regression model, neural network model and ensemble learning model. The best model selected is the bagging ensemble algorithm with ridge regression as the basic model. The model training unit is used to import the historical process parameters of Yirun into the regression model selected by the model selection unit corresponding to each tobacco leaf type for training according to the tobacco leaf type. The model evaluation and optimization unit is used to evaluate the regression models corresponding to each tobacco leaf type according to the prediction effect of each trained regression model on the test dataset, and optimize the model parameters of each regression model by adjusting the evaluation results of different model parameters, and save the optimization results of the regression models corresponding to each tobacco leaf type in the model library. The model decision unit is used to make decisions based on the real-time process parameters of the first-stage nutrient solution, using an optimized regression model matched with the tobacco leaf type corresponding to the real-time process parameters of the first-stage nutrient solution, to obtain the predicted results of the moisture content and / or temperature at the outlet of the first-stage nutrient solution. The main PLC control module further includes a control mode determination unit, and the client includes a current tobacco leaf type determination unit, a result prediction unit, and a prediction timer, wherein: The control mode determination unit is used to determine the control mode based on the material presence status and / or production status of the inlet. The control mode includes result feedback mode and feedforward intelligent control mode. If there is no material at the inlet or the production is in the head or tail state, the control mode is determined to be result feedback control mode. If there is material at the inlet and the production is in progress, or if the machine learning moisture model and / or the machine learning temperature model in the client is not operating normally or the network is disconnected, the control mode is determined to be feedforward intelligent control mode. The PID controller is specifically used to perform PID analysis on the moisture and / or temperature at the humidifier outlet based on the prediction of the humidifier outlet moisture and / or temperature by the result prediction unit and the corresponding set value, so as to adjust the amount of water and / or steam entering the humidifier tube. The current tobacco leaf type determination unit is used to query the latest real-time process parameter record from the real-time data table of the database in the feedforward intelligent control mode, and determine the current tobacco leaf type based on the query result. The result prediction unit is used to import the last updated machine learning moisture model and / or machine learning temperature model into the model library according to the current tobacco leaf type, and use the imported machine learning moisture model and / or machine learning temperature model to predict the moisture and / or temperature at the outlet of a nutrient solution. The prediction timer is used to time the prediction time of the machine learning moisture model and / or the machine learning temperature model to repeat the cyclic control process. The model training unit uses an online self-learning method to train the regression model selected by the model selection unit. Specifically, the model training unit is used to: query the corresponding data from the historical data table according to the tobacco leaf type to form a modeling dataset, perform dynamic modeling, and obtain a dynamic model. The model evaluation and optimization unit is specifically used to: evaluate the dynamic model, and if the evaluation criteria are met, update the newly created model to the model library; The model evaluation and optimization unit is also equipped with a dynamic modeling timer to keep track of the dynamic modeling time and to cycle through the dynamic modeling process.

2. The intelligent control system for leaf re-drying and moistening moisture and temperature according to claim 1, characterized in that, The PID controller is specifically used to: take the outlet moisture setpoint and the predicted value of the outlet moisture obtained by the machine learning moisture model as input, perform PID control on the outlet moisture of the first humidifier, calculate the corrected water addition amount, and adjust the opening of the water valve according to the calculation result of the corrected water addition amount to adjust the amount of water entering the humidifier tube. The PID controller is also used to take the outlet temperature setpoint and the predicted value of the first-stage outlet temperature obtained by the machine learning temperature model as inputs, perform PID control on the first-stage outlet temperature, calculate the corrected steam quantity, and adjust the opening of the steam valve according to the calculation result of the corrected steam quantity to regulate the amount of steam entering the humidifier.

3. The intelligent control system for leaf re-drying and moistening moisture and temperature according to claim 2, characterized in that, The main PLC control module also includes a data optimization unit, used to smooth the collected real-time process parameters of the first lubrication system, and / or to change the speed of the integral constant of the PID controller, and / or to optimize the calculation results of the corrected water addition, and / or to optimize the calculation results of the corrected steam addition.

4. The intelligent control system for leaf re-drying and moistening moisture and temperature according to claim 1, characterized in that, The data acquisition PLC control module uses a queue method to perform data stacking, and performs spatiotemporal alignment preprocessing on the data to obtain the historical process parameters of the first run.

5. The intelligent control system for leaf re-drying and moistening moisture and temperature according to claim 1, characterized in that, The data acquisition unit is also used to clean the acquired historical process parameters and / or real-time process parameters of the first-run process to obtain valid data, and to smooth the valid data in the historical process parameters and / or real-time process parameters of the first-run process.

6. The intelligent control system for leaf re-drying and moistening moisture and temperature according to claim 1, characterized in that, The main PLC control module includes a programmable logic controller (PLC) of model S7-1500, the data acquisition PLC control module includes a programmable logic controller (PLC) of model S7-400, and the database uses a MySQL database.

7. A method for intelligent control of leaf drying, re-drying, and temperature control using the system described in any one of claims 1-6, characterized in that, Includes the following steps: Obtain historical and real-time process parameters of Yirun; The outlet moisture and / or temperature of the first-run water source are predicted using the real-time process parameters of the first-run water source and the machine learning moisture model and / or machine learning temperature model constructed based on the historical process parameters of the first-run water source. Based on the real-time process parameters of the first lubricant and the predicted results of the lubricant outlet moisture and / or temperature, the lubricant outlet moisture and / or temperature are controlled.

Images