Cigar tobacco dry ice expansion process parameter screening method and application thereof

Abstract

The invention relates to a method for screening dry ice expansion process parameters of cigar tobacco leaves and application of the method. The method comprises the following steps: analyzing the consistency of physical and chemical detection indexes of tobacco shreds by adopting a mean value weighted variable coefficient and a variance contribution weighted variable coefficient; analyzing the consistency of the smoke panel results by adopting a Kendall harmony coefficient; performing correlation analysis on dry ice expansion process intervention conditions and smoke panel results; performing correlation analysis on dry ice expansion process intervention conditions and physical and chemical detection indexes; analyzing correlation between the physical and chemical detection indexes of the tobacco shreds and smoking evaluation results; and evaluating and analyzing a machine learning algorithm. According to the screening method, a set of optimal intervention condition combination screening method which has high consistency and is based on smoke panel test results and physicochemical test results is constructed.

Description

Technical Field

[0001] The present invention relates to the field of tobacco leaf processing, and in particular to a method for screening process parameters of dry ice expansion of cigar tobacco leaves and an application thereof. Background Art

[0002] Dry ice expanded tobacco technology is a key tobacco production process in the industry. It involves impregnating cut tobacco with liquid carbon dioxide, forming dry ice inside and on its surface. The tobacco then flows rapidly through a venturi tube into a sublimation tube. The carbon dioxide rapidly sublimates in a hot stream of air at approximately 350°C, causing the tobacco to expand. Carbon dioxide is colorless, odorless, and non-toxic. It is abundant, inexpensive, and has a low vaporization point, enabling expansion rates of over 70% for cut tobacco. This technology also helps improve the physical quality of cut tobacco (such as leaf structure and fill value), reducing tobacco consumption and thus cigarette production costs. It also alters the chemical composition of cut tobacco (such as nicotine, total sugars, and reducing sugars), enhancing smoking safety. It also improves the sensory qualities of cigarettes (such as aroma, aroma volume, off-flavors, and irritation), enhancing tobacco quality, improving the applicability of lower-grade tobacco, fully utilizing tobacco resources, and increasing the economic benefits of cigarette production. The dry ice method for producing expanded cut tobacco is of great significance to the development of the tobacco industry. However, there are few reports on the research and application of dry ice expansion of cigar tobacco in the existing technology.

[0003] While existing technical solutions exist to improve the quality of cut-to-length cigars by adjusting the filler ratio, these solutions demonstrate that dry ice expansion treatment significantly enhances mouth, nose, and throat comfort, reduces smoke concentration and strength, and improves combustibility. However, the following challenges remain: First, the dry ice expansion process and parameters used for traditional cigarettes cannot be directly applied to cigars. Differences exist between traditional cigarettes and cigars in terms of variety characteristics, agricultural production techniques, and industrial processing methods, resulting in significant differences in the tobacco leaf condition. Therefore, the dry ice expansion process for cigar cut tobacco requires process parameter adjustments tailored to the specific characteristics of cigars. Second, cigar cut tobacco loses a certain amount of aroma quality and volume after dry ice expansion treatment. Preliminary dry ice expansion tests on cigar cut tobacco, as well as those on many other cigarette cut tobacco products, demonstrate that while impurities, smoke concentration, and strength are reduced, aroma quality and volume are also reduced. Therefore, achieving a comprehensive balance between these sensory quality indicators is a pressing issue. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide a method for screening process parameters of dry ice expansion of cigar tobacco leaves and its application.

[0005] In order to achieve the purpose of the invention, the present invention adopts the following technical solutions:

[0006] In a first aspect, the present invention provides a method for screening process parameters for dry ice expansion of cigar tobacco leaves, the method comprising the following steps:

[0007] S1. Consistency analysis of tobacco physicochemical test indicators: The mean-weighted coefficient of variation and variance-weighted coefficient of variation were used to quantitatively analyze the fluctuation characteristics of the physicochemical test indicators of cigar tobacco samples.

[0008] S2. Consistency analysis of smoking evaluation results: The consistency of smoking evaluation results was evaluated and analyzed using the Kendall concordance coefficient;

[0009] S3. Correlation analysis between dry ice expansion process intervention conditions and adsorption evaluation results;

[0010] S4. Correlation analysis between dry ice expansion process intervention conditions and physical and chemical testing indicators;

[0011] S5. Correlation analysis between tobacco physicochemical test indicators and smoking evaluation results;

[0012] S6. Machine Learning Algorithm Evaluation and Analysis: A machine learning model is used to model the physical and chemical testing indicators of cut tobacco, smoking evaluation results, and intervention conditions, and to screen for the optimal dry ice expansion process parameters for cigar tobacco leaves.

[0013] There is no distinction between the order of step S1 and step S2.

[0014] In the present invention, the dual-frame indicator fluctuation characteristic quantification method can ensure the consistency of sample detection, thereby providing more effective support for the auxiliary adjustment of scoring; and through the multi-dimensional consistency evaluation method, the final score is made more objective and accurate; finally, the combined features are used to perform multi-model training of the machine learning algorithm to obtain high-dimensional evaluation results, which greatly helps the selection of the optimal intervention conditions, thereby constructing a set of optimal intervention condition combination screening methods based on the evaluation results and physical and chemical test results.

[0015] Preferably, the intervention conditions of the dry ice expansion process include immersion time, hot air temperature and hot air speed.

[0016] Preferably, the physical and chemical testing indicators in step S1 include filling height, filling value, dry weight of cut tobacco and equilibrium moisture content.

[0017] Preferably, the calculation formula of the mean weighted coefficient of variation is as follows:

[0018]

[0019] Where MWCV is the mean weighted coefficient of variation, w i is the mean of the i-th feature, is the coefficient of variation of the i-th indicator, σ i is the standard deviation, μi is the mean.

[0020] Preferably, the calculation formula of the variance contribution weighted coefficient of variation is as follows:

[0021]

[0022] Where MWCV is the weighted coefficient of variation of variance contribution, is the coefficient of variation of the index, σ i is the standard deviation, μ i is the mean.

[0023] In the samples of the present invention, different physical and chemical test indicators adopt different dimensions. By adopting a dual evaluation framework of mean weighted coefficient of variation (MWCV) and variance contribution weighted coefficient of variation (VCWCV), the volatility characteristics of indicators are quantified in multiple dimensions. MWCV reduces dimensional interference by normalizing the means of different physical and chemical test indicators, while VCWCV amplifies discrete information through variance. The combination of the two can avoid the blind spots of single evaluation.

[0024] Preferably, the evaluation indicators in step S2 include quality characteristics and style characteristics.

[0025] Preferably, the quality characteristics include aroma quality, aroma quantity, off-flavor, pungency, aftertaste, burnability and gray; the style characteristics include concentration, strength and aroma type.

[0026] Preferably, step S2 specifically includes analyzing the authority and weight of different judges after evaluating and analyzing the consistency of the evaluation results using the Kendall concordance coefficient.

[0027] Preferably, the authority and weight analysis includes: standardizing and reducing the dimension of the evaluation and scoring results, calculating the authority and weight, and performing weighted averaging on the results to obtain the sample scores of the intervention conditions.

[0028] Preferably, the standardization is performed using a StandardScaler function.

[0029] Preferably, the dimension reduction method adopts principal component analysis.

[0030] Preferably, the method for calculating authority is to project the evaluation and scoring data of each judge into the principal component space of principal component analysis (PCA), calculate the square value after projection, that is, the contribution of each sample on each principal component, sum the square projection values ​​of all principal components, and obtain the total contribution of the judge, that is, the authority.

[0031] Preferably, the weight is calculated by calculating the total contribution of all judges and calculating the proportion of the total contribution of each judge, that is, the weight.

[0032] Preferably, the calculation formula for the sample score of the intervention condition is as follows:

[0033]

[0034] Where M represents the number of judges, N represents the evaluation index, which includes 7 quality characteristics and 3 style characteristics, a total of 10 indicators, and w i represents the weight of the i-th judge, s ij represents the score of the jth evaluation dimension evaluated by the i-th judge, where

[0035]

[0036] Preferably, the correlation analysis methods described in steps S3, S4 and S5 respectively and independently use Pearson correlation coefficient for analysis.

[0037] Preferably, the machine learning model described in step S6 includes any one of a LightGBM model, a multi-layer perceptron model or a support vector machine model, or a combination of at least two of them.

[0038] In the present invention, by utilizing combined features to perform machine learning multi-model training, high-dimensional evaluation results are obtained, and the weight-based results are fine-tuned to make the differences in the evaluation results more obvious, which greatly helps in the selection of optimal intervention conditions.

[0039] In a second aspect, the present invention provides a device for screening process parameters of dry ice expansion of cigar tobacco leaves, wherein the screening device is used to perform the screening method described in the first aspect.

[0040] Preferably, the device comprises:

[0041] Consistency analysis module for physical and chemical testing indicators of cut tobacco: This module uses the mean-weighted coefficient of variation and the variance-contribution-weighted coefficient of variation to quantitatively analyze the fluctuation characteristics of the physical and chemical testing indicators of cigar tobacco samples;

[0042] Consistency analysis module for smoking evaluation results: uses Kendall's harmony coefficient to evaluate and analyze the consistency of smoking evaluation results;

[0043] Correlation analysis module between dry ice expansion process intervention conditions and adsorption evaluation results;

[0044] Correlation analysis module between dry ice expansion process intervention conditions and physical and chemical testing indicators;

[0045] Correlation analysis module between tobacco physicochemical testing indicators and smoking evaluation results;

[0046] Machine learning algorithm evaluation module: a machine learning model is used to model the tobacco physicochemical detection index, evaluation result and intervention condition, and the optimal cigar leaf dry ice expansion process parameter is screened.

[0047] In a third aspect, the present application provides a computer readable storage medium, which stores a computer program or instructions, and the computer program or instructions are used to make the processor execute the screening method of the first aspect.

[0048] In a fourth aspect, the present application provides a computer device, which comprises a memory and a processor, and the memory stores a computer program or instructions, and the computer program or instructions are executed by the processor to realize the steps in the screening method of the first aspect.

[0049] In a fifth aspect, the present application provides a method for dry ice expansion of cigar leaf, which comprises dry ice expansion treatment of cigar leaf under the conditions that the immersion time is 48s, the hot air temperature is 310℃, and the hot air speed is 33m / s.

[0050] Compared with the prior art, the present application has the following beneficial effects:

[0051] In the present application, the index fluctuation characteristic quantification method of the double framework can guarantee the consistency of sample detection, thereby providing more effective support for the auxiliary adjustment of scoring; and the multi-dimensional consistency evaluation method makes the final score more objective and accurate; finally, the machine learning algorithm multi-model training is carried out by using the combined features, and a high-dimensional evaluation result is obtained, which greatly helps the selection of the optimal intervention condition, thereby constructing an optimal intervention condition combination screening method based on the evaluation result and the physicochemical detection result. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 is a correlation analysis diagram of dry ice expansion process intervention conditions and evaluation results;

[0053] Figure 2 is a data distribution diagram of the gray index under different intervention conditions;

[0054] Figure 3 is a data distribution diagram of the concentration index under different intervention conditions;

[0055] Figure 4 is a correlation analysis diagram of dry ice expansion process intervention conditions and physicochemical detection indexes;

[0056] Figure 5 is a 2-methylpyrazine index change trend thermodynamic diagram under the combined conditions of immersion time and hot air temperature;

[0057] Figure 6It is a thermodynamic diagram of the change trend of CO(%) index under the combination of immersion time and hot air temperature;

[0058] Figure 7 It is a thermodynamic diagram of the change trend of the total particulate matter (TPM) index under the combined conditions of immersion time and hot air temperature;

[0059] Figure 8 It is a correlation analysis matrix diagram of physical and chemical test indicators and absorption evaluation results;

[0060] Figure 9 It is a line graph showing the relationship between the absorption evaluation index and the key physical and chemical testing indexes. DETAILED DESCRIPTION

[0061] The technical solution of the present invention is further described below by way of specific embodiments. It should be understood by those skilled in the art that the embodiments are merely to help understand the present invention and should not be regarded as specific limitations of the present invention.

[0062] The cigar tobacco leaves used in the following examples are mixed tobacco leaves.

[0063] Example

[0064] This embodiment provides a method for screening process parameters for dry ice expansion of cigar tobacco leaves, which is as follows:

[0065] S1. Analysis of consistency of physical and chemical testing indicators of cut tobacco

[0066] The mean-weighted coefficient of variation and variance-contribution-weighted coefficient of variation were used to quantitatively analyze the fluctuation characteristics of four physical and chemical testing indicators of cigar tobacco leaf samples, including filling height, filling value, tobacco dry weight, and equilibrium moisture content. The mean-weighted coefficient of variation and variance-contribution-weighted coefficient of variation of each physical and chemical testing indicator of cigar tobacco leaf samples under different dry ice expansion process intervention conditions (immersion time, hot air temperature, and hot air speed) were calculated. The consistency results of different intervention condition combinations are shown in Table 1.

[0067] The calculation formula of the mean weighted coefficient of variation is as follows:

[0068]

[0069] Where MWCV is the mean weighted coefficient of variation, w i is the mean of the i-th feature, is the coefficient of variation of the i-th indicator, σ i is the standard deviation, μ i is the mean;

[0070] The calculation formula of the variance contribution weighted coefficient of variation is as follows:

[0071]

[0072] Where MWCV is the weighted coefficient of variation of variance contribution, is the coefficient of variation of the index, σ i is the standard deviation, μ i is the mean.

[0073] Table 1

[0074]

[0075] According to the data in the table, when the immersion time was 48 seconds, the temperature was 310°C, and the speed was 33 m / s (Group 3), both the MWCV (0.061) and VCWCV (0.094) reached their lowest values, indicating that the mean fluctuation and variance contribution of the test indicators under these conditions were effectively controlled, making it the most accurate and stable test result combination within the current experimental range. When the immersion time was 39 seconds, the temperature was 298°C, and the speed was 37 m / s (Group 2) and when the immersion time was 75 seconds, the temperature was 304°C, and the speed was 35 m / s (Group 6), the VCWCV was significantly higher than the other groups, indicating that the variance contribution weights of these two groups of parameters were higher.

[0076] The coefficient of variation threshold was set at no more than 40%. Based on the analysis results in Table 1, the detection consistency of each group of samples was obtained as shown in Table 2. According to the results in the table, the research data basically met the consistency requirements.

[0077] Table 2

[0078] Group Parameter combination MWCV VCWCV Consistency judgment 1 30s, 286℃, 34m / s 0.093362 0.140029 Fully compliant 2 39s, 298℃, 37m / s 0.122242 0.435124 Partially compliant 3 48s, 310℃, 33m / s 0.060873 0.09431 Fully compliant 4 57s, 280℃, 36m / s 0.086293 0.175542 Fully compliant 5 66s, 292℃, 32m / s 0.081596 0.148348 Fully compliant 6 75s, 304℃, 35m / s 0.090929 0.510713 Partially compliant

[0079] S2. Consistency analysis of smoking evaluation results

[0080] (1) A total of nine judges evaluated and scored the samples of the six groups under different intervention conditions in S1. The evaluation indicators included quality characteristics (aroma quality, aroma volume, impurities, pungency, aftertaste, burnability, and gray) and style characteristics (concentration, strength, and aroma type). The consistency of the evaluation results was evaluated and analyzed using the Kendall concordance coefficient. The consistency of the scores of each judge in each group was statistically analyzed. The results are shown in Table 3.

[0081] Table 3

[0082] Group Immersion time (s) Hot air temperature (℃) Hot air speed (m / s) consistency 4 57 280 36 0.285296 1 30 286 34 0.344147 5 66 292 32 0.203198 2 39 298 37 0.426976 6 75 304 35 0.454592 3 48 310 33 0.279348

[0083] According to the data in the table, the consistency among the judges is low, and the Kendall coefficients of some samples are lower than 0.5, indicating that there are certain differences in the judges' scoring of the samples.

[0084] (2) Based on the scoring consistency results, it is necessary to analyze the review authority of each judge to obtain the evaluation authority and weight of different judges, as follows:

[0085] ① Use the StandardScaler function to standardize the scoring data (mean is 0, variance is 1);

[0086] ②Use principal component analysis (PCA) to reduce the dimension of the standardized data;

[0087] ③Authority calculation: Project each judge's scoring data into the principal component space of PCA, calculate the square value after projection, that is, the contribution of each sample on each principal component, and sum the square projection values ​​of all principal components to obtain the total contribution of the expert, that is, the authority;

[0088] ④ Weight calculation: Calculate the total contribution of all judges and the proportion of each judge's total contribution, i.e., the weight;

[0089] ⑤Further calculate the average Euclidean distance between each judge's score and all other judges' scores to evaluate the specificity of each judge's score;

[0090] ⑥ Take a weighted average of the judges’ evaluation results and calculate the sample score;

[0091] The sample score for the intervention condition was calculated as follows:

[0092]

[0093] Where M represents the number of judges, N represents the evaluation index, which includes 7 quality characteristics and 3 style characteristics, a total of 10 indicators, and w i represents the weight of the i-th judge, s ij represents the score of the jth evaluation dimension evaluated by the i-th judge, where

[0094] The authority and weight results of each judge in the above ③, ④⑤ and the Euclidean distance of scores (expressed as average value) are shown in Table 4, and the sample scores in ⑤ are shown in Table 5.

[0095] Table 4

[0096] Group Hot air speed (m / s) consistency Scoring Euclidean distance A 52.14704 0.103466 1.850304 B 54.82065 0.108771 2.004847 C 40.57137 0.080499 1.549164 D 81.88361 0.162467 1.927684 E 83.73297 0.166137 1.900488 F 39.19074 0.077759 1.860914 G 37.12365 0.073658 1.853986 H 46.68674 0.092632 1.603312 I 67.84324 0.13461 1.626061

[0097] Table 5

[0098] Group Immersion time (s) Hot air temperature (℃) Hot air speed (m / s) Weighted score 2 39 298 37 49.17827 1 30 286 34 50.13352 6 75 304 35 50.44345 5 66 292 32 51.75919 4 57 280 36 52.63593 3 48 310 33 52.64625

[0099] The results in the table show that the highest weighted score was achieved when the immersion time was 48 seconds, the hot air temperature was 310°C, and the hot air speed was 33 m / s (Group 3). A result very close to the highest weighted score was achieved when the immersion time was 57 seconds, the hot air temperature was 280°C, and the hot air speed was 36 m / s (Group 4).

[0100] S3. Correlation analysis between dry ice expansion process intervention conditions and adsorption evaluation results

[0101] The Pearson correlation coefficient was used to analyze the correlation between the dry ice expansion process intervention conditions and the adsorption evaluation results. Figure 1 As shown in the figure, according to the analysis results, the correlation between each smoking index and the immersion time and hot air temperature is relatively low, and there is no obvious linear correlation. To further determine the correlation, the data distribution of each smoking index under each intervention condition combination is plotted. Taking the gray index as an example and ignoring the influence of wind speed, the distribution diagram is as follows Figure 2 and Figure 3 As shown in the figure, the gray fraction shows no regular change with increasing hot air temperature, and initially rises and then falls with increasing immersion time. The concentration data distribution under different conditions shows no clear trend, and neither does the aroma quality or aroma volume. Therefore, it is necessary to establish intermediate variables based on external characteristics to increase the correlation between intervention conditions and smoking results.

[0102] S4. Correlation analysis between dry ice expansion process intervention conditions and physical and chemical testing indicators

[0103] (1) The Pearson correlation coefficient was used to analyze the correlation between the dry ice expansion process intervention conditions and physical and chemical test indicators. The analysis results are as follows: Figure 4 As shown in the figure, -1 indicates a completely negative correlation, 1 indicates a completely positive correlation, and 0 indicates no linear relationship. It can be seen from the figure that in terms of physical related indicators, the filling value, filling height, tobacco dry weight and equilibrium moisture content are closely related to the intervention conditions; in terms of chemical indicators, the correlation between each indicator and the intervention conditions is significantly different. The relationship models between them and the smoking evaluation results were established respectively. Specifically, the 30 most important features of the three intervention conditions of impregnation time, hot air temperature and hot air speed were selected, and the union was taken as the physical and chemical indicators most closely related to the intervention conditions, as shown in Table 6.

[0104] Table 6

[0105]

[0106] According to the indicators in the table and combined Figure 4 It can be seen that the correlation between physical indicators and intervention conditions is higher than that between chemical indicators. The relevant features of the three intervention conditions are combined to obtain the final set of key physical and chemical detection indicators as shown in Table 7.

[0107] Table 7

[0108]

[0109]

[0110] According to the above statistical analysis data and the correlation analysis results in the figure, the fluctuation range of hot air speed in all samples is within 10% and is relatively continuous.

[0111] (2) Considering the hot air speed as a constant variable, a thermodynamic diagram of the changing trends of various physical and chemical indicators under different combinations of immersion time and hot air temperature is drawn as follows:

[0112] ① The relationship between 2-methylpyrazine, immersion time and hot air temperature is as follows Figure 5 As shown in the figure, it can be seen that with the increase of immersion time and hot air temperature, the content of the component first increases and then decreases, and there is an interval extreme value. When the immersion time and hot air temperature are in the middle value, the 2-methylpyrazine content reaches the peak.

[0113] ②The relationship between CO and impregnation time and hot air temperature is as follows Figure 6 As shown in Figure 2, it can be seen that the CO content continues to increase with the increase of hot air temperature.

[0114] ③The relationship between TPM, immersion time and hot air temperature is as follows Figure 7 As shown in the figure, it can be seen that due to the different combination conditions of variables, the mutual influence leads to different production and precipitation degrees of chemical components, which makes the TPM index have more than one extreme point with the increase of immersion time and hot air temperature.

[0115] In summary, it can be seen that there is an obvious relationship between the effects of different intervention condition combinations on physical and chemical detection indicators, but it is not a fragrance relationship because the model selected is a nonlinear model.

[0116] S5. Correlation analysis between tobacco physicochemical testing indicators and smoking evaluation results

[0117] (1) The Pearson correlation coefficient was used to analyze the correlation between the physical and chemical test indicators of tobacco and the smoking evaluation results, and a correlation matrix diagram of the physical and chemical test indicators and the smoking evaluation results was obtained, as shown in the figure: Figure 8 As shown in the figure, different physical and chemical testing indicators have different effects on the evaluation and scoring results. Therefore, for each evaluation indicator, the 40 most relevant physical and chemical testing features are counted from high to low, and then the number of times all physical and chemical testing features rank in the top 40 in the 10 evaluation dimensions is counted. The indicators that appear more than or equal to three times are selected as key physical and chemical testing indicators, totaling 42, as shown in Table 8.

[0118] Table 8

[0119]

[0120]

[0121] The key physical and chemical test indicators in Table 8 were compared with those in Table 7 using the intersection-and-union method to calculate coverage and assess the correlation between the two. The calculated intersection-and-union ratio for the key physical and chemical test indicators in Tables 7 and 8 was 12.8%, indicating that 12.8% of the physical and chemical test indicators most closely related to the intervention conditions and the smoking assessment results were identical. Therefore, physical and chemical test indicators can, to a certain extent, expand the relationship characteristics between smoking assessment results and intervention condition combinations.

[0122] (2) Draw a line graph showing the relationship between the absorption index and key physical and chemical testing indicators. Figure 9 As shown below:

[0123] ①As the physical and chemical indicators increase, the scores of the absorption indicators first increase and then decrease.

[0124] As can be seen from the figure, as the 2,6-lutidine content increases, the irritation index score first increases and then decreases.

[0125] ②As the physical and chemical indicators increase, the scores of the absorption indicators continue to rise.

[0126] As can be seen from the figure, as the pyridine content increases, the score of the aftertaste index shows an overall upward trend.

[0127] ③As the physical and chemical test indicators increase, the scores of the absorption index decrease first and then increase.

[0128] As can be seen from the figure, as the Pyridine content increases, the momentum index score first decreases and then increases.

[0129] ④As the physical and chemical test indicators increase, there is no obvious trend in the scores of the absorption index.

[0130] As can be seen from the figure, with the change of butyl acetate, the change of concentration index does not show an obvious trend response, which may be due to the subjective fluctuation of the evaluation results and the mutual influence between chemical indicators.

[0131] S6. Evaluation and Analysis of Machine Learning Algorithms

[0132] (1) Three nonlinear machine learning models, namely the LightGBM model, the multi-layer perceptron model (MLP) and the support vector machine model (SVM), were used to model the physical and chemical detection indicators, smoking evaluation results and intervention conditions of tobacco. After training, the test sample scores of the six groups involved in the program were further averaged to obtain the final prediction scores under the six intervention conditions, as shown in Table 9.

[0133] Table 9

[0134]

[0135]

[0136] According to the data in the table, the overall trends of the prediction scores of the three models are relatively consistent. The highest score is obtained when the immersion time is 48s, the hot air temperature is 310℃, and the hot air speed is 33m / s (Group 3), which is consistent with the weighted score results of the samples in S2.

[0137] (2) According to the prediction results of machine learning, the weighted average results of the authoritative results of the smoking evaluation results in S2 were weighted again to obtain the final score. The weight of the weighted average result based on the authoritative results of the smoking evaluation results was 0.8, and the weight of the machine learning algorithm score fine-tuning result based on the combined characteristics of physical and chemical test indicators and intervention conditions was 0.2. The results are shown in Table 10.

[0138] Table 10

[0139] Group Immersion time (s) Hot air temperature (℃) Hot air speed (m / s) Final score 2 39 298 37 48.16062 1 30 286 34 49.61482 6 75 304 35 50.06276 5 66 292 32 51.62735 4 57 280 36 52.39274 3 48 310 33 52.747

[0140] The results in the table show that the optimal combination of dry ice expansion process parameters for cigar tobacco leaves is an immersion time of 48 seconds, a hot air temperature of 310°C, and a hot air speed of 33 m / s (Group 3). This once again verifies that the screening method provided by the present invention can make the difference in evaluation results more obvious, greatly assisting in the selection of optimal intervention conditions for related processes, and constructing a highly consistent screening method for optimal intervention condition combinations based on smoking evaluation results and physical and chemical test results.

[0141] The applicant declares that the present invention is illustrated by the above-described embodiments, but the present invention is not limited to the above-described embodiments. This does not mean that the present invention must rely on the above-described embodiments in order to be implemented. Those skilled in the art should understand that any improvements to the present invention, equivalent replacements for raw materials in the present invention, additions of auxiliary ingredients, and selection of specific methods, etc., fall within the scope of protection and disclosure of the present invention.

[0142] The preferred embodiments of the present invention are described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, various simple modifications can be made to the technical solution of the present invention, and these simple modifications all fall within the scope of protection of the present invention.

[0143] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any appropriate manner without contradiction. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations.

Claims

1. A method for screening process parameters for dry ice expansion of cigar tobacco leaves, characterized in that: The method comprises the following steps: S1. Consistency analysis of tobacco physicochemical test indicators: The mean-weighted coefficient of variation and variance-weighted coefficient of variation were used to quantitatively analyze the fluctuation characteristics of the physicochemical test indicators of cigar tobacco samples. S2. Consistency analysis of smoking evaluation results: The consistency of smoking evaluation results was evaluated and analyzed using the Kendall concordance coefficient; S3. Correlation analysis between dry ice expansion process intervention conditions and adsorption evaluation results; S4. Correlation analysis between dry ice expansion process intervention conditions and physical and chemical testing indicators; S5. Correlation analysis between tobacco physicochemical test indicators and smoking evaluation results; S6. Machine Learning Algorithm Evaluation and Analysis: A machine learning model is used to model the physical and chemical testing indicators of cut tobacco, smoking evaluation results, and intervention conditions, and to screen for the optimal dry ice expansion process parameters for cigar tobacco leaves. There is no distinction between the order of step S1 and step S2.

2. The screening method according to claim 1, wherein The intervention conditions of the dry ice expansion process include immersion time, hot air temperature and hot air speed.

3. The screening method according to claim 1, wherein The physical and chemical testing indicators in step S1 include filling height, filling value, dry weight of cut tobacco and equilibrium moisture content; Preferably, the calculation formula of the mean weighted coefficient of variation is as follows: Where MWCV is the mean weighted coefficient of variation, w i is the mean of the i-th feature, is the coefficient of variation of the i-th indicator, σ i is the standard deviation, μ i is the mean; Preferably, the calculation formula of the variance contribution weighted coefficient of variation is as follows: Where MWCV is the weighted coefficient of variation of variance contribution, is the coefficient of variation of the index, σ i is the standard deviation, μ i is the mean.

4. The screening method according to claim 1, wherein The evaluation indicators in step S2 include quality characteristics and style characteristics; Preferably, the quality characteristics include aroma quality, aroma quantity, miscellaneous smell, pungency, aftertaste, burnability and gray; the style characteristics include concentration, strength and aroma type; Preferably, step S2 specifically includes analyzing the authority and weight of different judges after evaluating and analyzing the consistency of the evaluation results using Kendall's concordance coefficient; Preferably, the authority and weight analysis includes: standardizing and reducing the dimension of the evaluation and scoring results, calculating the authority and weight, and performing weighted averaging on the results to obtain the sample scores of the intervention conditions; Preferably, the calculation formula for the sample score of the intervention condition is as follows: In the formula, M represents the number of judges, N represents the evaluation indicators, including 7 quality characteristics and 3 style characteristics, a total of 10 indicators, w i represents the weight of the i-th judge, s ij represents the score of the jth evaluation dimension evaluated by the i-th judge, where 5. The screening method according to claim 1, wherein The correlation analysis methods described in steps S3, S4 and S5 are analyzed independently using the Pearson correlation coefficient.

6. The screening method according to claim 1 or 4, characterized in that The machine learning model described in step S6 includes any one of a LightGBM model, a multi-layer perceptron model or a support vector machine model, or a combination of at least two of them.

7. A device for screening process parameters of dry ice expansion of cigar tobacco leaves, characterized in that: The screening device is used to perform the screening method according to any one of claims 1 to 6.

8. The device according to claim 7, characterized in that The device comprises: Consistency analysis module for physical and chemical testing indicators of cut tobacco: This module uses the mean-weighted coefficient of variation and the variance-contribution-weighted coefficient of variation to quantitatively analyze the fluctuation characteristics of the physical and chemical testing indicators of cigar tobacco samples; Consistency analysis module for smoking evaluation results: uses Kendall's harmony coefficient to evaluate and analyze the consistency of smoking evaluation results; Correlation analysis module between dry ice expansion process intervention conditions and adsorption evaluation results; Correlation analysis module between dry ice expansion process intervention conditions and physical and chemical testing indicators; Correlation analysis module between tobacco physicochemical testing indicators and smoking evaluation results; Machine learning algorithm evaluation module: A machine learning model is used to model the physical and chemical testing indicators of tobacco, smoking evaluation results and intervention conditions, and to screen out the optimal dry ice expansion process parameters for cigar tobacco leaves.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program or instructions, and the computer program or instructions are used to enable a processor to implement the screening method according to any one of claims 1 to 6 when executed.

10. A computer device, characterized in that: The computer device includes a memory and a processor, wherein the memory stores a computer program or instructions, and when the computer program or instructions are executed by the processor, the steps in the screening method according to any one of claims 1 to 6 are implemented.

Images