Method for automatically controlling moisture in cold room of tobacco flake redrying machine based on intelligent algorithm
Through intelligent algorithms and neural network technology, the automatic control of the cold room moisture of the smoke refrigerator is realized, which solves the problem of quality instability caused by traditional manual operations, and improves the performance and quality consistency of the refrigerator.
Patent Information
- Application Number
- CN202411375626.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-29
- Publication Date
- 2024-11-26
AI Technical Summary
The moisture control in the cooling area of the traditional tobacco refrigerator relies on manual operation, resulting in unstable quality of the refrigerated roasting and difficult to meet the quality requirements of the cigarette industry.
Automatic control method of cold room moisture of smoke refrigerator using intelligent algorithms, collects parameters that affect moisture through the PLC system, uses OPC server to realize data transmission and storage, uses neural network algorithm to perform data training and prediction, and adjusts the drying zone temperature in real time to achieve Automatic moisture control.
The manual operation intensity is reduced, the performance indicators of the tobacco re-grilling machine are improved, and the stability and consistency of the re-grilling quality is ensured.
Smart Images

Figure CN119014565A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of tobacco leaf threshing and redrying line equipment in the tobacco industry, and in particular to a method for automatically controlling moisture in a cold room of a tobacco leaf redrying machine. Background Art
[0002] The tobacco leaf redrying machine is a key link in the leaf threshing and redrying production line. Its control level is directly related to the temperature and moisture content of the tobacco leaves, and has a decisive influence on the quality of the subsequent finished cigarette packages.
[0003] From the perspective of control, the tobacco slice re-drying machine has the characteristics of complex models, many variables, and great influence from the previous processing links. Because the traditional control method is difficult to implement on this equipment, the previous tobacco slice re-drying machine is basically operated in real time and manually controlled by operators based on their personal experience. The differences in experience and sense of responsibility of operators in different shifts will cause the instability of the leaf re-drying quality, resulting in the re-drying indicators unable to meet the subsequent process requirements of the cigarette industry enterprises, affecting the processing market and brand building of the re-drying enterprises. It is urgent to solve the system and method of realizing the automatic control of the moisture in the cooling zone of the tobacco slice re-drying machine and improving the performance indicators of the tobacco slice re-drying machine. Summary of the invention
[0004] The main technical problem solved by the present invention is to provide a method for automatically setting various parameters in the drying zone of a slice baking machine and automatically controlling the moisture in the cooling zone of the slice baking machine.
[0005] The technical solution of the present invention is: an automatic control method for moisture in the cold room of a tobacco sheet redrying machine based on an intelligent algorithm, including three steps: data collection, data transmission, and moisture control. Data collection is completed by the underlying control system PLC. The collected data are parameters that affect the moisture in the cold room of the tobacco sheet redrying machine. The data transmission between the underlying PLC control system and the control algorithm is realized through the OPC server. The collected data is stored in the database. After a certain amount of data is collected, data training and data prediction are performed through the control algorithm. The set temperature value of the drying area obtained by the control algorithm is transmitted to the PLC control system, and then the temperature of each drying area is controlled in real time through the tobacco sheet redrying machine control system to realize automatic control of moisture in the cold room of the tobacco sheet redrying machine. In this method, the moisture in the cold room refers to the moisture content of tobacco leaves in the cold room.
[0006] The data collection is mainly completed by the underlying control system PLC. The main content of the data collection is to collect various parameters that affect the moisture in the cold room of the tobacco slice redrying machine, specifically including: the actual temperature of each zone in the drying area of the tobacco slice redrying machine, the fan frequency of the circulation fan in each zone in the drying area, the fan frequency of the dehumidification fan in each zone in the drying area, the opening of the cold air inlet damper in each zone in the drying area, the opening of the dehumidification damper in each zone in the drying area, the dehumidification degree of each zone in the drying area, as well as the main transmission speed of the tobacco slice redrying machine, inlet moisture, inlet temperature, inlet flow, ambient temperature, ambient humidity, ambient air pressure, and actual value of moisture in the cold room.
[0007] The parameter data that affect the moisture content in the cold room of the tobacco redrying machine are collected using the stacking method. When the tobacco leaves of the same cross section of the drying machine arrive at the data acquisition instrument at the entrance of the drying machine, the PLC system collects the actual parameter data of the tobacco leaves of this section. As the equipment runs, the PLC system continues to collect the actual parameter data of the tobacco leaves of this section. The PLC program performs the stacking process of this parameter inside until the tobacco leaves of this cross section walk under the data acquisition instrument in the cold room, completing the data collection and stacking process. At this time, the PLC system uses the out-of-stack data as the data to be collected; among them, the main drive speed, ambient temperature, ambient humidity, ambient air pressure, and actual values of moisture content in the cold room do not need to be stacked, and these data are used as historical data for algorithm training.
[0008] Since the tobacco slice re-drying machine is relatively long, it takes about 3 to 4 minutes for the tobacco slices to go from the entrance to the cold room, so the data cannot be collected at the same time, and it is necessary to use the stacking method to collect the data and smooth the collected data. Take the collection of the moisture at the entrance of the tobacco slice re-drying machine as an example. When the tobacco leaves of the same cross section of the tobacco slice re-drying machine reach the moisture meter at the entrance of the slice drying machine, the PLC system collects the actual moisture of the tobacco leaves of this section. As the equipment runs, the PLC program performs the stacking process of this moisture inside, until about 3 minutes later (this time is determined by the specific characteristics of the equipment), the tobacco leaves of this cross section go to the bottom of the moisture meter in the cold room and the stacking is completed. At this time, the PLC system uses this out-of-stack data as the data to be collected, and the other data are the same. The specific stacking time is determined by the characteristics of the equipment. The basic principle is that the data collected by the sensor is stacked until it reaches the bottom of the moisture meter in the cold room. All parameters of tobacco leaves with the same cross section of the tobacco slice re-drying machine arrive at the bottom of the moisture meter in the cold room and then are stacked. The PLC system records the out-of-stack data at this time.
[0009] The data transmission is realized through the OPC server between the underlying PLC control system and the control algorithm, including the data transmission of the PLC control system to the control algorithm, and the parameter settings of the drying area of the tobacco sheet redrying machine obtained by the control algorithm are transmitted to the PLC control system. The collected data is stored in the database through data transmission, and data training is performed after a certain amount of data is collected. The specific amount of data collection is determined according to actual needs to meet the needs of automatic control accuracy.
[0010] The control algorithm process includes: data normalization, initialization of neural network parameters, historical data training, forward learning, and reverse updating of the weight value w of each node. ij and deviation value θ j , calculate the deviation △ between the predicted value and the actual value of the cold room moisture in historical data, calculate the set temperature value of each zone in the drying area, and transmit the data obtained by the control algorithm to the PLC system. The specific process is as follows:
[0011] Step 1: Data normalization
[0012] For the parameters in the collected database that affect the moisture in the cold room of the slice baking machine, normalize the data to the range of 0 to 1, with the maximum value less than 1 and not equal to 1, and the minimum value greater than 0 and not equal to 1;
[0013] Step 2: Historical data training
[0014] The parameter data that affects the moisture in the cold room of the tobacco redrying machine collected in the database is used as the input data x i (i is the number of variables corresponding to each data), including the actual temperature of each zone in the drying area, the fan frequency of the circulation fan in each zone in the drying area, the fan frequency of the dehumidification fan in each zone in the drying area, the opening of the cold air inlet damper in each zone in the drying area, the opening of the dehumidification damper in each zone in the drying area, the dehumidification degree of each zone in the drying area, the main transmission speed, inlet moisture, inlet temperature, inlet flow, ambient temperature, ambient humidity, ambient air pressure, and the moisture in the cooling zone (moisture content of the smoke sheet in the cooling zone) corresponding to the above data in the collected database is used as the output value for data training; the data training method is as follows:
[0015] First, initialize the neural network parameters
[0016] Using the neural network model, give each hidden layer node a weight value w ij and bias value θ j Assign a random number in the interval (-1, 1) respectively, set the mean square error e between the forward prediction data and the actual data, set the number of hidden layer nodes N, set the learning rate l, and set the maximum number of learning times M;
[0017] Secondly, positive learning
[0018] The input layer data x i Import the first neural node of the hidden layer for operation and calculate the first hidden layer node data And then through The first hidden layer data corresponding to the first layer node is obtained; and by analogy, the hidden layer data is imported into the neural node for operation to calculate the next hidden layer node data. And then through Get the second hidden layer data corresponding to the second layer nodes; repeat the above process, the number of calculations is the number of hidden layer nodes N, and the last layer is the output layer value This is the predicted value corresponding to this data, that is, the predicted moisture content in the cooling zone of the tobacco redrying machine O j ; Among them, w ij is the weight value of each hidden layer node, θ j is the bias value of each hidden layer node, and f(x) is the activation function;
[0019] Then, update the weight value w of each node in reverse ij and bias value θ j
[0020] Calculate the data error Err of the output node j =O j (1-O j )(T j -O j ), the error of hidden layer node j Then through w ij =w ij +Δw,Δw ij =(l)Err j O i ,θ j =θ j +Δθ,Δθ j =(l)Err j , update the weight and bias value of each node; where w jk is the weight of the node to the next layer node, Err k is the error of node k, i is the node in the previous layer, j is the node in this layer, k is the node in the next layer, O j is the predicted moisture value of the cold room, T j is the actual moisture value of the cold room smoke sheet corresponding to this data, l is the set learning rate;
[0021] When all historical data training is completed, the training process is repeated from the first historical data; within the training cycle, when the cold room moisture value data O is predicted j The moisture value of the cold room smoke sheet corresponding to this data is T j When the mean square error e reaches the set value range, or the actual number of training times reaches the set number of training times M, the training ends and the weight values w of each node after training are saved. ij and bias value θ j to the designated location;
[0022] Step 3: Calculate the deviation between the predicted value and the actual value of the cold room moisture in historical data
[0023] The parameter data that affects the moisture in the cold room of the tobacco redrying machine collected in the database is used as the input data x i (i is the number of data), including the actual temperature of each zone in the drying area, the fan frequency of the circulation fan in each zone in the drying area, the fan frequency of the dehumidification fan in each zone in the drying area, the opening of the cold air inlet damper in each zone in the drying area, the opening of the dehumidification damper in each zone in the drying area, the dehumidification degree of each zone in the drying area, the main transmission speed, inlet moisture, inlet temperature, inlet flow, ambient temperature, ambient humidity, ambient pressure, normalize the input data, and call the weight values w of the data training in the previous step ij and deviation value θ j , using the aforementioned “forward learning” method, the moisture in the cooling zone O j The data is forward predicted to obtain the average value of the actual value of the cold room moisture corresponding to each data in the database, and the predicted value of the cold room moisture corresponding to each data is forward predicted. j The average value of the two values is calculated.
[0024] Step 4: Calculate the set temperature value of each drying area
[0025] Read the actual temperature T of the bottom PLC system in each drying zone iA , calculate the upper limit value T of the predicted set temperature value of each zone i-h and the lower limit T i-l , T i-h =T iA +H,T i-l =T iA -H; then calculate the temperature setting value T for each drying zone ir , T ir =T i-l +(T i-h -T i-l )×j / N T ; Among them, H is the parameter for preliminary determination of high and low temperature values, M i is the fine-tuning value of the set temperature of each drying zone, N T is the number of temperature divisions, j = 1 ~ N T ;
[0026] The prediction input data includes: ir , the fan frequency of the circulation fan in each zone of the drying area, the fan frequency of the dehumidification fan in each zone of the drying area, the opening of the cold air inlet damper in each zone of the drying area, the opening of the dehumidification damper in each zone of the drying area, the dehumidification degree of each zone of the drying area, the main transmission speed, the inlet moisture, the inlet temperature, the inlet flow, the ambient temperature, the ambient humidity, the ambient air pressure; except T ir In addition, all data are collected in real time by the underlying PLC system;
[0027] Normalize the above data and call the weight values w obtained from the data training in the previous stepij and deviation value θ j , using the aforementioned “forward learning” method, the data is forward predicted, and the predicted value is reversely normalized to obtain the predicted cooling zone moisture O k , we get O = O k +△, where O is the corrected predicted value and △ is the corrected value.
[0028] The above method is a process of looping once j. This process loops for N T times, and obtain a set of data that is closest to the actual moisture in the cold room currently collected, that is, the tth data T with the smallest deviation between the predicted data and the actual data. is , T is =T i-l +(T i-h -T is-l )×t / N T +M i , T is That is, set the temperature value for drying zone i;
[0029] Step 5: Data obtained by the control algorithm is transmitted to the PLC system
[0030] The drying zone i temperature setting value T obtained by the above method is is The temperature is transmitted to the PLC control system through the OPC server and used as the temperature setting parameter value of each drying zone for real-time control of the tobacco redrying machine control system.
[0031] The present invention adopts the above scheme, thereby effectively realizing automatic moisture control of the cold room of the slice baking machine, greatly reducing the manual operation intensity, reducing the dependence of the performance indicators of the slice baking machine on manual operation experience, and improving various performance indicators of the slice baking machine. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] Figure 1 Schematic diagram of a flow chart of an embodiment of the present invention. DETAILED DESCRIPTION
[0033] The technical solution of the present invention is further described in detail below in conjunction with the specific embodiments and the accompanying drawings, but the present invention is not limited to the following technical solutions. The technologies or products not specified in the embodiments are all existing technologies or conventional products that can be purchased.
[0034] Embodiment 1: As shown in the attached Figure 1As shown, the automatic moisture control method of the cold room of the tobacco redrying machine based on the intelligent algorithm includes three steps: data collection, data transmission, and moisture control. The data collection is completed by the underlying control system PLC. The collected data are parameters that affect the moisture in the cold room of the tobacco redrying machine. The data transmission between the underlying PLC control system and the control algorithm is realized through the OPC server, and the collected data are stored in the database. After a certain amount of data is collected, data training and data prediction are performed through the control algorithm. The set temperature value of the drying zone obtained by the control algorithm is transmitted to the PLC control system, and then the temperature of each drying zone is controlled in real time through the tobacco redrying machine control system to realize automatic control of the moisture in the cold room of the tobacco redrying machine.
[0035] Data collection is mainly completed by the underlying control system PLC. The main content of data collection is to collect various parameters that affect the moisture content of the cold room of the tobacco slice redrying machine, including: the actual temperature of each zone in the drying area of the tobacco slice redrying machine, the fan frequency of the circulation fan in each zone of the drying area, the fan frequency of the dehumidification fan in each zone of the drying area, the opening of the cold air inlet damper in each zone of the drying area, the opening of the dehumidification damper in each zone of the drying area, the dehumidification degree of each zone in the drying area, as well as the main drive speed of the tobacco slice redrying machine, inlet moisture, inlet temperature, inlet flow, ambient temperature, ambient humidity, ambient air pressure, and actual moisture content in the cold room.
[0036] The data of various parameters that affect the moisture content of the cold room of the tobacco slice re-drying machine are collected by stacking. When the tobacco leaves of the same cross section of the slice roasting machine arrive at the data acquisition instrument at the entrance of the slice roasting machine, the PLC system collects the actual parameter data of the tobacco leaves of this section. As the equipment runs, the PLC system continues to collect the actual parameter data of the tobacco leaves of this section. The PLC program performs the stacking process of this parameter inside until the tobacco leaves of this cross section walk under the cold room data acquisition instrument to complete the data collection and stacking process. At this time, the PLC system uses this out-of-stack data as the data to be collected; among them, the main drive speed, ambient temperature, ambient humidity, ambient air pressure, and actual values of moisture content in the cold room do not need to be stacked, and these data are used as historical data for algorithm training. Since the tobacco slice re-drying machine is relatively long, it takes about 3 to 4 minutes for the tobacco leaves to go from the entrance to the cold room, so the data cannot be collected at the same time, and the stacking method needs to be used to collect the data and smooth the collected data to a certain extent. Taking the collection of moisture at the inlet of the tobacco redrying machine as an example, when the tobacco leaves of the same cross section of the tobacco redrying machine arrive at the moisture meter at the inlet of the drying machine, the PLC system collects the actual moisture of the tobacco leaves of this section. As the equipment runs, the PLC program performs the stacking process of this moisture inside. After about 3 minutes (this time is determined by the specific characteristics of the equipment), the tobacco leaves of this cross section reach the bottom of the moisture meter in the cold room and the stacking is completed. The PLC system then uses this out-of-stack data as the data to be collected, and the same applies to other data. The specific stacking time is determined by the characteristics of the equipment. The basic principle is that the data collected by the sensor is stacked until it reaches the bottom of the moisture meter in the cold room. All parameters of tobacco leaves with the same cross section of the tobacco redrying machine arrive at the bottom of the moisture meter in the cold room and then are stacked, and the PLC system records the out-of-stack data at this time.
[0037] Data transmission is achieved through the OPC server between the underlying PLC control system and the control algorithm, including the PLC control system data transmission to the control algorithm, and the tobacco slice re-roasting machine drying zone parameter settings obtained by the control algorithm are transmitted to the PLC control system. This method is mainly used to solve the universality of the algorithm. For different underlying control systems, the algorithm program does not need to be changed, and the software kepserver is used. The collected data is stored in the database through data transmission. Data training can be performed after the amount of collected data exceeds 50,000.
[0038] The control algorithm process includes: data normalization, initialization of neural network parameters, historical data training, forward learning, and reverse update of each node weight value w ij and deviation value θ j , calculate the deviation △ between the predicted value and the actual value of the cold room moisture in historical data, calculate the set temperature value of each zone in the drying area, and transmit the data obtained by the control algorithm to the PLC system. The specific process is as follows:
[0039] Step 1: Data normalization
[0040] For the parameters in the collected database that affect the moisture in the cold room of the slice baking machine, normalize the data to the range of 0 to 1, with the maximum value less than 1 and not equal to 1, and the minimum value greater than 0 and not equal to 1;
[0041] Step 2: Historical data training
[0042] The parameter data that affects the moisture in the cold room of the tobacco redrying machine collected in the database is used as the input data x i (i is the number of variables corresponding to each data), including the actual temperature of each zone in the drying area, the fan frequency of the circulation fan in each zone in the drying area, the fan frequency of the dehumidification fan in each zone in the drying area, the opening of the cold air inlet damper in each zone in the drying area, the opening of the dehumidification damper in each zone in the drying area, the dehumidification degree of each zone in the drying area, the main transmission speed, inlet moisture, inlet temperature, inlet flow, ambient temperature, ambient humidity, ambient air pressure, and the moisture in the cooling zone (moisture content of the smoke sheet in the cooling zone) corresponding to the above data in the collected database is used as the output value for data training; the data training method is as follows:
[0043] First, initialize the neural network parameters.
[0044] Using the neural network model, give each hidden layer node a weight value w ij and bias value θ j Assign a random number in the interval (-1, 1) respectively, set the mean square error e between the forward prediction data and the actual data, set the number of hidden layer nodes N, set the learning rate l, set the maximum number of learning times M, and adjust and determine the parameters according to the actual situation of the tobacco redrying machine.
[0045] Secondly, forward learning, the input layer data x i Import the first neural node of the hidden layer for operation and calculate the first hidden layer node data And then through The first hidden layer data corresponding to the first layer node is obtained; and by analogy, the hidden layer data is imported into the neural node for operation to calculate the next hidden layer node data. And then through Get the second hidden layer data corresponding to the second layer nodes; repeat the above process, the number of calculations is the number of hidden layer nodes N, and the last layer is the output layer value This is the predicted value corresponding to this data, that is, the predicted moisture content in the cooling zone of the tobacco redrying machine O j ; Among them, w ij is the weight value of each hidden layer node, θ j is the bias value of each hidden layer node, and f(x) is the activation function;
[0046] Then, update the weight value w of each node in reverse ij and bias value θ j
[0047] Calculate the data error Err of the output node j =O j (1-O j )(T j -O j ), the error of hidden layer node j Then through w ij =w ij +Δw,Δw ij =(l)Err j O i ,θ j =θ j +Δθ,Δθ j =(l)Err j , update the weight and bias value of each node; where w jk is the weight of the node to the next layer node, Err k is the error of node k, i is the node in the previous layer, j is the node in this layer, k is the node in the next layer, O j is the predicted moisture value of the cold room, T j is the actual cold room moisture value corresponding to this data, l is the set learning rate;
[0048] When all historical data training is completed, the training process is repeated from the first historical data; within the training cycle, when the cold room moisture value data O is predicted j The moisture value of the cold room smoke sheet corresponding to this data is T j When the mean square error e reaches the set value range, or the actual number of training times reaches the set number of training times M, the training ends and the weight values w of each node after training are saved. ij and bias value θ j to the designated location;
[0049] Step 3: Calculate the deviation between the predicted value and the actual value of the cold room moisture in historical data
[0050] The parameter data that affects the moisture in the cold room of the tobacco redrying machine collected in the database is used as the input data x i (i is the number of data), including the actual temperature of each zone in the drying area, the fan frequency of the circulation fan in each zone in the drying area, the fan frequency of the dehumidification fan in each zone in the drying area, the opening of the cold air inlet damper in each zone in the drying area, the opening of the dehumidification damper in each zone in the drying area, the dehumidification degree of each zone in the drying area, the main transmission speed, inlet moisture, inlet temperature, inlet flow, ambient temperature, ambient humidity, ambient pressure, normalize the input data, and call the weight values w of the data training in the previous step ij and deviation value θ j , using the aforementioned “forward learning” method, the moisture in the cooling zone O jThe data is forward predicted to obtain the average value of the actual value of the cold room moisture corresponding to each data in the database, and the predicted value of the cold room moisture corresponding to each data is forward predicted. j The average value of the two values is calculated.
[0051] Step 4: Calculate the set temperature value of each drying area
[0052] Read the actual temperature T of the bottom PLC system in each drying zone iA , calculate the upper limit value T of the predicted set temperature value of each zone i-h and the lower limit T i-l , T i-h =T iA +H,T i-l =T iA -H; then calculate the temperature setting value T for each drying zone ir , T ir =T i-l +(T i-h -T i-l )×j / N T ; Among them, H is the parameter for preliminary determination of high and low temperature values, M i is the fine-tuning value of the set temperature of each drying zone, N T is the number of temperature divisions, j = 1 ~ N T ;
[0053] The prediction input data includes: ir , the fan frequency of the circulation fan in each zone of the drying area, the fan frequency of the dehumidification fan in each zone of the drying area, the opening of the cold air inlet damper in each zone of the drying area, the opening of the dehumidification damper in each zone of the drying area, the dehumidification degree of each zone of the drying area, the main transmission speed, the inlet moisture, the inlet temperature, the inlet flow, the ambient temperature, the ambient humidity, the ambient air pressure; except T ir In addition, all data are collected in real time by the underlying PLC system;
[0054] Normalize the above data and call the weight values w obtained from the data training in the previous step ij and deviation value θ j , using the aforementioned “forward learning” method, the data is forward predicted, and the predicted value is reversely normalized to obtain the predicted cooling zone moisture O k , we get O = O k +△, where O is the corrected predicted value and △ is the corrected value.
[0055] The above method is a process of looping once j. This process loops for N T times, and obtain a set of data that is closest to the actual moisture in the cold room currently collected, that is, the tth data T with the smallest deviation between the predicted data and the actual data. is , T is=T i-l +(T i-h -T is-l )×t / N T , T is That is, set the temperature value for drying zone i;
[0056] Step 5: Data obtained by the control algorithm is transmitted to the PLC system
[0057] The drying zone i temperature setting value T obtained by the above method is is The temperature is transmitted to the PLC control system through the OPC server and used as the temperature setting parameter value of each drying zone for real-time control of the tobacco redrying machine control system.
[0058] In this embodiment, there are 4 drying zones, the mean square error between the forward prediction data and the actual data is e=0.01, the number of hidden layer nodes is N=10, the learning rate is l=0.01, the maximum number of learning times is M=30000, and when the number of learning times reaches 25466 times and the mean square error e of the actual data reaches 0.01, the training ends. 1-h =70℃、T 1-l =60℃, the high and low values of the temperature in the second drying zone are used to preliminarily determine the parameter T 2-h =75℃、T 1-l =65℃, the high and low values of the three drying zones are used to preliminarily determine the parameter T 3-h =80℃、T 1-l =70℃, the high and low values of the temperature in the four drying zones are used to preliminarily determine the parameter T 4-h =85℃、T 4-l =75℃, number of temperature divisions N T = 10. The following table data are respectively introduced into the algorithm of the established model, and the deviation between the corresponding predicted value and the actual value is calculated.
[0059]
[0060] From the above data, it can be seen that the deviation between the predicted value of the cold room moisture and the actual value of the cold room moisture in the fourth input value is the smallest. 1s =63℃, T 2s =68℃、T 3s =73℃、T 4s =78℃.
[0061] Embodiment 2: As shown in the attached Figure 1As shown, the automatic moisture control method of the cold room of the tobacco redrying machine based on the intelligent algorithm includes three steps: data collection, data transmission, and moisture control. The data collection is completed by the underlying control system PLC. The collected data are parameters that affect the moisture in the cold room of the tobacco redrying machine. The data transmission between the underlying PLC control system and the control algorithm is realized through the OPC server, and the collected data are stored in the database. After a certain amount of data is collected, data training and data prediction are performed through the control algorithm. The set temperature value of the drying zone obtained by the control algorithm is transmitted to the PLC control system, and then the temperature of each drying zone is controlled in real time through the tobacco redrying machine control system to realize automatic control of the moisture in the cold room of the tobacco redrying machine.
[0062] Data collection is mainly completed by the underlying control system PLC. The main content of data collection is to collect various parameters that affect the moisture content of the cold room of the tobacco slice redrying machine, including: the actual temperature of each zone in the drying area of the tobacco slice redrying machine, the fan frequency of the circulation fan in each zone of the drying area, the fan frequency of the dehumidification fan in each zone of the drying area, the opening of the cold air inlet damper in each zone of the drying area, the opening of the dehumidification damper in each zone of the drying area, the dehumidification degree of each zone in the drying area, as well as the main drive speed of the tobacco slice redrying machine, inlet moisture, inlet temperature, inlet flow, ambient temperature, ambient humidity, ambient air pressure, and actual moisture content in the cold room.
[0063] The data of various parameters that affect the moisture content of the cold room of the tobacco slice re-drying machine are collected by stacking. When the tobacco leaves of the same cross section of the slice roasting machine arrive at the data acquisition instrument at the entrance of the slice roasting machine, the PLC system collects the actual parameter data of the tobacco leaves of this section. As the equipment runs, the PLC system continues to collect the actual parameter data of the tobacco leaves of this section. The PLC program performs the stacking process of this parameter inside until the tobacco leaves of this cross section walk under the cold room data acquisition instrument to complete the data collection and stacking process. At this time, the PLC system uses this out-of-stack data as the data to be collected; among them, the main drive speed, ambient temperature, ambient humidity, ambient air pressure, and actual values of moisture content in the cold room do not need to be stacked, and these data are used as historical data for algorithm training. Since the tobacco slice re-drying machine is relatively long, it takes about 3 to 4 minutes for the tobacco leaves to go from the entrance to the cold room, so the data cannot be collected at the same time, and the stacking method needs to be used to collect the data and smooth the collected data to a certain extent. Taking the collection of moisture at the inlet of the tobacco redrying machine as an example, when the tobacco leaves of the same cross section of the tobacco redrying machine arrive at the moisture meter at the inlet of the drying machine, the PLC system collects the actual moisture of the tobacco leaves of this section. As the equipment runs, the PLC program performs the stacking process of this moisture inside. After about 3 minutes (this time is determined by the specific characteristics of the equipment), the tobacco leaves of this cross section reach the bottom of the moisture meter in the cold room and the stacking is completed. The PLC system then uses this out-of-stack data as the data to be collected, and the same applies to other data. The specific stacking time is determined by the characteristics of the equipment. The basic principle is that the data collected by the sensor is stacked until it reaches the bottom of the moisture meter in the cold room. All parameters of tobacco leaves with the same cross section of the tobacco redrying machine arrive at the bottom of the moisture meter in the cold room and then are stacked, and the PLC system records the out-of-stack data at this time.
[0064] Data transmission is achieved through the OPC server between the underlying PLC control system and the control algorithm, including the PLC control system data transmission to the control algorithm, and the tobacco slice re-roasting machine drying zone parameter settings obtained by the control algorithm are transmitted to the PLC control system. This method is mainly used to solve the universality of the algorithm. For different underlying control systems, the algorithm program does not need to be changed, and the software kepserver is used. The collected data is stored in the database through data transmission. Data training can be performed after the amount of collected data exceeds 50,000.
[0065] The control algorithm process includes: data normalization, initialization of neural network parameters, historical data training, forward learning, and reverse update of each node weight value w ij and deviation value θ j , calculate the deviation △ between the predicted value and the actual value of the cold room moisture in historical data, calculate the set temperature value of each zone in the drying area, and transmit the data obtained by the control algorithm to the PLC system. The specific process is as follows:
[0066] Step 1: Data normalization
[0067] For the parameters in the collected database that affect the moisture in the cold room of the slice baking machine, normalize the data to the range of 0 to 1, with the maximum value less than 1 and not equal to 1, and the minimum value greater than 0 and not equal to 1;
[0068] Step 2: Historical data training
[0069] The parameter data that affects the moisture in the cold room of the tobacco redrying machine collected in the database is used as the input data x i (i is the number of variables corresponding to each data), including the actual temperature of each zone in the drying area, the fan frequency of the circulation fan in each zone in the drying area, the fan frequency of the dehumidification fan in each zone in the drying area, the opening of the cold air inlet damper in each zone in the drying area, the opening of the dehumidification damper in each zone in the drying area, the dehumidification degree of each zone in the drying area, the main transmission speed, inlet moisture, inlet temperature, inlet flow, ambient temperature, ambient humidity, ambient air pressure, and the moisture in the cooling zone (moisture content of the smoke sheet in the cooling zone) corresponding to the above data in the collected database is used as the output value for data training; the data training method is as follows:
[0070] First, initialize the neural network parameters.
[0071] Using the neural network model, give each hidden layer node a weight value w ij and bias value θ j Assign a random number in the interval (-1, 1) respectively, set the mean square error e between the forward prediction data and the actual data, set the number of hidden layer nodes N, set the learning rate l, set the maximum number of learning times M, and adjust and determine the parameters according to the actual situation of the tobacco redrying machine.
[0072] Secondly, forward learning, the input layer data x i Import the first neural node of the hidden layer for operation and calculate the first hidden layer node data And then through The first hidden layer data corresponding to the first layer node is obtained; and by analogy, the hidden layer data is imported into the neural node for operation to calculate the next hidden layer node data. And then through Get the second hidden layer data corresponding to the second layer nodes; repeat the above process, the number of calculations is the number of hidden layer nodes N, and the last layer is the output layer value This is the predicted value corresponding to this data, that is, the predicted moisture content in the cooling zone of the tobacco redrying machine O j ; Among them, w ij is the weight value of each hidden layer node, θ j is the bias value of each hidden layer node, and f(x) is the activation function;
[0073] Then, update the weight value w of each node in reverse ij and bias value θ j
[0074] Calculate the data error Err of the output node j =O j (1-O j )(T j -O j ), the error of hidden layer node j Then through w ij =w ij +Δw,Δw ij =(l)Err j O i ,θ j =θ j +Δθ,Δθ j =(l)Err j , update the weight and bias value of each node; where w jk is the weight of the node to the next layer node, Err k is the error of node k, i is the node in the previous layer, j is the node in this layer, k is the node in the next layer, O j is the predicted moisture value of the cold room, T j is the actual cold room moisture value corresponding to this data, l is the set learning rate;
[0075] When all historical data training is completed, the training process is repeated from the first historical data; within the training cycle, when the cold room moisture value data O is predicted j The moisture value of the cold room smoke sheet corresponding to this data is T j When the mean square error e reaches the set value range, or the actual number of training times reaches the set number of training times M, the training ends and the weight values w of each node after training are saved. ij and bias value θ j to the designated location;
[0076] Step 3: Calculate the deviation between the predicted value and the actual value of the cold room moisture in historical data
[0077] The parameter data that affects the moisture in the cold room of the tobacco redrying machine collected in the database is used as the input data x i (i is the number of data), including the actual temperature of each zone in the drying area, the fan frequency of the circulation fan in each zone in the drying area, the fan frequency of the dehumidification fan in each zone in the drying area, the opening of the cold air inlet damper in each zone in the drying area, the opening of the dehumidification damper in each zone in the drying area, the dehumidification degree of each zone in the drying area, the main transmission speed, inlet moisture, inlet temperature, inlet flow, ambient temperature, ambient humidity, ambient pressure, normalize the input data, and call the weight values w of the data training in the previous step ij and deviation value θ j , using the aforementioned “forward learning” method, the moisture in the cooling zone O jThe data is forward predicted to obtain the average value of the actual value of the cold room moisture corresponding to each data in the database, and the predicted value of the cold room moisture corresponding to each data is forward predicted. j The average value of the two values is calculated.
[0078] Step 4: Calculate the set temperature value of each drying area
[0079] Read the actual temperature T of the bottom PLC system in each drying zone iA , calculate the upper limit value T of the predicted set temperature value of each zone i-h and the lower limit T i-l , T i-h =T iA +H,T i-l =T iA -H; then calculate the temperature setting value T for each drying zone ir , T ir =T i-l +(T i-h -T i-l )×j / N T ; Among them, H is the parameter for preliminary determination of high and low temperature values, M i is the fine-tuning value of the set temperature of each drying zone, N T is the number of temperature divisions, j = 1 ~ N T ;
[0080] The prediction input data includes: ir , the fan frequency of the circulation fan in each zone of the drying area, the fan frequency of the dehumidification fan in each zone of the drying area, the opening of the cold air inlet damper in each zone of the drying area, the opening of the dehumidification damper in each zone of the drying area, the dehumidification degree of each zone of the drying area, the main transmission speed, the inlet moisture, the inlet temperature, the inlet flow, the ambient temperature, the ambient humidity, the ambient air pressure; except T ir In addition, all data are collected in real time by the underlying PLC system;
[0081] Normalize the above data and call the weight values w obtained from the data training in the previous step ij and deviation value θ j , using the aforementioned “forward learning” method, the data is forward predicted, and the predicted value is reversely normalized to obtain the predicted cooling zone moisture O k , we get O = O k +△, where O is the corrected predicted value and △ is the corrected value.
[0082] The above method is a process of looping once j. This process loops for N T times, and obtain a set of data that is closest to the actual moisture in the cold room currently collected, that is, the tth data T with the smallest deviation between the predicted data and the actual data. is , T is=T i-l +(T i-h -T is-l )×t / N T , T is That is, set the temperature value for drying zone i;
[0083] Step 5: Data obtained by the control algorithm is transmitted to the PLC system
[0084] The drying zone i temperature setting value T obtained by the above method is is The temperature is transmitted to the PLC control system through the OPC server and used as the temperature setting parameter value of each drying zone for real-time control of the tobacco redrying machine control system.
[0085] In this embodiment, there are 5 drying zones, the mean square error between the forward prediction data and the actual data is e=0.01, the number of hidden layer nodes is N=10, the learning rate is l=0.01, and the maximum number of learning times is M=30000. When the number of learning times reaches 30000 times, the mean square error of the actual data e reaches 0.021, and the training ends. The high and low values of the temperature in the drying zone 1 are used to preliminarily determine the parameter T 1-h =68℃、T 1-l =58℃, the high and low values of the temperature in the second drying zone are used to preliminarily determine the parameter T 2-h =75℃、T 1-l =65℃, the high and low values of the three drying zones are used to preliminarily determine the parameter T 3-h =83℃、T 1-l =73℃, the high and low values of the temperature in the four drying zones are used to preliminarily determine the parameter T 4-h =86℃、T 4-l =76℃, the high and low values of the five drying zones are used to preliminarily determine the parameter T 5-h =80℃、T 5-l =70℃, number of temperature divisions N T = 10. The following table data are respectively introduced into the algorithm of the established model, and the deviation between the corresponding predicted value and the actual value is calculated.
[0086]
[0087] From the above data, it can be seen that the seventh substitution value has the smallest deviation from the predicted value of cold room moisture and the actual value of cold room moisture. 1s =64℃, T 2s =71℃、T 3s =79℃、T 4s =82℃、T 5s =76℃.
[0088] The method effectively realizes the automatic control of moisture in the cold room of the slice baking machine, greatly reduces the manual operation intensity, reduces the dependence of the performance indicators of the slice baking machine on manual operation experience, and improves various performance indicators of the slice baking machine.
[0089] The technical contents of the present invention are explained above, but the protection scope of the present invention is not limited to the said contents. Within the knowledge scope of ordinary technicians in this field, various changes can be made to the technical contents of the present invention without departing from the purpose of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for automatically controlling moisture in a cold room of a tobacco sheet redrying machine based on an intelligent algorithm, characterized in that: It includes three steps: data collection, data transmission, and moisture control. Data collection is completed by the underlying control system PLC. The collected data are parameters that affect the moisture in the cold room of the tobacco redrying machine. The data transmission between the underlying PLC control system and the control algorithm is realized through the OPC server, and the collected data are stored in the database. After a certain amount of data is collected, data training and data prediction are carried out through the control algorithm. The set temperature value of the drying area obtained by the control algorithm is transmitted to the PLC control system, and then the temperature of each drying area is controlled in real time through the tobacco redrying machine control system to realize automatic control of the moisture in the cold room of the tobacco redrying machine.
2. The method for automatically controlling moisture in a cold room of a tobacco sheet redrying machine based on an intelligent algorithm according to claim 1, characterized in that: The parameters that affect the moisture content of tobacco leaves in the cold room of the tobacco redrying machine include: the actual temperature of each zone in the drying area of the tobacco redrying machine, the fan frequency of the circulating fan in each zone in the drying area, the fan frequency of the dehumidification fan in each zone in the drying area, the opening of the cold air inlet damper in each zone in the drying area, the opening of the dehumidification damper in each zone in the drying area, the dehumidification degree of each zone in the drying area, as well as the main transmission speed of the tobacco redrying machine, inlet moisture, inlet temperature, inlet flow, ambient temperature, ambient humidity, ambient air pressure, and actual value of moisture content in the cold room.
3. The method for automatically controlling moisture in a cold room of a tobacco sheet redrying machine based on an intelligent algorithm according to claim 1 or 2, characterized in that: The parameter data that affect the moisture content in the cold room of the slice torrefaction machine are collected by stacking. When the tobacco leaves of the same cross section of the slice torrefaction machine arrive at the data acquisition instrument at the entrance of the slice torrefaction machine, the PLC system collects the actual parameter data of the tobacco leaves of this cross section. As the equipment runs, the PLC system continues to collect the actual parameter data of the tobacco leaves of this cross section. The stacking process of this parameter is performed inside the PLC program until the tobacco leaves of this cross section walk under the data acquisition instrument in the cold room, completing the data collection and stacking process. At this time, the PLC system uses the out-of-stack data as the collected data. Among them, the actual values of the main transmission speed, ambient temperature, ambient humidity, ambient air pressure, and moisture content in the cold room do not need to be stacked, and these data are used as historical data for algorithm training.
4. The method for automatically controlling moisture in a cold room of a tobacco sheet redrying machine based on an intelligent algorithm according to claim 1, characterized in that: The data transmission is realized by the OPC server between the underlying PLC control system and the control algorithm, including the transmission of PLC control system data to the control algorithm, and the transmission of various parameter settings of the drying zone of the tobacco sheet redrying machine obtained by the control algorithm to the PLC control system.
5. The automatic moisture control method for the cold room of the tobacco slice redrying machine based on intelligent algorithm according to claim 1, 2 or 3 is characterized in that: the process of the control algorithm includes: data normalization processing, initialization of neural network parameters, historical data training, forward learning, reverse update of each node weight value w ij and deviation value θ j , calculate the deviation △ between the predicted value and the actual value of the cold room moisture in historical data, calculate the set temperature value of each zone in the drying area, and transmit the data obtained by the control algorithm to the PLC system. The specific process is as follows: 5.1 Data Normalization For the parameters in the collected database that affect the moisture in the cold room of the baking machine, normalize the data to the range of 0 to 1, with the maximum value less than 1 and not equal to 1, and the minimum value greater than 0 and not equal to 0; 5.2 Historical Data Training The parameter data that affects the moisture in the cold room of the tobacco redrying machine collected in the database is used as the input data x i (i is the number of variables corresponding to each piece of data), and the moisture content of the cooling zone tobacco sheet corresponding to each piece of data in the collected database is used as the output value for data training; The data training method is as follows: 5.2.1 Initializing Neural Network Parameters Using the neural network model, give each hidden layer node a weight value w ij and bias value θ j Assign a random number in the interval (-1, 1) respectively, set the mean square error e between the forward prediction data and the actual data, set the number of hidden layer nodes N, set the learning rate l, and set the maximum number of learning times M; 5.2.2 Forward Learning The input layer data x i Import the first neural node of the hidden layer for operation and calculate the first hidden layer node data And then through The first hidden layer data corresponding to the first layer node is obtained; and by analogy, the hidden layer data is imported into the neural node for operation to calculate the next hidden layer node data. And then through Get the second hidden layer data corresponding to the second layer nodes; repeat the above process, the number of calculations is the number of hidden layer nodes N, and the last layer is the output layer value This is the predicted value corresponding to this data, that is, the predicted moisture content in the cooling zone of the tobacco redrying machine O j ; Among them, w ij is the weight value of each hidden layer node, θ j is the bias value of each hidden layer node, and f(x) is the activation function; 5.2.3 Reverse update of each node weight value w ij and bias value θ j Calculate the data error Err of the output node j =O j (1-O j )(T j -O j ), the error of hidden layer node j Then through w ij =w ij +Δw,Δw ij =(l)Err j O i ,θ j =θ j +Δθ,Δθ j =(l)Err j , update the weight and bias value of each node; among them, w jk is the weight of the node to the next node, Err k is the error of node k, i is the node in the previous layer, j is the node in this layer, k is the node in the next layer, O j is the predicted moisture value of the cold room, T j is the actual moisture value of the cold room smoke sheet corresponding to this data, l is the set learning rate; When all historical data training is completed, the training process is repeated from the first historical data; within the training cycle, when the cold room moisture value data O is predicted j The moisture value of the cold room smoke sheet corresponding to this data is T j When the mean square error e reaches the set value range, or the actual number of training times reaches the set number of training times M, the training ends and the weight values w of each node after training are saved. ij and bias value θ j to the designated location; 5.3 Calculate the deviation between the predicted value and the actual value of the cold room moisture in historical data △ The parameter data that affects the moisture in the cold room of the tobacco redrying machine collected in the database is used as the input data x i , normalize the input data and call the weight values w of the data training in the previous step ij and deviation value θ j , using the aforementioned "forward learning" method, the moisture in the cooling zone O j The data is forward predicted to obtain the average value of the actual value of the cold room moisture corresponding to each data in the database, and the predicted value of the cold room moisture corresponding to each data is forward predicted. j The average value of the two values is calculated. 5.4 Calculate the set temperature value of each zone in the drying area Read the actual temperature T of the bottom PLC system in each drying zone iA , calculate the upper limit value T of the predicted set temperature value of each zone i-h and the lower limit T i-l , T i-h =T iA +H,T i-l =T iA -H; then calculate the temperature setting value T for each drying zone ir , T ir =T i-l +(T i-h -T i-l )×j / N T ; Among them, H is the parameter for preliminary determination of high and low temperature values, M i is the fine-tuning value of the set temperature of each drying zone, N T is the number of temperature divisions, j = 1 ~ N T ; The prediction input data includes the following data: T ir , the fan frequency of the circulation fan in each zone of the drying area, the fan frequency of the dehumidification fan in each zone of the drying area, the opening of the cold air inlet damper in each zone of the drying area, the opening of the dehumidification damper in each zone of the drying area, the dehumidification degree of each zone of the drying area, the main transmission speed, the inlet moisture, the inlet temperature, the inlet flow, the ambient temperature, the ambient humidity, and the ambient air pressure; among them, except T ir In addition, all data are collected in real time by the underlying PLC system; Normalize the above data and call the weight values w obtained from the data training in the previous step ij and deviation value θ j , using the aforementioned "forward learning" method, the data is forward predicted, and the predicted value is reversely normalized to obtain the predicted cooling zone moisture O k , we get O = O k +△, where O is the corrected predicted value and △ is the corrected value; The above method is a process of looping once j. This process loops for N T times, and obtain a set of data that is closest to the actual moisture in the cold room currently collected, that is, the tth data T with the smallest deviation between the predicted data and the actual data. is , T is =T i-l +(T i-h -T is-l )×t / N T +M i , T is That is, set the temperature value for drying zone i; 5.5 Data obtained by the control algorithm is transmitted to the PLC system The drying zone i temperature setting value T obtained by the above method is is The temperature is transmitted to the PLC control system through the OPC server and used as the temperature setting parameter value of each drying zone for real-time control of the tobacco redrying machine control system.
Citation Information
Patent Citations
Real-time prediction method for boiler internal flow field thermal deviation model
CN110472274A
Tobacco shred drying moisture and temperature control prediction system and method
CN111887460A
Redryer moisture content control method based on formula parameter library
CN112263012A
Roller cut tobacco drying process parameter optimization method and system based on neural network
CN116611349A
Intelligent control system and method for threshing, redrying and moistening moisture and temperature
CN116719369A
