阅读说明：本技术 一种打叶复烤配方模块的设计方法 (Design method of threshing and redrying formula module ) 是由 郝贤伟 毕一鸣 田雨农 廖付 何文苗 李石头 帖金鑫 李永生 张立立 慕继瑞 赵 于 2020-05-28 设计创作，主要内容包括：本发明提供一种基于近红外光谱的打叶复烤配方模块的设计的方法,包括如下步骤：采集样品粉末的近红外光谱；将六项化学指标、部位指数和香型指数合并为烟叶内在质量表征指数；对光谱进行预处理,利用偏最小二乘方法建立近红外光谱与内在质量间的关联模型,其中,建模方法为PLS2；采集目标配方模块中的各等级烟叶,以及拟用于配方的其他各等级烟叶,粉碎过筛,采集样品粉末的近红外光谱；针对原目标配方中的每个等级烟叶,从拟替代烟叶集合中选出N个最相近烟叶；将烟叶集合作为替代样本池；利用线性规划求解,得到与原配方最接近的各等级烟叶组合及其比例；结合评吸小组感官评价,验证替代方案的有效性。(The invention provides a design method of a threshing and redrying formula module based on near infrared spectrum, which comprises the following steps: collecting a near infrared spectrum of sample powder; combining the six chemical indexes, the part indexes and the odor indexes into a tobacco leaf intrinsic quality characterization index; preprocessing the spectrum, and establishing a correlation model between the near infrared spectrum and the internal quality by using a partial least square method, wherein the modeling method is PLS 2; collecting tobacco leaves of all grades in a target formula module and other tobacco leaves of all grades to be used for a formula, crushing and sieving, and collecting a near infrared spectrum of sample powder; selecting N most similar tobacco leaves from a set of tobacco leaves to be replaced aiming at each grade of tobacco leaves in the original target formula; collecting tobacco leaves as a substitute sample pool; solving by utilizing linear programming to obtain the tobacco combination of each grade and the proportion thereof which are closest to the original formula; the effectiveness of the alternatives was verified in conjunction with sensory evaluation by the panel.)

1. A design method of a threshing and redrying formula module is characterized by comprising the following steps:

1) collecting various tobacco leaf samples, crushing and sieving, and collecting the near infrared spectrum of sample powder;

2) after preparing a sample from YC/T31-1996 tobacco and tobacco product sample preparation and moisture determination, respectively determining the contents of total sugar, reducing sugar, total plant alkaloid, chlorine, total nitrogen and potassium in the tobacco and tobacco products according to YC/T159-2002 tobacco and tobacco product water-soluble sugar determination, YC/T160 + 2002 tobacco and tobacco product total plant alkaloid determination, YC/T162 + 2011 tobacco and tobacco product chlorine determination, YC/T161 + 2002 tobacco and tobacco product total nitrogen determination, YC/T217 + 2007 tobacco and tobacco product potassium determination, and detecting the total sugar, nicotine, reducing sugar, chlorine, potassium and total nitrogen in the sample;

3) the part of the sample is calibrated by flue-cured tobacco grading standard: the quantitative value of the upper leaf, the middle leaf and the lower leaf is 1,2 and 3 respectively; calibrating the style of the sample fragrance type according to the faint scent type, the middle fragrance type and the strong fragrance type indexes, and calculating the corresponding fragrance type indexes;

4) combining the six chemical indexes, the part indexes and the odor indexes into a tobacco leaf intrinsic quality characterization index;

5) preprocessing the spectrum, and establishing a correlation model between the near infrared spectrum and the internal quality by using a partial least square method, wherein the modeling method is PLS 2;

6) collecting tobacco leaves of all grades in a target formula module and other tobacco leaves of all grades to be used for a formula, crushing and sieving, and collecting a near infrared spectrum of sample powder;

7) giving the intrinsic quality characterization index of each sample in (6) by using the model in (5);

8) selecting N closest tobacco leaves from the set of the tobacco leaves to be replaced according to the step (7) aiming at each grade of tobacco leaves in the original target formula;

9) collecting the tobacco leaves selected in the step (8) as a substitute sample pool;

10) solving by utilizing linear programming to obtain the tobacco combination of each grade and the proportion thereof which are closest to the original formula;

11) and (5) verifying the effectiveness of the alternative scheme by combining the indexes in the step (7) with sensory evaluation of a panel of smoking groups.

2. The design method of claim 1, wherein the spectrum collection range in step (1) is: 10000cm^-1-4000cm^-1Spectral resolution of 8cm^-1The scanning times are 64 times;

the characterization form of the spectrum in the step (5) is as follows:

wherein X represents a spectrum matrix, each row represents a spectrum, n represents a spectrum dimension, and m represents a sampleAnd (4) counting.

The chemical and site characterization form in the step (5) is as follows:

wherein Y1 represents the corresponding attribute label, i.e. the label value of total sugar, nicotine, reducing sugar, chlorine, potassium, total nitrogen, etc. or part in the chemistry, and m represents the number of samples;

the characterization form of the odor type in the step (5) is as follows:

Y_xx＝[y_m1y_m2y_m3]wherein, Y_XXRepresents the individual scent index score, y, of the sample_m1Represents the index of faint scent, y_m2Represents the middle fragrance index, y_m3Representing aroma index, and m represents sample number;

the calculation formula of the odor index is Y_{Fragrance type}＝[y_m1*2+y_m2-y_m3-1]/2 wherein Y is_{Fragrance type}Representing the comprehensive score of the odor type index;

in the step (5), the inherent quality of the tobacco leaves is characterized as Y ═ Y_{Total sugar}y_Nicotiney_{Reducing sugar}y_Chloriney_{Potassium salt}y_{Total nitrogen}y_{Location of a body part}y_{Fragrance type}]Wherein Y represents the comprehensive score of each index;

the spectrum preprocessing method in the step (5) is a first derivative (Savitzky-Golay) processing, the window width parameter is 11, and the polynomial order parameter is 2;

in the step (5), the modeling method of the intrinsic quality is PLS2, namely 8-dimensional intrinsic indexes Y and X are modeled, regression is carried out by using a PLS2 method, and the number f of the main factors is determined by using leave-one-out cross validation;

in the step (6), the spectrum acquisition mode is the same as that in the step (1);

in the step (7), the predicted value of each sample is a vector of 1 x 8 of total sugar, nicotine, reducing sugar, chlorine, potassium, total nitrogen, part and odor;

in the step (8), the method for screening the similar tobacco leaves of each raw tobacco of the target module comprises the following steps: all samples in (7) are normalized by columnAnd (4) recording the normalized index value of the ith raw cigarette of the target module as H_i＝[h_i1h_i2...h_i8]The normalized index value of the jth raw cigarette to be selected is L_i＝[l_i1l_i2...l_i8]The difference between the two isWherein abs is an absolute value; calculating all raw cigarettes to be selected and ith raw cigarette of a target module, and selecting N cigarettes with the minimum S value; here, N takes the value of 8.

3. The design method as claimed in claim 2, wherein in step (9), all the raw tobaccos in the target module are traversed, and the selected raw tobaccos are collected into a set (repeatedly selected and deleted), and the set has M samples.

4. The design method according to claim 2, wherein in the step (10), eight indexes of the target module are recorded as R ═ R₁r₂...r₈]The optimization goal is to obtain the proportionality coefficients C ═ C of M raw tobaccos₁c₂...c_M]So thatWherein, c_p≥0p＝1,2,...,M；

Wherein r is_kIs the value of the kth index of the target module, c_pIs the proportion of each raw cigarette in the formula_pkIs the value of the k index of p raw tobacco.

Technical Field

A design method of a threshing and redrying formula module is specifically a tobacco formula design method based on near infrared spectrum fitting and index evaluation, and belongs to the field of tobacco processing formula design.

Background

The quality stability of cigarette products is an important basis for the stable and healthy development of cigarette enterprises, and the quality stability of cigarette products mainly depends on the stability of the quality of tobacco leaf raw materials, but the quality of tobacco leaves is influenced by varieties, ecological conditions and production measures, and has larger quality difference between years. At present, cigarette industry enterprises with strong competitiveness in China mainly improve the quality stability of tobacco strips by a modularized formula mode of a threshing and redrying processing link so as to improve the guarantee capability of the quality stability of cigarette products.

The modularized formula refers to a tobacco sheet module which is formed by matching and designing tobacco leaves with different producing areas, different varieties and different grades according to a certain proportion and then threshing and redrying the tobacco leaves to meet the use requirements of cigarette products. In the actual formula working process, due to the influences of weather conditions, cultivation measures and the like, the fluctuation of the quality of key raw materials required by the modular formula design is large, a plurality of small-scale tobacco lamina modules with single scale are generated, tobacco leaves with similar quality are required to be searched for in the product formula, and the product formula can be frequently adjusted. Therefore, how to enlarge the scale of the tobacco slice module in the design link of the tobacco slice module, improve the controllability of the quality of the tobacco slice module and keep the relative stability of the quality of the brand cigarettes becomes a key problem of research.

At present, the design of a tobacco lamina formula module is developed from the analysis of indexes such as chemical components, sensory quality, smoke quality and the like by mainly depending on the experience of a formulator to the analysis of multidimensional quality indexes through a mathematical statistics method to realize the tobacco leaf formula design. For example, Shenyujun et al (application patent No. 200610077683.7) disclose a method for establishing a threshing and redrying formula module, which combines a plurality of tobacco leaves with the same or similar style characteristics and different grades according to a certain proportion to form a formula module with a certain function or application in a cigarette formula.

Plum-dongliang et al (application patent No. 200810046379.4) disclose a design method of a cigarette leaf group formula, which primarily selects representative tobacco leaves of main material tobacco, auxiliary material tobacco and filler according to the experience of formula personnel around the design target and the chemical component requirements of a product, sequentially screens similar tobacco leaves for basic main material tobacco categories, screens other raw materials for auxiliary material tobacco and filler according to compatibility, and searches a plurality of optional leaf group formulas meeting the requirements by using formula cost as a fitness evaluation function and using the proportion of the chemical components, the main material, the auxiliary material and the filler as constraint conditions through a genetic algorithm.

Guo Xinfeng et al (application patent number: 200910116521.2) disclose a design method of cigarette formula, which divides main material cigarette, sub-standard main material cigarette and non-standard main material cigarette by sugar-base ratio and comprehensive quality, and sequentially forms modules of the main material cigarette, the standard main material cigarette and the non-standard main material cigarette, establishes a mutual combination test of 3 types of modules, and finally completes the design of cigarette formula by screening. The method mainly adopts a mode of sensory evaluation or chemical assistance to classify the tobacco leaves and then carry out the combination in different proportions, the formulation personnel needs to carry out continuous sensory evaluation on the experimental results and adjust the formulation, the cognition on the style and the quality characteristics, and the selection and the judgment of the main materials, the auxiliary materials and the filling materials are still influenced by strong subjectivity, the physiology and the psychology of the formulation personnel and fixed formulation experience or thinking set.

Lijun et al (application patent No. 201510652701.8) disclose a method for designing a tobacco leaf formulation, which obtains characteristic parameters of tobacco leaf raw materials by detection, sets target values of the characteristic parameters, processes the characteristic parameters by a weighted absolute value distance analysis method to obtain weighted absolute values of distances between measured values and the target values of the characteristic parameters, sorts the tobacco leaf raw materials in a sequence from small to large according to the weighted absolute values of the distances, and obtains a new tobacco leaf formulation by module design and skeleton design.

Songceng Fei et al (application patent number: 201810991053.2) disclose a tobacco leaf formula design method based on Fermat point and combined tobacco leaves, the invention obtains each chemical index and derivative index data of tobacco leaves, takes multiple indexes as the coordinate position of each tobacco leaf in a high-dimensional formula geometric space, calculates the geometric Fermat point in the high-dimensional space according to the coordinates, and obtains the final tobacco leaf formula proportion by adjusting the proportion of each tobacco leaf in the formula to enable the formula point (high-dimensional space gravity center point) to coincide with the Fermat point. The method adopts the original formula target value as a reference system, determines the formula through the similarity of sensory and chemical data indexes and reference, has certain objectivity and simplicity compared with the traditional formula mode, improves the formula design efficiency, but does not establish strong relevance between the chemical data and the internal quality of the tobacco leaves, and cannot realize better formula quality although better matching of the indexes is met.

The near infrared spectrum method is a fast, efficient and low-cost green analysis technology which is rapidly developed in recent years, and the near infrared spectrum method utilizes the universal frequency vibration or rotation of chemical bonds such as C-H, N-H, O-H, C-C and the like in organic matters to obtain an absorption spectrum in a near infrared region in a diffuse reflection mode, has the excellent characteristics of large information amount, good reproducibility and comprehensive quality information, is the comprehensive embodiment of tobacco quality information, and contains key chemical components, quality characteristic information, tobacco physical states and part of aroma components with higher content, so that the near infrared spectrum method has application prospects in analysis, evaluation, control and formula design of the comprehensive quality of tobacco.

The invention discloses a tobacco leaf formula improvement method, which comprises the steps of obtaining a proportion coefficient of each tobacco leaf raw material in an original formula through calculation processing by a non-negative regression coefficient regression method on the basis of detecting and obtaining the original tobacco leaf formula and near infrared spectrum data of each tobacco leaf raw material contained in the original tobacco leaf formula, sequencing the tobacco leaf raw materials according to the proportion coefficient from large to small, and obtaining an improved tobacco leaf formula through module design and framework design.

Zhang Feng et al (application patent number: 201210344034.3) disclose a method for assisting cigarette formulation with SIMCA based on near infrared spectrum information, which is characterized in that near infrared spectrum modeling is performed, a tobacco leaf sample to be replaced is taken as a target, a sample to be detected is scanned to obtain near infrared spectrum data, a replacement rule is set according to the information of the replacement sample, the near infrared data of the sample to be replaced is compared with a digital model in a memory to obtain the replaceable tobacco leaf samples, sorting is performed according to mahalanobis distance, the samples with smaller mahalanobis distance are more similar, and sensory evaluation verification is performed.

Wulijun et al (application patent number: 201811038973.9) a cigarette leaf group formula maintenance method, which searches for substitute tobacco leaves by sample collection, chemical component detection, sensory evaluation, spectrum pretreatment, characteristic spectrum extraction and construction of a spectrum similarity matrix.

Plum stone and the like (application patent number 201610156519.8) disclose a maintenance method of a raw tobacco leaf group formula based on near infrared spectrum, the invention adopts a method of combining near infrared and chemical components, and tobacco leaf varieties in the last year and the present year are divided into a common grade and an individual grade according to the consistency degree of the near infrared spectrum and the chemical value of the tobacco leaf varieties in the last year and the present year; aiming at the common grade, the same proportion as the previous year is adopted; aiming at the individual grades, combining tobacco varieties collected this year into a mixed sample, enabling the near infrared spectrum and the chemical value of the mixed sample to be respectively consistent with the corresponding individual grades, and obtaining a preliminary original tobacco leaf group formula of this year by adopting the proportion of the corresponding individual grades in the original tobacco leaf group formula of the last year according to the proportion of the mixed sample; if the difference between the chemical value and the near infrared spectrum of the sample prepared in the year and the sample prepared in the last year is expected, the raw tobacco leaf group formula in the year is obtained.

In the existing literature reports, two main ideas of tobacco formula design are provided, namely, the design and adjustment are carried out by combining various data of tobacco leaves and human sensory evaluation; and secondly, calculating a substitute formula by utilizing the near infrared spectrum in a mode of minimum spectral error. In the first approach, there are a large number of human-based sensory evaluations as formulas or reference data, which are labor intensive and difficult to inherit and reproduce. In the second way, the corresponding calculation only takes the minimum error as the optimization target, and the utilization of the formula experience is lacked, and the calculation result is often difficult to interpret and has low practicability.

Disclosure of Invention

The invention aims to design a formula with the quality similar to that of a target redried strip tobacco module, and the technical key points of the invention are different from the technical key points of the document:

1) according to the invention, the formula design is carried out according to the near infrared spectrum, and no record of formula personnel is needed in the design process, and no sensory evaluation data of a user and other data (such as smoke data of each tobacco leaf and the like) which need to be acquired at a higher cost are used.

2) The method does not use spectrum similarity as an optimization target, establishes the correlation between the near infrared spectrum and specific tobacco leaf internal information, and adopts the tobacco leaf internal information as the optimization target.

3) In the invention, a 'similar alternation' rule is provided to constrain the solving process. Before fitting calculation, firstly, respectively screening similar tobacco leaves aiming at each tobacco leaf in the original formula to form a candidate tobacco leaf set. In the solving process, samples are only selected from the candidate tobacco leaf set, and the problems that the mathematical calculation is similar, the tobacco leaf composition difference is large and the sense is dissimilar are solved.

4) The method adopts a mode of index evaluation and sensory verification to ensure the repeatability and the practicability of the method.

Specifically, in order to achieve the above object, the present invention provides a method for designing a threshing and redrying formula module based on near infrared spectrum, comprising the steps of:

1) collecting various tobacco leaf samples, crushing and sieving, and collecting the near infrared spectrum of sample powder;

2) after preparing a sample from YC/T31-1996 tobacco and tobacco product sample preparation and moisture determination, respectively determining the contents of total sugar, reducing sugar, total plant alkaloid, chlorine, total nitrogen and potassium in the sample according to YC/T159-2002 tobacco and tobacco product water-soluble sugar determination, YC/T160-2002 tobacco and tobacco product total plant alkaloid determination, YC/T162-2011 tobacco and tobacco product chlorine determination, YC/T161-2002 tobacco and tobacco product total nitrogen determination, YC/T217-2007 tobacco and tobacco product potassium determination, and detecting the total sugar, nicotine, reducing sugar, chlorine, potassium and total nitrogen in the sample;

3) the part of the sample is calibrated by flue-cured tobacco grading standard: the quantitative value of the upper leaf, the middle leaf and the lower leaf is 1,2 and 3 respectively; calibrating the style of the sample fragrance type according to the faint scent type, the middle fragrance type and the strong fragrance type indexes, and calculating the corresponding fragrance type indexes;

4) combining the six chemical indexes, the part indexes and the odor indexes into a tobacco leaf intrinsic quality characterization index;

5) preprocessing the spectrum, and establishing a correlation model between the near infrared spectrum and the internal quality by using a partial least square method, wherein the modeling method is PLS 2;

6) collecting tobacco leaves of all grades in a target formula module and other tobacco leaves of all grades to be used for a formula, crushing and sieving, and collecting a near infrared spectrum of sample powder;

7) giving the intrinsic quality characterization index of each sample in (6) by using the model in (5);

8) selecting N closest tobacco leaves from the set of the tobacco leaves to be replaced according to the step (7) aiming at each grade of tobacco leaves in the original target formula;

9) collecting the tobacco leaves selected in the step (8) as a substitute sample pool;

10) solving by utilizing linear programming to obtain the tobacco combination of each grade and the proportion thereof which are closest to the original formula;

11) and (5) verifying the effectiveness of the alternative scheme by combining the indexes in the step (7) with sensory evaluation of a panel of smoking groups.

In some embodiments, the spectrum collection range in step (1) is: 10000cm^-1-4000cm^-1Spectral resolution of 8cm^-1Number of scansIt was 64 times.

The characterization form of the spectrum in the step (5) is as follows:

where X represents a spectral matrix, one spectrum per row, n represents the spectral dimension, and m represents the number of samples.

The chemical and site characterization form in the step (5) is as follows:

where Y1 represents the corresponding attribute label, such as total sugar, nicotine, reducing sugar, chlorine, potassium, total nitrogen, etc. in chemistry or the label value of the site, and m represents the number of samples.

The characterization form of the odor type in the step (5) is as follows:

Y_xx＝[y_m1y_m2y_m3]wherein, Y_XXRepresents the individual scent index score, y, of the sample_m1Represents the index of faint scent, y_m2Represents the middle fragrance index, y_m3Representing aroma index, and m represents sample number;

the calculation formula of the odor index is Y_{Fragrance type}＝[y_m1*2+y_m2-y_m3-1]/2 wherein Y is_{Fragrance type}Representing the comprehensive score of the odor type index;

in the step (5), the inherent quality of the tobacco leaves is characterized as Y ═ Y_{Total sugar}y_Nicotiney_{Reducing sugar}y_Chloriney_{Potassium salt}y_{Total nitrogen}y_{Location of a body part}y_{Fragrance type}]Wherein Y represents the comprehensive score of each index;

the spectrum preprocessing method in the step (5) is a first derivative (Savitzky-Golay) processing, the window width parameter is 11, and the polynomial order parameter is 2.

In the step (5), the modeling method of the intrinsic quality is PLS2, namely 8-dimensional intrinsic indexes Y and X are modeled, regression is carried out by using a PLS2 method, and the number f of the main factors is determined by using leave-one-out cross validation;

the spectrum collection mode in the step (6) is the same as that in the step (1).

In the step (7), the predicted value of each sample is a vector of 1 × 8 of the components of total sugar, nicotine, reducing sugar, chlorine, potassium, total nitrogen, part and aroma.

In the step (8), the method for screening the similar tobacco leaves of each raw tobacco of the target module comprises the following steps: all samples in (7) are normalized according to columns, and the normalized index value of the ith raw cigarette of the target module is recorded as H_i＝[h_i1h_i2... h_i8]The normalized index value of the jth raw cigarette to be selected is L_i＝[l_i1l_i2... l_i8]The difference between the two isWherein abs is the absolute value of the calculation. And calculating all the raw cigarettes to be selected and the ith raw cigarette of the target module, and selecting N cigarettes with the minimum S value. Here, N takes the value of 8.

In some embodiments, in step (9), all the raw tobaccos in the target module are traversed, and the selected raw tobaccos are collected into a set (repeatedly selected and deleted), so that M samples are obtained.

In some embodiments, in step (10), the target module eight criteria is denoted as R ═ R₁r₂... r₈]The optimization goal is to obtain the proportionality coefficients C ═ C of M raw tobaccos₁c₂... c_M]So thatWherein the content of the first and second substances,

wherein r is_kIs the value of the kth index of the target module, c_pIs the proportion of each raw cigarette in the formula_pkIs the value of the k index of p raw tobacco.

The method of the invention has the following advantages:

the evaluation information of the formula design is obtained quickly, and a large amount of evaluation by formula personnel is not needed in the design process, and sensory evaluation data of people and other data (such as smoke data of each tobacco leaf and the like) which needs to be obtained at a higher cost are not used.

The existing method for designing the formula by adopting a near infrared spectrum fitting mode does not consider the limitation on the admittance of the components of the formula, so that the calculation result of the fitting reaches the design target (the spectral error is minimum), but the components and the proportion participating in the formula have larger difference with the original formula, and the common sense of the formula and the sensory evaluation result are not met.

The internal quality characterization of chemical, part and odor type components is adopted to replace near infrared spectroscopy for optimization solution, and the result is more interpretable.

And the repeatability and the practicability of the method are guaranteed by adopting a mode of index evaluation and sensory verification.

Detailed Description

The leaf tobacco module in the middle of HD in a certain province is a module which is successfully explored through a formula for many years, and because the raw tobacco resources in places such as H producing areas, D producing areas and the like are limited, other city and county tobacco leaves in the same province are supposed to be matched to form a quality which achieves or approaches the quality of the original formula. The actual recipe for the target module is shown in the table below.

TABLE 1 target Module ("HD" middle Module) recipe make-up

In order to achieve the aim, C2FA1, C3FA1 and C2FC3 grades of tobacco leaf producing places such as L, Y, N and P of S city, M of Z city, J and X of B city, G and W of C city, F of G city and the like are selected. The target sample is the 'HD' middle blade cigarette module.

The near infrared spectrum predicted values of various indexes of raw tobacco and raw tobacco to be selected in the target formula module are shown in a table 2:

TABLE 2 near infrared spectrum prediction value of each index of conventional chemistry of each production area

For the H origin and D origin C2FA1, C3FA1 and C2FC3 in the target module, respectively, the similar raw tobacco samples are calculated. The calculation method comprises the following steps: normalizing the indexes of the target sample and the sample to be selected, calculating the sum of the absolute differences of the indexes of the sample to be selected and the target sample, and recording the sum asl_k,h_kRespectively representing the k index values of the target and the sample to be selected.

Taking C2FA1 from the D-origin as an example, the most similar 10 raw smoke ratings are shown in table 3:

TABLE 3C 2FA1 level similarity sample ordering with D Producer

It can be seen that the selected sample and the target D-origin C2FA1 belong to the same province S city in the administrative division, and the geographical locations are closer, and the soil, climate, etc. are similar, so the tobacco leaf quality has greater similarity. In the prior literature, the process of 'similar substitution' in practice is not considered in formula calculation, and screening is carried out in a large number of samples, the selected formula is possibly similar to a target in a set index, but the difference of the tobacco leaves is large, the sensory evaluation has similar parts, but obvious difference, and the practicability is not strong. In the invention, similar tobacco leaf screening is carried out on each tobacco leaf in the target formula, so that the condition is avoided.

All 6 samples in the original formula are traversed to obtain a candidate tobacco leaf set as shown in table 4:

TABLE 4 set of candidate tobacco leaves for original formula sample

Calculating by using the step (10), namely solving a constraint least square problem, neglecting samples with the proportion of less than 1.0% in the obtained optimal solution, and obtaining similar formula modules of the target module, as shown in table 5:

TABLE 5 alternative Module recipe composition for target Module

The difference between the target module and the calculated surrogate module for each index is shown in table 6:

TABLE 6 comparison of quality indicators for target and surrogate modules

Table 6 shows the comparison of the quality indexes of the target module and the substitution module, and it can be known from table 6 that the substitution module obtained according to the present invention has close intrinsic quality of chemical, part and odor components compared with the target module, thereby ensuring the repeatability and practicability of the method.