Reconstituted tobacco sensory quality score determination method and system based on absolute dry quantification before coating

Abstract

The invention relates to the technical field of computer deep learning, in particular to a reconstituted tobacco sensory quality score determination method and system based on absolute dry quantification before coating. The method comprises the following steps: acquiring absolute dry quantitation of to-be-detected reconstituted tobacco before coating; inputting the absolute dry quantity before coating into a trained sensory quality scoring model; and obtaining a sensory quality score prediction value which is output by the sensory quality score model and is calculated based on the absolute dry quantification before coating, and determining the sensory quality score prediction value as a sensory quality score value corresponding to the to-be-detected reconstituted tobacco. The invention aims to solve the problem of how to improve the product stability in the paper-making reconstituted tobacco production process.

Description

Technical Field

[0001] The present application relates to the field of computer deep learning technology, and in particular to a method and system for determining a sensory quality score of reconstituted tobacco leaves based on absolute dry weight before coating. Background Art

[0002] Reconstituted tobacco is an important part of the tobacco industry, and its quality directly affects the quality of tobacco products. The national standard GB 5606.4-2005 "Cigarettes Part 4: Sensory Technical Requirements" stipulates several rules for judging the sensory quality of cigarettes, and evaluates the sensory quality of cigarettes from aspects such as the appearance, smoke, and aroma of tobacco leaves.

[0003] However, manual smoking evaluation usually involves sampling several cigarettes from a production batch, testing the parameters related to the sensory quality of the cigarettes using chemical experiments, and then combining the subjective judgment of professional smoking testers on the appearance, smoke, aroma and other aspects of the tobacco leaves to score the sensory quality of the cigarettes. If the score is unqualified, all cigarettes in the production batch will be judged as unqualified.

[0004] Related technical solution 1, Chinese patent publication number CN117422330A discloses a method for evaluating the quality of ordinary cigarette tobacco, which discloses the following indicators for evaluating the quality of tobacco: 1. Leaf quality indicators (including leaf moisture content, leaf filling value, leaf whole rate and leaf broken rate); 2. Cigarette physical indicators (including cigarette unit weight, cigarette unit weight standard deviation, cigarette draw resistance, cigarette draw resistance standard deviation, cigarette filter ventilation rate and cigarette filter ventilation rate standard deviation); 3. Smoke indicators (including tar release extremes, carbon monoxide release extremes and smoke nicotine release extremes); 4. Sensory indicators (including gloss, aroma, harmony, miscellaneous smell, irritation and aftertaste). The score of each indicator corresponds to a different weight. The final comprehensive evaluation score of tobacco is obtained by multiplying each indicator by its own weight and adding them together.

[0005] Related technical solution 2, Chinese patent publication number CN110458474A discloses a comprehensive evaluation method for the physical quality of cigarette rolling, which discloses the following 7 indicators for evaluating the quality of cut tobacco: support weight, draw resistance, hardness, ventilation rate, end cut tobacco, dust content and empty head rate. Through data collection, data standardization processing, and principal component analysis of standardized data, a batch of cigarette comprehensive physical quality evaluation model is established, and the cigarette rolling quality is quantitatively scored through the evaluation model.

[0006] However, the above-mentioned related technical solutions all focus on analyzing the indicators of tobacco itself, and do not involve the research on the indicators and / or process steps that are correlated with the sensory quality of reconstituted tobacco in the preparation process of cigarettes, especially the preparation process of reconstituted tobacco.

[0007] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0008] The main purpose of the present application is to provide a method for determining the sensory quality score of reconstituted tobacco based on the absolute dry weight before coating, aiming to solve the problem of how to improve the product stability in the papermaking process of reconstituted tobacco production.

[0009] To achieve the above-mentioned purpose, the present application provides a method for determining the sensory quality score of reconstituted tobacco leaves based on absolute dry weight before coating, the method comprising:

[0010] Obtaining the absolute dry weight of the reconstituted tobacco leaves to be tested before coating;

[0011] Inputting the absolute dry weight before coating into the trained sensory quality scoring model;

[0012] The predicted sensory quality score value output by the sensory quality score model and calculated based on the absolute dry weight before coating is obtained, and the predicted sensory quality score value is determined as the sensory quality score value corresponding to the reconstituted tobacco leaf to be tested.

[0013] Optionally, the trained sensory quality scoring model includes a sensory quality scoring calculation module for absolute dry weight before coating, and the training process of the sensory quality scoring model includes:

[0014] Obtain the absolute dry quantitative sample data set before coating;

[0015] The absolute dry quantitative data before coating is enhanced.

[0016] Based on the data-enhanced absolute dry quantitative sample data set before coating and the corresponding sensory quality score data set, a correlation data table is constructed;

[0017] Input the associated data table into the model with lightGBM architecture for training, and calculate the AUC value of the model after each training is completed;

[0018] After the AUC value reaches the preset AUC threshold, the training is completed.

[0019] Optionally, before the step of inputting the associated data table into a model with a lightGBM architecture for training and calculating the AUC value of the model after each training is completed, the step further includes:

[0020] Oversampling the abnormal absolute dry weight samples before coating that are not in the standard absolute dry weight interval before coating in the associated data table, and undersampling the normal absolute dry weight samples before coating that are in the standard absolute dry weight interval before coating, so as to balance the number of samples between the abnormal absolute dry weight samples before coating and the normal absolute dry weight samples before coating;

[0021] The step of inputting the associated data table into a model with a lightGBM architecture for training, and calculating the AUC value of the model after each training is completed includes:

[0022] The balanced association data table is input into the model with lightGBM architecture for training.

[0023] Optionally, the model adjusts the maximum depth and the number of leaf nodes in the model hyperparameters during training.

[0024] Optionally, after the step of obtaining the predicted sensory quality score value output by the sensory quality score model and calculated based on the absolute dry weight before coating, and determining the predicted sensory quality score value as the sensory quality score value corresponding to the reconstituted tobacco to be tested, the step further includes:

[0025] During the process of executing the papermaking method reconstituted tobacco process, obtaining the sensory quality score value associated with the reconstituted tobacco in the current production line at every preset period;

[0026] Determine whether the sensory quality score is within a normal value range;

[0027] When it is detected that the sensory quality score value is not in the normal value range, determining whether the absolute dry weight before coating in the current production line is in the standard absolute dry weight before coating range;

[0028] If not, the abnormal target absolute dry quantitative process equipment parameters before coating are determined and output among the absolute dry quantitative process equipment parameters before coating.

[0029] Optionally, the abnormal target absolute dry quantitative process equipment parameter before coating includes at least one of sizing amount, retention rate, and base uniformity.

[0030] In addition, to achieve the above-mentioned purpose, the present application also provides a papermaking method reconstituted tobacco system, characterized in that the papermaking method reconstituted tobacco system includes: a memory, a processor, and a program for determining the sensory quality score of reconstituted tobacco based on the absolute dry weight before coating, which is stored on the memory and can be run on the processor, and when the program for determining the sensory quality score of reconstituted tobacco based on the absolute dry weight before coating is executed by the processor, the steps of the method for determining the sensory quality score of reconstituted tobacco based on the absolute dry weight before coating as described above are implemented.

[0031] This application has at least the following beneficial effects:

[0032] 1. The process parameters of reconstituted tobacco with strong correlation with sensory quality scores in the production process of papermaking reconstituted tobacco were explored, as well as the equipment parameters associated with the process parameters of reconstituted tobacco, so as to achieve the correlation between the sensory quality scores and the process parameters of reconstituted tobacco;

[0033] 2. Convert the sensory quality score obtained by traditional manual smoking evaluation into a real-time calculation by the system based on process parameters, so as to realize the efficient application of sensory quality score in the industrial production of papermaking reconstituted tobacco leaves;

[0034] 3. Through the collection of reconstituted tobacco process parameters that have a strong correlation with the sensory quality score in the production process of papermaking reconstituted tobacco, the parameters are pre-processed and input into the sensory quality score model to simulate the predicted sensory quality score close to the actual manual smoking score, thereby realizing real-time simulation monitoring of the sensory quality score in the production process of reconstituted tobacco and ensuring the output stability of reconstituted tobacco in the production process of papermaking reconstituted tobacco. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a schematic diagram of the architecture of the hardware operating environment of the papermaking process reconstituted tobacco system involved in the embodiment of the present application;

[0036] Figure 2 This is a flow chart of the first embodiment of the method for determining the sensory quality score of reconstituted tobacco leaves based on the absolute dry weight before coating of the present application;

[0037] Figure 3 It is a schematic diagram of the fitting curve of the absolute dry weight and sensory evaluation score before coating involved in the embodiments of the present application;

[0038] Figure 4 This is a schematic diagram of the association between the associated process parameters and the sensory quality scores of reconstituted tobacco leaves selected from the association table involved in the embodiments of the present application;

[0039] Figure 5 This is a flow chart of a second embodiment of the method for determining the sensory quality score of reconstituted tobacco leaves based on the absolute dry weight before coating of the present application.

[0040] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0041] In order to better understand the above technical solution, exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0042] As an implementation plan, Figure 1 This is a schematic diagram of the architecture of the hardware operating environment of the papermaking reconstituted tobacco system involved in the embodiment of the present application.

[0043] like Figure 1 As shown, the papermaking method reconstituted tobacco leaf system may include: a processor 1001, such as a CPU, a memory 1005, a user interface 1003, a network interface 1004, and a communication bus 1002. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 1005 may be a high-speed RAM memory, or a stable memory (non-volatile memory), such as a disk memory. The memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0044] Those skilled in the art will understand that Figure 1 The architecture of the reconstituted tobacco system for papermaking shown in the figure does not constitute a limitation on the reconstituted tobacco system for papermaking, and may include more or less components than those shown in the figure, or combine certain components, or arrange the components differently.

[0045] like Figure 1 As shown, the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module, and a program for determining the sensory quality score of reconstituted tobacco leaves based on the absolute dry basis weight before coating. The operating system is a program for managing and controlling the hardware and software resources of the papermaking process reconstituted tobacco leaf system, and the program for determining the sensory quality score of reconstituted tobacco leaves based on the absolute dry basis weight before coating and other software or programs.

[0046] exist Figure 1In the papermaking reconstituted tobacco system shown, the user interface 1003 is mainly used to connect to the terminal and communicate data with the terminal; the network interface 1004 is mainly used for the background server and communicates data with the background server; the processor 1001 can be used to call the program for determining the sensory quality score of the reconstituted tobacco based on the absolute dry weight before coating stored in the memory 1005.

[0047] In this embodiment, the papermaking process reconstituted tobacco system includes: a memory 1005, a processor 1001, and a reconstituted tobacco sensory quality score determination program based on the absolute dry basis weight before coating stored in the memory and executable on the processor, wherein:

[0048] When the processor 1001 calls the program for determining the sensory quality score of the reconstituted tobacco leaves based on the absolute dry weight before coating stored in the memory 1005, the following operations are performed:

[0049] Obtaining the absolute dry weight of the reconstituted tobacco leaves to be tested before coating;

[0050] Inputting the absolute dry weight before coating into the trained sensory quality scoring model;

[0051] The predicted sensory quality score value output by the sensory quality score model and calculated based on the absolute dry weight before coating is obtained, and the predicted sensory quality score value is determined as the sensory quality score value corresponding to the reconstituted tobacco leaf to be tested.

[0052] When the processor 1001 calls the program for determining the sensory quality score of the reconstituted tobacco leaves based on the absolute dry weight before coating stored in the memory 1005, the following operations are performed:

[0053] Obtain the absolute dry quantitative sample data set before coating;

[0054] The absolute dry quantitative data before coating is enhanced.

[0055] Based on the data-enhanced absolute dry quantitative sample data set before coating and the corresponding sensory quality score data set, a correlation data table is constructed;

[0056] Input the associated data table into the model with lightGBM architecture for training, and calculate the AUC value of the model after each training is completed;

[0057] After the AUC value reaches the preset AUC threshold, the training is completed.

[0058] When the processor 1001 calls the program for determining the sensory quality score of the reconstituted tobacco leaves based on the absolute dry weight before coating stored in the memory 1005, the following operations are performed:

[0059] Oversampling the abnormal absolute dry weight samples before coating that are not in the standard absolute dry weight interval before coating in the associated data table, and undersampling the normal absolute dry weight samples before coating that are in the standard absolute dry weight interval before coating, so as to balance the number of samples between the abnormal absolute dry weight samples before coating and the normal absolute dry weight samples before coating;

[0060] The balanced association data table is input into the model with lightGBM architecture for training.

[0061] When the processor 1001 calls the program for determining the sensory quality score of the reconstituted tobacco leaves based on the absolute dry weight before coating stored in the memory 1005, the following operations are performed:

[0062] During the process of executing the papermaking method reconstituted tobacco process, obtaining the sensory quality score value associated with the reconstituted tobacco in the current production line at every preset period;

[0063] Determine whether the sensory quality score is within a normal value range;

[0064] When it is detected that the sensory quality score value is not in the normal value range, determining whether the absolute dry weight before coating in the current production line is in the standard absolute dry weight before coating range;

[0065] If not, the abnormal target absolute dry quantitative process equipment parameters before coating are determined and output among the absolute dry quantitative process equipment parameters before coating.

[0066] Based on the hardware architecture of the above-mentioned papermaking reconstituted tobacco system based on computer deep learning technology, an embodiment of the method for determining the sensory quality score of reconstituted tobacco based on the absolute dry weight before coating is proposed in the present application.

[0067] Reference Figure 2 In a first embodiment, the method for determining the sensory quality score of reconstituted tobacco leaves based on the absolute dry weight before coating comprises the following steps:

[0068] Step S10, obtaining the absolute dry weight of the reconstituted tobacco leaves to be tested before coating;

[0069] In this embodiment, firstly, the absolute dry weight before coating of the reconstituted tobacco leaves to be tested is obtained;

[0070] It should be noted that the reconstituted tobacco to be tested refers to the reconstituted tobacco that has gone through the reconstituted tobacco process and is waiting to be tested to see if its quality meets the sensory quality standards. Optionally, the reconstituted tobacco to be tested can be selected from a product batch previously produced in the production batch as the reconstituted tobacco to be tested.

[0071] The traditional way to score the sensory quality of reconstituted tobacco is to manually smoke the cigarettes made from the reconstituted tobacco and then evaluate and score them. Among the process parameters of the reconstituted tobacco itself, the absolute dry weight before coating that affects the sensory quality score is still unclear.

[0072] In this embodiment, in order to study which parameters are strongly correlated with the sensory quality score of papermaking reconstituted tobacco, corresponding statistics are performed based on the existing production database Uniformence and combined with the actual sensory quality manual smoking scores, so as to determine the fitting curves in each process parameter.

[0073] For example, see Figure 3 The figure shows the fitting curve of the absolute dry weight and sensory evaluation score before coating. The following relationship can be obtained through the fitting curve:

[0074] When the absolute dry weight before coating is greater than 49g / m 2 When the absolute dry weight before coating was 0.0447 W / m, the sensory quality score was negatively correlated.

[0075] After determining the fitting curve, Minitab was further used to perform correlation analysis on the key process parameters involved, referring to the corresponding table between the correlation coefficient and the correlation strength shown in Table 1:

[0076] Table 1. Correspondence between correlation coefficient and correlation strength

[0077] Correlation coefficient Strength of association 0.8-10 Very strong correlation 0.6-0.8 Strong correlation 0.4-0.6 Moderately strong correlation 0.2-0.4 Weak correlation 0.0-0.2 Very weak or no correlation

[0078] Based on the relationship correspondence table given in Table 1, we screened out parameters with correlation coefficients close to 0.8 or above from various process parameters as process parameters with strong correlation.

[0079] Based on the relationship correspondence table given in Table 1, the correlation coefficient of the absolute dry weight before coating is 0.809, which is a process parameter with strong correlation.

[0080] Optionally, the absolute dry weight before coating is detected by Kr-85 (Krypton 85) rays installed inside the QCS. The principle is: Kr-85 (Krypton 85) rays penetrate the paper to form ray changes, which are converted into current signals through circuit conversion amplifiers. OPCsmart is used to read the absolute dry weight current signal before coating of the QCS client, convert it into specific values ​​and display it on the DCS human-machine interface, and collect the values ​​to the background production database Uniformance for call.

[0081] Step S20, inputting the absolute dry weight before coating into the trained sensory quality scoring model;

[0082] Furthermore, after determining that the associated process parameter with a strong correlation with the sensory quality score value is the absolute dry weight before coating, during the papermaking process of reconstituted tobacco leaves, the system collects the absolute dry weight before coating in the reconstituted tobacco leaves in real time, and inputs the absolute dry weight before coating into the trained sensory quality score calculation model constructed based on the deep learning model. The sensory quality score calculation module based on the absolute dry weight before coating in the sensory quality score calculation model simulates and predicts the sensory quality score value corresponding to the absolute dry weight before coating.

[0083] In order to ensure that the sensory quality score calculated by the model is close enough to the sensory quality score after manual smoking, during the training process, we collected a large number of reconstituted tobacco leaves produced in historical batches and scored them by manual smoking. We obtained a correlation table between the absolute dry weight before coating and the sensory quality score of a large number of reconstituted tobacco leaves. Figure 4 This is a schematic diagram of the correlation between the absolute dry weight of reconstituted tobacco leaves before coating and the sensory quality score selected from the correlation table (the specific values ​​of the sensory quality score are omitted due to industrial secrets). Using the data in this correlation table, multiple deep learning training and algorithm optimization were carried out. After testing, the accuracy of the final model reached 92.3%.

[0084] Step S30, obtaining the predicted sensory quality score value output by the sensory quality score model and calculated based on the absolute dry weight before coating, and determining the predicted sensory quality score value as the sensory quality score value corresponding to the reconstituted tobacco leaf to be tested.

[0085] In this embodiment, a predicted sensory quality score that is as consistent as possible with the actual smoking score is calculated based on the absolute dry quantitative calculation before coating, and the predicted sensory quality score is determined as the sensory quality score corresponding to the reconstituted tobacco to be tested. Thus, by monitoring the sensory quality score, the absolute dry quantitative calculation link before coating that affects the sensory quality score in the production process of the reconstituted tobacco is monitored in real time. When the sensory quality score is abnormal, the absolute dry quantitative calculation link before coating that affects the sensory quality score is checked immediately.

[0086] Further and optionally, after obtaining the sensory quality score corresponding to the reconstituted tobacco leaves to be tested, it can be determined whether the reconstituted tobacco leaves to be tested are within a preset normal value range. When it is detected that the sensory quality score is not within the normal value range, it is then detected whether the absolute dry weight before coating associated with the sensory quality score is abnormal.

[0087] In addition, it should be noted that, in this embodiment, after we found the absolute dry weight before coating that has a strong correlation with the sensory quality score, we did not directly monitor based on the numerical value of the absolute dry weight before coating. The reason is that there are many quality monitoring indicators in the production process of reconstituted tobacco leaves, but the general monitoring indicators are only for evaluating the output grade of reconstituted tobacco leaves, which does not affect whether the reconstituted tobacco leaves are qualified; and the sensory quality score value is an indicator with a veto nature as to whether the production of reconstituted tobacco leaves is qualified.

[0088] In the actual papermaking process of reconstituted tobacco, we allow a certain fluctuation in the numerical range of the absolute dry weight before coating. Therefore, even if the absolute dry weight before coating of reconstituted tobacco in a batch deviates from the preset numerical range, but does not cause the sensory quality score monitored by the system to deviate from the normal value range, we still regard the output of this batch of reconstituted tobacco as qualified, and do not regard the fluctuating absolute dry weight before coating as an abnormal process parameter. In other words, it is only meaningful to judge whether the absolute dry weight before coating that affects the sensory quality score is abnormal when the sensory quality score is monitored to be abnormal.

[0089] In the technical solution provided in this embodiment, the absolute dry weight before coating that is strongly correlated with the sensory quality score in the process parameters of reconstituted tobacco leaves by the papermaking method is found, the absolute dry weight before coating is input into the trained sensory quality score model, and the sensory quality score prediction value is calculated based on the absolute dry weight before coating and then used as the sensory quality score value corresponding to the reconstituted tobacco leaves to be tested, thereby achieving efficient application of the sensory quality score in the industrial production of reconstituted tobacco leaves by the papermaking method and ensuring the output stability of reconstituted tobacco leaves in the production process of reconstituted tobacco leaves by the papermaking method.

[0090] Second embodiment

[0091] Based on the first embodiment, in this embodiment, the trained sensory quality scoring model includes a sensory quality scoring calculation module for absolute dry weight before coating, and the training process of the sensory quality scoring model includes:

[0092] Step S100, obtaining an absolute dry quantitative sample data set before coating;

[0093] Step S200, performing data enhancement on the absolute dry quantitative data before coating,

[0094] Step S300, constructing a correlation data table based on the absolute dry quantitative sample data set before coating and the corresponding sensory quality score data set after data enhancement;

[0095] Step S400, input the associated data table into a model with a lightGBM architecture for training, and calculate the AUC value of the model after each training is completed;

[0096] Step S500: After the AUC value reaches the preset AUC threshold, the training is completed.

[0097] In this embodiment, during the training process, we found that after the original normalization and standardization were used to pre-process the absolute dry quantitative sample data set before coating, the trained model had the problem of poor robustness. When the input absolute dry quantitative value before coating was significantly different from the value in the original sample data set, the predicted value of the sensory quality score was significantly different from the sensory quality score obtained by the actual evaluation.

[0098] Therefore, in this embodiment, data enhancement is used to preprocess the absolute dry quantitative sample data set before coating to increase the data diversity and data training amount of the model, and based on the enhanced updated absolute dry quantitative data set before coating and the sensory quality score data set, an associated data table is constructed to improve the robustness of the model.

[0099] Based on the data-enhanced absolute dry quantitative sample data set before coating and the corresponding sensory quality score data set, an associated data table was constructed, and the associated data table was input into the model with the lightGBM architecture for training, and the AUC value of the model was calculated after each training was completed.

[0100] In this embodiment, AUC (Area Under Curve) is used to evaluate the performance index of the sensory quality scoring model. After each training is completed, the performance index of the current model, that is, the AUC value, can be calculated by calling the AUC function in Python. If the AUC value does not reach the preset AUC threshold after the training is completed, the hyperparameters in each model are adjusted and the training continues.

[0101] When the AUC value of the deep learning model with the lightGBM architecture in the sensory quality score calculation module of the absolute dry weight before coating reaches the preset AUC threshold, the training is terminated to obtain the trained sensory quality score calculation module of the absolute dry weight before coating.

[0102] Furthermore, in this embodiment, before step S400, the following steps are further included:

[0103] Step S600, oversampling the abnormal absolute dryness before coating quantitative samples that are not in the standard absolute dryness before coating quantitative interval in the associated data table, and undersampling the normal absolute dryness before coating quantitative samples that are in the standard absolute dryness before coating quantitative interval, so as to balance the number of samples between the abnormal absolute dryness before coating quantitative samples and the normal absolute dryness before coating quantitative samples;

[0104] The step S400 includes:

[0105] Step S700: input the balanced associated data table into a model with a lightGBM architecture for training.

[0106] Further, by Figure 3 It can be seen that due to the large number of data feature types of the absolute dry weight before coating in the associated data table, the computing power overhead of the model is greatly increased during the training process, and the prediction of the model during the training process is concentrated on predicting that the absolute dry weight before coating is 50-50.7g / m 2 Therefore, a data balancing method is used to balance the number of samples between the abnormal absolute dry quantitative samples before coating and the normal absolute dry quantitative samples before coating to avoid overfitting of the model.

[0107] Finally, the associated data table after balancing again is input into the model with lightGBM architecture for training.

[0108] Furthermore, in this embodiment, the adjusted hyperparameters include the maximum depth and the number of leaf nodes.

[0109] The maximum depth refers to the depth of the longest path from the root node to the leaf node in the tree model. A deeper tree can better fit the training data, but it is easy to cause overfitting, making the model perform well on the training set but poorly on the test set.

[0110] The number of leaf nodes refers to the maximum number of leaf nodes in a tree. If the number of leaf nodes is too large, the tree may overfit the training data.

[0111] Preferably, after comparative experiments, the maximum depth and the number of leaf nodes can be set to -1 and 30 respectively.

[0112] When the AUC value of the deep learning model with the lightGBM architecture in the sensory quality score calculation module of the absolute dry weight before coating reaches the preset AUC threshold, the training is terminated to obtain the trained sensory quality score calculation module of the absolute dry weight before coating.

[0113] Third embodiment

[0114] Based on any embodiment, refer to Figure 5 In this embodiment, after step S30, the following steps are further included:

[0115] Step S40, during the process of executing the papermaking method reconstituted tobacco process, obtaining the sensory quality score value associated with the reconstituted tobacco in the current production line at every preset period;

[0116] Step S50, determining whether the sensory quality score is within a normal value range;

[0117] Step S60, when it is detected that the sensory quality score value is not in the normal value range, determining whether the absolute dry weight before coating in the current production line is in the standard absolute dry weight before coating range;

[0118] Step S70: If not, determine and output the abnormal target absolute dry quantitative process equipment parameters before coating among the absolute dry quantitative process equipment parameters before coating.

[0119] In this embodiment, after the sensory quality score value corresponding to the absolute dry weight before coating is calculated by the deep learning model based on the first embodiment, during the execution of the papermaking method reconstituted tobacco process, the sensory quality score value associated with the reconstituted tobacco in the current production line is obtained every preset period.

[0120] Optionally, the preset period may be to obtain data once every 5 minutes.

[0121] After obtaining the sensory quality score value, determine whether the numerical interval of the sensory quality score value is in the normal value interval. When it is detected that the sensory quality score value is not in the normal value interval, it is determined that the absolute dry weight before coating of the reconstituted tobacco to be tested is abnormal, and the absolute dry weight before coating is used as the target tobacco process parameter to be tested. When it is determined that the absolute dry weight before coating is not in its corresponding standard absolute dry weight before coating interval, the abnormal target absolute dry weight before coating process equipment parameters among the absolute dry weight process equipment parameters before coating are determined and output.

[0122] Optionally, the normal value interval of the sensory quality score can be set to a fixed value of ±1; wherein the fixed value can be set according to the specific processing scenario of the reconstituted tobacco leaves.

[0123] Further and optionally, in the present embodiment, when it is determined that the absolute dry weight before coating is not within the normal value range, the absolute dry weight before coating corresponds to one or several target reconstituted tobacco equipment parameters, and the target reconstituted tobacco equipment parameters refer to the parameters corresponding to the equipment that affects the absolute dry weight before coating in the process flow of reconstituted tobacco by the papermaking method.

[0124] The target absolute dry quantitative process equipment parameters before coating specifically include at least one of sizing amount, retention rate, and base uniformity.

[0125] The sizing amount refers to the slurry flow rate required to meet the quantitative requirements. Optionally, a flow meter can be set in the slurry pipeline in the coating process, and the system can collect the sizing amount by collecting the flow data on the flow meter. Among them, the sizing amount is positively correlated with the absolute dry weight before coating within a certain numerical range.

[0126] Retention rate (also known as filler retention rate) refers to the proportion of fillers or fibers retained in paper during the papermaking process, that is, the percentage of fillers in paper to fillers in the slurry. Retention rate is positively correlated with the absolute dry weight before coating.

[0127] In addition, the retention rate is a calculated value, calculated as follows:

[0128] Retention rate = absolute dry weight of retained slurry / absolute dry weight of slurry * 100%

[0129] Among them, the absolute dry amount of retained slurry and the absolute dry amount of slurry are both the collection amounts of the system.

[0130] The uniformity of the sheet base refers to the quantitative value of the consistency of the sheet base material of the reconstituted tobacco leaves on the entire paper web. The uniformity of the sheet base is monitored in real time by a horizontal scanning frame and adjusted in linkage by the quality control system. The better the uniformity of the sheet base, the more stable the absolute dry weight before coating. On the contrary, the worse the uniformity, the greater the quantitative fluctuation, and the extreme difference will affect the back-end processing.

[0131] When an abnormality is detected in the value of the absolute dry weight before coating, based on the above-mentioned target absolute dry weight before coating process equipment parameters and the positive / negative correlation between these operating parameters and the absolute dry weight before coating, it is detected whether the above-mentioned target absolute dry weight before coating process equipment parameters are too large / too small, and when it is found that the target absolute dry weight before coating process equipment parameters are too large / too small, the parameters are quickly adjusted to regulate the absolute dry weight before coating of papermaking reconstituted tobacco.

[0132] In the technical solution provided in this embodiment, after the sensory quality score value corresponding to the absolute dry weight before coating is calculated by the deep learning model, when an abnormal sensory quality score value is detected during the execution of the papermaking process of reconstituted tobacco, it is determined whether the absolute dry weight before coating is abnormal, and when the absolute dry weight before coating is abnormal, the abnormal equipment parameters related to the absolute dry weight before coating are output, thereby realizing real-time monitoring of the key parameters affecting the sensory quality score in the production process of papermaking reconstituted tobacco, and timely feedback when an abnormality occurs, thereby ensuring product stability in the papermaking production process of reconstituted tobacco.

[0133] In addition, it can be understood by a person skilled in the art that all or part of the processes in the method for implementing the above embodiment can be completed by instructing the relevant hardware through a computer program. The computer program includes program instructions, and the computer program can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the papermaking process reconstituted tobacco system to implement the process steps of the embodiment of the above method.

[0134] Therefore, the present application also provides a computer-readable storage medium, which stores a program for determining the sensory quality score of reconstituted tobacco leaves based on the absolute dry weight before coating. When the program for determining the sensory quality score of reconstituted tobacco leaves based on the absolute dry weight before coating is executed by a processor, it implements the various steps of the method for determining the sensory quality score of reconstituted tobacco leaves based on the absolute dry weight before coating as described in the above embodiment.

[0135] The computer-readable storage medium may be a USB flash drive, a mobile hard disk, a read-only memory (ROM), a magnetic disk, or an optical disk, etc., which may store program codes.

[0136] It should be noted that since the storage medium provided in the embodiment of the present application is the storage medium used to implement the method of the embodiment of the present application, based on the method introduced in the embodiment of the present application, the person skilled in the art can understand the specific structure and deformation of the storage medium, so it is not repeated here. All storage media used in the method of the embodiment of the present application belong to the scope of protection of this application.

[0137] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0138] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0139] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0140] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0141] It should be noted that in the claims, any reference signs placed between brackets shall not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The present application may be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third etc. does not indicate any order. These words may be interpreted as names.

[0142] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0143] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for determining the sensory quality score of reconstituted tobacco leaves based on absolute dry weight before coating, characterized in that: The method comprises the following steps: Obtaining the absolute dry weight of the reconstituted tobacco leaves to be tested before coating; Inputting the absolute dry weight before coating into the trained sensory quality scoring model; The predicted sensory quality score value output by the sensory quality score model and calculated based on the absolute dry weight before coating is obtained, and the predicted sensory quality score value is determined as the sensory quality score value corresponding to the reconstituted tobacco leaf to be tested.

2. The method according to claim 1, characterized in that The trained sensory quality scoring model includes a sensory quality scoring calculation module for absolute dry weight before coating, and the training process of the sensory quality scoring model includes: Obtain the absolute dry quantitative sample data set before coating; The absolute dry quantitative data before coating is enhanced. Based on the data-enhanced absolute dry quantitative sample data set before coating and the corresponding sensory quality score data set, a correlation data table is constructed; Input the associated data table into the model with lightGBM architecture for training, and calculate the AUC value of the model after each training is completed; After the AUC value reaches the preset AUC threshold, the training is completed.

3. The method according to claim 2, characterized in that Before the step of inputting the associated data table into the model with the lightGBM architecture for training and calculating the AUC value of the model after each training is completed, the method further includes: Oversampling the abnormal absolute dry weight samples before coating that are not in the standard absolute dry weight interval before coating in the associated data table, and undersampling the normal absolute dry weight samples before coating that are in the standard absolute dry weight interval before coating, so as to balance the number of samples between the abnormal absolute dry weight samples before coating and the normal absolute dry weight samples before coating; The step of inputting the associated data table into a model with a lightGBM architecture for training, and calculating the AUC value of the model after each training is completed includes: The balanced association data table is input into the model with lightGBM architecture for training.

4. The method according to claim 2, characterized in that The model adjusts the maximum depth and number of leaf nodes in the model hyperparameters during the training process.

5. The method according to any one of claims 1 to 4, characterized in that: After the step of obtaining the predicted sensory quality score value output by the sensory quality score model and calculated based on the absolute dry weight before coating, and determining the predicted sensory quality score value as the sensory quality score value corresponding to the reconstituted tobacco to be tested, the method further includes: During the process of executing the papermaking method reconstituted tobacco process, obtaining the sensory quality score value associated with the reconstituted tobacco in the current production line at every preset period; Determine whether the sensory quality score is within a normal value range; When it is detected that the sensory quality score value is not in the normal value range, determining whether the absolute dry weight before coating in the current production line is in the standard absolute dry weight before coating range; If not, determine and output the abnormal target absolute dry quantitative process equipment parameters before coating among the absolute dry quantitative process equipment parameters before coating.

6. The method according to claim 5, characterized in that The abnormal target absolute dry quantitative process equipment parameter before coating includes at least one of sizing amount, retention rate, and base uniformity.

7. A papermaking process reconstituted tobacco system, characterized in that: The papermaking reconstituted tobacco system comprises: a memory, a processor, and a program for determining the sensory quality score of the reconstituted tobacco based on the absolute dry weight before coating, which is stored in the memory and can be run on the processor. When the program for determining the sensory quality score of the reconstituted tobacco based on the absolute dry weight before coating is executed by the processor, the steps of the method for determining the sensory quality score of the reconstituted tobacco based on the absolute dry weight before coating as described in any one of claims 1 to 6 are implemented.

Images