A method for determining tobacco leaf formulation based on similarity of key chemical indicators

Abstract

The invention discloses a method for determining a tobacco leaf formula based on similarity of key chemical indicators, comprising the following steps: determining key chemical indicators related to sensory quality; obtaining key chemical indicator values ​​and quantities of a tobacco leaf formula of the previous year, namely a target formula, and key chemical indicator values ​​and quantities of alternative tobacco leaves of the current year; normalizing the values ​​of each indicator of the alternative tobacco leaves and tobacco leaves with the target formula; dividing the normalized sample data into a target sample, a mean sample, a mandatory sample, and a to-be-selected sample by comparing the production area and grade of each tobacco leaf sample with the target formula; selecting a number of samples close to the target sample from the to-be-selected samples, and forming a total sample together with the mandatory sample; taking minimizing the difference between the normalized indicator values ​​of the to-be-selected tobacco leaves and the normalized indicator values ​​of the mean samples in the total sample as an objective function, taking the usage conditions of the mandatory samples and the alternative samples as constraints, solving the objective function, and obtaining a combination of tobacco leaves of each grade and their proportions that are closest to the target formula.

Description

Technical Field

[0001] The present invention relates to a method for determining tobacco leaf formula based on the similarity of key chemical indicators. Specifically, the method utilizes key chemical indicators that are strongly correlated with sensory quality in combination with mathematical statistics methods to perform formula simulation design, and belongs to the field of tobacco leaf formula design. Background Art

[0002] The continued development of national cigarette brands has placed higher demands on the cost and quality of tobacco leaf raw materials. However, high-quality tobacco resources are limited. Industrial enterprises have a higher demand for middle-grade tobacco than for upper and lower-grade tobacco, but the actual output of high-quality middle-grade tobacco is low, at approximately 30%. Furthermore, tobacco leaf quality is primarily influenced by the environment, variety, and production practices, with the environment having the greatest impact. The quality of tobacco leaves varies from one producing area to another, with domestic industrial enterprises showing strong demand for tobacco leaves from high-quality producing areas. These factors, coupled with the "big brands, big markets" development strategy, have further accentuated the supply and demand imbalance for tobacco leaves, hindering the sustainable development of the tobacco industry.

[0003] Tobacco leaf blending involves rationally combining tobacco leaves from various production areas, grades, and styles to achieve the desired quality for cigarette production. The blends are then threshed and re-roasted to create tobacco sheets suitable for the tobacco industry. Tobacco leaf blending, when raw materials are scarce, can enhance the value of tobacco leaves by combining various types of tobacco, making it a preferred method for tobacco leaf quality improvement by many industrial enterprises. Furthermore, tobacco leaf blending can be adjusted based on the quality of the tobacco leaves of the current year to achieve consistent quality from year to year, thus maintaining the stability of the raw material quality of the product.

[0004] Currently, tobacco leaf formula design has evolved from relying primarily on the formulator's experience to combining analysis of indicators such as chemical composition, sensory quality, and smoke quality. There are also design methods that consider formula scale. For example, Chinese patent application publication number CN111671123A discloses a leaf-threshing and redrying formula design method. This method first constructs a formula efficacy positioning model; determines suitability evaluation weights, and establishes a method for allocating to-be-allocated raw materials. The efficacy positioning model is then used to predict the category of the to-be-allocated raw materials. The suitability evaluation weights are used to calculate the suitability scores and suitability indices for the to-be-allocated raw materials, and the formula design for the to-be-allocated raw materials is performed according to the allocation principle. This method uses conventional chemical composition and sensory quality as operational information, and is significantly influenced by subjective factors. For example, Chinese patent application publication number CN111543668A discloses a method for designing leaf threshing and redrying formulas. The method uses partial least squares to establish a correlation model between near-infrared spectra and intrinsic quality. Six chemical indicators, a near-infrared predicted site index, and an aroma index are combined into an intrinsic quality characterization index for tobacco leaves. A similarity algorithm is then used to select the N most similar tobacco leaves from the set of proposed replacement leaves. Using linear programming, a combination of tobacco leaves of various grades and their proportions that are closest to the original formula are obtained. However, the use of linear programming does not fully consider the scale and quantity of the formula. The actual designed formula may deviate from the actual operation of the formula. Furthermore, the model needs to be maintained and updated annually during the formulation process, leaving room for improvement in its prediction accuracy. Furthermore, selecting only the most similar quality leaves makes it difficult to meet module scale requirements. Summary of the Invention

[0005] The purpose of the present invention is to solve the problem that tobacco leaf formula design is highly subjective and only focuses on the similarity of target quality indicators while ignoring the actual output quantity of the formula. A method for determining tobacco leaf formula using key chemical indicators related to sensory quality indicators is provided.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] A method for determining tobacco leaf formulation based on similarity of key chemical indicators, comprising:

[0008] Identify key chemical indicators related to sensory quality;

[0009] Obtain the key chemical index values ​​and quantities of each component of the previous year's tobacco leaf formula, as well as the key chemical index values ​​and inventory quantities of the current year's alternative tobacco leaves, and use the previous year's tobacco leaf formula as the target formula;

[0010] Normalizing the key chemical index values ​​of the candidate tobacco leaves and the target blend tobacco leaves to obtain normalized index data of each tobacco leaf sample;

[0011] By comparing the production area and grade of each tobacco leaf sample with the target formula, the normalized index data of each tobacco leaf sample is divided into target samples, mean samples of target samples, mandatory samples and candidate samples; wherein the mandatory samples are samples of the candidate tobacco leaves that have the same subject as the target formula;

[0012] For each target sample, select several samples that are closest to it from the candidate samples; combine all the selected samples to obtain replacement samples; the replacement samples and the required samples constitute the total sample;

[0013] Based on the comparison between the inventory quantity of the mandatory sample and the quantity of the same item in the target formula, set the usage quantity conditions of the mandatory sample; and set the usage quantity conditions of the alternative sample based on the inventory quantity of the alternative sample;

[0014] Taking minimizing the difference between the normalized index value of the candidate tobacco leaves in the total sample and the normalized index value of the mean sample as the objective function, an optimization model is constructed with the usage conditions of the required samples and the alternative samples as constraints. The optimization model is solved to obtain the combination of tobacco leaves of various grades and their proportions that are closest to the target formula.

[0015] Furthermore, the key chemical indicators are determined according to the following method:

[0016] Collect various tobacco leaf samples, cut each type of tobacco leaf sample into shreds and mix them evenly, take a portion of them and grind and sieve them for chemical composition determination to obtain chemical indicators, and roll the remaining portion into cigarettes for sensory quality evaluation to obtain sensory quality indicators;

[0017] Conduct a simple correlation analysis between the chemical indicators and sensory quality indicators of various tobacco leaves obtained, and select chemical indicators with significant differences in correlation coefficients with at least a set proportion of sensory quality indicators as primary key indicators;

[0018] The sensory quality indexes of each type of tobacco leaves were summed up to obtain the total smoking score. Based on the total smoking score, all tobacco leaves were divided into three categories: usable, expandable, and unusable.

[0019] Perform variance analysis on the primary key indicators of the three types of tobacco leaves, and use those with significant differences in variance analysis as secondary key indicators;

[0020] Compare the 95% confidence intervals of the secondary key indicators that have reached significant differences in variance analysis among the three types of tobacco leaves, and select the indicators with less overlap and clear boundaries between the three types of tobacco leaves as the tertiary key indicators;

[0021] Factor analysis was performed on the three-level key indicators to determine the key chemical indicators.

[0022] Furthermore, the factor analysis of the three-level key indicators is performed to determine the key chemical indicators, including:

[0023] Factor analysis was conducted on the three-level key indicators, and the factor with a cumulative variance explanation rate of >90% after rotation was taken as the representative factor;

[0024] By analyzing the load coefficients of the representative factors by various third-level key indicators, the specific indicators mainly reflected by each representative factor are determined, and the determined specific indicators are used as key chemical indicators.

[0025] Furthermore, the key chemical indicators include: nicotine, total sugar, reducing sugar, sugar-alkali ratio, total nitrogen, starch, ash, scopoletin and methyl palmitate.

[0026] Furthermore, for each target sample, several samples closest to it are selected from the candidate samples, including:

[0027] Calculate the similarity between each target sample and all candidate samples, and select the first n candidate samples with the smallest similarity values.

[0028] Furthermore, the similarity between the target sample and the candidate sample is calculated according to the following method:

[0029] Assume that the normalized index value of the target sample i-th tobacco leaf is H i =[h i1 h i2 ...h iK ], the normalized index value of the jth tobacco leaf in the selected sample is L i =[l j1 l j2 ...l jK ], then the similarity S between the target sample i-th tobacco leaf and the candidate sample j-th tobacco leaf ij Calculated according to the following formula:

[0030]

[0031] Among them, K is the number of normalized index values, l jk is the kth normalized index value of the jth tobacco leaf in the selected sample, h ik is the kth normalized index value of the i-th tobacco leaf of the target sample.

[0032] Furthermore, the objective function is:

[0033]

[0034] Where Q is the difference between the normalized index value of the selected tobacco leaves in the total sample and the normalized index value of the mean sample; K is the number of normalized index values; r k is the kth normalized index value of the mean sample; c pis the proportional coefficient of the pth tobacco leaf, p=1,2,...,m+q, m+q is the total number of samples, m is the number of replacement samples, q is the number of required samples, c p ≥0, z p is the dosage of the pth tobacco leaf to be solved, y pk is the kth normalized index value of the pth tobacco leaf.

[0035] Furthermore, the setting of the usage condition of the mandatory sample based on the comparison between the inventory quantity of the mandatory sample and the quantity of the same subject of the target formula includes:

[0036] Compare the inventory quantity of a certain tobacco leaf in the mandatory sample with the quantity of the same subject in the target formula. If the quantity of a certain subject in the mandatory sample is ≤ the quantity of the corresponding subject in the target formula, the entire inventory is required to be used. If the quantity of a certain subject in the mandatory sample is ≥ the quantity of the corresponding subject in the target formula, the usage amount can be adjusted.

[0037] Furthermore, the replacement sample requires that the usage ratio of the total sample quantity is not less than 1% and is less than the inventory quantity.

[0038] Furthermore, the said available, expandable, and unavailable are specifically: available means the total score of the evaluation and absorption is ≥60 points, expandable means 30 points < total score of the evaluation and absorption <60 points, and unavailable means the total score of the evaluation and absorption is ≤30 points.

[0039] Compared with the prior art, the present invention has the following beneficial technical effects:

[0040] (1) Using objective chemical indicators related to sensory quality for formula design can solve the problem of strong subjectivity of the formula, achieve a balance between objective indicators and subjective evaluation, and is practical;

[0041] (2) The use of objective data can stably monitor the changes in tobacco leaf quality from year to year, meet the actual operation of the formula, respect the continuity of tobacco leaf procurement from year to year, improve the efficiency of formula design, and achieve annual stability of tobacco leaf module quality by controlling the consistency of key quality indicators. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0043] The present invention will be further described below in conjunction with specific examples. The following examples are only used to more clearly illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention.

[0044] The present invention provides a method for determining tobacco leaf formula based on the similarity of key chemical indicators. The method detects chemical indicators, obtains sensory evaluation data, establishes correlation analysis to determine key chemical indicators, and obtains the tobacco leaf inventory quantity. The main formula is determined by matching the target formula production area and grade. The main formula is used as a calculation unit, and together with the remaining tobacco leaf units, it constitutes a formula calculation candidate library. The method uses mathematical statistics methods to simulate and calculate formulas similar to the target.

[0045] like Figure 1 As shown, the method specifically includes:

[0046] Step 1: Identify key chemical indicators related to sensory quality;

[0047] 1) Collect samples of various upper tobacco leaves. Remove all stems and cut into shredded tobacco (0.9 ± 0.1 mm wide). Mix each type of shredded tobacco thoroughly, then collect 80 g of each type, dry at 60°C, pulverize, and pass through a 40-mesh sieve. Prepare laboratory samples for chemical composition analysis according to YC / T31-1996 for chemical composition determination. The remaining shredded tobacco was rolled into cigarettes for sensory quality evaluation.

[0048] 2) Routine chemical composition and trace element detection: Total alkaloids, total sugars, reducing sugars, total nitrogen, potassium, chlorine, pH, and starch were detected using the Zhejiang Zhongyan near-infrared mathematical model; calcium: refer to YC / T 174-2003; sulfur: refer to YC / T 284-2009; magnesium: refer to YC / T 175-2003; phosphorus: refer to DB53 / T 357-2011; boron refer to [1] Wang Zhirong, Wang Yunhua. Study on the decolorization method for determining available boron in soil by azomethine method [J]. Soil and Fertilizer, 1991(2):3.; ether extract: refer to YC / T 176-2003; polyphenols: refer to YC / T 202-2006; free amino acids: refer to YC / T 282-2009. Glandular hair secretion components: refer to Fu Qiujuan, Du Yongmei, Liu Xinmin, et al. Determination of tobacco cedartrienediol by ultra-high performance liquid chromatography. Chinese Tobacco Science, 2017, 38(3): 67-73.; Non-volatile organic acids and higher fatty acids: refer to Liu Baizhan, Xu Liang, Hu Bianxia, ​​et al. Analysis of non-volatile organic acids and certain higher fatty acids in cigarettes. Tobacco Science and Technology, 2000(1): 25-27.

[0049] The chemical indicators detected are shown in the following table:

[0050] Table 1 Chemical index categories and index names

[0051]

[0052] 3) Sensory quality evaluation

[0053] Zhejiang China Tobacco organized smoking evaluation experts to conduct sensory quality evaluation on cigarettes made from various types of collected tobacco leaves based on Zhejiang China Tobacco corporate standards to obtain sensory quality indicators.

[0054] Sensory quality indicators include fineness, clarity, aroma, impurities, smoothness, pungency, aftertaste, smoke concentration, and strength.

[0055] The sensory quality indicators of each type of tobacco leaf are summed to obtain the total smoking score for that type of tobacco leaf. Based on the total smoking score, tobacco leaves can be divided into three categories: usable (≥60 points), expandable (30-60 points), and unusable (≤30 points).

[0056] 4) Primary key indicator screening

[0057] Simple correlation analysis was performed using spss23 statistical analysis software, and the chemical composition indexes with correlation coefficients reaching significant differences (P < 0.05) with 60% of the sensory quality indexes were selected as primary key indicators;

[0058] Table 2 Simple correlation analysis between main chemical components and sensory quality indicators

[0059]

[0060]

[0061] Note: *: P < 0.05, **: P < 0.01, aroma (total) = delicate feeling + clarity + aroma amount

[0062] A simple correlation analysis was conducted between 47 chemical component indicators in each category and 10 sensory quality indicators. The results are shown in Table 2. Indicators with 6 or more significant correlations with sensory quality indicators were selected as primary key indicators. Among them, total sugar, reducing sugar, starch, sugar-alkali ratio, difference between two sugars, sucrose ester, methyl palmitate, unsaturated fatty acid methyl ester, methyl stearate and other indicators were positively correlated with most sensory quality indicators, while nicotine, total nitrogen, ash, total magnesium, scopoletin, ether extract and malic acid were negatively correlated with sensory quality indicators.

[0063] Based on the above analysis, a total of 16 indicators including nicotine, total sugar, reducing sugar, sugar-alkali ratio, difference between two sugars, total nitrogen, starch, ash, total magnesium, malic acid, ether extract, sucrose ester, cypermethrin, methyl palmitate, unsaturated fatty acid methyl ester and methyl stearate were selected as primary key indicators.

[0064] 5) Secondary key indicator screening

[0065] According to Zhejiang China Tobacco's overall evaluation of various types of upper tobacco raw materials, the sensory quality indicators obtained from the sensory quality evaluation of each type of tobacco raw materials are added together to obtain the total smoking score. Based on the total smoking score, all tobacco raw materials are divided into three categories: usable, expandable, and unusable.

[0066] A variance analysis was performed on the values ​​of 16 primary key indicators according to the three categories of raw materials: "usable, expandable, and unusable". The results are shown in Table 3. The indicators with significant differences (P < 0.05) in the variance analysis were selected as secondary key indicators.

[0067] Table 3 Variance analysis of the preliminary selection indicators of three types of upper leaf raw materials for the overall evaluation of Zhejiang Zhongyan

[0068]

[0069]

[0070] Table 3 shows that the three types of raw materials showed extremely significant differences in nicotine, total sugar, reducing sugar, sugar-alkali ratio, difference between the two sugars, total nitrogen, starch, ash, total magnesium, scopolamine, malic acid, methyl palmitate, unsaturated fatty acid methyl esters, and methyl stearate. Ether extracts and sucrose esters showed no differences and were not selected as secondary key indicators. Finally, 14 secondary key indicators were selected through variance analysis: nicotine, total sugar, reducing sugar, sugar-alkali ratio, difference between the two sugars, total nitrogen, starch, ash, total magnesium, malic acid, scopolamine, methyl palmitate, unsaturated fatty acid methyl esters, and methyl stearate.

[0071] 6) Three-level key indicator screening

[0072] The 95% confidence intervals of chemical indicators with significant differences in variance analysis of the three types of raw materials were compared, and the three types of raw materials with less overlapping intervals, better distinction and clear boundaries between indicator categories were selected as the third-level key indicators.

[0073] Table 4 95% confidence intervals for secondary key indicators of three types of upper tobacco leaf raw materials in the overall evaluation of Zhejiang China Tobacco

[0074]

[0075]

[0076] Table 4 shows that for the seven indicators (nicotine, reducing sugars, starch, total sugars, sugar-alkali ratio, difference between the two sugars, and nicotine), there was little overlap between the three categories, indicating good differentiation between the groups. The "usable" category was well differentiated from the "expandable" category and the "unusable" category, while there was significant overlap between the "expandable" category and the "unusable" category, indicating poor differentiation. For the ash indicator, there was overlap between the three categories, indicating moderate differentiation between the groups. For the two indicators (total nitrogen and methyl palmitate), there was good differentiation between the "usable" category and the "unusable" and "expandable" categories, while the differentiation between the "unusable" category and the "expandable" category was relatively poor. The four indicators (total magnesium, methyl stearate, unsaturated fatty acid methyl esters, and malic acid) showed some ability to differentiate between the three raw material categories, but there was significant overlap between the categories, indicating poor differentiation. A 95% confidence interval indicates that 95% of the sample indicator values ​​fall within this interval, indicating good differentiation and clear boundaries between the indicator categories, demonstrating good overall evaluation and classification of the raw materials. Therefore, a total of 10 indicators, including nicotine, reducing sugar, starch, total sugar, ash, sugar-alkali ratio, difference between the two sugars, cypermethrin, total nitrogen and methyl palmitate, were further screened as third-level key indicators, and the indicator screening analysis and confirmation were continued.

[0077] 7) Final key indicator screening

[0078] Factor analysis was performed on the three-level key indicators to eliminate collinearity issues. Factors with a cumulative variance explanation rate of >90% after rotation were used as representative factors. The loading coefficients of the various indicators on the representative factors were analyzed to identify the specific indicators reflected by each factor. These specific indicators were used as key chemical indicators for formula design calculations.

[0079] Ten third-level key indicators were selected as candidate indicators and factor analysis was performed to extract representative factors. The final key chemical indicators were determined through factor analysis. Table 5 shows that the factor analysis extracted five factors. The variance explanation rates of these five factors after rotation were 25.40%, 22.60%, 18.90%, 15.13%, and 12.77%, respectively. The cumulative variance explanation rate after rotation was 94.804%, indicating that the five factors explained 94.804% of the total data.

[0080] Table 5 Variance explained by factor analysis

[0081]

[0082] As shown in Table 6, the loading coefficients of each indicator on the five factors reveal that Factor 1 primarily reflects ash content, nicotine, and the sugar-alkali ratio; Factor 2 primarily reflects total sugar, reducing sugar, and ash content; Factor 3 reflects scopolamine and methyl palmitate; Factor 4 reflects total nitrogen; and Factor 5 reflects starch. Therefore, nine indicators—ash content, nicotine, sugar-alkali ratio, total sugar, reducing sugar, scopolamine, methyl palmitate, total nitrogen, and starch—were selected as key chemical indicators for formulation design.

[0083] Table 6 Factor loading coefficients after rotation

[0084]

[0085] Note: *p<0.05 **p<0.01

[0086] The correlation analysis between the 9 selected key chemical indicators and the total score of the inhalation evaluation was performed, and the results are shown in Table 7. As can be seen from Table 7, the correlation coefficients between the 9 selected indicators and the total score of the inhalation evaluation all reached a significant level, indicating that there is a strong linear relationship between the 9 selected indicators and the total score of the inhalation evaluation.

[0087] Step 2: Obtain the key chemical index values ​​and quantities of each component of the tobacco leaf formula of the previous year, as well as the key chemical index values ​​and inventory quantities of the alternative tobacco leaves (inventory resources) of the current year, and use the tobacco leaf formula of the previous year as the target formula;

[0088] Among them, the subject represents a certain grade in a certain production area, and the same subject means that the tobacco production area and grade are the same.

[0089] Table 8 Key chemical indicators, quantities and proportions of target formula

[0090]

[0091]

[0092] Select the same subjects as the previous year and compare the number of the same subjects. If the number of a certain subject in the current year is ≤ the number of subjects corresponding to the target formula, it constitutes the "main formula". If the number of a certain subject in the current year is ≥ the number of subjects corresponding to the target formula, the quantity of the previous year will be included in the "main formula" as the first-level alternative library for formula design. The remaining quantity and non-corresponding subjects constitute the second-level alternative library for formula design selection, as shown in Table 9, Table 10, and Table 11.

[0093] Table 9 Comparison of inventory resources and target formula resources

[0094] serial number Production area grade Inventory resources (dan) Target formula quantity (dan) Gap quantity (dan) 1 Huidong, Liangshan B2FA1 9000 13275 -4275 2 Guiyang, Chenzhou B2FA1 2432 6064 -3632 3 Sanmenxia Lushi B2FA1 3539.5 5690 -2150.5 4 Luoyang Luoning B2FA1 3000 5647 -2647 5 Enshi Lichuan B2FA1 4573 5521 -948 6 Enshi Jianshi B2FA1 6188.5 4140 2048.5 7 Dechang, Liangshan B2FA1 3000 3871 -871 8 Dejiang, Tongren B2FA1 4538 3455 1083 9 Ningxiang, Changsha B2FA1 2824 3314 -490 10 Sanmenxia Lingbao B2FA1 2000 2880 -880 11 Nanyang Fangcheng B2FA1 1098.5 2534 -1435.5 12 Xuyong, Luzhou B2FA1 4000 1845 2155 13 Phoenix in western Hunan B2FA1 — 2736 -2736 14 Yongzhou Jianghua B2FA1 — 1315 -1315 — total B2FA1 46193.5 58236 -16093.5

[0095] Table 10: First-level candidate library for formulation design

[0096]

[0097]

[0098] Table 11 Secondary candidate library for formulation design

[0099]

[0100]

[0101] Step 3: normalizing the key chemical index values ​​of the candidate tobacco leaves and the target blend tobacco leaves to obtain normalized index data of each tobacco leaf sample;

[0102] All samples (x) of the 9 indicators are normalized (y) using the maximum and minimum normalization method. The specific formula is as follows (where y max =1,y min =0):

[0103]

[0104] The index value data of the candidate tobacco leaves and the target blend tobacco leaves were normalized according to the above-mentioned maximum and minimum normalization method, and the normalized data of 61 tobacco leaf samples were obtained, as shown in Table 12.

[0105] Table 12 Normalized data of tobacco leaf samples

[0106]

[0107]

[0108] Step 4: By comparing the production area and grade of each tobacco leaf sample with the target formulation, the normalized index data of each tobacco leaf sample is divided into target samples, mean samples of target samples, required samples, and samples to be selected;

[0109] The origin and grade information of the 61 tobacco leaf samples in Table 12 were compared with the target samples, and the normalized sample data was divided into four categories: samples numbered 47-60 were in the target formula, i.e., target samples (t = 14), and numbered 61 was the weighted mean of the target samples (mean sample, 1); numbers 14, 17, 20, 22, 26, 28, 30, 34, 37, 12, 24, and 39 were alternative samples with the same subject as the target formula, i.e., required samples (q = 12), and the remaining numbered tobacco leaves were samples to be selected (w = 34). The details are shown in Table 13.

[0110] Table 13 Information table of required samples and alternative samples

[0111]

[0112]

[0113] Step 5: For each target sample, select several samples that are closest to it from the candidate samples; combine all the selected samples to obtain replacement samples; the replacement samples and the required samples constitute the total sample;

[0114] Calculate the similarity value S (see Table 14) between each target sample and the 34 candidate samples, and select the first n candidate samples with the smallest similarity; and repeat the calculation for the 14 target samples, combine the selected sample sets and delete the duplicate samples to obtain the replacement sample pool set M.

[0115] The calculation method of the similarity value S between the target sample and the sample to be selected is:

[0116] The normalized index value of the target sample i-th tobacco leaf is H i =[h i1 h i2 ...h i9 ], the normalized index value of the jth tobacco leaf in the selected sample is L i =[l j1 l j2 ...l j9 ], the difference between the two is defined as:

[0117]

[0118] Among them, S ij is the similarity between the target sample i-th tobacco leaf and the candidate sample j-th tobacco leaf, S ij The smaller it is, the higher the similarity is; jk is the kth normalized index value of the jth tobacco leaf in the selected sample, h ik is the kth normalized index value of the i-th tobacco leaf of the target sample.

[0119] Compute all candidate samples against the target formulation's i-th tobacco leaf, and select n samples with the smallest S values. In this example, n is 6. Traverse all target samples and aggregate the selected candidate samples into a set (delete duplicates), resulting in a total of m samples as the selected tobacco leaf replacement sample pool set M.

[0120] Through Matlab programming, it is calculated that the set M contains 26 samples (m=26), which are numbered 1-4, 8-11, 13, 15-16, 18-19, 23, 25, 27, 29, 31-33, 40-42, and 44-46. The 12 mandatory samples and 26 alternative samples constitute the total sample.

[0121] Table 14 Similarity S between 34 candidate samples and 14 target samples

[0122]

[0123]

[0124] Step 6: Based on the comparison between the inventory quantity of the mandatory sample and the quantity of the same item in the target formula, set the usage quantity conditions of the mandatory sample; and set the usage quantity conditions of the alternative sample based on the inventory quantity of the alternative sample;

[0125] Set the usage conditions of mandatory samples according to actual conditions, for example, mandatory samples must be a certain value or a certain range.

[0126] In this embodiment, samples No. 14, 17, 20, 22, 26, 28, 30, 34, and 37 are mandatory samples and all stocks are used, while samples No. 12, 24, and 39 are mandatory samples and the usage amount is adjustable; set M is the pool of samples to be selected and the usage amount is arbitrary.

[0127] In addition, to ensure practical operability, the usage of the final selected alternative samples is required to account for no less than 1% of the total number of samples and be less than the inventory.

[0128] Step 7: Taking minimizing the difference between the normalized index value of the tobacco leaves to be selected in the total sample and the normalized index value of the mean sample as the objective function, and constructing an optimization model with the usage conditions of the required samples and the alternative samples as constraints, the optimization model is solved to obtain the combination of tobacco leaves of various grades and their proportions that are closest to the target formula.

[0129] The objective function is:

[0130]

[0131] Among them, r k is the kth normalized index value of the mean sample; c p is the proportion coefficient of the pth tobacco leaf in the total sample, p=1,2,...,38, c p ≥0, z p is the dosage of the pth tobacco leaf to be solved, y pk is the normalized value of the kth indicator of the pth tobacco leaf.

[0132] The fmincon function in Matlab was used to solve the nonlinear programming problem and the combination and proportion of tobacco leaves of various grades closest to the original formula were obtained.

[0133] To ensure that nonlinear programming can obtain z p For the global optimal solution, a penalty function is introduced in the optimization iteration. The sequential quadratic programming (SQP) algorithm is used here. The maximum iteration is set to 10,000 times, and the minimum calculated value of 5,000 random initial values ​​is used as the final solution. The tobacco leaf replacement plan of the target module is obtained, and the weighted average of each indicator is calculated.

[0134] A total of 23 tobacco leaves were optimized and selected, including 9 mandatory all-stock samples, 3 mandatory samples with adjustable dosage, and 11 samples with arbitrary dosage in sample pool M, namely 2, 13, 15, 18, 23, 25, 27, 31, 40-42, with a total dosage of 68795.2.8, as shown in Table 15.

[0135] Table 15 Calculation of replacement module recipe information

[0136] serial number Production area grade Dosage percentage 14 Luoyang Luoning B2FA1 3000.0 4.36％ 17 Nanyang Fangcheng B2FA1 1098.5 1.60％ 20 Sanmenxia Lingbao B2FA1 2000.0 2.91％ 22 Sanmenxia Lushi B2FA1 3539.5 5.14％ 26 Enshi Lichuan B2FA1 4573.0 6.65％ 28 Guiyang, Chenzhou B2FA1 2432.0 3.54％ 30 Ningxiang, Changsha B2FA1 2824.0 4.10％ 34 Dechang, Liangshan B2FA1 3000.0 4.36％ 37 Huidong, Liangshan B2FA1 9000.0 13.08％ 12 Dejiang, Tongren B2FA1 4538.0 6.60％ 24 Enshi Jianshi B2FA1 6188.5 9.00％ 39 Xuyong, Luzhou B2FA1 4000.0 5.81％ 2 Xuancheng, Southern Anhui B2FA1 968.0 1.41％ 13 Dejiang, Tongren B3FA1 825.4 1.20％ 15 Luoyang Luoning B3FA1 2254.8 3.28％ 18 Nanyang Fangcheng B3FA1 1211.9 1.76％ 23 Sanmenxia Lushi B3FA1 1475.4 2.14％ 25 Enshi Jianshi B3FA1 6186.4 8.99％ 27 Enshi Lichuan B3FA1 3544.3 5.15％ 31 Ningxiang, Changsha B3FA1 2620.6 3.81％ 40 Chuxiong Mouding B2FA1 941.5 1.37％ 41 Dali Auspicious Clouds B2FA1 1347.6 1.96％ 42 Honghe Luxi B2FA1 1225.8 1.78％ total — — 68795.2 100.00％

[0137] Based on 9 chemical indicators and combined with the sensory quality evaluation of the smoking panel, the calculated tobacco replacement plan was fine-tuned to obtain the final tobacco replacement plan for this module.

[0138] Table 16 Comparison of target module and alternative module indicators

[0139]

[0140] It can be seen from Table 16 that, except for methyl palmitate, the relative errors of other indicators are all less than 10%.

[0141] Table 17 Comparison of sensory quality between target formula and alternative formula

[0142]

[0143]

[0144] Sensory evaluation of the target and alternative formulations revealed that, as shown in Table 17, sensory quality indicators other than aftertaste were not significantly different from those of the alternative formulation (p>0.05). The target formulation scored better in terms of stimulation and aftertaste, but slightly lower in terms of aroma. Combining near-infrared predicted quality indicators with sensory evaluation results, the alternative formulations were similar to the target formulation in terms of intrinsic quality indicators for chemical composition, body part, and aroma profile, and exhibited a high degree of similarity in sensory indicators, demonstrating the practical value of this method in identifying alternative tobacco formulations.

[0145] The present invention has been disclosed above with preferred embodiments, which are not intended to limit the present invention. Any technical solutions obtained by adopting equivalent replacement or equivalent transformation solutions fall within the protection scope of the present invention.

Claims

1. A method for determining tobacco leaf formulation based on similarity of key chemical indicators, characterized in that: include: Identify key chemical indicators related to sensory quality; Obtain the key chemical index values ​​and quantities of each component of the previous year's tobacco leaf formula, as well as the key chemical index values ​​and inventory quantities of the current year's alternative tobacco leaves, and use the previous year's tobacco leaf formula as the target formula; Normalizing the key chemical index values ​​of the candidate tobacco leaves and the target blend tobacco leaves to obtain normalized index data of each tobacco leaf sample; By comparing the production area and grade of each tobacco leaf sample with the target formula, the normalized index data of each tobacco leaf sample is divided into target samples, mean samples of target samples, mandatory samples and candidate samples; wherein the mandatory samples are samples of the candidate tobacco leaves that have the same subject as the target formula; For each target sample, select several samples that are closest to it from the candidate samples; combine all the selected samples to obtain replacement samples; the replacement samples and the required samples constitute the total sample; Based on the comparison between the inventory quantity of the mandatory sample and the quantity of the same item in the target formula, set the usage quantity conditions of the mandatory sample; and set the usage quantity conditions of the alternative sample based on the inventory quantity of the alternative sample; Taking minimizing the difference between the normalized index value of the candidate sample and the normalized index value of the mean sample in the total sample as the objective function, and taking the usage conditions of the required samples and the replacement samples as the constraints to construct an optimization model, the optimization model is solved to obtain the combination and proportion of tobacco leaves of various grades that are closest to the target formula; The key chemical indicators are determined according to the following method: Collect various tobacco leaf samples, cut each type of tobacco leaf sample into shreds and mix them evenly, take a portion of them and grind and sieve them for chemical composition determination to obtain chemical indicators, and roll the remaining portion into cigarettes for sensory quality evaluation to obtain sensory quality indicators; Conduct a simple correlation analysis between the chemical indicators and sensory quality indicators of various tobacco leaves obtained, and select chemical indicators with significant differences in correlation coefficients with at least a set proportion of sensory quality indicators as primary key indicators; The sensory quality indexes of each type of tobacco leaves were summed up to obtain the total smoking score. Based on the total smoking score, all tobacco leaves were divided into three categories: usable, expandable, and unusable. Perform variance analysis on the primary key indicators of the three types of tobacco leaves, and use those with significant differences in variance analysis as secondary key indicators; Compare the 95% confidence intervals of the secondary key indicators that have reached significant differences in the variance analysis of the three types of tobacco leaves, and select the indicators with less overlap and clear boundaries between the three types of tobacco leaves as the tertiary key indicators; Conduct factor analysis on the three-level key indicators to determine the key chemical indicators; The factor analysis of the three-level key indicators is performed to determine the key chemical indicators, including: Factor analysis was conducted on the three-level key indicators, and the factors with a cumulative variance explanation rate of >90% after rotation were taken as representative factors; By analyzing the load coefficients of the three-level key indicators on the representative factors, the specific indicators mainly reflected by each representative factor are determined, and the determined specific indicators are used as key chemical indicators; The objective function is: ； Where Q is the difference between the normalized index value of the selected sample and the normalized index value of the mean sample in the total sample; K is the number of normalized index values; is the kth normalized index value of the mean sample; is the proportional coefficient of the pth tobacco leaf, , is the total number of samples, is the number of replacement samples, is the required number of samples, , , is the dosage of the pth tobacco leaf to be solved, is the kth normalized index value of the pth tobacco leaf.

2. The method for determining tobacco leaf formulation based on similarity of key chemical indicators according to claim 1, characterized in that: The key chemical indicators include: nicotine, total sugar, reducing sugar, sugar-alkali ratio, total nitrogen, starch, ash, scopoletin and methyl palmitate.

3. The method for determining tobacco leaf formulation based on similarity of key chemical indicators according to claim 1, characterized in that: For each target sample, several samples closest to it are selected from the candidate samples, including: Calculate the similarity between each target sample and all candidate samples, and select the first n candidate samples with the smallest similarity values.

4. The method for determining tobacco leaf formulation based on similarity of key chemical indicators according to claim 3, characterized in that: The similarity between the target sample and the selected sample is calculated according to the following method: Assume that the normalized index value of the target sample i-th tobacco leaf is , the normalized index value of the jth tobacco leaf in the selected sample is , then the similarity between the target sample i-th tobacco leaf and the candidate sample j-th tobacco leaf is Calculated according to the following formula: ； Among them, K is the number of normalized index values, is the kth normalized index value of the jth tobacco leaf in the sample to be selected, is the kth normalized index value of the i-th tobacco leaf of the target sample.

5. The method for determining tobacco leaf formulation based on similarity of key chemical indicators according to claim 1, characterized in that: The usage conditions of the mandatory samples are set based on the comparison between the inventory quantity of the mandatory samples and the quantity of the same items in the target formula, including: Compare the inventory quantity of a certain tobacco leaf in the mandatory sample with the quantity of the same subject in the target formula. If the quantity of a certain subject in the mandatory sample is ≤ the quantity of the corresponding subject in the target formula, the entire inventory is required to be used. If the quantity of a certain subject in the mandatory sample is ≥ the quantity of the corresponding subject in the target formula, the usage amount can be adjusted.

6. The method for determining tobacco leaf formulation based on similarity of key chemical indicators according to claim 1, characterized in that: The proportion of replacement samples required to be used is not less than 1% of the total sample quantity and is less than the inventory quantity.

7. The method for determining tobacco leaf formulation based on similarity of key chemical indicators according to claim 2, characterized in that: Available if the total score is ≥60 points, expandable if 30 points < total score <60 points, unavailable if the total score is ≤30 points.

Images