Cigarette comprehensive quality instant evaluation method based on independence weight analysis and machine learning

Abstract

The invention discloses a cigarette comprehensive quality real-time evaluation method based on independence weight analysis and machine learning, and the main design concept of the invention lies in that objective weighting is carried out by adopting an independence weight method, so that the influence of subjective prejudice or man-made interference on a cigarette comprehensive quality evaluation result is avoided, and the evaluation accuracy is improved. Scientific preciseness is ensured, meanwhile, a solid theoretical basis in the industry is provided, and the accuracy and reliability of an empowerment result are fully ensured. Moreover, after the cigarette comprehensive quality measurement value is calculated, modeling is carried out on the cigarette comprehensive quality measurement value by adopting a machine learning algorithm, and according to a constructed cigarette comprehensive quality measurement value prediction model, the cigarette comprehensive quality measurement value can be predicted directly from raw and auxiliary material parameters and cigarette parameters of an unknown cigarette sample. According to the method, the cigarette comprehensive quality measurement value is predicted to evaluate the cigarette comprehensive quality, the finished cigarette does not need to be subjected to spot check in the production process, and a reliable and feasible scheme is provided for real-time evaluation of the cigarette comprehensive quality.

Description

Technical Field

[0001] The present invention relates to the field of cigarette manufacturing, and in particular to a method for instant evaluation of comprehensive cigarette quality based on independence weight analysis and machine learning. Background Art

[0002] The quality evaluation of cigarette products is a key link in the tobacco industry chain. The quality of cigarettes is directly related to the market competitiveness and economic benefits of cigarette companies. The evaluation of cigarette quality is a multi-dimensional process involving multiple key aspects, including: cigarette ash characteristic indicators, smoke chemical indicators, sensory evaluation indicators, etc. Relying on only one aspect of the evaluation index cannot fully reflect the overall quality of cigarettes; therefore, it is of great significance to establish a scientific evaluation method for the comprehensive quality of cigarettes.

[0003] At present, the main analysis and evaluation method provided by the industry is to use the hierarchical analysis method to determine the respective weights of cigarette quality indicators. This method needs to rely on the experience of decision makers to judge, and depends on the influence of human factors, which may lead to deviations; there is also a comprehensive analysis method of the importance of cigarette combustion quality indicators designed by a singular detection method, so as to obtain the importance ranking and weight of indicators, but this method does not consider the comprehensive smoke chemical indicators and sensory evaluation indicators; and, based on the hierarchical analysis method, entropy weight method, standard deviation method and CRITIC method, a comprehensive evaluation method of cigarette physical quality based on combined weighting and efficacy coefficient method is established, but this type of scheme using the hierarchical analysis method is likely to be interfered by human factors. Summary of the invention

[0004] In view of the above, the present invention aims to provide an instant evaluation method for the comprehensive quality of cigarettes based on independence weight analysis and machine learning, so as to overcome the problems that the above-mentioned existing comprehensive quality evaluation methods of cigarettes need to rely on subjective weighting, are more dependent on work experience, and the comprehensive quality evaluation of cigarettes has a lag defect.

[0005] The technical solution adopted by the present invention is as follows:

[0006] The present invention provides a method for instant evaluation of comprehensive quality of cigarettes based on independence weight analysis and machine learning, which includes:

[0007] Collecting cigarette quality indicators, wherein the cigarette quality indicators include: cigarette ash characteristic indicators, smoke chemical indicators, and sensory evaluation indicators;

[0008] De-dimensionalizing the cigarette quality index;

[0009] Based on the independence weight analysis method, the weight of each of the cigarette quality indicators is determined, including calculation using the following formula:

[0010]

[0011] Among them, R i represents the multiple correlation coefficient; X j represents the jth cigarette quality index after dedimensionalization; H represents X minus X j The residual matrix of Represents the result of averaging X; the inverse of the complex correlation coefficient As the score, the weight value W is obtained after normalization i ;

[0012] Based on the weights, the comprehensive quality metric of the cigarette is calculated, including using the following formula:

[0013]

[0014] Where, Y represents the comprehensive quality measurement value of cigarettes; W i Indicates the weight value of cigarette quality index; Z i It represents the de-dimensionalized cigarette quality index;

[0015] The collected raw and auxiliary material parameters and cigarette parameters corresponding to the cigarette quality index are used as independent variables, and the comprehensive quality metric value of the cigarette is used as the dependent variable. A machine learning algorithm is used to establish and train a prediction model for the comprehensive quality metric value of the cigarette, and the output of the model includes the importance ranking results of the raw and auxiliary material parameters and the cigarette parameters; wherein the raw and auxiliary material parameters include at least one of the following: tobacco shred parameters, cigarette paper parameters, tipping paper parameters, forming paper parameters, and filter rod parameters; the cigarette parameters include at least one of the following: cigarette circumference, cigarette weight, standard draw resistance, closed draw resistance, cigarette paper ventilation rate, filter tip ventilation rate, total ventilation rate, cigarette length, and short cigarette length;

[0016] The raw and auxiliary material parameters and cigarette parameters of the actual tested cigarette samples are input into the cigarette comprehensive quality metric value prediction model to obtain the cigarette comprehensive quality metric value of the actual tested cigarette samples.

[0017] In at least one possible implementation, the dedimensionalization process is calculated using the following formula:

[0018]

[0019] Among them, Z i represents the dimensionless cigarette quality index; y i Indicates the cigarette quality index obtained by testing cigarette samples; Represents the average value of each cigarette quality index.

[0020] In at least one possible implementation, the cigarette ash characteristic index includes at least one of the following: gray color, crack rate, ash shrinkage rate, carbon line uniformity, carbon line width, and ash holding power;

[0021] The smoke chemical index includes at least one of the following: tar release, nicotine release, and CO release;

[0022] The sensory evaluation index includes at least one of the following: aroma, harmony, miscellaneous smell, irritation, and aftertaste.

[0023] In at least one possible implementation, the tobacco parameters include: physical parameters and chemical parameters;

[0024] Wherein, the physical parameters include at least one of the following: leaf weight, bulked weight, thin thread or flake weight, and stem weight;

[0025] The chemical parameters include at least one of the following: total sugar, reducing sugar, total alkaloids, total nitrogen, nitrate, potassium, chlorine content, ratio of two sugars, sugar-alkali ratio, nitrogen-alkali ratio, potassium-chlorine ratio.

[0026] In at least one possible implementation, the cigarette paper parameters include at least one of the following: fiber type, basis weight, air permeability, citrate content, potassium ion content, sodium ion content, and ash content;

[0027] The tipping paper parameters include at least one of the following: air permeability, overlap length;

[0028] The forming paper parameters include at least one of the following: basis weight, air permeability;

[0029] The filter rod parameters include at least one of the following: length of the non-groove section of the filter rod, length of the groove section of the filter rod, circumference of the filter rod, pressure drop of a single filter rod, and specification of the tow.

[0030] In at least one possible implementation, the machine learning algorithm includes any one of the following: a random forest algorithm, a multivariate linear regression algorithm, and a neural network algorithm.

[0031] Compared with the prior art, the main design concept of the present invention is that an independent weighting method is used for objective weighting, thereby avoiding the influence of subjective bias or human interference on the comprehensive quality evaluation results of cigarettes. While ensuring scientific rigor, it has a solid theoretical basis in the industry and fully guarantees the accuracy and reliability of the weighting results. Moreover, after calculating the comprehensive quality metric value of cigarettes, a machine learning algorithm is used to model the comprehensive quality metric value of cigarettes. According to the constructed comprehensive quality metric value prediction model of cigarettes, the comprehensive quality metric value of cigarettes can be directly predicted from the raw and auxiliary material parameters and cigarette parameters of unknown cigarette samples to conduct comprehensive quality evaluation of cigarettes. There is no need to conduct random inspections of finished cigarettes during the production process, which provides a reliable and feasible solution for instant evaluation of the comprehensive quality of cigarettes. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be further described below with reference to the accompanying drawings, in which:

[0033] Figure 1 A schematic flow chart of a method for instant evaluation of comprehensive cigarette quality based on independence weight analysis and machine learning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0034] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be interpreted as limiting the present invention.

[0035] The present invention proposes an embodiment of a method for instant evaluation of comprehensive quality of cigarettes based on independence weight analysis and machine learning. Specifically, Figure 1 shown, including:

[0036] Step S1, collecting cigarette quality indicators, wherein the cigarette quality indicators include: cigarette ash characteristic indicators, smoke chemical indicators, and sensory evaluation indicators;

[0037] It can also be expanded here that the cigarette ash characteristic indicators can include at least one of the following: gray color, crack rate, ash shrinkage rate, carbon line uniformity, carbon line width, and ash holding capacity; the smoke chemical indicators can include at least one of the following: tar release, nicotine release, and CO release; the sensory evaluation indicators can include at least one of the following: aroma, harmony, miscellaneous smell, irritation, and aftertaste.

[0038] Step S2, performing dimensionless processing on the cigarette quality index;

[0039] In actual operation, the dimensionless processing process can be calculated using the following formula:

[0040]

[0041] Among them, Z i represents the dimensionless cigarette quality index; y i Indicates the cigarette quality index obtained by testing cigarette samples; Represents the average value of each cigarette quality index.

[0042] Step S3, based on the independence weight analysis method, determining the weight of each of the cigarette quality indicators, including calculating using the following formula:

[0043]

[0044] Among them, R i represents the multiple correlation coefficient; X j represents the jth cigarette quality index after dedimensionalization; H represents X minus X j The residual matrix of represents the result after X is averaged; furthermore, the inverse of the complex correlation coefficient As the score, the weight value W is obtained after normalization i .

[0045] Step S4: Based on the weight, the comprehensive quality metric value of the cigarette is calculated, including using the following formula:

[0046]

[0047] Where, Y represents the comprehensive quality measurement value of cigarettes; W i Indicates the weight value of cigarette quality index; Z i It represents the de-dimensionalized cigarette quality index;

[0048] Step S5, using the collected raw and auxiliary material parameters and cigarette parameters corresponding to the cigarette quality index as independent variables, and the cigarette comprehensive quality metric value as the dependent variable, using a machine learning algorithm to establish and train a cigarette comprehensive quality metric value prediction model, and making the output of the model include the importance ranking results of the raw and auxiliary material parameters and the cigarette parameters; wherein the raw and auxiliary material parameters include at least one of the following: tobacco shred parameters, cigarette paper parameters, tipping paper parameters, forming paper parameters, and filter rod parameters; the cigarette parameters include at least one of the following: cigarette circumference, cigarette weight, standard draw resistance, closed draw resistance, cigarette paper ventilation rate, filter tip ventilation rate, total ventilation rate, cigarette length, and short cigarette length;

[0049] It can be understood that the machine learning algorithms that can be used include but are not limited to: random forest algorithm, multivariate linear regression algorithm, neural network algorithm, etc.

[0050] In addition, specifically, the tobacco parameters include: physical parameters and chemical parameters; the physical parameters include at least one of the following: leaf weight, expanded weight, thin wire or thin sheet weight, stem weight; the chemical parameters include at least one of the following: total sugar, reducing sugar, total alkaloids, total nitrogen, nitrate, potassium, chlorine content, ratio of two sugars, sugar-alkali ratio, nitrogen-alkali ratio, potassium-chlorine ratio.

[0051] The cigarette paper parameters include at least one of the following: fiber type, basis weight, air permeability, citrate content, potassium ion content, sodium ion content, and ash content; the tipping paper parameters include at least one of the following: air permeability and overlap length; the forming paper parameters include at least one of the following: basis weight and air permeability; the filter rod parameters include at least one of the following: length of the non-groove section of the filter rod, length of the groove section of the filter rod, circumference of the filter rod, pressure drop of a single filter rod, and tow specifications.

[0052] Step S6: actually testing the raw and auxiliary material parameters and cigarette parameters of the cigarette sample, and inputting them into the cigarette comprehensive quality metric value prediction model to obtain the cigarette comprehensive quality metric value of the actually tested cigarette sample.

[0053] Those skilled in the art can understand that the above-mentioned cigarette weight, total ventilation rate of cigarettes, cigarette paper basis weight, tipping paper air permeability in tipping paper parameters, cigarette paper air permeability in cigarette paper parameters, forming paper air permeability in forming paper parameters, cigarette paper citrate content in cigarette paper parameters, cigarette paper sodium ion content, cigarette paper potassium ion content, tar release in smoke chemical indicators, nicotine release in smoke chemical indicators, CO release in smoke chemical indicators, etc., can be measured according to various established standards in the industry; and the cigarette ash characteristic indicators can be measured with reference to existing technologies and research in the industry.

[0054] Based on the above solution description, two specific implementation examples are provided below for understanding and reference.

[0055] First embodiment:

[0056] This embodiment provides a method for instant evaluation of comprehensive quality of cigarettes based on independence weight analysis and machine learning algorithm, comprising the following steps:

[0057] 1. Take 58 cigarettes as samples and collect the cigarette quality index data of each sample, including tar release, nicotine release, CO release, gray scale, crack rate, ash shrinkage rate, carbon line width, and carbon line uniformity.

[0058] 2. In order to eliminate the interference of different dimensions of cigarette quality indicators on subsequent data processing, the cigarette quality indicators are dedimensionalized and the calculation results are shown in Table 1.

[0059] Table 1 Calculation results of dedimensionalized cigarette quality indicators

[0060]

[0061]

[0062] 3. De-dimensionalize the calculation results of each cigarette quality index in step 2 (Z i ), input into SPSSAU data processing system, and use independence weight method to calculate Z i The calculation results of their respective weights are shown in Table 2.

[0063] Table 2 Calculation results of the independence weight method

[0064] Cigarette quality indicators Multiple correlation coefficient R The inverse of the multiple correlation coefficient is 1 / R Weight <![CDATA[Z 焦油释放量 ]]> 0.880 1.137 8.39％ <![CDATA[Z 烟碱释放量 ]]> 0.829 1.206 8.90％ <![CDATA[Z CO释放量 ]]> 0.715 1.398 10.32％ <![CDATA[Z 灰度 ]]> 0.442 2.264 16.71％ <![CDATA[Z 裂口率 ]]> 0.359 2.783 20.54％ <![CDATA[Z 缩灰率 ]]> 0.376 2.663 19.65％ <![CDATA[Z 炭线宽度 ]]> 0.952 1.050 7.75％ <![CDATA[Z 炭线整齐度 ]]> 0.952 1.050 7.75％

[0065] 4. Combining the dedimensionalized calculation results of cigarette quality indicators in Table 1 and the independence weight calculation results in Table 2, the comprehensive quality measurement value Y of cigarettes is calculated using the following formula. The calculation results are shown in the measured value column in Table 3.

[0066] Y=W 焦油释放量 ×Z 焦油释放量 +W 烟碱释放量 ×Z 烟碱释放量 +WCO 释放量 ×ZCO 释放量 +W 灰度 ×Z 灰度 +W 裂口率 ×Z 裂口率 +W 缩灰率 ×Z 缩灰率 +W 炭线宽度 ×Z 炭线宽度 +W 炭线整齐度 ×Z 炭线整齐度

[0067] 5. Collect the raw and auxiliary material parameters and cigarette parameters corresponding to the cigarette quality indicators (cigarette paper air permeability, cigarette paper basis weight, cigarette paper citrate content, cigarette paper potassium ion content, cigarette paper sodium ion content, tar paper air permeability, forming paper air permeability, cigarette circumference, cigarette weight, total ventilation rate) as independent variables; the comprehensive quality metric value of cigarettes calculated in step 4 is used as the dependent variable. Imported into the IBM SPSS Modeler 18.0 data processing system, the machine learning algorithm selects the random forest algorithm, and after setting parameters including the number of models to be constructed, the sample size and the maximum number of nodes (where the number of models is set to 100, the sample size is set to 1.0, and the minimum number of child nodes is set to 2), the prediction model of the comprehensive quality metric value of cigarettes is constructed. The internal prediction value and the measured value results of the prediction model of the comprehensive quality metric value of cigarettes are shown in Table 3.

[0068] Table 3 Internal prediction values ​​and measured values ​​of the cigarette comprehensive quality metric prediction model

[0069]

[0070]

[0071] Table 4 Comparison analysis of the internal prediction values ​​and measured values ​​of the cigarette comprehensive quality metric prediction model

[0072] Average Error Mean absolute error Standard Deviation Linear correlation Mean square error 0.027 1.049 1.298 0.956 3.445

[0073] As can be seen from Table 4, the linear correlation fit R between the internal predicted value and the measured value of the cigarette comprehensive quality metric prediction model is 0.956, indicating that the constructed cigarette comprehensive quality metric prediction model has good prediction ability for cigarette comprehensive quality metric values.

[0074] 6. Take 4 unknown cigarette samples, collect data on cigarette weight, cigarette circumference, total ventilation rate, cigarette paper permeability, cigarette paper basis weight, cigarette paper citrate content, cigarette paper potassium ion content, cigarette paper sodium ion content, tipping paper permeability, and forming paper permeability, and then use the cigarette comprehensive quality metric value prediction model obtained in step 5 to predict the cigarette comprehensive quality metric values ​​of the 4 unknown cigarette samples. After obtaining the predicted values ​​of the cigarette comprehensive quality metric values, they are compared and analyzed with the measured values. The results are shown in Tables 5 to 6.

[0075] Table 5 Prediction results of comprehensive cigarette quality metrics of unknown cigarette samples

[0076] Unknown sample number Measured value Predicted value Residual YZ-1 88.92 89.97 -1.05 YZ-2 90.21 89.37 0.84 YZ-3 91.82 93.61 -1.79 YZ-4 92.39 91.02 1.37

[0077] Table 6 Analysis of prediction results of comprehensive cigarette quality metrics of unknown cigarette samples

[0078] Minimum error Maximum error Average Error Mean absolute error Standard Deviation Mean square error -1.79 1.37 0.158 1.263 1.515 1.724

[0079] It can be seen from Tables 5 and 6 that the prediction residuals of the comprehensive quality metric values ​​of cigarettes are all within 2.0, and the average absolute error is only 1.263, indicating that the constructed prediction model for the comprehensive quality metric value of cigarettes can accurately predict the comprehensive quality metric value of cigarettes of unknown cigarette samples. The comprehensive quality metric value of cigarettes can be directly predicted from the raw and auxiliary material parameters and cigarette parameters of unknown cigarette samples to evaluate the comprehensive quality of cigarettes, providing a method for instant evaluation of the comprehensive quality of cigarettes.

[0080] In addition, this embodiment also analyzes the importance of the raw and auxiliary material parameters and cigarette parameters of cigarettes to the comprehensive quality metric value of cigarettes, and the process is as follows:

[0081] The total number of cigarette samples for the training model was set to 58, and the number of random forest decision trees constructed was 100. When each decision tree splits at a node, the method with the largest entropy increase of the independent variable in this training model is used as the optimal split method. The overall entropy increase of each feature independent variable in the random forest model is calculated, and the independent variables are sorted according to the size of the overall entropy increase. The independent variable with a larger overall entropy increase is more important. The importance ranking results of cigarette raw and auxiliary material parameters and cigarette parameters are shown in Table 7.

[0082] Table 7 Importance ranking of cigarette raw and auxiliary material parameters and cigarette parameters

[0083] variable Important Values Importance Cigarette paper citrate content 146.85 1 Cigarette paper quantity 92.32 2 Total ventilation rate 77.05 3 Cigarette paper air permeability 75.08 4 Potassium ion content of cigarette paper 60.80 5 Tipping paper air permeability 58.40 6 Sodium ion content of cigarette paper 41.41 7 Forming paper air permeability 16.32 8

[0084] As can be seen from Table 7, in the random forest model, the importance of cigarette raw and auxiliary material parameters and cigarette parameters to the comprehensive quality metric of cigarettes is ranked as follows: cigarette paper citrate content > cigarette paper basis weight > total ventilation rate > cigarette paper air permeability > cigarette paper potassium ion content > cork paper air permeability > cigarette paper sodium ion content > forming paper air permeability.

[0085] Second embodiment

[0086] This embodiment provides a method for instant evaluation of comprehensive quality of cigarettes based on independence weight analysis and machine learning algorithm, comprising the following steps:

[0087] 1. Take 100 cigarettes as samples and collect the cigarette quality index data of each sample, including tar release, nicotine release, and CO release.

[0088] 2. In order to eliminate the interference of different dimensions of cigarette quality indicators on subsequent data processing, the cigarette quality indicators are dedimensionalized and the calculation results are shown in Table 9.

[0089] 3. De-dimensionalize the calculation results of each cigarette quality index in step 2 (Z i ), input into SPSSAU data processing system, and use independence weight method to calculate Z i The respective weights and calculation results are shown in Table 8.

[0090] Table 8 Calculation results of independence weight method

[0091] Cigarette quality indicators Multiple correlation coefficient R The inverse of the multiple correlation coefficient is 1 / R Weight <![CDATA[Z 焦油释放量 ]]> 0.959 1.043 30.92％ <![CDATA[Z 烟碱释放量 ]]> 0.953 1.049 31.10％ <![CDATA[Z CO释放量 ]]> 0.781 1.281 37.98％

[0092] 4. Combining the dedimensionalized calculation results of cigarette quality indicators in Table 9 and the independence weight calculation results in Table 8, the comprehensive quality measurement value Y of cigarettes is calculated using the following formula. The calculation results are shown in Table 9.

[0093] Y=W 焦油释放量 ×Z 焦油释放量 +W 烟碱释放量 ×Z烟碱释放量 +W CO释放量 ×Z CO释放量

[0094] Table 9 Dimensionless calculation results of cigarette quality indicators and comprehensive cigarette quality measurement values

[0095]

[0096]

[0097]

[0098]

[0099] 5. Collect the raw and auxiliary material parameters and cigarette parameters corresponding to the cigarette quality indicators (citrate content of cigarette paper, potassium ion content of cigarette paper, sodium ion content of cigarette paper, leaf weight, expanded weight, thin weight, thin sheet weight, stem weight, cigarette length, cigarette circumference, cigarette weight, total ventilation rate) as independent variables; the comprehensive quality metric value of cigarettes calculated in step 4 is used as the dependent variable. Imported into the IBM SPSS Modeler 18.0 data processing system, the machine learning algorithm selects the random forest algorithm, and after setting parameters including the number of models to be constructed, the sample size and the maximum number of nodes (where the number of models is set to 100, the sample size is set to 1.0, and the minimum number of child nodes is set to 2), the prediction model of the comprehensive quality metric value of cigarettes is constructed. The comparison results of the internal prediction value and the measured value of the prediction model of the comprehensive quality metric value of cigarettes are shown in Table 10.

[0100] Table 10 Comparison analysis of the internal prediction values ​​and measured values ​​of the cigarette comprehensive quality metric prediction model

[0101] Average Error Mean absolute error Standard Deviation Linear correlation Mean square error -0.023 1.398 2.376 0.963 5.243

[0102] As can be seen from Table 10, the linear correlation fit R between the internal predicted value and the measured value of the cigarette comprehensive quality metric prediction model is 0.963, indicating that the constructed cigarette comprehensive quality metric prediction model has good prediction ability for the cigarette comprehensive quality metric.

[0103] 6. Take 5 unknown cigarette samples, collect data on cigarette paper citrate content, cigarette paper potassium ion content, cigarette paper sodium ion content, leaf shred weight, expanded shred weight, thin shred weight, thin sheet weight, stem shred weight, cigarette length, cigarette circumference, cigarette weight, and total ventilation rate, and then use the cigarette comprehensive quality metric value prediction model obtained in step 5 to predict the cigarette comprehensive quality metric values ​​of the 5 unknown cigarette samples. After obtaining the predicted values ​​of the cigarette comprehensive quality metric values, they are compared and analyzed with the measured values. The results are shown in Tables 11 to 12.

[0104] Table 11 Prediction results of comprehensive cigarette quality metrics of unknown cigarette samples

[0105] Unknown sample number Measured value Predicted value Residual YZ-1 94.74 95.99 -1.25 YZ-2 95.40 96.70 -1.30 YZ-3 96.81 97.91 -1.10 YZ-4 89.80 89.67 0.13 YZ-5 98.65 97.01 1.64

[0106] Table 12 Analysis of prediction results of comprehensive cigarette quality metrics of unknown cigarette samples

[0107] Minimum error Maximum error Average Error Mean absolute error Standard Deviation Mean square error -1.64 1.30 -0.377 1.084 1.270 1.432

[0108] It can be seen from Tables 11 and 12 that the prediction residuals of the comprehensive quality metric values ​​of cigarettes are all within 2.0, and the average absolute error is only 1.084, indicating that the constructed prediction model for the comprehensive quality metric value of cigarettes can accurately predict the comprehensive quality metric value of cigarettes of unknown cigarette samples. The comprehensive quality metric value of cigarettes can be directly predicted from the raw and auxiliary material parameters and cigarette parameters of unknown cigarette samples to evaluate the comprehensive quality of cigarettes, providing a method for instant evaluation of the comprehensive quality of cigarettes.

[0109] In addition, this embodiment also analyzes the importance of the raw and auxiliary material parameters and cigarette parameters of cigarettes to the comprehensive quality metric value of cigarettes, and the process is as follows:

[0110] The total number of cigarette samples for the training model was set to 100, and the number of random forest decision trees constructed was set to 100. When each decision tree splits at a node, the method with the largest entropy increase of the independent variable in this training model is used as the optimal split method. The overall entropy increase of each feature independent variable in the random forest model is calculated, and the independent variables are sorted according to the size of the overall entropy increase. The independent variable with a larger overall entropy increase is more important. The importance ranking results of cigarette raw and auxiliary material parameters and cigarette parameters are shown in Table 13.

[0111] Table 13 Importance ranking of cigarette raw and auxiliary material parameters and cigarette parameters

[0112] variable Important Values Importance Total ventilation rate 834.48 1 Cigarette circumference 804.61 2 Potassium ion content of cigarette paper 610.07 3 Cigarette paper citrate content 546.36 4 Cigarette weight 328.01 5 Bulking weight 150.85 6 Sodium ion content of cigarette paper 121.57 7 Leaf weight 49.61 8 Thin wire weight 47.25 9 Stem weight 21.94 10

[0113] As can be seen from Table 13, in the random forest model, the importance of cigarette raw and auxiliary material parameters and cigarette parameters to the comprehensive quality metric of cigarettes is ranked as follows: total ventilation rate > cigarette circumference > cigarette paper potassium ion content > cigarette paper citrate content > cigarette weight > expanded wire weight > cigarette paper sodium ion content > leaf wire weight > thin wire weight > stem wire weight.

[0114] In summary, the main design concept of the present invention is to use the independence weight method for objective weighting, thereby avoiding the influence of subjective bias or human interference on the comprehensive quality evaluation results of cigarettes. While ensuring scientific rigor, it has a solid theoretical basis in the industry and fully guarantees the accuracy and reliability of the weighting results. Moreover, after calculating the comprehensive quality metric value of cigarettes, a machine learning algorithm is used to model the comprehensive quality metric value of cigarettes. According to the constructed comprehensive quality metric value prediction model of cigarettes, the comprehensive quality metric value of cigarettes can be directly predicted from the raw and auxiliary material parameters and cigarette parameters of unknown cigarette samples to conduct comprehensive quality evaluation of cigarettes. There is no need to conduct random inspections of finished cigarettes during the production process, which provides a reliable and feasible solution for instant evaluation of the comprehensive quality of cigarettes.

[0115] If the expressions expressing orientation are mentioned in the embodiments of the present invention, they are relative concepts based on the embodiments. In addition, "at least one" means one or more, and "plurality" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, wherein a, b, c can be single or multiple.

[0116] The above describes in detail the structure, features and effects of the present invention based on the embodiments shown in the drawings, but the above is only a preferred embodiment of the present invention. It should be noted that the technical features involved in the above embodiments and their preferred methods can be reasonably combined and matched into a variety of equivalent schemes by those skilled in the art without departing from or changing the design ideas and technical effects of the present invention; therefore, the present invention is not limited to the scope of implementation shown in the drawings, and all changes made in accordance with the concept of the present invention, or modifications to equivalent embodiments with equivalent changes, which still do not exceed the spirit covered by the specification and drawings, should be within the protection scope of the present invention.

Claims

1. A method for instant evaluation of comprehensive quality of cigarettes based on independence weight analysis and machine learning, characterized in that: include: Collecting cigarette quality indicators, wherein the cigarette quality indicators include: cigarette ash characteristic indicators, smoke chemical indicators, and sensory evaluation indicators; De-dimensionalizing the cigarette quality index; Based on the independence weight analysis method, the weight of each of the cigarette quality indicators is determined, including calculation using the following formula: Among them, R i represents the multiple correlation coefficient; X j represents the jth cigarette quality index after dedimensionalization; H represents X minus X j The residual matrix of Represents the result of averaging X; the inverse of the complex correlation coefficient As the score, the weight value W is obtained after normalization i ; Based on the weights, the comprehensive quality metric of the cigarette is calculated, including using the following formula: Where, Y represents the comprehensive quality measurement value of cigarettes; W i Indicates the weight value of cigarette quality index; Z i It represents the de-dimensionalized cigarette quality index; The collected raw and auxiliary material parameters and cigarette parameters corresponding to the cigarette quality index are used as independent variables, and the comprehensive quality metric value of the cigarette is used as the dependent variable. A machine learning algorithm is used to establish and train a prediction model for the comprehensive quality metric value of the cigarette, and the output of the model includes the importance ranking results of the raw and auxiliary material parameters and the cigarette parameters; wherein the raw and auxiliary material parameters include at least one of the following: tobacco shred parameters, cigarette paper parameters, tipping paper parameters, forming paper parameters, and filter rod parameters; the cigarette parameters include at least one of the following: cigarette circumference, cigarette weight, standard draw resistance, closed draw resistance, cigarette paper ventilation rate, filter tip ventilation rate, total ventilation rate, cigarette length, and short cigarette length; The raw and auxiliary material parameters and cigarette parameters of the actual tested cigarette samples are input into the cigarette comprehensive quality metric value prediction model to obtain the cigarette comprehensive quality metric value of the actual tested cigarette samples.

2. The instant evaluation method for comprehensive cigarette quality based on independence weight analysis and machine learning according to claim 1 is characterized in that: The dedimensionalization process is calculated using the following formula: Among them, Z i represents the dimensionless cigarette quality index; y i Indicates the cigarette quality index obtained by testing cigarette samples; Represents the average value of each cigarette quality index.

3. The instant evaluation method for comprehensive cigarette quality based on independence weight analysis and machine learning according to claim 1, characterized in that: The cigarette ash characteristic index includes at least one of the following: gray color, crack rate, ash shrinkage rate, carbon line uniformity, carbon line width, and ash holding power; The smoke chemical index includes at least one of the following: tar release, nicotine release, and CO release; The sensory evaluation index includes at least one of the following: aroma, harmony, miscellaneous smell, irritation, and aftertaste.

4. The instant evaluation method for comprehensive cigarette quality based on independence weight analysis and machine learning according to claim 1, characterized in that: The tobacco parameters include: physical parameters and chemical parameters; Wherein, the physical parameters include at least one of the following: leaf weight, bulked weight, thin thread or flake weight, and stem weight; The chemical parameters include at least one of the following: total sugar, reducing sugar, total alkaloids, total nitrogen, nitrate, potassium, chlorine content, ratio of two sugars, sugar-alkali ratio, nitrogen-alkali ratio, potassium-chlorine ratio.

5. The instant evaluation method for comprehensive cigarette quality based on independence weight analysis and machine learning according to claim 1, characterized in that: The cigarette paper parameters include at least one of the following: fiber type, basis weight, air permeability, citrate content, potassium ion content, sodium ion content, and ash content; The tipping paper parameters include at least one of the following: air permeability, overlap length; The forming paper parameters include at least one of the following: basis weight, air permeability; The filter rod parameters include at least one of the following: length of the non-groove section of the filter rod, length of the groove section of the filter rod, circumference of the filter rod, pressure drop of a single filter rod, and specification of the tow.

6. The instant evaluation method for comprehensive cigarette quality based on independence weight analysis and machine learning according to any one of claims 1 to 5, characterized in that: The machine learning algorithm includes any one of the following: random forest algorithm, multivariate linear regression algorithm, neural network algorithm.

Images