Multivariable collaborative head material moisture control method based on cut tobacco dryer process association prediction

Abstract

The invention provides a multivariable collaborative head material moisture control method based on cut-tobacco dryer process association prediction, which relates to the technical field of control and comprises the following steps: acquiring production data of a previous process of a cut-tobacco dryer, respectively calculating mean value characteristics of outlet moisture content parameters of a head material part and a whole batch, and writing calculation results into a database; obtaining cut tobacco brand information and process parameters at the current production moment, forming a current working condition feature vector, calculating a current outlet moisture content change trend by using linear fitting, merging the current outlet moisture content change trend into the current working condition feature vector, and calculating an optimal control quantity set value under the current working condition based on a pre-trained K-means model; based on the optimal control quantity set value, using an adjustable coefficient to correct the control quantity of the frequency of a hot air fan, the opening degree of a moisture removal air door and the material scale flow; the corrected control quantity is issued to a PLC through OPC service to achieve cooperative control over the cut-tobacco drier; and updating a K-means model based on the optimal working condition feature vector data subset.

Description

Technical Field

[0001] The invention relates to control technology, and in particular to a multivariable coordinated head material moisture control method based on process correlation prediction of a tofu drying machine. Background Art

[0002] Drying tobacco is an important process in the tobacco production line of cigarette factories. Its main purpose is to control the moisture content of the dried tobacco within a certain range. At the same time, high-temperature heating is used to make the tobacco dry and expand quickly. On the one hand, it can remove green odors and free alkali, activate the aroma permeability of the tobacco leaves, and make the tobacco taste mellower. On the other hand, it can quickly shape the tobacco leaves, giving them good curl and looseness, thereby improving their elasticity, combustibility and filling capacity.

[0003] However, the process of drying the first-cut tobacco leaves (referring to the process in which the moisture content of the tobacco leaves at the outlet rises to the lower boundary of the set value) is highly dependent on manual labor, which causes instability in the initial stage of production and easily produces more dry ends, resulting in an increase in the amount of crushing and a waste of raw materials. In addition, there are deficiencies in the process parameters of the first-cut tobacco leaves set through manual experience, which can easily cause instability in the moisture content of the entire batch. Summary of the invention

[0004] The embodiment of the present invention provides a multivariable collaborative head material moisture control method based on the correlation prediction of the tofu drying machine process, which can solve the problems in the prior art.

[0005] According to a first aspect of the embodiments of the present invention,

[0006] A multivariable collaborative head material moisture control method based on the correlation prediction of the drying machine process is provided, including:

[0007] The production data of the front process of the tobacco drying machine is obtained, and after eliminating abnormal data, the mean characteristics of the outlet moisture content parameters of the head material part and the whole batch are calculated respectively, and the calculation results are written into the database, and the historical benchmark value of the outlet moisture content of the front process is obtained from the database, and the deviation between the outlet moisture content of the front process of the current corresponding batch and the benchmark value is calculated to obtain the basic correction value of the control parameter of the tobacco drying machine; the tobacco brand information and process parameters at the current production moment are obtained to form the current working condition feature vector, and the current outlet moisture content change trend is calculated by linear fitting and merged into the current working condition feature vector, and the optimal control quantity setting value under the current working condition is calculated based on the pre-trained K-means model in combination with the basic correction value of the control parameter of the tobacco drying machine;

[0008] The difference between the inlet moisture content and the historical average is obtained online, the theoretical dehydration amount is calculated, and the control amount of the hot air fan frequency, the dehumidification damper opening and the material scale flow is corrected using the adjustable coefficient based on the optimal control amount setting value, so as to increase the control frequency and reduce the control range to stabilize the outlet moisture content; the corrected control amount is sent to the PLC through the OPC service to realize the coordinated control of the tofu drying machine;

[0009] In the idle period at the end of each production batch, the K-means model update module is started regularly to obtain the production data of the tofu drying machine. The characteristic parameters associated with the moisture content at the outlet of the tofu drying machine are selected based on the minimum mutual information coefficient analysis method, and the characteristic parameters are subjected to delayed alignment processing to form a feature vector data set with time consistency; the feature vector preferred data set is screened according to the outlet moisture content deviation, and the feature vector preferred data set is sorted based on the CPK index to obtain the optimal operating condition feature vector data subset, and the K-means model is updated based on the optimal operating condition feature vector data subset.

[0010] In an optional embodiment,

[0011] The optimal control quantity setting value under the current working condition is calculated based on the pre-trained K-means model, including:

[0012] Collect the non-controlled process parameters of the tobacco drying machine in real time during production, and obtain the tobacco brand parameters at the current production moment;

[0013] Construct a feature vector including non-controlled process parameters, tobacco brand parameters, decision variables and evaluation indicators; the decision variables include the hot air fan frequency and the output opening of the dehumidification damper, and the evaluation indicator is the deviation between the actual value and the set value of the moisture content at the outlet of the tobacco drying machine;

[0014] The linear fitting method is used to calculate the change trend of the moisture content at the outlet of the current tofu drying machine, and the calculated change trend is merged into the constructed feature vector to form a complete feature vector of the current working condition;

[0015] The current working condition feature vector is input into the pre-trained K-means model for calculation. By calculating the cosine similarity between the current working condition feature vector and the centers of each cluster in the clustering model, the K clusters with the highest similarity are selected, and the average coordinates of their center points are used as the optimal setting values ​​of the control quantities of the hot air fan frequency and the output opening of the moisture exhaust damper under the current working condition.

[0016] In the iterative cycle, the acquired production process parameter data is used for incremental training to update the K-means model, and the parameters of the updated K-means model are written into the database for the next control cycle to call.

[0017] In an optional embodiment,

[0018] Calculating the cosine similarity between the current operating condition feature vector and each cluster center in the clustering model includes:

[0019]

[0020] Among them, A represents the current working condition feature vector, B represents the clustering parameter vector, and n n Represents the number of vectors; A i , B i They represent the i-th current operating condition feature vector and the i-th clustering parameter vector respectively.

[0021] In an optional embodiment,

[0022] Training a K-means model involves:

[0023] Obtain the characteristic vector of evaluation indexes in the production process of the tobacco drying machine, including non-controlled process parameters, tobacco brand parameters, decision variables hot air fan frequency and dehumidification damper output opening, and deviation between the actual value and the set value of moisture content at the outlet of the tobacco drying machine, and arrange the indicators in the characteristic vector in ascending order;

[0024] Calculate the sum of the distances from each data point in the sorted feature vector to the first randomly selected centroid, re-sort all data points according to the sum of the distances, select the data point with the smallest sum of distances as the second centroid, and repeat the process of centroid selection and distance calculation until a preset number of centroids are selected;

[0025] Using all the obtained centroids as cluster centers, each data point in the feature vector is divided into the cluster where the nearest cluster center is located, and the preliminary clustering results are obtained;

[0026] Calculate the average distance between each data point and other data points in its cluster and the average distance between each data point and other cluster centers, thereby calculating the silhouette coefficient of each data point, and evaluate the clustering effect under the current number of clusters through the mean of the silhouette coefficient;

[0027] The clustering effects under different clustering numbers are compared by the mean of the silhouette coefficient to obtain the optimal clustering number. The clustering process is then re-executed with this clustering number until the final clustering result of the process parameters of the tofu drying machine is obtained, completing the training of the K-means model.

[0028] In an optional embodiment,

[0029] The clustering effect under the current number of clusters is evaluated by the mean of the silhouette coefficient, including:

[0030]

[0031]

[0032] Among them, s(x i ) represents the i-th sample x i The silhouette coefficient, a(x i ) represents the sample x iThe average distance between all other sample points in the cluster, b(x i ) represents x i The average distance from the center of the non-cluster, N represents the number of samples, q(x i ) represents the sample x i The corresponding center of mass, x j represents the jth sample, and q(k) represents the kth centroid.

[0033] In an optional embodiment,

[0034] Obtain the difference between the inlet moisture content and the historical average online, calculate the theoretical dehydration amount, and use the adjustable coefficient based on the optimal control amount setting value to correct the control amount of the hot air fan frequency, dehumidification damper opening and material scale flow, increase the control frequency and reduce the control range to stabilize the outlet moisture content; send the corrected control amount to the PLC through the OPC service to realize the coordinated control of the tofu drying machine, including:

[0035] Using big data analysis technology, the theoretical dehydration amount under the current working conditions is calculated based on the difference between the moisture content of the current tobacco at the inlet of the tobacco drying machine and the average moisture content of multiple batches of tobacco in history, the difference between the moisture content of the inlet and outlet of the tobacco drying machine, and the adjustable coefficient;

[0036] The moisture content test value at the outlet of the tofu drying machine is compared with the theoretical dehydration calculation value, and the moisture content compensation correction value is calculated. The correction value is multiplied by the dehydration coefficient calculated by multivariate linear regression to obtain the final control value of the hot air fan frequency, the dehumidification damper output opening and the scale set flow rate;

[0037] By increasing the control adjustment frequency and reducing the control amplitude, the stabilization process of the moisture content at the outlet of the tofu drying machine can be accelerated;

[0038] The calculated hot air fan frequency, dehumidification damper output opening and scale set flow control quantity are sent to the PLC program segment through the OPC server to realize automatic control of the production process of the tofu drying machine.

[0039] According to a second aspect of the embodiments of the present invention,

[0040] Provide a multivariable collaborative raw material moisture control system based on the correlation prediction of the drying machine process, including:

[0041] The first unit is used to obtain the production data of the front process of the tobacco drying machine, and after eliminating abnormal data, calculate the mean characteristics of the outlet moisture content parameters of the head material part and the whole batch respectively, and write the calculation results into the database, and obtain the historical front process outlet moisture content benchmark value from the database, calculate the deviation between the outlet moisture content of the current corresponding batch of front process and the benchmark value to obtain the basic correction amount of the control parameter of the tobacco drying machine; obtain the tobacco brand information and process parameters at the current production moment to form the current working condition feature vector, use linear fitting to calculate the current outlet moisture content change trend and merge it into the current working condition feature vector, and calculate the optimal control amount setting value under the current working condition based on the pre-trained K-means model in combination with the basic correction amount of the tobacco drying machine control parameter;

[0042] The second unit is used to obtain the difference between the inlet moisture content and the historical average online, calculate the theoretical dehydration amount, and use the adjustable coefficient based on the optimal control amount setting value to correct the control amount of the hot air fan frequency, the dehumidification damper opening and the material scale flow, increase the control frequency, and reduce the control range to stabilize the outlet moisture content; send the corrected control amount to the PLC through the OPC service to realize the coordinated control of the tofu drying machine;

[0043] The third unit is used to start the K-means model update module in an idle period at the end of each production batch, obtain the production data of the tofu drying machine, select the characteristic parameters related to the moisture content at the outlet of the tofu drying machine based on the minimum mutual information coefficient analysis method, and perform delayed alignment processing on the characteristic parameters to form a feature vector data set with time consistency; select the feature vector preferred data set according to the outlet moisture content deviation, sort the feature vector preferred data set based on the CPK index, obtain the optimal operating condition feature vector data subset, and update the K-means model based on the optimal operating condition feature vector data subset.

[0044] According to a third aspect of the embodiments of the present invention,

[0045] An electronic device is provided, comprising:

[0046] processor;

[0047] a memory for storing processor-executable instructions;

[0048] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0049] According to a fourth aspect of the embodiments of the present invention,

[0050] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0051] In view of the control characteristics of the wire drying head material, the present invention adopts a wire drying machine multivariable collaborative head material moisture control method based on process correlation prediction to reduce the dry head amount of wire drying head material. The control method based on process correlation prediction can predict the properties of the incoming material in advance and set more accurate control quantities. It has stronger predictive ability and robustness, and can greatly reduce the dry head amount and reduce production consumption while ensuring the stability of the overall outlet moisture.

[0052] The drying process in the silk production has complex heat and mass transfer characteristics, which reflects the various problems involving specific production details in the process, such as unstable steam quality, unstable tobacco flow, large fluctuations in inlet moisture content, changes in ambient temperature and humidity, etc. The moisture content needs to be manually adjusted based on the operator's own experience, which often leads to problems such as difficulty in adjusting the moisture content of the dried silk head, inaccurate control, large moisture fluctuations, and large amount of dry ends. Therefore, it is urgent to improve the original control mode and adopt intelligent control methods to improve the control level, realize unmanned operation, improve the quality of tobacco, and reduce the amount of dry ends, thereby reducing costs and energy consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a flow chart of a multivariable collaborative head material moisture control method based on the correlation prediction of the tofu drying machine process according to an embodiment of the present invention;

[0054] Figure 2 Shown is a flow chart of the inventive method;

[0055] Figure 3 This is a diagram illustrating the data acquisition module;

[0056] Figure 4 This is an illustration of the multivariable collaborative moisture control module for the drying machine head material based on process correlation prediction;

[0057] Figure 5 This is a diagram illustrating the model's self-update module;

[0058] Figure 6 It is a structural schematic diagram of a multivariable collaborative raw material moisture control system based on the correlation prediction of the tofu drying machine process according to an embodiment of the present invention. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0061] Figure 1 FIG. 1 is a flow chart of a multivariable collaborative head material moisture control method based on the correlation prediction of the tofu drying machine process according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0062] S101. Obtain the production data of the front process of the tobacco drying machine, and after eliminating abnormal data, calculate the mean characteristics of the outlet moisture content parameters of the head material part and the whole batch respectively, and write the calculation results into the database, and obtain the historical outlet moisture content benchmark value of the front process from the database, calculate the deviation between the outlet moisture content of the current corresponding batch of the front process and the benchmark value to obtain the basic correction amount of the control parameter of the tobacco drying machine; obtain the tobacco brand information and process parameters at the current production moment to form the current working condition feature vector, use linear fitting to calculate the current outlet moisture content change trend and merge it into the current working condition feature vector, and calculate the optimal control amount setting value under the current working condition based on the pre-trained K-means model in combination with the basic correction amount of the tobacco drying machine control parameter;

[0063] S102. Obtain the difference between the inlet moisture content and the historical average online, calculate the theoretical dehydration amount, and use the adjustable coefficient based on the optimal control amount setting value to correct the control amount of the hot air fan frequency, the dehumidification damper opening, and the material scale flow rate, increase the control frequency, and reduce the control range to stabilize the outlet moisture content; send the corrected control amount to the PLC through the OPC service to realize the coordinated control of the tofu drying machine;

[0064] S103. During the idle period at the end of each production batch, the K-means model update module is started regularly to obtain the production data of the tofu drying machine, select the characteristic parameters related to the moisture content at the outlet of the tofu drying machine based on the minimum mutual information coefficient analysis method, and perform delayed alignment processing on the characteristic parameters to form a feature vector data set with time consistency; select the feature vector preferred data set according to the outlet moisture content deviation, sort the feature vector preferred data set based on the CPK index, obtain the optimal operating condition feature vector data subset, and update the K-means model based on the optimal operating condition feature vector data subset.

[0065] For example, the tobacco industry has conducted a lot of research on the stability of the moisture content of the dried cut tobacco material outlet. The main purpose is to achieve a smoother PID automatic adjustment capability for the moisture content control of the dried cut tobacco material outlet by improving mechanical equipment, optimizing PLC control program segments, stabilizing loose moisture regain and moisture content of the feeding outlet section, etc.

[0066] However, when the parameters such as the incoming material flow rate, incoming material moisture, steam quality, hot air temperature, hot air fan frequency, ambient temperature and humidity in the wire drying section fluctuate greatly, manual adjustment based on experience is still required. In addition, during the wire drying process, there are many factors that affect the moisture content of the head material outlet. These factors are coupled with each other and have strong time lag and uncertainty. The control method based solely on traditional "PID" + "manual experience" cannot meet the requirements of high timeliness, adaptability and high stability.

[0067] Some studies have also started from this point of view, applying machine learning, model predictive control and other methods to the moisture control of the head material of the tobacco drying machine. However, judging from the current implementation results in the tobacco industry, the application effect of the control mode developed by the pure data model still needs to be improved. It is necessary to further study and apply advanced control technology, use process correlation prediction combined with intelligent control methods to improve the control level and effect of the drying head material process, and thus achieve the reduction of dry head amount.

[0068] By implementing a control system based on process correlation prediction for the multivariable collaborative end material moisture control method for tobacco dryers, real-time and effective production control of tobacco dryer end material production can be achieved. After deployment, intelligent, fast and stable control of the moisture content of tobacco end material outlets of tobacco dryers can be achieved, reducing the amount of dry ends, and further realizing rapid adaptive intelligent control of tobacco dryer production process indicators, reducing manual intervention and adjustment, and reducing production costs.

[0069] In an optional embodiment,

[0070] The optimal control quantity setting value under the current working condition is calculated based on the pre-trained K-means model, including:

[0071] Collect the non-controlled process parameters of the tobacco drying machine in real time during production, and obtain the tobacco brand parameters at the current production moment;

[0072] Construct a feature vector including non-controlled process parameters, tobacco brand parameters, decision variables and evaluation indicators; the decision variables include the hot air fan frequency and the output opening of the dehumidification damper, and the evaluation indicator is the deviation between the actual value and the set value of the moisture content at the outlet of the tobacco drying machine;

[0073] The linear fitting method is used to calculate the change trend of the moisture content at the outlet of the current tofu drying machine, and the calculated change trend is merged into the constructed feature vector to form a complete feature vector of the current working condition;

[0074] The current working condition feature vector is input into the pre-trained K-means model for calculation. By calculating the cosine similarity between the current working condition feature vector and the centers of each cluster in the clustering model, the K clusters with the highest similarity are selected, and the average coordinates of their center points are used as the optimal setting values ​​of the control quantities of the hot air fan frequency and the output opening of the moisture exhaust damper under the current working condition.

[0075] In the iterative cycle, the acquired production process parameter data is used for incremental training to update the K-means model, and the parameters of the updated K-means model are written into the database for the next control cycle to call.

[0076] In an optional embodiment,

[0077] Calculating the cosine similarity between the current operating condition feature vector and each cluster center in the clustering model includes:

[0078]

[0079] Among them, A represents the current working condition feature vector, B represents the clustering parameter vector, and n n Represents the number of vectors; A i , B i They represent the i-th current operating condition feature vector and the i-th clustering parameter vector respectively.

[0080] For example, the specific implementation of selecting the optimal operating condition data consists of the following steps:

[0081] 1. Data cleaning. This step removes invalid data and classifies data of different raw material grades (inlet moisture content, material flow, steam flow, cumulative material flow, steam pressure, steam dryness and pre-feeding outlet moisture content of the same batch);

[0082] 2. Data processing. This step conducts statistical tests and MIC mutual information correlation maximum value analysis on multiple batches of production data and combines the expert experience of the actual process flow to determine a group of process characteristics that are highly correlated with the outlet moisture content. Based on the process and delay analysis, the detection delay between the inlet and outlet of each process is determined, and the data delay is aligned to form a feature vector (process parameter) data set with time consistency; then the data is screened, where the difference between the set value of the outlet moisture content of the silk drying machine and the actual value of the outlet moisture content of the silk drying machine is the target value (evaluation index), and the interval from the target value minus 1.5 to the target value plus 1.5 is the threshold range. The feature vector data set is screened to form a feature vector (inlet moisture content, material flow rate, steam flow rate, cumulative material flow rate, steam pressure, steam dryness and the same batch of pre-feeding outlet moisture content) data set with effective control effect.

[0083] 3. Analyze the data set in step 2, examine the data CPK index from the SPC system, and filter the data with CPK greater than the set threshold. Then sort in reverse order according to the CPK size, and take out the optimal operating condition data from the top 20 production batches and save them in the database.

[0084] In an optional embodiment,

[0085] Training a K-means model involves:

[0086] Obtain the characteristic vector of the evaluation index of the deviation between the actual value and the set value of the moisture content at the outlet of the tobacco drying machine in the production process of the tobacco drying machine, including the decision variables of the non-controlled process parameters, tobacco brand parameters, hot air fan frequency and moisture exhaust door output opening, and the evaluation index of the deviation between the actual value and the set value of the moisture content at the outlet of the tobacco drying machine, and arrange the indicators in the characteristic vector in ascending order;

[0087] Calculate the sum of the distances from each data point in the sorted feature vector to the first randomly selected centroid, re-sort all data points according to the sum of the distances, select the data point with the smallest sum of distances as the second centroid, and repeat the process of centroid selection and distance calculation until a preset number of centroids are selected;

[0088] Using all the obtained centroids as cluster centers, each data point in the feature vector is divided into the cluster where the nearest cluster center is located, and the preliminary clustering results are obtained;

[0089] Calculate the average distance between each data point and other data points in its cluster and the average distance between each data point and other cluster centers, thereby calculating the silhouette coefficient of each data point, and evaluate the clustering effect under the current number of clusters through the mean of the silhouette coefficient;

[0090] The clustering effects under different clustering numbers are compared by the mean of the silhouette coefficient to obtain the optimal clustering number. The clustering process is then re-executed with this clustering number until the final clustering result of the process parameters of the tofu drying machine is obtained, completing the training of the K-means model.

[0091] For example, in the real-time production control stage, the module will collect production process parameter information in real time and identify tobacco brand information online, which is recorded as B_1.

[0092] Based on the cigarette brand information B1 at the current moment, the current process parameter value is obtained online to form a process parameter characteristic vector, which is recorded as c_1=(X_1, B_1, Y_1, Z_1), where X_1 is the non-controlled process parameter x_1, x_2, ...x_n, B_1 is the cigarette brand parameter, Y_1 is the decision variable (hot air fan frequency, dehumidification damper output opening, etc.), and Z_1 is the evaluation index (the deviation between the actual outlet moisture content and the set value).

[0093] Use linear fitting to calculate the current outlet moisture change trend ρ, increase or decrease, merge the calculation result with the features in the second step, c_1 = (X_1, B_1, Y_1, Z_(1,)ρ), and convert it into a feature vector after processing. Use the Kmeans algorithm to train the acquired data set and store the results of daily training in the database.

[0094] Load the trained Kmeans model parameters from the database, input the features obtained in the third step into the model for calculation, and obtain the optimal matching values ​​of the hot air fan frequency setting value and other control quantities under the current working conditions.

[0095] In an optional embodiment,

[0096] The clustering effect under the current number of clusters is evaluated by the mean of the silhouette coefficient, including:

[0097]

[0098] Among them, s(x i ) represents the i-th sample x i The silhouette coefficient, a(x i ) represents the sample x i The average distance between all other sample points in the cluster, b(x i ) represents x i The average distance from the center of the non-cluster, N represents the number of samples, q(x i ) represents the sample x i The corresponding center of mass, x j represents the jth sample, and q(k) represents the kth centroid.

[0099] In an optional embodiment,

[0100] Obtain the difference between the inlet moisture content and the historical average online, calculate the theoretical dehydration amount, and use the adjustable coefficient based on the optimal control amount setting value to correct the control amount of the hot air fan frequency, dehumidification damper opening and material scale flow, increase the control frequency and reduce the control range to stabilize the outlet moisture content; send the corrected control amount to the PLC through the OPC service to realize the coordinated control of the tofu drying machine, including:

[0101] Using big data analysis technology, the theoretical dehydration amount under the current working conditions is calculated based on the difference between the moisture content of the current tobacco at the inlet of the tobacco drying machine and the average moisture content of multiple batches of tobacco in history, the difference between the moisture content of the inlet and outlet of the tobacco drying machine, and the adjustable coefficient;

[0102] The moisture content test value at the outlet of the tofu drying machine is compared with the theoretical dehydration calculation value, and the moisture content compensation correction value is calculated. The correction value is multiplied by the dehydration coefficient calculated by multivariate linear regression to obtain the final control value of the hot air fan frequency, the dehumidification damper output opening and the scale set flow rate;

[0103] By increasing the control adjustment frequency and reducing the control amplitude, the stabilization process of the moisture content at the outlet of the tofu drying machine can be accelerated;

[0104] The calculated hot air fan frequency, dehumidification damper output opening and scale set flow control quantity are sent to the PLC program segment through the OPC server to realize automatic control of the tofu drying machine production process.

[0105] For example, the matching degree of the final hot air fan frequency given by the above clustering to the current working condition is investigated, the coordinated control of the dehumidification damper output opening and the scale set flow rate is implemented, and the theoretical dehydration amount of the current working condition information is learned and calculated using big data technology: Where x is the current inlet moisture content, is the average of the inlet moisture content of multiple batches in history, Δx is the difference of the inlet moisture content, Δy is the difference of the outlet moisture content, Δy / Δx is the calculation of the dehydration amount, and λ is the adjustable coefficient. When the tobacco reaches the outlet moisture detection point, the compensation value is calculated based on the inlet moisture difference for correction, and then multiplied by an adjustable coefficient ρ (the dehydration coefficient calculated by multivariate linear regression) to obtain the final hot air fan frequency, dehumidification damper output opening and scale set flow control amount. At the same time, increase the control adjustment frequency and reduce the control amplitude to quickly stabilize the outlet moisture.

[0106] The final hot air fan frequency, dehumidification damper output opening and scale set flow control quantity Y are obtained. The process control is realized by sending the calculated hot air fan frequency, dehumidification damper output opening and scale set flow control quantity Y to the PLC program segment through OPC serve to realize the control of the raw material production process.

[0107] Each batch of data is self-updated and consists of automatic data cleaning, data processing, and characteristic process parameter data clustering. The module is started regularly at the end of each batch of production. After the module is running, it automatically obtains all process data produced by the silk drying process on that day, and automatically performs data cleaning, data processing, model feature vector data set construction and other operations on the acquired data. The current batch data set is trained to obtain the model update and save it.

[0108] The present invention also provides another embodiment,

[0109] The control system constructed and integrated by the method described in the present invention includes three modules, namely a data acquisition module, a process correlation prediction collaborative control module, and a model update module. The three are integrated to form a control system, which is deployed on a server to realize real-time intelligent control of the tofu drying machine. Figure 2 Shown is a flow chart of the inventive method. Figure 3 This is a diagram illustrating the data acquisition module. Figure 4 This is an illustration of the multivariable collaborative moisture control module for the drying machine head material based on process correlation prediction. Figure 5The method of the present invention also includes the analysis of industrial process data in the early stage of system construction and the data processing method of the process correlation prediction model involved.

[0110] The self-adaptive intelligent control method of a silk drying machine based on process-related predictive control of the present invention has a main control loop that obtains raw material characteristic information and outlet moisture content setting value standards and production process parameters issued by the production management system in real time through a data acquisition module, and identifies required parameter information online (parameters include: inlet moisture content, material flow rate, steam flow rate, cumulative material flow rate, steam pressure, steam dryness and pre-feeding outlet moisture content of the same batch), clusters the working conditions using the Kmeans model according to the stage CPK size in production, and then infers and calculates the optimal hot air fan frequency and dehumidification damper output opening control amount at the current moment, performs self-learning correction through feedback of the silk drying machine outlet information, calculates the optimal hot air fan frequency setting value and dehumidification damper output opening, and then performs multivariable collaborative control according to the moisture rising slope and the time to reach the set value, calculates the collaborative optimization regulation of the dehumidification damper opening and the scale set flow rate, and finally performs 0PC Serve communicates with the PLC network inside the tobacco factory and sends the calculated control quantity to the PLC program segment to control the hot air fan frequency, dehumidification damper output opening and scale set flow rate, ensuring that the control quantity in the head material stage adapts to the current working conditions, thereby reducing the amount of dry head generated in the head material stage and reducing costs.

[0111] After the system integrating the method described in the present invention is deployed on the server, the communication between the system and the PLC is established through OPC Serve, and the production stage data is collected in real time. After the optimal working condition data selection, the front process data update, the Kmeans reasoning prediction control quantity, the multivariable collaborative control, the clustering model update and other modules, the control quantity of the hot air fan frequency and the output opening of the dehumidification damper and the scale set flow obtained by reasoning prediction are sent to the PLC program segment for implementation control and the program is further corrected in combination with the multivariable collaborative control. Before and after implementation, the non-steady-state time of the head material part and the dry head loss of the head material are checked. The data before and after implementation are analyzed on a month-on-month basis. The results show that the system integrating the method described in the present invention effectively improves the control quality of the head material part of the moisture content of the drying silk outlet. After the implementation of the present invention, the non-steady-state time of the drying silk is reduced from 151s to 128s, an overall reduction of 15.2%; the non-steady-state dry head amount is reduced from 11.495 to 7.8, a reduction of 32%, and the control effect is significantly improved. Effective intelligent control of the tobacco drying machine's process parameters has been achieved, which has reduced the time of the non-steady-state stage of the tobacco drying head material, reduced tobacco loss, improved tobacco utilization, and increased the company's production efficiency.

[0112] Figure 6 Schematic diagram of the structure of a multivariable collaborative head material moisture control system based on the correlation prediction of the tofu drying machine process according to an embodiment of the present invention. Figure 6 As shown, the system comprises:

[0113] The first unit is used to obtain the production data of the front process of the tobacco drying machine, and after eliminating abnormal data, calculate the mean characteristics of the outlet moisture content parameters of the head material part and the whole batch respectively, and write the calculation results into the database, and obtain the historical front process outlet moisture content benchmark value from the database, calculate the deviation between the outlet moisture content of the current corresponding batch of front process and the benchmark value to obtain the basic correction amount of the control parameter of the tobacco drying machine; obtain the tobacco brand information and process parameters at the current production moment to form the current working condition feature vector, use linear fitting to calculate the current outlet moisture content change trend and merge it into the current working condition feature vector, and calculate the optimal control amount setting value under the current working condition based on the pre-trained K-means model in combination with the basic correction amount of the tobacco drying machine control parameter;

[0114] The second unit is used to obtain the difference between the inlet moisture content and the historical average online, calculate the theoretical dehydration amount, and use the adjustable coefficient based on the optimal control amount setting value to correct the control amount of the hot air fan frequency, dehumidification damper opening and material scale flow, increase the control frequency, and reduce the control range to stabilize the outlet moisture content; send the corrected control amount to the PLC through the OPC service to realize the coordinated control of the tofu drying machine;

[0115] The third unit is used to start the K-means model update module in an idle period at the end of each production batch, obtain the production data of the tofu drying machine, select the characteristic parameters related to the moisture content at the outlet of the tofu drying machine based on the minimum mutual information coefficient analysis method, and perform delayed alignment processing on the characteristic parameters to form a feature vector data set with time consistency; select the feature vector preferred data set according to the outlet moisture content deviation, sort the feature vector preferred data set based on the CPK index, obtain the optimal operating condition feature vector data subset, and update the K-means model based on the optimal operating condition feature vector data subset.

[0116] According to a third aspect of the embodiments of the present invention,

[0117] An electronic device is provided, comprising:

[0118] processor;

[0119] a memory for storing processor-executable instructions;

[0120] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0121] According to a fourth aspect of the embodiments of the present invention,

[0122] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0123] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multivariable collaborative head material moisture control method based on process correlation prediction of a tofu drying machine, characterized in that: include: The production data of the front process of the tobacco drying machine is obtained, and after eliminating abnormal data, the mean characteristics of the outlet moisture content parameters of the head material part and the whole batch are calculated respectively, and the calculation results are written into the database, and the historical benchmark value of the outlet moisture content of the front process is obtained from the database, and the deviation between the outlet moisture content of the front process of the current corresponding batch and the benchmark value is calculated to obtain the basic correction value of the control parameter of the tobacco drying machine; the tobacco brand information and process parameters at the current production moment are obtained to form the current working condition feature vector, and the current outlet moisture content change trend is calculated by linear fitting and merged into the current working condition feature vector, and the optimal control quantity setting value under the current working condition is calculated based on the pre-trained K-means model in combination with the basic correction value of the control parameter of the tobacco drying machine; The difference between the inlet moisture content and the historical average is obtained online to calculate the theoretical dehydration amount. Based on the optimal control amount setting value, the adjustable coefficient is used to correct the control amount of the hot air fan frequency, the dehumidification damper opening and the material scale flow, increase the control frequency and reduce the control range to stabilize the outlet moisture content; The corrected control quantity is sent to PLC through OPC service to realize the coordinated control of the tofu drying machine; In the idle period at the end of each production batch, the K-means model update module is started regularly to obtain the production data of the tofu drying machine. Based on the minimum mutual information coefficient analysis method, the characteristic parameters associated with the moisture content at the outlet of the tofu drying machine are selected, and the characteristic parameters are aligned with time delay to form a feature vector data set with time consistency. The optimal feature vector data set is screened according to the outlet moisture content deviation, the optimal feature vector data set is sorted based on the CPK index, the optimal operating condition feature vector data subset is obtained, and the K-means model is updated based on the optimal operating condition feature vector data subset.

2. The method according to claim 1, characterized in that The optimal control quantity setting value under the current working condition is calculated based on the pre-trained K-means model, including: Collect the non-controlled process parameters of the tobacco drying machine in real time during production, and obtain the tobacco brand parameters at the current production moment; Construct a feature vector including non-controlled process parameters, tobacco brand parameters, decision variables and evaluation indicators; the decision variables include the hot air fan frequency and the output opening of the dehumidification damper, and the evaluation indicator is the deviation between the actual value and the set value of the moisture content at the outlet of the tobacco drying machine; The linear fitting method is used to calculate the change trend of the moisture content at the outlet of the current tofu drying machine, and the calculated change trend is merged into the constructed feature vector to form a complete feature vector of the current working condition; The current working condition feature vector is input into the pre-trained K-means model for calculation. By calculating the cosine similarity between the current working condition feature vector and the centers of each cluster in the clustering model, the K clusters with the highest similarity are selected, and the average coordinates of their center points are used as the optimal setting values ​​of the control quantities of the hot air fan frequency and the output opening of the moisture exhaust damper under the current working condition. In the iterative cycle, the acquired production process parameter data is used for incremental training to update the K-means model, and the parameters of the updated K-means model are written into the database for the next control cycle to call.

3. The method according to claim 2, characterized in that Calculating the cosine similarity between the current operating condition feature vector and each cluster center in the clustering model includes: Where A represents the current operating condition feature vector, B represents the clustering parameter vector, and n represents the number of vectors; A i , B i They represent the i-th current operating condition feature vector and the i-th clustering parameter vector respectively.

4. The method according to claim 3, characterized in that Training a K-means model involves: Obtain the characteristic vector of evaluation indexes in the production process of the tobacco drying machine, including non-controlled process parameters, tobacco brand parameters, decision variables hot air fan frequency and dehumidification damper output opening, and deviation between the actual value and the set value of moisture content at the outlet of the tobacco drying machine, and arrange the indicators in the characteristic vector in ascending order; Calculate the sum of the distances from each data point in the sorted feature vector to the first randomly selected centroid, re-sort all data points according to the sum of the distances, select the data point with the smallest sum of distances as the second centroid, and repeat the process of centroid selection and distance calculation until a preset number of centroids are selected; Using all the obtained centroids as cluster centers, each data point in the feature vector is divided into the cluster where the nearest cluster center is located, and the preliminary clustering results are obtained; Calculate the average distance between each data point and other data points in its cluster and the average distance between each data point and other cluster centers, thereby calculating the silhouette coefficient of each data point, and evaluate the clustering effect under the current number of clusters through the mean of the silhouette coefficient; The clustering effects under different clustering numbers are compared by the mean of the silhouette coefficient to obtain the optimal clustering number. The clustering process is then re-executed with this clustering number until the final clustering result of the process parameters of the tofu drying machine is obtained, completing the training of the K-means model.

5. The method according to claim 4, characterized in that The clustering effect under the current number of clusters is evaluated by the mean of the silhouette coefficient, including: Among them, s(x i ) represents the i-th sample x i The silhouette coefficient, a(x i ) represents the sample x i The average distance between all other sample points in the cluster, b(x i ) represents x i The average distance from the center of the non-cluster, N represents the number of samples, q(x i ) represents the sample x i The corresponding center of mass, x j represents the jth sample, and q(k) represents the kth centroid.

6. The method according to claim 1, characterized in that The difference between the inlet moisture content and the historical average is obtained online, the theoretical dehydration amount is calculated, and the control amount of the hot air fan frequency, the dehumidification damper opening and the material scale flow is corrected using the adjustable coefficient based on the optimal control amount setting value, so as to increase the control frequency and reduce the control range to stabilize the outlet moisture content; The corrected control quantity is sent to the PLC through the OPC service to realize the coordinated control of the tofu drying machine, including: Using big data analysis technology, the theoretical dehydration amount under the current working conditions is calculated based on the difference between the moisture content of the current tobacco at the inlet of the tobacco drying machine and the average moisture content of multiple batches of tobacco in history, the difference between the moisture content of the inlet and outlet of the tobacco drying machine, and the adjustable coefficient; The moisture content test value at the outlet of the tofu drying machine is compared with the theoretical dehydration calculation value, and the moisture content compensation correction value is calculated. The correction value is multiplied by the dehydration coefficient calculated by multivariate linear regression to obtain the final control value of the hot air fan frequency, the dehumidification damper output opening and the scale set flow rate; By increasing the control adjustment frequency and reducing the control amplitude, the stabilization process of the moisture content at the outlet of the tofu drying machine can be accelerated; The calculated hot air fan frequency, dehumidification damper output opening and scale set flow control quantity are sent to the PLC program segment through the 0PC server to realize automatic control of the production process of the tofu drying machine.

7. A multivariable coordinated head material moisture control system based on process correlation prediction of a tofu drying machine, used to implement the method described in any one of claims 1 to 6, characterized in that: include: The first unit is used to obtain the production data of the front process of the tobacco drying machine, and after eliminating abnormal data, calculate the mean characteristics of the outlet moisture content parameters of the head material part and the whole batch respectively, and write the calculation results into the database, and obtain the historical front process outlet moisture content benchmark value from the database, calculate the deviation between the outlet moisture content of the current corresponding batch of front process and the benchmark value to obtain the basic correction amount of the control parameter of the tobacco drying machine; obtain the tobacco brand information and process parameters at the current production moment to form the current working condition feature vector, use linear fitting to calculate the current outlet moisture content change trend and merge it into the current working condition feature vector, and calculate the optimal control amount setting value under the current working condition based on the pre-trained K-means model in combination with the basic correction amount of the tobacco drying machine control parameter; The second unit is used to obtain the difference between the inlet moisture content and the historical average online, calculate the theoretical dehydration amount, and use the adjustable coefficient to correct the control amount of the hot air fan frequency, dehumidification damper opening and material scale flow based on the optimal control amount setting value, increase the control frequency, and reduce the control range to stabilize the outlet moisture content; The corrected control quantity is sent to PLC through OPC service to realize the coordinated control of the tofu drying machine; The third unit is used to start the K-means model update module in the idle period at the end of each production batch, obtain the production data of the tofu drying machine, select the characteristic parameters related to the moisture content of the tofu drying machine outlet based on the minimum mutual information coefficient analysis method, and perform delayed alignment processing on the characteristic parameters to form a feature vector data set with time consistency; The optimal feature vector data set is screened according to the outlet moisture content deviation, the optimal feature vector data set is sorted based on the CPK index, the optimal operating condition feature vector data subset is obtained, and the K-means model is updated based on the optimal operating condition feature vector data subset.

8. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images