Cigarette tobacco blend composition analysis method based on fusion graphs

Abstract

Disclosed in the present invention is a cigarette tobacco blend composition analysis method based on fusion graphs, comprising the following steps: (1) preparing a cigarette sample to be analyzed and single-grade tobacco leaf samples; (2) constructing fusion graphs of said cigarette sample and the single-grade tobacco leaf samples; and (3) performing analysis on the basis of the fusion graphs to obtain tobacco leaf composition and proportions of the cigarette to be analyzed. The method of the present invention can complete the composition analysis of a finished cigarette within a few minutes to obtain clear blend composition and proportions, is objective and efficient, has high universality, good repeatability, high sensitivity, and has unique advantages in finished cigarette analysis in the tobacco industry.

Description

A method for analyzing tobacco leaf composition based on fusion graphs Technical Field

[0001] The present invention belongs to the technical field of tobacco, and in particular relates to a method for analyzing the composition of tobacco leaves based on a fusion atlas. Background Art

[0002] The quality and style of cigarettes are primarily determined by product designers through the blending of tobacco leaves from different origins, varieties, and grades. This typically requires relying on formula experience and sensory evaluation. From a stockpile of hundreds of raw material grades, 10-20 types of tobacco are manually selected and formulated in varying proportions. This results in extremely complex leaf composition, making it difficult to manually analyze the composition of unknown tobacco leaves. Being able to analyze leaf composition through instrumental testing, objective data, and scientific techniques would be crucial for analyzing competitive cigarettes and designing leaf compositions.

[0003] In recent years, near-infrared (NIR) technology has been widely used in tobacco leaf quality analysis and evaluation, as well as formulation design and maintenance, due to its rapidity, high efficiency, and rich quality information. However, NIR analysis only considers the correlation between tobacco leaf quality and leaf formulation under "static conditions," ignoring the quality characteristics of tobacco products under combustion conditions during consumption. Consequently, it cannot truly represent the smoking quality of tobacco leaves, specifically the quality characteristics of the smoke.

[0004] Thermogravimetric analysis (TG / DTA) provides stable reaction conditions under programmed temperature, making it an ideal experimental tool for tobacco pyrolysis research. Derivative thermogravimetry, also known as derivative thermogravimetry, is a technique derived from thermogravimetry. This technique records the first-order derivative of the TG curve with respect to temperature or time; the resulting result is a derivative thermogravimetric curve, or DTG curve. The DTG curve's characteristics include: it accurately reflects the starting temperature, maximum reaction rate, and end temperature of each weight loss stage; the area of ​​each peak on the DTG curve is proportional to the corresponding sample weight loss on the TG curve; and the DTG curve can clearly distinguish certain thermal processes where steps are not clearly visible on the TG curve. The key advantage of thermogravimetry is its high quantitativeness, accurately measuring both the mass change and the rate of change. This characteristic allows thermogravimetry to be used to study any substance that undergoes a mass change when heated.

[0005] The analysis of tobacco leaf composition currently relies on a combination of methods, including cut tobacco chemical composition analysis, smoke chemical composition analysis, and sensory evaluation. This approach is labor-intensive and highly subjective, and the conclusions drawn are vague and of limited reference value.

[0006] In order to solve the above problems, the present invention is proposed.

[0007] Summary of the Invention

[0008] In order to solve the current problems in the analysis of tobacco leaf group composition, the present invention constructs a fusion graph that can comprehensively reflect the quality of tobacco by fusing near-infrared spectra and thermal analysis graphs. Furthermore, a fusion graph difference correlation model is established to simulate and evaluate the conformity between the formula analysis composition and the actual leaf group formula. In order to improve the versatility of tobacco leaf group analysis and the work efficiency of analysts, the present invention uses fusion graphs to analyze and characterize the quality information of cigarette tobacco. The present invention designs a fusion graph difference correlation model and a formula analysis combined optimization algorithm to automatically search for the tobacco leaf ratio of the cigarette leaf group formula, and uses objective data to analyze the composition of the cigarette leaf group. It can clearly obtain the leaf group formula composition and proportion, which is of great significance to the analysis of competing cigarettes and the design of leaf group formulas.

[0009] The present invention proposes a method for analyzing the composition of tobacco leaves based on a fusion atlas. The specific steps are to analyze competing cigarettes (cigarettes to be analyzed) based on the ability of the fusion atlas to characterize tobacco quality, and analyze their specific tobacco leaf composition and formula ratio.

[0010] The technical solutions of the present invention are as follows:

[0011] A method for analyzing tobacco leaf composition based on a fusion atlas comprises the following steps: (1) preparing cigarette samples to be analyzed and single-grade tobacco leaf samples; (2) constructing a fusion atlas of the cigarette samples to be analyzed and the single-grade tobacco leaf samples; and (3) analyzing the fusion atlas to obtain the tobacco leaf composition and proportion of the cigarette to be analyzed.

[0012] Preferably, in step (1), there is one cigarette sample to be analyzed, and no fewer than fifty single-grade tobacco leaf samples are selected; each sample is placed in a constant temperature and humidity environment at (22±1)°C and a relative humidity of (60±2)% for equilibrium for no fewer than 48 hours. Generally, the analyzed cigarette sample is no less than 5g, and the sample is crushed to a mesh size of no less than 100 mesh; generally, no fewer than fifty typical single-grade tobacco leaf samples are selected, and the tobacco leaf sample information must cover different grades, different origins, and different parts, and the smoking taste of the typical single-grade tobacco leaf samples varies greatly; the tobacco leaf sample is no less than 5g, and the sample is crushed to a mesh size of no less than 100 mesh.

[0013] Preferably, in step (2), the steps for constructing the fusion atlas of the cigarette sample to be analyzed and the single-grade tobacco leaf sample are as follows:

[0014] (21) Near infrared spectrum acquisition: Weigh 3 g of sample powder and place it in a sample cup, and scan 4000-9000 cm -1 The spectrum of each band was repeated 10 times for each sample and the spectrum average was taken;

[0015] (22) Thermal analysis spectrum acquisition: The samples were placed in a thermogravimetric crucible and heated. The program was as follows: initial temperature 50°C, heating rate 10°C / min; end temperature 900°C, constant temperature at 900°C for 5 minutes; the protective gas and reaction gas were both nitrogen, and the test was conducted under the condition of a flow rate of 20mL / min; before sample analysis, the thermogravimetric analyzer was set to maintain the temperature at 900°C for 10 minutes to allow the impurities in the selected thermogravimetric alumina crucible furnace to be completely discharged, and an empty crucible was used as a reference. The instrument balance sensitivity of the thermogravimetric analyzer is not less than 0.1μg, and the curve resolution is not less than 50 million resolution points. The test results are taken as temperature (°C) as the X-axis and mass (%) as the Y-axis, the TG data is exported and the first-order derivative of the temperature is calculated to obtain the differential weight loss DTG data and form a thermal analysis spectrum;

[0016] (23) Fusion spectrum construction: The sample near-infrared spectrum vector F m and thermal analysis spectrum vector F n Multiply to get matrix A m×n =F m ×F n , and perform maximum pooling V on it according to the row vector m =MAX n (A m×n )=[v1 v2 … v i ], then V m Perform logarithmic normalization to obtain the fusion map matrix of the sample We obtain a fusion atlas matrix Y of the cigarette sample to be analyzed, and a fusion atlas matrix X of n single-level tobacco leaves = [X1 X2 … X n ], n is the number of single-grade tobacco leaf samples.

[0017] Preferably, the specific steps of analyzing the fusion map in step (3) and obtaining the tobacco composition and proportion of the cigarette to be analyzed are as follows:

[0018] (31) Formula ratio coding: The formula ratio of each single grade of tobacco leaves is coded as a real number R = [r1 r2 … r n ], n is the number of single-grade tobacco leaf samples;

[0019] (32) Randomly initialize the encoding matrix R; the r value is initialized to a real value between 0 and 1, and the sum of the values ​​of each encoding matrix should be 1: establish a search space in a range of more than 10 times the number of tobacco leaves in the recipe, and randomly initialize the encoding matrix, i.e., R1, R2, ...; since the number of tobacco leaves in the recipe is generally between 10 and 20, establish a search space in a range of 10 times, and randomly initialize 200 encoding matrices, i.e., R1, R2, ..., R 200 ;

[0020] (33) Calculate the cigarette fusion map matrix Z after combining single-grade tobacco leaves according to the recipe ratio R;

[0021] (34) Calling the fusion graph difference association model to calculate the difference e between Z and Y;

[0022] (35) Convert the difference value e into a probability value P(e);

[0023] (36) According to the probability value, several formula ratios are selected to participate in the next iteration, and two schemes are randomly selected for linear recombination: r (1) =r1+a*(r1-r2), and the reorganized real number coding matrix R1 is obtained (1) 、R2 (1) ,…; where a is a proportional factor, generated by a random number uniformly distributed in [-d, 1+d]; d is the value that limits the range of recombination;

[0024] (37) Repeat steps (32)-(34) to perform iterative search and iteratively calculate e (2) 、e (3) 、e (4) 、e (5) ,…, until e is less than a certain value;

[0025] (38) Sort by probability value P(e) from large to small and take several formulation ratios; that is, obtain the tobacco leaf composition and ratio of the cigarette to be analyzed.

[0026] Preferably, step (32) should ensure that the sum of the values ​​of each encoding matrix is ​​1, and the initialization formula is as follows:

[0027] Preferably, step (33) calculates the cigarette fusion atlas matrix Z after the single-grade tobacco leaves are combined according to the recipe ratio R, and the calculation formula is as follows: Z i =X′×R i ; Among them, R i is the i-th random coding matrix, X is the tobacco leaf fusion atlas matrix, Z i According to the formula ratio R i The formula fusion graph matrix composed of.

[0028] Preferably, the calculation formula of the difference degree e in step (34) is as follows: Among them, Y is the fusion map matrix of the cigarette to be analyzed, Z is the fusion map matrix of the tobacco leaves according to the proportion of the formula, and ∑ is the covariance matrix of Y and Z.

[0029] Preferably, step (35) converts the difference e value into a probability value P(e) between 0 and 1, and the calculation formula is as follows:

[0030] Preferably, the value of d in step (36) is 0.2-0.3.

[0031] Preferably, step (37) iterates the calculation until e<0.0001.

[0032] The present invention has the following beneficial effects:

[0033] 1. The method of the present invention is designed to integrate the graph difference association model and the formula analysis combination optimization algorithm, automatically search the tobacco leaf ratio of the cigarette leaf group formula, and can complete the composition analysis of any finished cigarette to be analyzed on the market within a few minutes, and can obtain clear formula composition and proportion values. It is objective, efficient, and versatile, with good repeatability and high sensitivity, and has unique advantages in the analysis of finished cigarettes in the tobacco industry.

[0034] 2. The method of the present invention avoids the wet chemical operation methods such as the analysis of the chemical composition of a large amount of tobacco cut and the chemical composition analysis of smoke in conventional cigarette leaf group analysis, and switches to dry chemical operation. The operation is simple, the sample amount used is extremely small, within 10 mg, it is non-toxic and harmless, does not cause any harm to the operator, and does not cause any environmental pollution.

[0035] 3. While providing fusion spectra of near-infrared and thermal analysis quality of finished cigarettes and single-grade tobacco leaves, the method of the present invention can significantly reduce the workload and the number of experiments. It provides concrete formula design goals, rich data support and digital technical means for cigarette product development, and realizes automatic search and objective evaluation of formula design schemes, which can effectively avoid the influence of subjective factors and differential characterization brought about by traditional reliance on expert experience and sensory evaluation. DETAILED DESCRIPTION

[0036] The present invention is further illustrated below by way of examples, but is not intended to be limiting. Experimental procedures not specifically specified in the examples generally followed conventional conditions, those described in manuals, or those recommended by the manufacturers. The general equipment, materials, and reagents used were all commercially available unless otherwise specified. The raw materials required in the following examples and comparative examples were all commercially available.

[0037] Example: A method for analyzing the composition and proportion of tobacco leaf components of a finished product sample of a well-known domestic brand of cigarettes (cigarettes to be analyzed) is as follows:

[0038] (1) One finished cigarette sample of a well-known domestic brand (cigarette to be analyzed) and 50 single-grade tobacco samples of different origins, different parts, and different grades (5 grams each) (single-grade tobacco leaves) were selected. Both the cigarette to be analyzed and the single-grade tobacco leaves were sieved through a 100-mesh sieve and equilibrated in a constant temperature and humidity environment of (22±1)℃ and (60±2)% for 48 hours.

[0039] (2) Weigh 3g of sample powder and place it in a sample cup. Scan the sample at 4000-9000cm -1 The spectrum of the band was repeated 10 times for each sample and the spectrum average was taken to obtain the near-infrared spectrum of the sample;

[0040] (3) Before thermogravimetric analysis of the sample, set the thermogravimetric analyzer to 900°C for 10 minutes to exhaust the impurities in the furnace, and use an empty crucible as a reference. Weigh (5.00±0.05) mg of the sample and place it in a thermogravimetric platinum crucible. The heating program is: initial temperature 50°C, heating rate 10°C / min, end temperature 900°C, constant temperature at 900°C for 5 minutes, and the protective gas and reaction gas are both nitrogen, with a flow rate of 20 mL / min. The test is performed with temperature (°C) as the X-axis and mass (%) as the Y-axis. The TG data is exported and the first-order derivative of the temperature is calculated to obtain the differential weight loss DTG data and form a thermal analysis spectrum;

[0041] (3) Calculate the fusion map matrix Y of the cigarette to be analyzed, and the single-level tobacco leaf fusion map matrix X = [X1 X2 … X 50 ], the number of fusion map matrix variables is m;

[0042] Table 1: Cigarette fusion graph matrix Y to be analyzed

[0043] Table 2: Fusion map matrix X of 50 single-grade tobacco leaves

[0044] (4) Set the formula ratio real number encoding matrix R = [r1 r2 … r 50 ]; among them, r1, r2…r 50 They represent the usage ratios of 50 single-grade tobacco leaves in the recipe, as shown in Table 3:

[0045] Table 3: Recipe ratio real number encoding matrix R

[0046] (5) Change r in the above table i Randomly initialize to a real value between 0 and 1, and i Normalization is performed to ensure that the sum of the proportion values ​​of each tobacco leaf is 1. The formula is: n is 50;

[0047] Table 4: Random initialization results of the recipe ratio real number encoding matrix R

[0048] (6) According to the above method, 200 real number coding matrices R1, R2, ..., R are initialized at the same time. 200 , thus establishing the formula composition analysis search space, as shown in the following table:

[0049] Table 5: Recipe ratio real number encoding initialization

[0050] (7) The single-grade tobacco leaf fusion map matrix X is calculated according to the formula ratio real number coding matrix R to form the combined cigarette fusion map matrix Z; Z i =X′×R i ;

[0051] Table 6: Combined cigarette fusion map matrix Z

[0052] (8) Calling the fusion graph difference association model: The difference e between Z and Y was calculated to evaluate the conformity of the analysis of the tobacco leaf composition, as shown in Table 7 below:

[0053] Table 7: Conformity of analysis of tobacco leaf composition (difference between Z and Y)

[0054] (9) Convert the difference e into a probability value P(e): As shown in Table 8 below:

[0055] Table 8: Conversion of difference into probability values

[0056] (10) Randomly screen the first 100 candidate solutions for formula ratio according to the probability value, and perform linear reorganization on the formula ratios of the candidate solutions, r (1) =r1+a*(r1-r2); where a is a proportional factor, generated by a random number uniformly distributed in [-d, 1+d]. To limit the recombination range to a small value, d is set to 0.25.

[0057] Get the reorganized real number coding matrix R1 (1) 、R2 (1) ,…,R 200 (1) , as shown in Table 9 below:

[0058] Table 9: Real number coding of the formula ratio after the first reorganization

[0059] (11) According to the real number coding of the reorganized formula ratio, calculate the combined cigarette fusion map matrix Z (1) , call the fusion graph difference association model to calculate the difference between Z and Y e (1) , iterative calculation e (2) 、e (3) 、e (4) 、e (5) ,…, until e=0.000095<0.0001.

[0060] (12) Sort by probability value P(e) from large to small and output the top 5 candidate formula ratio solutions, as shown in Table 10 below:

[0061] Table 10: Recipe analysis results and P(e) values ​​(select the top 5 candidates with the highest probability)

[0062] (13) According to the first five candidate formula ratios in the table above, the tobacco leaves with a formula ratio of 0 are filtered out to obtain the composition and ratio of the complete tobacco leaf formula, as shown in Tables 11-15 below:

[0063] Table 11: R 172 Corresponding leaf group formula

[0064] Table 12: R 26 Corresponding leaf group formula

[0065] Table 13: R 23 Corresponding leaf group formula

[0066] Table 14: R 31 Corresponding leaf group formula

[0067] Table 15: R 156 Corresponding leaf group formula

[0068] Verification experiment: According to the five recipes shown in Tables 11-15 above, the single-grade tobacco samples involved were blended into cigarette cut tobacco. Nine sensory evaluation experts were organized to conduct sensory evaluation and score the sensory quality differences between the blended cut tobacco samples and the cut tobacco samples of the cigarettes to be analyzed. The average value was taken as the actual evaluation value of the quality difference, and the quality difference was qualitatively characterized after rounding off. The scoring gradient is set as shown in Table 16 below:

[0069] Table 16: Sensory quality difference scoring gradient settings

[0070] The results of the evaluation are shown in Table 17 below:

[0071] Table 17: Sensory evaluation verification results

[0072] As can be seen from Table 17, the formula ratio candidate R 172 The conformity with the cigarette to be analyzed is 72.05%, and there is no difference in the sensory evaluation results; candidate R 26 、R 23 、R 31 、R 156The conformity with the cigarettes to be analyzed was 13.25%, 4.45%, 2.20% and 1.15% respectively, and the sensory evaluation results were slightly different.

[0073] The above-described embodiments merely represent several implementation methods of the present invention. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent. It should be noted that a person skilled in the art would be able to make various modifications and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the patent for this invention shall be determined by the appended claims.

Claims

1. A method for analyzing tobacco leaf composition based on fusion atlas, characterized in that: The method comprises the following steps: (1) preparing cigarette samples and single-grade tobacco leaf samples to be analyzed; (2) constructing a fusion map of the cigarette samples and single-grade tobacco leaf samples to be analyzed; and (3) analyzing the fusion map to obtain the tobacco leaf composition and proportion of the cigarettes to be analyzed.

2. The method according to claim 1, characterized in that Step (1) There is one cigarette sample to be analyzed, and no less than fifty single-grade tobacco leaf samples are selected; each sample is placed in a constant temperature and humidity environment at (22±1)°C and a relative humidity of (60±2)% for equilibrium for no less than 48 hours.

3. The method according to claim 1, characterized in that Step (2) The steps for constructing the fusion atlas of the cigarette sample to be analyzed and the single-grade tobacco leaf sample are as follows: (21) Near infrared spectrum acquisition: Weigh 3 g of sample powder and place it in a sample cup, and scan 4000-9000 cm -1 The spectrum of each band was repeated 10 times for each sample and the spectrum average was taken; (22) Thermal analysis spectrum acquisition: The samples were placed in a thermogravimetric crucible and heated up with the following program: initial temperature 50°C, heating rate 10°C / min; end temperature 900°C, constant temperature at 900°C for 5 min; the protective gas and reaction gas were both nitrogen, and the flow rate was 20 mL / min. The test results were plotted with temperature (°C) as the X-axis and mass (%) as the Y-axis. The TG data were derived and the first-order derivative of the temperature was calculated to obtain the differential weight loss (DTG) data and form a thermal analysis spectrum. (23) Fusion spectrum construction: The sample near-infrared spectrum vector F m and thermal analysis spectrum vector F n Multiply to get matrix A m×n =F m ×F n , and perform maximum pooling V on it according to the row vector m =MAX n (A m×n )=[v1v2…v i ], then V m Logarithmic normalization is performed to obtain the fusion map matrix of the sample We obtain a fusion atlas matrix Y of the cigarette sample to be analyzed, and a fusion atlas matrix X of n single-level tobacco leaves = [X1X2…X n ], n is the number of single-grade tobacco leaf samples.

4. The method according to claim 1, characterized in that The specific steps of step (3) analyzing the fusion map and obtaining the tobacco composition and proportion of the cigarette to be analyzed are as follows: (31) Formula ratio coding: The formula ratio of each single grade of tobacco leaves is coded as a real number R = [r1r2…r n ], n is the number of single-grade tobacco leaf samples; (32) Randomly initialize the encoding matrix R; the value of r is initialized to a real value between 0 and 1, and the sum of the values ​​of each encoding matrix should be 1: establish a search space in a range of more than 10 times the number of tobacco leaves in the recipe, and randomly initialize the encoding matrix, i.e., R1, R2, ...; (33) Calculate the cigarette fusion map matrix Z after combining single-grade tobacco leaves according to the recipe ratio R; (34) Calling the fusion graph difference association model to calculate the difference e between Z and Y; (35) Convert the difference value e into a probability value P(e); (36) According to the probability value, several formula ratios are selected to participate in the next iteration, and two schemes are randomly selected for linear recombination: r (1) =r1+a*(r1-r2), and the reorganized real number coding matrix R1 is obtained (1) 、R2 (1) ,…; where a is a proportional factor, generated by a random number uniformly distributed in [-d, 1+d]; d is the value that limits the range of recombination; (37) Repeat steps (32)-(34) to perform iterative search and iteratively calculate e (2) 、e (3) 、e (4) 、e (5) ,…, until e is less than a certain value; (38) Sort by probability value P(e) from large to small and take several formulation ratios; that is, obtain the tobacco leaf composition and ratio of the cigarette to be analyzed.

5. The method according to claim 4, characterized in that: Step (32) should ensure that the sum of the values ​​of each encoding matrix should be 1. The initialization formula is as follows:

6. The method according to claim 4, characterized in that: Step (33) calculates the cigarette fusion atlas matrix Z after combining single-grade tobacco leaves according to the recipe ratio R. The calculation formula is as follows: Z i =X′×R i ; Among them, R i is the i-th random coding matrix, X is the tobacco leaf fusion atlas matrix, Z i According to the formula ratio R i The formula fusion graph matrix composed of.

7. The method according to claim 4, characterized in that: The calculation formula of the difference degree e in step (34) is as follows: Among them, Y is the fusion map matrix of the cigarette to be analyzed, Z is the fusion map matrix of the tobacco leaves in proportion, and Σ is the covariance matrix of Y and Z.

8. The method according to claim 4, characterized in that: Step (35) converts the difference e value into a probability value P(e) between 0 and 1, and the calculation formula is as follows:

9. The method according to claim 4, characterized in that: The value of d in step (36) is 0.2-0.

3.

10. The method according to claim 4, characterized in that: Step (37) is iterated until e<0.0001.

Images