A tobacco packing balancing strategy and system
By building a total weight and deviation prediction model, combining BP neural network and random forest algorithm, the tobacco packing plan is adjusted in real time, which solves the problem of prediction deviation that cannot be eliminated in traditional methods and achieves high accuracy and material balance in tobacco packing.
Patent Information
- Application Number
- CN202411478812.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-22
- Publication Date
- 2025-09-05
- Estimated Expiration
- 2044-10-22
AI Technical Summary
In the existing technology, traditional machine learning methods have biases that cannot be eliminated when predicting the weight of tobacco packaging, resulting in the inability to accurately control the weight of the last box of tobacco and effectively solve the problem of material balance during packaging.
By constructing a total weight prediction model and a deviation prediction model, combined with the BP neural network and random forest algorithm, the total weight of the tobacco packing and the predicted deviation value are determined, and the packing plan is adjusted in real time to maintain balance.
It achieves high-precision control of tobacco packing, eliminates forecast deviations, ensures balanced storage of material weight in the tail box, and improves material distribution efficiency and quality stability.
Smart Images
Figure CN119059050B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of material storage technology, and more specifically, to a tobacco packing balancing strategy and system. Background Art
[0002] The tobacco-making process, from the time raw tobacco boxes leave the warehouse to the time finished tobacco is loaded into the cabinet, involves numerous critical stages. This complex, lengthy process leads to numerous points of waste. Lean production, cost reduction, and efficiency improvement are both effective and pressing needs for factories to improve their competitiveness. The packing and warehousing of tobacco leaves in elevated warehouses is a key step in the tobacco-making process. The weight of the final box directly impacts the quality of the finished tobacco. If the final box contains insufficient tobacco, it will dry out, delay its release from storage, and affect material distribution efficiency.
[0003] In existing technology, traditional machine learning methods predict the weight of tobacco boxes in the high-bay warehouse before they are packed. This prediction is then used to balance the weight distribution of the final boxes. However, this existing technology suffers from problems such as inability to eliminate prediction bias and fluctuations in the prediction, which can lead to inaccurate weight control of the final box. Therefore, traditional machine learning methods cannot effectively address the issue of balanced material storage during packing. Summary of the Invention
[0004] The present application provides a tobacco packing balance strategy and system, which determines the tobacco packing plan by considering the total weight of the tobacco packing and the predicted deviation value of the tobacco, so that the predicted deviation is taken into account during packing, and the packing plan is adjusted in real time when the predicted deviation fluctuates to maintain packing balance.
[0005] This application provides a tobacco packing balancing strategy, including:
[0006] Determine the total weight of the cut tobacco obtained from the current batch of cut tobacco through the cutting process;
[0007] Obtaining a predicted deviation value of the first cut tobacco of the current batch of cut tobacco;
[0008] The tobacco packing plan for the current batch of cut tobacco is determined based on the total weight of the cut tobacco packing and the predicted deviation value of the first cut tobacco.
[0009] Preferably, the total weight prediction model is used to determine the total weight of the tobacco cuts obtained from the current batch of cut tobacco through the cutting process.
[0010] Preferably, obtaining the first tobacco cut prediction deviation value of the current batch of cut tobacco specifically includes:
[0011] Obtain the predicted deviation values of the silk threads of the current batch of silk threads of all belt scales before the last stage in the silk-making process;
[0012] Obtaining a second predicted deviation value of cut tobacco corresponding to the total weight of the cut tobacco box;
[0013] The first cut tobacco prediction deviation value is determined based on all the cut tobacco prediction deviation values and the second cut tobacco prediction deviation value.
[0014] Preferably, the maximum value among all the cut tobacco prediction deviation values and the second cut tobacco prediction deviation values is used as the first cut tobacco prediction deviation value.
[0015] Preferably, each belt scale corresponds to a leaf thread prediction model, and the absolute value of the difference between the predicted value of the leaf thread prediction model of each belt scale and the corresponding actual value is used as the leaf thread prediction deviation value of the belt scale.
[0016] The present application also provides a tobacco packing balancing system, comprising a total weight determination module, a first deviation acquisition module, and a solution determination module;
[0017] The total weight determination module is used to determine the total weight of the tobacco cuts obtained by the current batch of tobacco cuts through the tobacco cutting process;
[0018] The first deviation obtaining module is used to obtain the first tobacco cut tobacco prediction deviation value of the current batch of cut tobacco;
[0019] The scheme determination module is used to determine the tobacco packing scheme for the current batch of tobacco leaves based on the total weight of the tobacco packing and the first tobacco cut prediction deviation value.
[0020] Preferably, the total weight determination module uses a total weight prediction model to determine the total weight of the tobacco cuts obtained from the current batch of cut tobacco through the cutting process.
[0021] Preferably, the first deviation obtaining module includes a second deviation obtaining module, a third deviation obtaining module and a deviation value determining module;
[0022] The second deviation obtaining module is used to obtain the predicted deviation values of the leaf silk of all belt scales before the last section in the silk making process of the current batch of leaf silk;
[0023] The third deviation obtaining module is used to obtain a second cut tobacco prediction deviation value corresponding to the total weight of the cut tobacco box;
[0024] The deviation value determination module is used to determine the first cut tobacco prediction deviation value based on all cut tobacco prediction deviation values and the second cut tobacco prediction deviation value.
[0025] Preferably, the deviation value determination module is configured to take the maximum value among all the cut tobacco prediction deviation values and the second cut tobacco prediction deviation values as the first cut tobacco prediction deviation value.
[0026] Preferably, the second deviation obtaining module is used to use the absolute value of the difference between the predicted value of the blade prediction model of each work section and the corresponding actual value as the blade prediction deviation value of the belt scale.
[0027] Other features and advantages of the present application will become apparent from the following detailed description of exemplary embodiments of the present application with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the application and, together with the description, serve to explain the principles of the application.
[0029] Figure 1 A flow chart of the tobacco packing balancing strategy provided in this application;
[0030] Figure 2 An example of a silk making process flow and its input data provided for this application;
[0031] Figure 3 A schematic diagram of the total weight prediction model provided for this application;
[0032] Figure 4 An example of obtaining the predicted deviation value of blade silk for all belt scales provided in this application;
[0033] Figure 5 This is a structural diagram of the tobacco packing and balancing system provided in this application. DETAILED DESCRIPTION
[0034] Various exemplary embodiments of the present application will now be described in detail with reference to the accompanying drawings. It should be noted that unless otherwise specifically stated, the relative arrangements of components and steps, numerical expressions and numerical values set forth in these embodiments do not limit the scope of the present application.
[0035] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present disclosure, its application, or uses.
[0036] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.
[0037] In all examples shown and discussed herein, any specific values should be interpreted as merely exemplary and not limiting. Therefore, other examples of the exemplary embodiments may have different values.
[0038] The present application provides a tobacco packing balance strategy and system, which determines the tobacco packing plan by considering the total weight of the tobacco packing and the predicted deviation value of the tobacco, so that the predicted deviation is taken into account during packing, and the packing plan is adjusted in real time when the predicted deviation fluctuates to maintain packing balance.
[0039] like Figure 1 As shown, the tobacco packing balancing strategy provided by this application includes:
[0040] S110: Determine the total weight of the cut tobacco obtained by the current batch of cut tobacco through the cut tobacco processing process.
[0041] As an embodiment, the total weight prediction model is used to determine the total weight of the tobacco cuts obtained by the current batch of tobacco cuts through the tobacco cutting process.
[0042] Specifically, when building a total weight prediction model, we first determine the model's input data by drawing a tobacco production process flow chart. Then, we identify some input data within the process flow chart, including factors related to material weight at each stage and the e-weight of each belt scale in the process, representing the total weight of the tobacco package. Input data also includes material brand names, etc.
[0043] Among them, the factors related to material weight mainly include material moisture measurement point, temperature measurement point, storage cabinet time, temperature and humidity, etc.
[0044] Figure 2 An example of a silk-making process and its input data is given. This example includes a five-stage production line. Four belt scales are located before the last stage, measuring the material weight: belt scale 1, belt scale 2, belt scale 3, and belt scale 4. Belt scale 5, located after the last stage, measures the cumulative weight of all packaged materials. The factor related to the material weight before belt scale 1 is defined as X. 1N Variable group; the factor related to the material weight after belt scale 1 and before belt scale 2 is defined as X 2N Variable group; and so on, define X 3N Variable group, X 4N Variable group and X 5N Variable group. The variables contained in each variable group can be deleted or added based on the actual situation on site.
[0045] In this example, for X 1N Variable group, X 11 The variable is the cumulative amount of water added to loosen and moisten leaves; X 12 The variable is the cumulative value of the scale before loosening and moistening the leaves; X 13 The variable is the outlet temperature of loose leaf; X 14 The variable is the moisture value of loose leaf outlet; X 15 The variable is the moisture value of the leaf before cutting; X 16The variable is the storage time; X 17 The variable is the cabinet change time; X 18 The variable is atmospheric temperature; X 19 The variable is atmospheric humidity.
[0046] For X 2N Variable group, X 21 The variable is the cumulative amount of material and water added once; X 22 The variable is the accumulated value before adding material once; X 23 The variable is the outlet temperature of the primary feed; X 24 The variable is the moisture value of the primary feed outlet; X 25 The variable is the moisture value of the primary feed inlet; X 26 The variable is the storage time; X 27 The variable is the cabinet change time from the pre-distribution cabinet to the distribution cabinet; X 28 The variable is atmospheric temperature; X 29 The variable is atmospheric humidity.
[0047] X 3N Variable group, X 4N Variable group and X 5N The factor selection method of the variable group is consistent.
[0048] Figure 2 In the figure, the weighing weight (i.e., cumulative weight) of each belt scale is recorded as Y1, Y2, Y3, Y4, and Y5 respectively.
[0049] Among them, X 1N Variable group, X 2N Variable group, X 3N Variable group, X 4N Variable group, X 5N The variable group, Y1, Y2, Y3, and Y4, are the input data of the total weight prediction model. These data are known data. The predicted value X of Y5 is the output data of the total weight prediction model.
[0050] After determining the input and output data of the total weight prediction model, as an embodiment, a total weight prediction model G1 is established based on the BP neural network algorithm, such as Figure 3 shown.
[0051] As an embodiment, for the parameters of the BP neural network model, if the number of input layer variables is less than 20, the number of hidden layers is set to 2 and the learning rate is assigned to 0.08; if the number of input layer variables is less than 50 and more than 20, the number of hidden layers is set to 3 and the learning rate is assigned to 0.12; if the number of input layer variables is greater than 50, the number of hidden layers is set to 4 and the learning rate is assigned to 0.15.
[0052] Before training the gross weight prediction model G1, a preset number of batches (e.g., 1000) of historical data are exported from the historical database, excluding data from production failure batches and test production batches. This historical data is then preprocessed to form a batch-based data format. The preprocessed historical data is then divided into a training set and a validation set according to a preset ratio. Finally, the model is trained using the training and validation sets to obtain the trained gross weight prediction model G1.
[0053] In this step, the X 1N Variable group, X 2N Variable group, X 3N Variable group, X 4N Variable group, X 5N The variable group, Y1, Y2, Y3, and Y4, are input into the total weight prediction model G1 to obtain the corresponding total weight X of the tobacco packaging.
[0054] S120: Obtain the first tobacco cut prediction deviation value of the current batch of cut tobacco.
[0055] As an embodiment, obtaining the first cut tobacco prediction deviation value of the current batch of cut tobacco includes:
[0056] S1201: Obtain the predicted deviation values of the silk threads of the current batch of silk threads of all belt scales before the last stage in the silk making process.
[0057] like Figure 2 As shown, a belt scale is provided after each work section, and before the last work section, each belt scale corresponds to a leaf silk prediction model, and the absolute value of the difference between the predicted value of the leaf silk prediction model of each belt scale and the corresponding actual value is used as the leaf silk prediction deviation value of the belt scale.
[0058] As an embodiment, a random forest algorithm is used to construct a leaf silk prediction model.
[0059] like Figure 4 As shown, in Figure 2 In the example shown, for the belt scale 1, the blade prediction model G 21 The input data is material brand and X 1N The variable group, the model output is the predicted value of the weighing weight Y1 of belt scale 1. For belt scale 2, the leaf prediction model G 22 The input data is material brand, X 1N Variable group, belt scale 1 weighing weight Y1, X 2N The variable group, the model output is the predicted value of the weighing weight Y2 of belt scale 2. For belt scale 3, the leaf prediction model G 23 The input data is material brand, X 1N Variable group, belt scale 1 weighing weight Y1, X2N Variable group, belt scale 2 weighing weight Y2, X 3N The variable group, the model output is the predicted value of the weighing weight Y3 of belt scale 3. For belt scale 4, the leaf prediction model G 24 The input data is material brand, X 1N Variable group, belt scale 1 weighing weight Y1, X 2N Variable group, belt scale 2 weighing weight Y2, X 3N Variable group, belt scale 3 weighing weight Y3, X 4N Variable group, the model output is the predicted value of the weighing weight Y4 of belt scale 4.
[0060] The above pre-processed historical data is used to train the leaf silk prediction model of each belt scale.
[0061] In this step, the relevant input data of the current batch of leaf silk are input into each leaf silk prediction model respectively, and the corresponding prediction value A is obtained respectively. 21 、A 22 、A 23 、A 24 , the predicted value A 21 、A 22 、A 23 、A 24 The corresponding actual weighing weight B 21 、B 22 、B 23 、B 24 The absolute value of the difference is used as the blade prediction deviation value C of each belt scale 21 、C 22 、C 23 、C 24 :
[0062] C 21 =|B 21 -A 21 |
[0063] C 22 =|B 22 -A 22 |
[0064] C 23 =|B 23 -A 23 |
[0065] C 24 =|B 24 -A 24 |
[0066] S1202: Obtain a second cut tobacco prediction deviation value corresponding to the total weight of the cut tobacco box.
[0067] As an embodiment, the deviation prediction model G3 is used to obtain the second cut tobacco prediction deviation value. As an embodiment, the deviation prediction model G3 is constructed using a random forest algorithm.
[0068] To construct the deviation prediction model G3, first, input data from a preset number (e.g., 200) batches of recently produced tobacco is fed into the total weight prediction model G1 to obtain a predicted value X. The absolute value of the difference between the predicted value and the actual value is used as the predicted deviation value for the total tobacco package weight, which serves as the output data of the deviation prediction model. The dataset formed by the recently produced input data and the corresponding predicted deviation values serves as the training set for training the deviation prediction model G3.
[0069] Input the input data of the current batch of tobacco shreds into the deviation prediction model G3 to obtain the second tobacco shreds prediction deviation value C of the current batch of tobacco shreds 25 .
[0070] S1203: Determine a first cut tobacco prediction deviation value W based on all cut tobacco prediction deviation values and the second cut tobacco prediction deviation value.
[0071] As an embodiment, the maximum value among all the predicted deviation values of the cut tobacco and the second cut tobacco predicted deviation values is used as the first cut tobacco predicted deviation value, that is, the maximum value function is used to obtain the first cut tobacco predicted deviation value W=MAX(C 21 ,C 22 ,C 23 ,C 24 ,C 25 ).
[0072] S130: Determine a tobacco packing plan for the current batch of cut tobacco based on the total tobacco packing weight X and the first cut tobacco prediction deviation value W.
[0073] From the above, we can see that the range of the predicted total weight of tobacco packaging is X±W.
[0074] In the tobacco packing plan, try to make the loading weight of the last box greater than 1 / 2 of the full box loading weight, and try to avoid the situation where there is very little material. In addition, make sure that the rear box before the last box is as full as possible, but it does not have to be full. The loading weight of the rear box can be set to be greater than 3 / 4 of the full box loading weight. The numerical value f of the rear box can be set to a fixed value based on the actual process conditions, or a range value can be set and automatically determined by the system. The f value should not be too large or too small. Generally, the suitable value range of f value is 2-5. The box before the rear box should be full, and the full box loading weight of each box is set to the process specified value q.
[0075] As an embodiment, when determining the tobacco packing plan for the current batch of shredded tobacco, the total number of boxes n, the number f of rear boxes before the last box, and the loading weight of the rear boxes and the last box are determined based on the relationship between the first tobacco prediction deviation value W, the total weight X of the tobacco packing and the process specified value q.
[0076] Preferably, if the first tobacco prediction deviation value W is greater than the threshold determined by the total weight X of the tobacco box, it is also determined whether to adjust the number f of the rear boxes, the loading weight of the rear boxes and the last box based on the relationship between the loading weight of the rear boxes and the total weight X of the tobacco box.
[0077] An example of this preferred embodiment is given below to further illustrate the process of determining the above-mentioned tobacco packing solution.
[0078] (1) If There are two situations:
[0079] (1-1) If but The loading weight of the rear box of the f box is The predicted loading weight center value of the last box is (qw). If the prediction deviation value ±W is taken into account, the predicted loading weight range of the last box is between (q-2w) and q.
[0080] Among them, Round(x) is the rounding function.
[0081] (1-2) If or or but Finally, the loading weight of the rear box of box f is The predicted loading weight center value of the last box is (qw). If the prediction deviation value ±W is taken into account, the predicted loading weight range of the last box is between (q-2w) and q.
[0082] (2) If There are two situations:
[0083] (2-1) If but The loading weight of the rear box of the f box is There are two cases according to the V value:
[0084] (2-1-1) If The f value remains unchanged, and the center value of the predicted loading weight of the last box is (qw). If the prediction deviation value ±W is taken into account, the predicted loading weight range of the last box is between (q-2w) and q.
[0085] (2-1-2) If Then the f value is added by 1, that is, f=f+1, and the formula Recalculate the V value until it meets So far, the predicted loading weight center value of the last box is (qw). If the predicted deviation value ±W is taken into account, the predicted loading weight range of the last box is between (q-2w) and q.
[0086] (2-2) If or or but The loading weight of the rear box of the f box is There are two cases according to the V value:
[0087] (2-2-1) If The f value remains unchanged, and the center value of the predicted loading weight of the last box is (qw). If the prediction deviation value ±W is taken into account, the predicted loading weight range of the last box is between (q-2w) and q.
[0088] (2-2-2) If Then the f value is added by 1, that is, f=f+1, and the formula Recalculate the V value until it meets So far, the predicted loading weight center value of the last box is (qw). If the predicted deviation value ±W is taken into account, the predicted loading weight range of the last box is between (q-2w) and q.
[0089] (3) If There is no need to divide it into two cases. It can be inferred that it must satisfy or or but At the same time, the last box is limited to The weight of the rear box of the f box is preset. There are two cases according to the V value:
[0090] (3-1) If The f value remains unchanged, and the predicted loading weight center value of the last box is If we consider the existence of prediction deviation The predicted loading weight of the last box ranges from 0 to q.
[0091] (3-2) If Then the f value is added by 1, that is, f=f+1, and the formula Recalculate the V value until it meets So far, the predicted loading weight center value of the last box is If we consider the existence of prediction deviation The predicted loading weight of the last box is between 0 and q, and there is a very small probability that the last box will be loaded less than materials.
[0092] From (3-1) and (3-2), it can be concluded that since the method for obtaining the above-mentioned first tobacco prediction deviation value W conforms to the normal distribution or is similar to the normal distribution, it can be deduced that although the numerical range of the predicted loading weight of the last box is between 0 and q, most of the data are normally distributed or similar to the normal distribution with 2q as the center line, and the vast majority of the data are close to the center value 2q, rather than close to 0 or q, which ensures that the last box of material is almost in a half-box state.
[0093] The following three examples illustrate the tobacco packing solution using actual data:
[0094] Assume that the process specified value q = 200 kg and set f = 2.
[0095] Example 1: Assuming X = 8680kg, W = 50kg, the predicted total weight range of tobacco packaging is 8680±50kg, that is, 8630kg~8730kg, which meets Conditions, and also meet Conditions, using the above (1-1) to balance the packing materials, then The loading weight of the rear box of box 2 is The center value of the predicted loading weight of the last box is 150kg. Taking into account the predicted deviation value of ±50kg, the predicted loading weight range of the last box is between 100kg and 200kg.
[0096] Example 2: Assuming X = 8390kg, W = 80kg, the predicted total weight range of tobacco packaging is 8490±80kg, that is, 8410kg~8570kg, which meets Conditions, and also meet Conditions, using the above (2-1) to balance the packing materials, then
[0097] The loading weight of the rear box of box 2 is Then it satisfies Conditions, use the above (2-1-2) to balance the packing materials, then add 1 to the f value, that is, f=f+1=3, and recalculate the V value using the formula, satisfy The loading weight of the rear box of the three boxes is 163kg, and the predicted loading weight center value of the last box is 120kg. If the predicted deviation value of ±80kg is taken into account, the predicted loading weight range of the last box is between 40kg and 200kg.
[0098] Example 3: Assuming X = 8370kg, W = 110kg, the predicted total weight range of tobacco packaging is 8370±110kg, that is, 8260kg~8480kg, which meets Conditions, while meeting or or Conditions, using the above (3) to carry out packing material balance, then
[0099]
[0100] The loading weight of the rear box of box 2 is Then it satisfies Conditions, use the above (3-2) to balance the packing materials, then add 1 to the f value, that is, f=f+1=3, and recalculate the V value using the formula, Dissatisfied Then add 1 to the f value, that is, f=f+1=4, and recalculate using the formula satisfy The load weight of the rear box of the 4 boxes is 167.5kg, and the predicted load weight center value of the last box is 100kg. Considering that the predicted deviation value is 110kg greater than The predicted loading weight of the last box is between 0kg and 200kg, and the weight data meets the following requirements: For a normal distribution or quasi-normal distribution with a central value, there is a very small probability that the last box will be loaded with less than 50 kg of materials.
[0101] Based on the above, the present application also provides a tobacco packing balance system. Figure 5 As shown, it includes a total weight determination module 510 , a first deviation acquisition module 520 and a solution determination module 530 .
[0102] The total weight determination module 510 is used to determine the total weight of the cut tobacco obtained by the current batch of cut tobacco through the cut tobacco processing process.
[0103] The first deviation obtaining module 520 is used to obtain the first predicted deviation value of the current batch of shredded tobacco.
[0104] The solution determination module 530 is used to determine the tobacco packing solution for the current batch of cut tobacco based on the total weight of the cut tobacco packing and the first cut tobacco prediction deviation value.
[0105] Preferably, the total weight determination module 510 uses a total weight prediction model to determine the total weight of the cut tobacco obtained by the current batch of cut tobacco through the cut tobacco making process.
[0106] Preferably, the first deviation obtaining module 520 includes a second deviation obtaining module 5201 , a third deviation obtaining module 5202 and a deviation value determining module 5203 .
[0107] The second deviation obtaining module 5201 is used to obtain the predicted deviation values of the silk leaves of all belt scales before the last working section in the silk making process of the current batch of silk leaves.
[0108] The third deviation obtaining module 5202 is used to obtain a second cut tobacco prediction deviation value corresponding to the total weight of the cut tobacco package.
[0109] The deviation value determination module 5203 is used to determine the first cut tobacco prediction deviation value based on all cut tobacco prediction deviation values and the second cut tobacco prediction deviation value.
[0110] Preferably, the deviation value determination module 5203 is configured to take the maximum value of all the cut tobacco prediction deviation values and the second cut tobacco prediction deviation values as the first cut tobacco prediction deviation value.
[0111] Preferably, the second deviation obtaining module 5201 is used to use the absolute value of the difference between the predicted value of the blade silk prediction model of each work section and the corresponding actual value as the blade silk prediction deviation value of the belt scale.
[0112] Preferably, the solution determination module 530 is used to determine the total number of boxes, the number of rear boxes before the last box, and the loading weight of the rear boxes and the last box based on the relationship between the first tobacco prediction deviation value, the total weight of the tobacco box and the process specified value.
[0113] Preferably, the scheme determination module 530 is also used to determine whether to adjust the number of rear boxes, the loading weight of the rear boxes and the last box based on the relationship between the loading weight of the rear boxes and the total weight of the tobacco boxes when the first tobacco prediction deviation value is greater than a threshold determined by the total weight of the tobacco boxes.
[0114] The beneficial effects of this application are as follows:
[0115] 1. This application eliminates the problems of upper limit of prediction accuracy and inability to eliminate prediction bias when predicting the total weight of tobacco packaging based on traditional machine learning methods through a combination of multiple models. It also solves the problem that the bias fluctuation in traditional machine learning methods causes the weight of the last box of tobacco to be unable to be accurately controlled.
[0116] 2. This application uses a BP neural network model to predict the total weight of tobacco packaging. The model is fitted based on five groups of variables related to material weight factors at the front end and the weights of four front-end belt scales as neural network inputs, and is based on reverse transmission to continuously reduce the prediction error, ultimately improving the accuracy of the predicted center value.
[0117] 3. This application uses a random forest model algorithm to fit the tobacco leaf prediction model of the front-end belt scale, and simultaneously uses the random forest model algorithm and new data to fit the deviation prediction model. It uses multiple dimensions and multiple sets of new and old data for cross-training to improve the accuracy of the predicted data coverage and error range. This application is based on the five predicted deviation values fitted by the random forest algorithm and then takes the maximum value. This ensures that the first tobacco leaf prediction deviation value covers the maximum possible range of possible fluctuations in the value, but it is also a reliable conclusion based on the model and data dimensions to assist in the development of mathematical methods in the weight balance storage strategy of the tail box material.
[0118] 4. This application uses three major scenarios and three levels of conditions to determine the tobacco packing plan. First, the three relationships between w and q are used for judgment. Then, the relationship between three rounding functions is used for judgment. Finally, the relationship between V and q is used to determine whether the f value needs to be adjusted. This optimized design method, based on these three levels of judgment, can achieve a highly accurate weight-balanced storage strategy for the final box of the packing material.
[0119] Although some specific embodiments of the present application have been described in detail by way of example, it will be understood by those skilled in the art that the above examples are for illustration only and are not intended to limit the scope of the present application. It will be understood by those skilled in the art that modifications may be made to the above embodiments without departing from the scope and spirit of the present application. The scope of the present application is defined by the appended claims.
Claims
1. A method for balancing the packing of cut tobacco, characterized in that: include: The total weight prediction model is used to determine the predicted value of the total weight of the tobacco box obtained by the current batch of tobacco leaves through the tobacco making process; Obtaining the first tobacco cutlet prediction deviation value of the current batch of tobacco cutlets, including: obtaining the tobacco cutlet prediction deviation values of all belt scales before the last work section in the tobacco cutting process of the current batch of tobacco cutlets; obtaining the second tobacco cutlet prediction deviation value corresponding to the predicted value of the total weight of the tobacco cutlet packaging; determining the first tobacco cutlet prediction deviation value based on all the tobacco cutlet prediction deviation values and the second tobacco cutlet prediction deviation value; wherein, the method for obtaining the tobacco cutlet prediction deviation value includes: each belt scale corresponds to a tobacco cutlet prediction model, and the absolute value of the difference between the predicted value of the tobacco cutlet prediction model of each belt scale and the corresponding actual value is used as the tobacco cutlet prediction deviation value of the belt scale; the second tobacco cutlet prediction deviation value is the absolute value of the difference between the predicted value of the total weight of the tobacco cutlet packaging and the corresponding actual value; A tobacco packing plan for the current batch of cut tobacco leaves is determined based on the predicted value of the total weight of the cut tobacco packing and the first cut tobacco prediction deviation value.
2. The method for balancing the packing of cut tobacco according to claim 1, characterized in that: The maximum value among all the cut tobacco prediction deviation values and the second cut tobacco prediction deviation value is used as the first cut tobacco prediction deviation value.
3. A tobacco packing balancing system, characterized in that: It includes a total weight determination module, a first deviation acquisition module and a solution determination module; The total weight determination module is used to determine the predicted value of the total weight of the cut tobacco box obtained by the current batch of cut tobacco through the cut tobacco processing process using the total weight prediction model; The first deviation obtaining module is used to obtain the first cut tobacco prediction deviation value of the current batch of cut tobacco; The first deviation obtaining module specifically includes: a second deviation obtaining module, a third deviation obtaining module and a deviation value determining module; The second deviation obtaining module is used to obtain the leaf silk prediction deviation values of all belt scales before the last work section in the silk making process of the current batch of leaf silk; wherein the leaf silk prediction deviation value is obtained in the following manner: each belt scale corresponds to a leaf silk prediction model, and the absolute value of the difference between the predicted value of the leaf silk prediction model of each belt scale and the corresponding actual value is used as the leaf silk prediction deviation value of the belt scale; The third deviation obtaining module is used to obtain a second cut tobacco prediction deviation value corresponding to the predicted value of the cut tobacco packing total weight; the second cut tobacco prediction deviation value is the absolute value of the difference between the predicted value of the cut tobacco packing total weight and the corresponding actual value; The deviation value determination module is used to determine the first cut tobacco prediction deviation value based on all cut tobacco prediction deviation values and the second cut tobacco prediction deviation value; The scheme determination module is used to determine the tobacco packing scheme of the current batch of cut tobacco leaves based on the predicted value of the total weight of the cut tobacco packing and the first cut tobacco prediction deviation value.
4. The tobacco packing balancing system according to claim 3, characterized in that: The deviation value determination module is configured to take the maximum value of all the cut tobacco prediction deviation values and the second cut tobacco prediction deviation value as the first cut tobacco prediction deviation value.
Citation Information
Patent Citations
Differential boxing method
CN102145759A
Method for judging uniformity of tobacco shred blending process
CN112155242A