Method for replacing tobacco in cigarette formula by means of fusion of near-infrared spectrums and pyrolysis components

Through the multimodal fusion algorithm of the convolutional neural network, the near-infrared spectrum is combined with thermal cracking component data to form a feature matrix, solving the problem of quality stability in tobacco leaf formula substitution, and realizing intelligent judgment and recommendation of tobacco leaf substitution to ensure that The quality and style of cigarettes are stable.

WO2024227311A1PCT designated stage expired Publication Date: 2024-11-07CHINA TOBACCO YUNNAN IND
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
PCT/CN2023/099683
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-04
Filing Date
2023-06-12
Publication Date
2024-11-07

AI Technical Summary

Technical Problem

The existing technology lacks objective, scientific and efficient methods in the replacement of tobacco leaf formulas, which makes it difficult to guarantee the quality and style stability of cigarettes, and near-infrared analysis technology fails to fully characterize the thermal transfer behavior and thermal cracking products of tobacco leaves under combustion conditions.

Method used

A multimodal fusion algorithm based on convolutional neural network is used to merge the near-infrared spectrum of tobacco leaves with thermal cracking component data to form a feature matrix, and whether the stock tobacco leaves can replace the tobacco leaves in the cigarette formula is determined by the Mahayana distance value.

Benefits of technology

It realizes an objective, scientific and efficient evaluation of the quality of tobacco leaves, ensures the stability of the quality and style of cigarettes, and provides intelligent tobacco leaves replacement recommendations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2023099683_07112024_PF_FP_ABST
    Figure CN2023099683_07112024_PF_FP_ABST
Patent Text Reader

Abstract

Disclosed in the present invention is a method for replacing a tobacco in a cigarette formula by means of fusion of near-infrared spectrums and pyrolysis components, which comprises the following steps: (1) processing near-infrared characteristic spectral band data and pyrolysis product characteristic component data of a tobacco in a cigarette formula by using a multi-modal fusion algorithm based on a convolutional neural network, so as to obtain a fusion characteristic matrix X for the tobacco in the cigarette formula; (2) processing near-infrared characteristic spectral band data and pyrolysis product characteristic component data of different stock tobaccos by using the multi-modal fusion algorithm based on a convolutional neural network, so as to obtain fusion characteristic matrices Y of the stock tobaccos; (3) calculating Mahalanobis distance values d of the matrix X and the matrices Y; and (4) selecting a stock tobacco of which the Mahalanobis distance value d is not greater than a certain value as a replacement tobacco for the tobacco in the cigarette formula. The method of the present invention is objective, scientific and efficient, and can determine whether stock tobaccos can replace tobaccos in cigarette formulas and provide intelligent recommendations for said replacements, thereby completely ensuring the stability of the quality and style of cigarettes.
Need to check novelty before this filing date? Find Prior Art

Description

A method for replacing tobacco leaves in cigarette formulas by integrating near-infrared spectroscopy pyrolysis components Technical Field

[0001] The present invention belongs to the technical field of cigarettes, and in particular relates to a method for determining tobacco leaf substitution in cigarette formulas by integrating near-infrared spectroscopy pyrolysis components. 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. When a particular tobacco leaf runs out, manual selection of similar-quality leaves from a stockpile of hundreds of grades is necessary, relying on recipe experience and sensory evaluation. This is highly subjective and labor-intensive, and due to the lack of objective, scientific, and efficient technical means, it is difficult to fully guarantee the stability of cigarette quality and style. Using instrumental testing methods, objective data, and scientific techniques to assist in blend substitution would be crucial for improving cigarette quality stability.

[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, efficiency, and information-rich nature. However, NIR analysis only considers the correlation between tobacco leaf quality and leaf formulation under static conditions, ignoring the heat transfer behavior and pyrolysis products during the combustion process of cigarette consumption. This technology cannot truly represent the smoking quality of tobacco leaves, specifically the quality characteristics of the smoke. Therefore, NIR-based formulation replacement technology has certain limitations.

[0004] To comprehensively characterize tobacco smoking quality, the present invention proposes a method combining near-infrared spectral data from tobacco leaves with pyrolysis-gas chromatography-mass spectrometry (Py-GC / MS) to analyze key chemical components during tobacco pyrolysis. This method provides a comprehensive and accurate reflection of the smoke quality characteristics of tobacco leaf raw materials used in cigarettes.

[0005] Summary of the Invention

[0006] The present invention uses a multimodal fusion algorithm based on a convolutional neural network to merge and project the quality information of tobacco leaves, namely the near-infrared spectrum and the thermal cracking components, into the same multimodal representation space, thereby forming a feature matrix that can fully characterize the smoking quality of tobacco leaves. On this basis, the quality similarity measurement of tobacco leaves in the formula and tobacco leaves in stock is measured and calculated to realize the judgment and intelligent recommendation of whether tobacco leaves in stock can replace tobacco leaves in cigarette formulas.

[0007] The technical solutions of the present invention are as follows:

[0008] A method for replacing tobacco leaves in cigarette formulas by integrating near-infrared spectroscopy pyrolysis components comprises the following steps:

[0009] (1) The near-infrared characteristic spectrum data and the characteristic component data of the pyrolysis products of the tobacco leaves in the cigarette formula are combined with a multimodal fusion algorithm based on a convolutional neural network to obtain the fusion feature matrix X of the tobacco leaves in the cigarette formula;

[0010] (2) The near-infrared characteristic spectrum data and pyrolysis product characteristic component data of different stock tobacco leaves are combined with a multimodal fusion algorithm based on a convolutional neural network to obtain the stock tobacco leaf fusion feature matrix Y;

[0011] (3) Calculate the Mahalanobis distance d between matrices X and Y;

[0012] (4) Selecting stock tobacco leaves with a Mahalanobis distance value d not greater than a certain value as replacement tobacco leaves for tobacco leaves in cigarette formulas.

[0013] Preferably, the content of the main cracking products in the tobacco leaf combustion process is used as the characteristic value, and the characteristic spectrum segments of the tobacco leaf near-infrared spectrum are screened respectively in combination with the absorption peak position of each molecular functional group in the near-infrared spectrum.

[0014] Preferably, the selected tobacco leaf near infrared spectrum characteristic spectrum is: 4350-4450cm -1 , 4500-4700cm -1 , 4800-4900cm -1 , 5050-5100cm -1 , 5220-5300cm -1 , 6050-6250cm -1 , 6400-6600cm -1 , 6950-7220cm -1 ; The main corresponding cracking products are nicotine, neophytadiene, aldehydes, alcohols, acids, ketones, phenols, benzenes and other substances.

[0015] Preferably, the cracking conditions are: in an atmosphere of nitrogen and oxygen mixed gas containing 9% oxygen, the initial temperature is 30° C., maintained for 5 seconds, increased to 900° C. at a rate of 30° C. / s, and maintained for 5 seconds.

[0016] Preferably, GC-MS is used to analyze the thermal cracking products; the GC-MS conditions are: Agilent DB-5MS (30m 250μm 0.25μm), rear injection port temperature 220°C, injection volume 1μL, split ratio 100:1; temperature program, initial temperature 40°C, hold 3min, increase to 240°C at 10°C / min, then increase to 280°C at a rate of 20°C / min, hold 5min; carrier gas: He, flow rate 1.5mL / min; electron impact (EI) ion source, electron energy 70eV, transfer line temperature 280°C, ion source temperature 230°C, quadrupole temperature 150°C, scan mode full scan; mass scan range 35-450amu, solvent delay 5min; after GC-MS analysis and detection, the Nist17 standard spectral library is used for qualitative analysis, substances with a matching degree of more than 80% are selected, and the peak area normalization method is used to semi-quantitate the cracking products.

[0017] Preferably, a convolutional neural network algorithm is used for feature fusion, and the steps are as follows:

[0018] (a) Near-infrared fully connected layer f m and the fully connected layer of thermal cracking characteristic components f n Multiply together to form a two-dimensional feature fusion transformation matrix A m×n ;

[0019] (b) To A m×n Perform maximum pooling to form m feature nodes as the new fusion feature layer B m ;

[0020] (c) Apply the softmax layer to B m Logarithmic normalization is performed to return the variables to the range [0, 1] to obtain the final fusion feature layer: the fusion feature matrix X of tobacco leaves in cigarette formula and the fusion feature matrix Y of tobacco leaves in stock.

[0021] Preferably, the Mahalanobis distance calculation formula is as follows: Where X is the fusion feature matrix of tobacco leaves in cigarette formula, Y is the fusion feature matrix of tobacco leaves in stock, and Σ is the covariance matrix of X and Y.

[0022] Preferably, stock tobacco leaves with a Mahalanobis distance value d not greater than 0.0500 are selected as substitute tobacco leaves for tobacco leaves in cigarette formulations.

[0023] Beneficial effects of the present invention:

[0024] The present invention's tobacco leaf replacement method for cigarette blends, integrating near-infrared spectral pyrolysis components, utilizes a multimodal fusion algorithm based on convolutional neural networks for the first time. This algorithm combines characteristic near-infrared spectral segments of tobacco leaves with pyrolysis product data and projects them into a single multimodal representation space, forming a feature matrix that fully characterizes the smoking quality of the tobacco leaves. Furthermore, the Mahalanobis distance is used to measure the quality similarity between tobacco leaves in the cigarette blend and those in stock, enabling intelligent judgment and recommendation of tobacco leaf substitutions in cigarette blends. This objective, scientific, and efficient method can determine whether stock tobacco leaves can be substituted for tobacco leaves in cigarette blends and intelligently recommend alternatives, fully ensuring the stability of cigarette quality and flavor. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] FIG1 is a schematic diagram of a multimodal fusion algorithm based on a convolutional neural network according to the present invention.

[0026] FIG2 is a diagram showing the screening of characteristic spectrum segments of the tobacco leaf near-infrared spectrum according to an embodiment.

[0027] FIG3 is a comparison of the differences in thermal cracking products of the tobacco leaf formula before and after replacement in the embodiment.

[0028] FIG4 is a comparison of sensory evaluation differences of the tobacco leaf formulations before and after replacement of the present invention. DETAILED DESCRIPTION

[0029] The present invention will be further described in detail below through specific embodiments, but this should not be construed as limiting the scope of the present invention to the following examples. Various substitutions and modifications made according to common technical knowledge and customary means in the art without departing from the spirit of the present invention are intended to be included within the scope of the present invention.

[0030] Example: Determining whether stockpiled tobacco leaves can replace tobacco leaves in cigarette formulations.

[0031] The present invention adopts four cigarette formulations, and the tobacco leaf information in these four cigarette formulations is shown in Table 1 below.

[0032] Table 1 Tobacco leaf information in four tobacco recipes

[0033] The steps for determining whether stock tobacco leaves can replace tobacco leaves in four cigarette formulas are as shown in Figure 1, that is, the method for replacing tobacco leaves in cigarette formulas by integrating near-infrared spectroscopy pyrolysis components, as follows:

[0034] (1) The near-infrared spectral characteristic band data and the characteristic component data of the pyrolysis products of the tobacco leaves in the cigarette formula are combined with a multimodal fusion algorithm based on a convolutional neural network to obtain the tobacco leaf quality characteristic matrix X in the cigarette formula; specifically:

[0035] The tobacco leaf sample of each cigarette formula was evenly divided into two parts; one part was used for near infrared spectroscopy analysis, and the other part was used for thermal cracking product analysis;

[0036] Near-infrared spectroscopy analysis of tobacco leaves: Place the tobacco leaf sample in a 40°C oven to dry for 2 hours, grind it through a 40-mesh sieve with a cyclone mill, weigh 3g of tobacco leaf powder and place it in a sample cup. Scan the 4000-9000cm -1 The spectrum of the band was repeated 10 times for each sample and the spectrum average was taken; the following near-infrared spectral characteristic bands were also scanned: 4350-4450cm -1 , 4500-4700cm -1 , 4800-4900cm -1 , 5050-5100cm -1 , 5220-5300cm -1 , 6050-6250cm -1 , 6400-6600cm -1 , 6950-7220cm -1 The above-mentioned near-infrared spectral characteristic bands are mainly the absorption peak positions of the molecular functional groups in the near-infrared spectra of the following cracking products during tobacco leaf combustion: nicotine, neophytadiene, aldehydes, alcohols, acids, ketones, phenols, benzenes, etc., as shown in Figure 2;

[0037] Analysis of tobacco leaf pyrolysis products: Weigh 0.1 g of sample (accurate to 0.01 mg) and place 1 μL into a pyrolysis tube. In an atmosphere of nitrogen and oxygen mixed with 9% oxygen, the initial temperature was 30°C, maintained for 5 seconds, then increased to 900°C at 30°C / s and maintained for 5 seconds. The pyrolysis products were analyzed by GC-MS. Agilent DB-5MS (30 m 250 μm The HPLC-MS / MS was as follows: 1.0447 nm (0.25 μm), rear injection port temperature 220°C, injection volume 1 μL, split ratio 100:1; temperature program: initial temperature 40°C, hold 5 s, increase to 900°C at 30°C / s, hold 5 s; carrier gas: He, flow rate 1.5 mL / min; electron impact (EI) ion source, electron energy 70 eV, transfer line temperature 280°C, ion source temperature 230°C, quadrupole temperature 150°C, scan mode: full scan; mass scan range 35-450 amu, solvent delay 5 min; GC-MS analysis and detection, using the Nist17 standard library for qualitative analysis, selecting substances with a matching degree of more than 80%, and semi-quantitative analysis of the pyrolysis products by peak area normalization, as shown in Table 2.

[0038] Table 2 Relative percentage of key characteristic components of tobacco leaves in cigarette formulations (%)

[0039] The convolutional neural network algorithm is used to fuse the near-infrared characteristic spectrum data and pyrolysis characteristic component data of tobacco samples in the recipe; first, the near-infrared characteristic spectrum fully connected layer f m and the fully connected layer of thermal cracking characteristic components f n Multiply together to form a two-dimensional feature fusion transformation matrix A m×n ; Then, for A m×n Perform maximum pooling to form m feature nodes as the new fusion feature layer B m ; Finally, apply the softmax layer to B m Logarithmic normalization is performed to compress the variables and return them to the range [0, 1], thereby obtaining the final fusion feature matrix X, as shown in Table 3.

[0040] Table 3 Tobacco leaf fusion feature matrix X in the recipe

[0041] (2) The same steps as step (1) are followed for the 12 different stock tobacco leaves to obtain the fusion feature matrix Y of the stock tobacco leaves. The 12 different stock tobacco leaves are shown in Table 4.

[0042] (3) Calculate the Mahalanobis distance d of the thermal property quality matrix of the tobacco leaves in the formula and the tobacco leaves in stock. The results are shown in Table 4.

[0043] (4) As can be seen from Table 4, only C-3 has a Mahalanobis distance greater than 0.0500 between the tobacco leaves in the formula and the tobacco leaves in stock, which is 0.0589. This tobacco leaf in stock cannot be used as a substitute for the tobacco leaves in the cigarette formula. The other 11 tobacco leaves in stock can be used as substitutes for the tobacco leaves in the four formulas.

[0044] Table 4 Calculation results of Mahalanobis distance d of the fusion feature matrix of tobacco leaves in the formula and tobacco leaves in stock

[0045] In order to verify the validity of the determination results of tobacco leaf replacement, the differences in thermal cracking products and sensory evaluation of the tobacco leaf replacement were compared before and after the replacement.

[0046] The differences in pyrolysis products of the four cigarette formulations before and after tobacco leaf replacement were compared, as shown in Figure 3. As can be seen from Figure 3, the differences in thermogravimetric and pyrolysis products between the cigarettes with the tobacco leaf replacement formulations and those before replacement were small, with only a slight difference in "Recommended Scheme 3 of Formulation C" (C-3).

[0047] Furthermore, sensory evaluation was used to assess the differences before and after tobacco leaf substitution. Using the sensory quality difference evaluation method for the formulations described in Table 5, sensory evaluations were conducted on cigarette samples before and after tobacco leaf substitution. The results are shown in Figure 4. As can be seen from Figure 4, only "Recommended Solution 3 of Formulation C" (C-3) showed a significant difference in evaluation (the Mahalanobis distance between the tobacco leaves before and after substitution was 0.0589), which is consistent with the judgment method of the present invention. This demonstrates that the judgment and recommendation methods of the present invention are consistent with the actual sensory evaluation results, and therefore the judgment method of the present invention is objective, scientific, and efficient.

[0048] Table 5 Evaluation of sensory quality differences of formulas

[0049] illustrate:

[0050] No difference: The quality difference is not perceived and the substituted formula can be used in cigarette production;

[0051] Slight difference: The quality difference cannot be accurately and stably perceived, and the substituted formula can be used in cigarette production;

[0052] ● There are differences: There are quality differences, and the replaced formula cannot be used in cigarette production.

[0053] The embodiments are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that the present invention is susceptible to various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A method for replacing tobacco leaves of cigarette blends by integrating near infrared spectroscopy pyrolysis components, characterized in that: The steps include: (1) The near-infrared characteristic spectrum data and the characteristic component data of the pyrolysis products of the tobacco leaves in the cigarette formula are combined with a multimodal fusion algorithm based on a convolutional neural network to obtain a fusion feature matrix X of the tobacco leaves in the cigarette formula; (2) The near-infrared characteristic spectrum data and the characteristic component data of pyrolysis products of different stock tobacco leaves are combined with a multimodal fusion algorithm based on a convolutional neural network to obtain the stock tobacco leaf fusion feature matrix Y; (3) Calculate the Mahalanobis distance d between matrices X and Y; (4) Selecting stock tobacco leaves whose Mahalanobis distance value d is not greater than a certain value as substitute tobacco leaves for tobacco leaves in cigarette formula.

2. The method for replacing tobacco leaves of cigarette blends integrating near-infrared spectroscopy pyrolysis components according to claim 1, characterized in that: Taking the contents of main pyrolysis products in tobacco leaf combustion process as characteristic values ​​and combining with the absorption peak positions of various molecular functional groups in near-infrared spectrum, the characteristic spectral bands of tobacco leaf near-infrared spectrum were screened respectively.

3. The method for replacing tobacco leaves of cigarette blends integrating near infrared spectroscopy pyrolysis components according to claim 2, characterized in that: The near infrared spectrum characteristic spectrum of the selected tobacco leaves is: 4350-4450cm -1 , 4500-4700cm -1 , 4800-4900cm -1 , 5050-5100cm -1 , 5220-5300cm -1 , 6050-6250cm -1 , 6400-6600cm -1 , 6950-7220cm -1 .

4. The method for replacing tobacco leaves of cigarette blends integrating near-infrared spectroscopy pyrolysis components according to claim 1, characterized in that: The thermal cracking conditions were as follows: in a nitrogen-oxygen mixed gas atmosphere containing 9% oxygen, the initial temperature was 30°C, maintained for 5 seconds, increased to 900°C at a rate of 30°C / s, and maintained for 5 seconds.

5. The method for replacing tobacco leaves of cigarette blends integrating near-infrared spectroscopy pyrolysis components according to claim 4, characterized in that: The pyrolysis products were analyzed by GC-MS; the GC-MS conditions were: Agilent DB-5MS (30m 250μm 0.25μm), rear injection port temperature 220℃, injection volume 1μL, split ratio 100:1; temperature program, initial temperature 40℃, maintained for 3min, increased to 240℃ at 10℃ / min, then increased to 280℃ at a rate of 20℃ / min, maintained for 5min; Carrier gas: He, flow rate 1.5mL / min; electron impact (EI) ion source, electron energy 70eV, transmission line temperature 280°C, ion source temperature 230°C, quadrupole temperature 150°C, scanning mode is full scan; mass scanning range is 35-450amu, solvent delay 5min; after GC-MS analysis and detection, the Nist17 standard spectral library was used for qualitative analysis, substances with a matching degree of more than 80% were selected, and the peak area normalization method was used to semi-quantitate the pyrolysis products.

6. The method for replacing tobacco leaves of cigarette blends integrating near infrared spectroscopy pyrolysis components according to claim 1, characterized in that: The convolutional neural network algorithm is used for feature fusion. The steps are as follows: (a) Near infrared fully connected layer f m and the fully connected layer of thermal cracking characteristic components f n Multiply them together to form a two-dimensional feature fusion transformation matrix A m×n ; (b) For A m×n Perform maximum pooling to form m feature nodes as the new fusion feature layer B m ; (c) Apply the softmax layer to B m Logarithmic normalization is performed to return the variables to between [0,1] to obtain the final fusion feature layer: The tobacco leaf fusion feature matrix X in cigarette formula and the tobacco leaf fusion feature matrix Y in stock are as follows:

7. The method for replacing tobacco leaves of cigarette blends integrating near infrared spectroscopy pyrolysis components according to claim 1, characterized in that: The Mahalanobis distance calculation formula is as follows: Among them, X is the fusion feature matrix of tobacco leaves in cigarette formula, Y is the fusion feature matrix of stock tobacco leaves, and Σ is the covariance matrix of X and Y.

8. The method for replacing tobacco leaves of cigarette blends integrating near-infrared spectroscopy pyrolysis components according to claim 1, characterized in that: The stock tobacco leaves with a Mahalanobis distance value d not greater than 0.0500 are selected as substitute tobacco leaves for tobacco leaves in cigarette formulas.

Citation Information

Patent Citations

  • Analog tobacco leaf searching method based on tobacco leaf chemical composition

    CN101251523A

  • Method for assisting cigarette formula by adopting SIMCA (Soft Independent Modeling of Class Analogy) based on Near-infrared spectral information

    CN102866127A

  • Multi-model method for evaluating similarity of tobacco materials based on tobacco aroma components

    CN106645530A

  • Characterization method of quality change in tobacco shred processing based on near infrared technology

    CN110174371A

  • Method for detecting cigarette sensory characteristic consistency based on near infrared spectroscopy

    CN112782115A