Method for evaluating cigarette product quality difference based on multi-dimensional space chemical fingerprint spectrum

Abstract

The invention discloses a method for evaluating cigarette product quality difference based on a multi-dimensional space chemical fingerprint spectrum, and relates to the field of cigarette quality evaluation.The method comprises the following steps that near-infrared diffuse reflection spectrum data of a sample is collected, and a near-infrared discrimination classification model is established after optimization treatment; establishing a near-infrared correction model by taking a control sample; introducing the collected sample into the near-infrared correction model of the control sample for calculation, and predicting and analyzing the chemical component contents of the comparison sample and the control sample; the method comprises the following steps: classifying samples according to different batches, applying a principal component analysis method, drawing a spectrum classification effect picture, calculating the distribution condition of each category of samples, drawing a principal component score picture, collecting sample gas chromatography fingerprint data by adopting HS-GC-MS combination, and establishing a gas chromatography discrimination classification model after optimization treatment; the method comprises the following steps: determining conventional chemical indexes, polyphenol content, ash content, pH value and volatile components of selected cigarette samples, and performing variance analysis, correlation analysis and discriminant analysis.

Description

Technical Field

[0001] The invention relates to the field of cigarette quality evaluation, and in particular to a method for evaluating the quality differences of cigarette products based on multidimensional space chemical fingerprints. Background Art

[0002] Cigarettes are an extremely complex chemical multi-component system. The quality of cigarettes and the formation of their style characteristics are the result of the synergistic effects of the internal chemical components of cigarettes (including leaf group formula and added flavors and fragrances). In order to judge the product quality of cigarettes, the existing conventional chemical indicators and data processing methods are often difficult to achieve a comprehensive system analysis due to the relatively discrete indicators and insufficient information, and it is difficult to obtain satisfactory results. Especially for cigarette products produced on different production lines of the same brand, the differences are relatively subtle. The existing methods usually use analytical methods such as HPLC, LC-MS, GC-MS to detect one or some of its chemical components, and use the detected chemical components to evaluate its quality.

[0003] In the difference evaluation of cigarettes produced on different production lines, nearly a hundred parameter indicators are involved. The parameters are both interrelated and mutually restricted. The possibility that all indicators on two production lines are exactly the same is very small. It is difficult to fully evaluate the quality of cigarettes by using a single indicator or a few indicators to evaluate the quality of the multi-component system of cigarettes. It is not objective to compare individual indicators of cigarettes alone.

[0004] Moreover, since cigarettes are a smoking product, the interaction between the multiple components of tobacco itself and the various artificially added flavors and fragrances makes smoking evaluation more subjectivist, making it difficult to objectively reflect the intrinsic quality of cigarettes; the differences between cigarettes in the same series or the same product on different production lines are small, especially low-content substances in cigarettes are difficult to accurately detect using a single analytical method, and even more difficult to evaluate and distinguish. Summary of the invention

[0005] In order to solve a series of problems in actual production, such as the difficulty in judging the difference between the same products by using a single or several chemical detection indicators, the great subjective influence of sensory evaluation, and the difficulty in evaluating and distinguishing, the present invention provides a method for evaluating the quality differences of cigarette products based on multidimensional space chemical fingerprints.

[0006] The present invention is implemented by the following technologies: The present invention provides a method for evaluating the quality difference of cigarette products based on multidimensional space chemical fingerprint, comprising the following steps: a. Establish near infrared discrimination and classification model Collect near-infrared diffuse reflectance spectral data of the samples. Specifically, prepare the samples with reference to GB / T31-1996, crush the prepared samples with a cyclone mill, pass them through a 60-mesh sieve, and put them into a sealed bag for later use. Preheat the spectrometer at room temperature for 2 hours, put the samples into a 5 cm quartz cup with more than half the capacity, gently press it to be flat, place it on a rotating table, and collect the near-infrared diffuse reflectance spectral data of the samples.

[0007] After optimization processing, a near-infrared discrimination and classification model was established. The optimization processing included signal noise reduction, outlier detection, spectral preprocessing, modeling spectral region selection and determination of the maximum number of principal components.

[0008] Take the control sample to establish the near-infrared calibration model, import the collected sample into the near-infrared calibration model of the control sample for calculation, and predict and analyze the chemical composition content of the comparative sample and the control sample.

[0009] b. Establishing chromatographic discrimination and classification model The HS-GC-MS was used to collect the gas chromatography fingerprint data of the sample. Specifically, the tobacco of 100 cigarettes was fully mixed and balanced at constant temperature and humidity for 48 hours, and then stored for use. The sample was placed in a headspace bottle, and 1 μL of 25 mg / mL naphthalene ethanol solution was added as an internal standard. The headspace bottle was placed in the headspace injector. The experiment found that if the temperature was too high, the color of the tobacco became darker after the headspace experiment, and a certain degree of burning occurred. If the temperature was too low, less volatile components were produced, the viscosity and sample volume of the sample were large, and the required equilibrium time was longer. Increasing the heating temperature can increase the value (sensitivity) of the headspace component. The higher the temperature, the more the amount detected will increase. However, if the temperature is too high, it is easy to cause the sample vial to burst or chemical reaction between the measured component and the septum material of the sealed container, resulting in the risk of component loss. Moreover, at low concentrations, there is a linear relationship between sample concentration and the amount detected in the headspace, but when the concentration is too high, there is no linear relationship between the two. Therefore, the method of increasing the sample amount is not used to improve the sensitivity of headspace analysis. Therefore, the three factors affecting headspace analysis, namely, headspace heating temperature, headspace equilibrium time and sample amount, were taken for orthogonal test with three factors and three levels. The injection time was 1 min, the injection volume was 3 mL; the injection loop temperature was 140°C; the transfer line temperature was 150°C, the injection was automatic, and GC / MS analysis was performed to determine the headspace heating temperature, heating time and sample amount; repeated measurements were performed under the optimal conditions, and the relative standard deviations of the 6 measurements of 31 components were calculated.

[0010] Optimally, the optimal conditions for setting the headspace sampler were a sample size of 0.6 g, a heating temperature of 130 °C, and a heating time of 30 min.

[0011] A gas chromatography discrimination and classification model was established after optimization processing, which included signal noise reduction, outlier detection, spectrum preprocessing, modeling spectral region selection and determination of the maximum number of principal components.

[0012] c. Index detection The selected cigarette samples were tested for conventional chemical indices, polyphenol content, ash content and pH value, and variance analysis, correlation analysis and discriminant analysis were performed; The indicators of the selected cigarette samples were tested for volatile components, and variance analysis, correlation analysis and discriminant analysis were performed.

[0013] Perform pattern recognition and component difference analysis on the models of steps a and b, perform statistics on the chemical composition indicators of the original production line and the new production line, and calculate the relative standard deviation (RSD) value; Analyze the impact of PLS-DA separation contribution, analyze the component indicators of the products of the two production lines, draw PLS-DA score diagrams and class centers, chemical indicator factor distribution diagrams and score loading diagrams, and analyze the characteristic chemical indicators of the quality differences of the two product lines of Xinyuan; compare the differences in the intrinsic chemical composition of the products of the two production lines, calculate the number of principal components of the PLS-DA class model of the main chemical composition indicators based on the indicators that are closer to the class center, calculate the contribution rate of each indicator to the classification of the Xinyuan line, and then select the indicators that can characterize the differences between the two product lines; The component data was normalized, different types of cigarette samples were classified, independent sample t-test or linear regression analysis was used, chemical indicators with significant differences were selected, correlation analysis was performed, chemical components with smaller significant differences were replaced, PLS-DA analysis was performed, and characteristic indicators were used to characterize the differences of cigarettes.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The method designed by the present invention is a method for evaluating the quality difference of cigarette products based on multidimensional space chemical fingerprints. By applying the overall information of near-infrared spectroscopy and chromatographic fingerprints or either of them, the principal component analysis combined with Mahalanobis distance (PCA-MD) analysis method is used to perform difference analysis on different products, and the products are studied to have significant differences in intrinsic quality. The method uses chemometrics software, partial least squares-discriminant analysis (PLS-DA), regression analysis, correlation analysis, and variance analysis methods to screen and analyze the characteristic indicators of the main chemical component indicators and volatile components, and screen out characteristic indicators that play an important role in the difference evaluation of the new original production line. According to the research results, this method can not only study the difference between the two groups, but also be used to prove the difference in ingredients between cigarette products. It can further screen out the characteristic indicators of product differences, which serves as an important basis for subsequent process adjustments and further promotes the research on the intrinsic quality of cigarettes. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A diagram showing the classification effect of the A new original line spectrum class model in Example 1.

[0016] Figure 2 It represents the principal component score diagram of the A new original line spectrum class model in Example 1.

[0017] Figure 3 The total ion flow diagram of the A original production line in Example 1 is shown.

[0018] Figure 4 The total ion current diagram of the new production line A in Example 1 is shown, wherein a represents the total ion current diagram of the new production line A in April; b represents the total ion current diagram of the new production line A in May.

[0019] Figure 5 It represents the PLS-DA score diagram of chemical indicators of A Xinyuan production line in Example 1.

[0020] Figure 6 It represents the distribution diagram of the type center and chemical index factors of the A new original production line in Example 1.

[0021] Figure 7 It represents the loading diagram of conventional chemical index scores of A Xinyuan production line in Example 1; in the figure, the upper left part from left to right are sugar-nitrogen ratio, total sugar, and Schmuck value, and the middle right part from left to right are scopoletin and rutin.

[0022] Figure 8 Represents the number of principal components of the PLS-DA model in Example 1.

[0023] Fig. 9 It represents the PLS-DA score diagram of the characteristic index of the new original production line A in Example 1.

[0024] Fig.10 It represents the distribution diagram of the type center and characteristic index factors of the new original production line A in Example 1.

[0025] Fig.11 It represents the characteristic index score loading diagram of A new original production line in Example 1.

[0026] Fig.12 Represents the number of principal components of the characteristic indicators of the PLS-DA model in Example 1.

[0027] Fig.13 A diagram showing the classification effect of the chromatographic model of the A Xinyuan production line in Example 1.

[0028] Fig.14 It represents the principal component score diagram of the chromatographic model of the A-Xinyuan production line in Example 1.

[0029] Fig.15The figure shows the PLS-DA score diagram of the volatile component index of the A new original production line in Example 1.

[0030] Fig.16 The figure shows the distribution diagram of the class center and volatile component index factors of the A new original production line in Example 1; due to the limitation of the software display, the figure shows overlapping errors. From top to bottom in the middle are dihydro-2-methyl-3 (2H) -furanone, 5-hydroxymethyl-furfural, 5-methyl-2 furfural, the center of the middle is tricyclic olefins (tricyclic terpenes), the left side of the middle is 4-cyclopentene-1,3-dione, 1-(2-furanyl)-ethanone, 6-methyl-5-furfuryl-3-one, the lower left side is α-dihydrodamascone, the upper overlapping right side of the middle is benzyl alcohol, and the lower side is 1,3-propylene glycol.

[0031] Fig.17 The figure shows the load diagram of the volatile component index scores of the A new original production line in Example 1; due to the limitation of the software display, the figure shows overlapping errors. From top to bottom in the middle are dihydro-2-methyl-3 (2H) -furanone and 5-hydroxymethyl-furfural, and from left to right in the overlapping area in the middle are furfural, 1- (2-furanyl) -ethanone, β-damascenone, 5-methyl-2 furfural, the center of the middle is tricyclic olefin (tricyclic terpene), the lower left side is 6-methyl-5-furfuryl-3-one, the lower left side is α-dihydrodamascenone, the upper right overlapping area is 1,3-propylene glycol, benzyl alcohol, the middle right overlapping area is nicotine, pyridine, benzaldehyde from left to right, and the lower right text overlapping area is 2-acetylpyrrole and 2,3-butanediol from left to right.

[0032] Fig.18 It represents the number of principal components of volatile components in the PLS-DA model in Example 1.

[0033] Fig.19 The figure shows the PLS-DA score graph of the characteristic volatile component index of the A new original production line in Example 1.

[0034] Fig. 20 It represents the distribution diagram of the class center and characteristic volatile component index factors of the A new original production line in Example 1.

[0035] Fig.21 The figure shows the characteristic volatile component index score loading diagram of the A new original production line in Example 1; due to the limitation of the software display, the figure shows overlapping errors, the upper middle part is dihydro-2-methyl-3 (2H) -furanone, the lower middle part is 6-methyl-5-furfuryl-3-one, the upper right part of the middle overlap is benzyl alcohol, and the lower part is 1,3-propylene glycol.

[0036] Fig. 22 Represents the number of principal components of the PLS-DA model in Example 1.

[0037] Fig.23The diagram shows the classification effect of the B spectral model in Example 2.

[0038] Fig.24 It represents the principal component score diagram of the B spectral model in Example 2.

[0039] Fig.25 It represents the PLS-DA score diagram of chemical indicators of B Xinyuan production line in Example 2.

[0040] Fig.26 It represents the distribution diagram of the class center and chemical index factors of the B new original production line in Example 2.

[0041] Fig. 27 It represents the chemical index score loading diagram in Example 2.

[0042] Fig.28 Represents the number of principal components of the PLS-DA model in Example 2.

[0043] Fig.29 It represents the PLS-DA score diagram of 8 chemical indicators of B Xinyuan production line in Example 2.

[0044] Fig.30 It represents the distribution diagram of the type center and 8 chemical index factors of the B new original production line in Example 2.

[0045] Fig.31 It represents the loading diagram of the scores of 8 chemical indicators in Example 2.

[0046] Fig.32 It represents the number of principal components of 8 chemical indicators in the PLS-DA model in Example 2.

[0047] Fig.33 A diagram showing the classification effect of the B chromatographic model in Example 2.

[0048] Fig.34 It represents the principal component score diagram of the B chromatographic model in Example 2.

[0049] Fig.35 It represents the PLS-DA score diagram of the volatile component index of the B new original production line in Example 2.

[0050] Fig.36The diagram shows the distribution of the class center and volatile component index factors of the B new original production line in Example 2; the diagram is limited by the software display and the display overlaps. From left to right in the middle are β-dihydrodamascone, trans-menthone, solanone, 5-hydroxymethylfurfural, menthone, 2,3-dihydro-3,5-dihydroxy-6-methyl-4H-pyran-4-one, geranyl acetone, megastigmatrienone B, benzaldehyde, dihydroactinolactone, megastigmatrienone D. On the right side is 1-(2-furanyl)-ethanone. On the lower left is 1-(1H-pyrrol-2-yl)ethanone. On the lower right are phenylacetaldehyde and 5-methylfurfural.

[0051] Fig.37 The load diagram of the volatile component index score in Example 2 is shown; due to the limitation of the software display, the figure shows overlap. The upper left part is trans-menthone, β-dihydrodamascone, 5-hydroxymethylfurfural, solanone, menthone, triglyceride, 2,3-dihydro-3,5-dihydroxy-6-methyl-4H-pyran-4-one. The overlapping part in the middle is (from left to right) geranyl acetone, megastigmatrienone B, dihydroactin, megastigmatrienone D, benzaldehyde, pyridine. The far right is 1-(2-furanyl)-ethanone.

[0052] Fig.38 It represents the number of principal components of volatile components in the PLS-DA model in Example 2.

[0053] Fig.39 The figure shows the PLS-DA score chart of 15 volatile component indicators of the B Xinyuan production line in Example 2.

[0054] Fig.40 The figure shows the distribution diagram of the class center and 15 volatile component index factors of the B new original production line in Example 2; due to the limitation of the software display, the figure is not fully displayed, the upper left part is β-dihydrodamascone, the middle left part is trans-menthone, the overlapping part in the middle is geranyl acetone, pyridine, and the right side is 1-(2-furyl)-ethanone.

[0055] Fig.41 The load diagram of the scores of 15 volatile components in Example 2 is shown; the figure is limited by the software display and is not fully displayed. The upper left part is β-dihydrodamascone, solanone, 5-5-hydroxymethylfurfural, geranylacetone, benzaldehyde, megastigmatrienone B, megastigmatrienone D, trans-menthone, furfuryl alcohol, pyridine. The far right side is 2,3-butanediol, 1-(2-furanyl)-ethanone, 5-methylfurfural, 2-cyclopenten-1-one.

[0056] Fig.42 Represents the number of principal components of the PLS-DA model in Example 2. DETAILED DESCRIPTION

[0057] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings. Example 1

[0058] A method for evaluating the quality differences of cigarette products based on multidimensional spatial chemical fingerprints, taking cigarette sample A as an example, comprises the following steps: a. Establish near infrared discrimination and classification model Collect the near-infrared diffuse reflectance spectral data of the samples, and use near-infrared spectrometer and headspace-gas chromatography-mass spectrometry technology to collect the overall spectrum information of the samples. Specifically, prepare the sample according to GB / T31-1996, then crush it with a cyclone mill, pass it through a 60-mesh sieve, and put it into a sealed bag for standby use. Preheat the FT-NIR spectrometer at room temperature for 2 hours, put the sample into a 5 cm quartz cup with more than half of the capacity, press it lightly to flatness, and place it on a rotating table. Call RESULTTM integrated software to collect the near-infrared diffuse reflectance spectral data of the sample, and set the parameters as follows: the spectral range is 10000-4000 cm -1 , resolution 8cm -1 ; Scanned 78 times; Spectrum format is *.spc; The principal component analysis-Mahalanobis distance (PCA-MD) method was used to optimize the spectrum to achieve the best state. The basic parameters were set as follows: the maximum number of principal components was determined to be 10, and the modeling spectral region was selected to be 10000-4000cm -1 The spectral preprocessing was standardized, the signal noise reduction was performed by the second-order derivative + Norris filter (7, 5), the modeling algorithm used the PCA-MD method, and the near-infrared discrimination and classification model was established after optimization.

[0059] ab, establish near infrared correction model The products produced by the original production line were taken as control samples. In this embodiment, the samples were aged tobacco leaves stored in the raw material warehouse, and near-infrared calibration models of 12 chemical components of aged tobacco leaves were established. The calibration model optimization parameters are shown in Table 1. The collected near-infrared diffuse reflectance spectrum of cigarette sample A was imported into the near-infrared correction model of the control sample for calculation, the chemical composition content of the sample and the control sample was predicted and analyzed, and the product differences between the cigarette sample A produced by the new production line and the original production line were evaluated.

[0060]

[0061] ac, difference evaluation The samples were classified into different batches and the original production line (i.e., control samples) and the new production line according to the production time. The principal component analysis method was applied to draw the spectrum classification effect diagram, calculate the distribution of samples in each category and draw the principal component score diagram, such as Figure 1 , 2As shown in the figure, the spectral information classification of the original production line and the new production line can be clearly divided into two parts, because the near-infrared spectrum can characterize the changes in the overall physical and chemical information of the sample. Therefore, the intrinsic quality of product A produced by the two production lines has significant differences.

[0062] b. Establishing gas chromatography discrimination and classification model HS-GC-MS was used to collect the gas chromatography fingerprint data of the samples. Specifically, shredded tobacco from 100 cigarettes was fully mixed and placed in a constant temperature and humidity chamber for 48 hours, where the temperature in the constant temperature and humidity chamber was 22°C and the relative humidity was 60%RH. The shredded tobacco was placed in a brown ground-mouth sample bottle for standby use. The sample was placed in a 20 mL headspace bottle, and 1 μL of 25 mg / mL naphthaleneethanol solution was added as an internal standard. The headspace bottle was placed in a headspace injector with an injection volume of 3 mL. As shown in Tables 2 to 4, the headspace heating temperature (A), heating time (B) and sample size (C) were taken for a three-factor three-level orthogonal test. The injection time was 1 min and the injection volume was 3. mL, injection loop temperature of 140°C; transfer line temperature of 150°C, automatic injection, GC / MS analysis, GC / MS conditions are: injection port temperature of 260°C, carrier gas of He, flow rate of 1 mL / min, split ratio of 10:1, GC / MS interface temperature of 280°C, ion source of EI source, electron energy of 70 eV, scan range of 35-455 amu, the furnace temperature rising conditions are 50℃ (1min)-3℃ / min-150℃ (1min)-10℃ / min-250℃ (1min), that is, keep at 50℃ for 1min, heat to 150℃ at a rate of 3℃ / min, and keep at 150℃ for 1min, then heat to 250℃ at a rate of 10℃ / min, and keep at 250℃ for 1min; the standard spectral library is NIST05, WILEY275 spectral library; the total peak area of ​​31 components in cigarettes is detected, and the optimal level combination is obtained by calculating the sum K of the test indicators of each factor at each level and the average k of the test indicators.

[0063]

[0064]

[0065]

[0066] From the results of the orthogonal experiment, the total peak area of ​​31 components is the largest obtained in Experiment 7, and the level combination is A3B1C3. The sum K of the test indicators of the three factors at each level and the average k of the test indicators are calculated (the results are shown in Table 4). From the results, it can be seen that the optimal level combination of the three factors A, B and C is A3B1C3. This optimal level combination is consistent with the experimental results; therefore, A3B1C3 is selected as the optimal level combination of this experiment, that is, the headspace sampler setting conditions are heating temperature 130℃, heating time 30min, and injection time 1 min.

[0067] Repeated measurements were performed under the optimal conditions, and the relative standard deviations of the six measurements of the 31 components were calculated. Specifically, six samples were accurately weighed and placed in headspace vials, and repeated measurements were performed under the conditions of a headspace temperature of 130 °C, a heating time of 30 min, and a sample size of 0.6 g. The relative standard deviations (RSDs) of the six measurements of the 31 components were calculated, as shown in Table 5.

[0068] From the results in the table, it can be seen that the relative standard deviations of 6 measurements of 15 components are less than 10%, the relative standard deviations of 16 components are in the range of 10-20%, and the average relative standard deviation of 6 measurements is 10.1%. The results show that the precision of determining volatile components in cigarette tobacco by this method is good and can meet the requirements of cigarette quality evaluation.

[0069] The volatile components in the samples produced by the original production line and the new production line were measured, and the chemical fingerprints of the above samples were extracted.

[0070] The volatile components of A on the new and old production lines were measured. The total ion currents of samples A on the new and old production lines are shown in the figure below: Figure 3 , 4 shown.

[0071] The optimization process was carried out, the maximum number of principal components was determined to be 10, the data preprocessing was standardized, the PCA-MD method was used as the modeling algorithm, the gas chromatography discrimination classification model was established, the spectral classification effect diagram was drawn, the distribution of samples in each category was calculated, and the principal component score diagram was drawn, as shown in Fig.13 , 14 As shown; c. Index detection ca, SPSS19.0 software was used to measure the indicators of 10 conventional chemical components, 4 polyphenol contents, ash content and pH value of the selected cigarette samples, and variance analysis, correlation analysis and discriminant analysis were performed.

[0072] A statistical analysis was conducted on the chemical composition indicators of the original production line and the new production line. The results are shown in Table 6 below.

[0073]

[0074] As can be seen from Table 6, compared with the new production line, the original production line has higher contents of nicotine, total nitrogen, potassium-chloride ratio, rutin and scopoletin than the new production line. From the relative standard deviations of the original and new production lines, it can be seen that except for the relative standard deviation RSD values ​​of sugar-alkali ratio, sugar-nitrogen ratio and pH value, the original production line is higher than the new production line, and the other indicators are lower than the new production line. Among them, the RSD value of scopoletin is the largest in the original and new production lines, which are 15.63% and 17.44% respectively. The RSD values ​​of potassium-chloride ratio, chlorogenic acid, rutin, scopoletin and total polyphenols in the new production line are significantly higher than those in the original production line. The results show that the differences in the contents of various indicators in the new production line are large.

[0075] The PLS-DA method (partial least squares discriminant analysis) was used to analyze the main chemical components of the products of the original production line and the new production line A, and to analyze the characteristic chemical indicators of the quality differences between the two product lines. Figure 5 As shown, the samples from the two production lines are clearly distinguishable, indicating that the products from the two production lines have significant differences in their intrinsic chemical composition.

[0076] like Figure 6 , 7 As shown in the figure, the cluster centers of the two production lines are far apart, proving that the Xinyuan production line has significant differences in intrinsic chemical composition. Different chemical composition indicators are distributed near the two cluster centers. The indicator closer to the cluster center indicates that the indicator has a greater contribution to the classification.

[0077] from Figure 8 As shown in Table 7, the cumulative contribution rate of the first principal component to the new original line classification is 81.91%, which can explain most of the information of the model. Through the contribution of the first principal component index, the contribution rate of each chemical index to the new original line classification is calculated. The calculation results are shown in Table 7.

[0078]

[0079] It can be seen from the table that nicotine and other five chemical components have a certain positive contribution to the new line, while chloride ions and other 11 chemical components have a certain contribution to the original line.

[0080] From the above analysis results, it can be seen that all the 16 indicators detected have more or less contributed to the classification of the new original line. However, due to the large number of chemical indicators detected, the complexity and difficulty of the analysis are increased, and it is difficult to find the differences in the chemical composition of the products of the new original production line. Therefore, it is necessary to screen the various chemical indicators obtained by analysis, and then select the chemical indicators that can characterize the differences of the new original line A products.

[0081] The samples of the original production line were taken as category 1, and the samples of the new production line were taken as category 2. The 16 indicators including nicotine, total nitrogen, total sugar, reducing sugar, sugar-alkali ratio, sugar-nitrogen ratio, Schmuck value, K2O, chlorine, potassium-chlorine ratio, ash, pH value, rutin, chlorogenic acid, scopoletin and total polyphenols were analyzed by Logarithmetic. 10 After normalization, SPSS19.0 software was used to perform linear regression analysis with the original production line and the new production line and classification (1,2) as dependent variables and 16 chemical composition indicators as independent variables to explore the impact of different indicators on classification. The results are shown in Table 8.

[0082]

[0083] As shown in Table 8, when the original production line was classified as Category 1 and the new production line as Category 2, linear regression analysis showed that within the 95% confidence interval, there were significant differences in the eight indicators of nicotine, reducing sugar, K2O, potassium-chlorine ratio, ash, chlorogenic acid, rutin and total polyphenols. In addition, the collinearity tolerance was relatively small for nicotine, while the tolerances for the other indicators were relatively large, which were 0.462, 0.289, 0.489, 0.366, 0.215, 0.3334 and 0.188, respectively. This shows that the classification of the original production line and the new production line has an important influence on these eight indicators.

[0084] The correlation analysis of the selected 8 indicators with other indicators is shown in Table 9. Nicotine has no significant correlation with the potassium-chlorine ratio, but has significant correlation with other indicators; reducing sugar has no significant correlation with the potassium-chlorine ratio, ash, total polyphenols, chlorine and pH value, but has significant correlation with other indicators; K2O has no significant correlation with chlorogenic acid and scopoletin, but has significant correlation with other indicators; potassium-chlorine ratio has no significant correlation with rutin and sugar-alkali ratio, but has significant correlation with other indicators; ash has no significant correlation with scopoletin, but has significant correlation with other indicators; chlorogenic acid has no significant correlation with rutin, total sugar and Schmuck's value, but has significant correlation with other indicators; rutin has no significant correlation with total sugar and scopoletin, but has significant correlation with other indicators; total polyphenols has no significant correlation with total sugar, total nitrogen, sugar-nitrogen ratio and Schmuck's value, but has significant correlation with other indicators.

[0085] From the correlation analysis, it can be seen that the correlations between the above-mentioned 8 indicators are mostly not significant, while the correlations between them and the remaining indicators are mostly significant. This further verifies that there is no collinearity between the 8 selected indicators, and that the 8 selected indicators can replace most of the information of the remaining indicators, revealing that the screening of the above 8 indicators plays an important role in the classification and differentiation of the original production line and the new production line.

[0086]

[0087] Eight indicators with significant differences were selected as characteristic indicators for further analysis of the differences in chemical composition between the two production lines of Xinyuan.

[0088] Repeat the above steps and use the PLS-DA method (partial least squares discriminant analysis) to analyze the eight main chemical components of the products of the original production line and the new production line A, and analyze the characteristic chemical indicators of the quality differences between the new and original product lines. Fig. 9 As shown in the figure, the samples of the two production lines are clearly distinguished, indicating that the products of the two production lines have significant differences in their intrinsic chemical composition. This shows that the selected characteristic indicators can accurately characterize the differences of the new original line products. The above characteristic indicator screening method is reasonable, and the selected characteristic indicators can be used for the difference analysis of the new original line.

[0089] like Fig.10 , 11 As shown in the figure, the cluster centers of the two production lines are far apart, proving that the Xinyuan production line has significant differences in intrinsic chemical composition. Different chemical composition indicators are distributed near the two cluster centers. The indicator closer to the cluster center indicates that the indicator has a greater contribution to the classification.

[0090] from Fig.12 As shown in Table 10, the cumulative contribution rate of the first principal component to the new original line classification is 87.50%, which can explain most of the information of the model. Through the contribution of the first principal component index, the contribution rate of each chemical index to the new original line classification is calculated. The calculation results are shown in Table 10.

[0091]

[0092] It can be seen from the table that 5 chemical components such as ash have important contributions to the new yarn, and 3 chemical components such as nicotine have important contributions to the original yarn.

[0093] cb, SPSS19.0 software was used to measure the indicators of 35 volatile components of the selected cigarette samples, and variance analysis, correlation analysis and discriminant analysis were performed.

[0094] A statistical analysis was conducted on the chemical composition indicators of the original production line and the new production line. The results are shown in Tables 11 and 12 below.

[0095]

[0096]

[0097] As can be seen from Table 12, the four volatile components of butanediol, 6-methyl-3-furanmethanol, 2,4-dihydroxy-2,6-dimethyl-3(2H)-furan-3-one and phenylethyl alcohol were not detected in the original production line. Among the other 31 chemical component indicators tested, the average nicotine content in the original production line and the new production line was the highest at 127.72 and 142.16, respectively, and the highest standard deviations were 17.77 and 20.73, respectively; the average content of megastigmatrienol was the lowest at 0.30 and 0.7, respectively. 1, while the standard deviation of 3-pentylfuran was the lowest at 0.04 and 0.05 respectively; From the coefficient of variation RSD values ​​of each intrinsic chemical component index of the original production line and the new production line in Table 6, it can be seen that the coefficient of variation of 6-methyl-3-furfural, 3-3-butanediol, triglyceride, megastigmatrienone and formic acid in the original production line is relatively large, and the coefficient of variation of 6-methyl-6-heptene-ethanone, megastigmatrienone, 3-pentyl-furan, 3-3-butanediol and dihydro-3-methyl-3 (2H)-furanone in the new production line is relatively large; The above indicators differed greatly between the original production line and the new production line. As the average content of chemical components increased, the corresponding standard deviation increased, but the overall change was not significant, ranging from 0.010 to 21.00, indicating that there was no significant difference in the content of volatile components between cigarette samples from the original production line and the new production line.

[0098] The PLS-DA method (partial least squares discriminant analysis) was used to analyze the characteristic volatile components of the products of the original production line and the new production line A, and to analyze the characteristic volatile component indicators of the quality differences between the new and original product lines. Fig.15 As shown, the samples from the two production lines are clearly distinguished, indicating that the products from the two production lines have significant differences in characteristic volatile components.

[0099] like Fig.16 , 17 As shown in the figure, the cluster centers of the two production lines are far apart, proving that the Xinyuan production line has significant differences in characteristic volatile components. Different characteristic volatile component indicators are distributed near the two cluster centers. The indicator closer to the cluster center indicates that the indicator has a greater contribution to the classification.

[0100] from Fig.18 As shown in Table 13, the cumulative contribution rate of the first principal component to the new original line classification is 84.04%, which can explain most of the information of the model. Through the contribution of the first three principal component indicators, the contribution rate of each chemical indicator to the new original line classification is calculated. The calculation results are shown in Table 13.

[0101]

[0102] It can be seen from the table that 16 volatile components such as isobutyraldehyde have a certain positive contribution to the new line, and 13 volatile components such as propylene glycol have a certain contribution to the original line.

[0103] From the above analysis results, it can be seen that all 31 volatile indicators detected have more or less contributed to the classification of the new original line. However, due to the large number of volatile components detected, the complexity and difficulty of the analysis are increased, and it is difficult to find the differences in the volatile components of the new original line products. Therefore, it is necessary to screen the various volatile component indicators obtained by analysis, and then select the volatile characteristic indicators that can characterize the differences of the new original line A products.

[0104] The 31 volatile component indicators of the original production line and the new production line were calculated by Log 10 After normalization, SPSS19.0 software was used to perform one-way ANOVA on the treated chemical indicators according to the independent sample T test method. The results are shown in Table 14.

[0105] As shown in Table 14, since the number of samples of the original production line (60) and the new production line (192) are different, the variances of the original production line and the new production line are different. From the single sample T test, it can be seen that under the condition of 95% test confidence interval, there is no significant difference between the original production line and the new production line in 3-vinylfuran, furfuryl alcohol, 5-methyl-2-furfural, 5-hydroxymethyl-furfural, triglyceride, tricyclic olefin (tricyclic terpene) and geranyl acetone, that is, the significant P values ​​are 0.352, 0.722, 0.070, 0.810, 0.521, 0.518 and 0.521 respectively. Therefore, the above 7 indicators have no important role in the difference analysis between the original production line and the new production line, and the remaining indicators have important roles in the difference between the original production line and the new production line.

[0106] The remaining 24 chemical composition indicators are divided according to different functional group properties, see Table 15.

[0107]

[0108] A correlation analysis was conducted on 24 indicators that play an important role in the difference analysis between the original production line and the new production line. The results are shown in Table 16.

[0109]

[0110]

[0111] The chemical components with correlation greater than 0.50 are listed in Table 17.

[0112]

[0113] As can be seen from Table 16, there is a certain correlation between the 24 chemical components. The statistical analysis of the chemical components with strong correlation shows that isobutyraldehyde (Z1) has a very significant positive correlation with hexanal (Z9) and α-dihydrodamascone (Z31), and a very significant negative correlation with 2,3-butanediol (Z8) and 2,3-dihydro-3,5-dihydroxy-6-methyl-4H-pyran-4-one (Z25); acetic acid (Z4) and furfural (Z11) and 1-(3-furanyl)-ethanone (Z14) have a very significant positive correlation with acetic acid (Z4). There is a very significant positive correlation between 1,3-propanediol (Z6) and 3-pentyl-furan (Z20) and benzyl alcohol (Z21); there is a very significant positive correlation between pyridine (Z7) and benzyl alcohol (Z21); there is a very significant positive correlation between 2,3-butanediol (Z8) and 3-acetylpyrrole (Z23) and 2,3-dihydro-3,5-dihydroxy-6-methyl-4H-pyran-4-one (Z25); there is a very significant positive correlation between furfural (Z11) It has a very significant positive correlation with 4-cyclopentene-1,3-dione (Z13) and 1-(3-furanyl)-ethanone (Z14); 4-cyclopentene-1,3-dione (Z13) has a very significant positive correlation with 3-furanyl)-ethanone (Z14); benzaldehyde (Z16) has a very significant positive correlation with phenylacetaldehyde (Z22) and 3-acetylpyrrole (Z23) and neophytadiene (Z35); 3-pentyl-furan (Z20) has a very significant positive correlation with 2,3-Dihydro-3,5-dihydroxy-6-methyl-4H-pyran-4-one (Z25) had a very significant positive correlation; 3-Acetylpyrrole (Z23) was very significantly positively correlated with 2,3-dihydro-3,5-dihydroxy-6-methyl-4H-pyran-4-one (Z25); 2,3-Dihydro-3,5-dihydroxy-6-methyl-4H-pyran-4-one (Z25) was very significantly positively correlated with nicotine (Z28).

[0114] Therefore, Z13 was used to replace Z14 and Z11, Z1 was used to replace Z9 and Z31, Z16 was used to replace Z22 and Z23, Z25 was used to replace Z20 and Z28, and Z21 was used to replace Z7. Fifteen indicators including isobutyraldehyde, formic acid, acetic acid, 1,3-propylene glycol, 2,3-butanediol, dihydro-3-methyl-3 (2H)-furanone, 4-cyclopentene-1,3-dione, benzaldehyde, 6-methyl-5-furfuryl-3-one, benzyl alcohol, 2,3-dihydro-3,5-dihydroxy-6-methyl-4H-pyran-4-one, ionene, β-damascenone, megantrienone and neophytadiene were screened out for the difference analysis of characteristic volatile component indicators between the original and new production lines.

[0115] Fifteen indicators with significant differences were selected as characteristic indicators for further analysis of the differences in volatile components of the two production lines of Xinyuan.

[0116] Repeat the above steps and use the PLS-DA method (partial least squares discriminant analysis) to analyze the 15 volatile components of the products of the original production line and the new production line A, and analyze the characteristic chemical indicators of the quality differences between the two product lines.

[0117] like Fig.19 As shown in the figure, the samples of the two production lines are clearly distinguished, indicating that the products of the two production lines have significant differences in volatile components. This shows that the selected characteristic indicators can accurately characterize the differences of the new original line products. The above characteristic indicator screening method is reasonable, and the selected characteristic indicators can be used for the difference analysis of the new original line.

[0118] like Fig. 20 , 21 As shown in the figure, the class centers of the two production lines are far apart, proving that the volatile components of the Xinyuan production line are significantly different. Different volatile component indicators are distributed near the two class centers. The indicator closer to the class center indicates that the indicator has a greater contribution to the classification.

[0119] from Fig. 22 As shown in Table 18, the cumulative contribution rate of the first principal component to the new original line classification reaches 85.00%, which can explain most of the information of the model. Through the contribution of the first principal component index, the contribution rate of each volatile component index to the new original line classification is calculated. The calculation results are shown in Table 18.

[0120]

[0121] It can be seen from Table 18 that 8 chemical components such as 1,3-propylene glycol have important contributions to the new line A. 7 chemical components such as isobutyraldehyde have important contributions to the original line A.

[0122] In summary, the overall information of near infrared spectroscopy was used to conduct a difference analysis on the new original line A products through principal component analysis combined with Mahalanobis distance (PCA-MD) analysis method. The results showed that there was a significant difference in intrinsic quality between the new production line and the original production line A products.

[0123] The overall information of chromatographic fingerprint was used to analyze the difference of the new original line A products through principal component analysis combined with Mahalanobis distance (PCA-MD) analysis method. The results showed that there were significant differences in the intrinsic quality of the new production line and the original production line A products.

[0124] Both spectral fingerprints and chromatographic fingerprints can be used to evaluate the overall differentiation of the new original line A products.

[0125] Using chemometrics software such as SPSS 19.0 and SIMCA 11.5, partial least squares-discriminant analysis (PLS-DA), regression analysis, correlation analysis, and variance analysis methods, characteristic index screening and difference analysis were performed on the main chemical composition indicators and volatile components, and characteristic indicators that play an important role in the difference evaluation of the new original production line were screened out. The verification results show that the screened characteristic indicators can be applied to the difference evaluation of the new original line products. Example 2

[0126] A method for evaluating the quality differences of cigarette products based on multidimensional spatial chemical fingerprints. Taking cigarette sample B as an example, the operation method and parameter settings are consistent with those in Example 1.

[0127] a. Establish near infrared discrimination and classification model Draw a spectrum classification effect diagram, calculate the distribution of samples in each category and draw a principal component score diagram, such as Fig.23 , 24 As shown in the figure, the spectral information classification of the original production line and the new production line can be clearly divided into two parts, because the near-infrared spectrum can characterize the changes in the overall physical and chemical information of the sample. Therefore, the intrinsic quality of product B produced by the two production lines has significant differences.

[0128] b. Establishing gas chromatography discrimination and classification model HS-GC-MS was used to collect gas chromatography fingerprint data of the samples, and the headspace sampler was set with a heating temperature of 130 °C, a heating time of 30 min, and an injection time of 1 min.

[0129] Draw a spectrum classification effect diagram, calculate the distribution of samples in each category and draw a principal component score diagram, such as Fig.33 , 34 shown.

[0130] c. Index detection ca, SPSS19.0 software was used to measure the indicators of 10 conventional chemical components, 4 polyphenol contents, ash content and pH value of the selected cigarette samples, and variance analysis, correlation analysis and discriminant analysis were performed.

[0131] A statistical analysis was conducted on the chemical composition indicators of the original production line and the new production line. The results are shown in Table 19 below.

[0132]

[0133] As can be seen from Table 19, compared with the new production line, except for the average content of potassium-chlorine ratio, sugar-alkali ratio, K2O, rutin and total polyphenols, the contents of other indicators are lower than those of the new production line. From the relative standard deviations of the original and new production lines, it can be seen that except for the relative standard deviation RSD values ​​of total sugar and scopoletin, the original production line is higher than the new production line, and the other indicators are lower than the new production line. Among them, the indicator with the largest RSD value (12.08%) of the original production line is scopoletin, and the indicator with the largest RSD value (12.47%) of the new production line is chlorogenic acid. The results show that the content of each indicator in the new production line is quite different.

[0134] The PLS-DA method (partial least squares discriminant analysis) was used to analyze the main chemical components of the products of the original production line and the new production line B, and to analyze the characteristic chemical indicators of the quality differences between the two product lines. Fig.25 As shown, the samples from the two production lines are clearly distinguishable, indicating that the products from the two production lines have significant differences in their intrinsic chemical composition.

[0135] like Fig.26 , 27 As shown in the figure, the cluster centers of the two production lines are far apart, proving that the Xinyuan production line has significant differences in intrinsic chemical composition. Different chemical composition indicators are distributed near the two cluster centers. The indicator closer to the cluster center indicates that the indicator has a greater contribution to the classification.

[0136] from Fig.28 As shown in Table 20, the cumulative contribution rate of the first two principal components to the new original line classification reaches 80.88%, which can explain most of the information of the model. Through the contribution of the first two principal component indicators, the contribution rate of each chemical indicator to the new original line classification is calculated. The calculation results are shown in Table 20.

[0137]

[0138] It can be seen from the table that 11 chemical components such as reducing sugar have a certain contribution to the new line B. Five chemical components such as rutin have a certain contribution to the original line B.

[0139] From the above analysis results, it can be seen that all the 16 indicators detected have more or less contributed to the classification of the new original line. However, due to the large number of chemical indicators detected, the complexity and difficulty of the analysis are increased, and it is difficult to find the differences in the chemical composition of the products of the new original production line. Therefore, it is necessary to screen the various chemical indicators obtained by analysis, and then select the chemical indicators that can characterize the differences of the new original line B products.

[0140] The samples of the original production line were taken as category 1, and the samples of the new production line were taken as category 2. The 16 indicators including nicotine, total nitrogen, total sugar, reducing sugar, sugar-alkali ratio, sugar-nitrogen ratio, Schmuck value, K2O, chlorine, potassium-chlorine ratio, ash, pH value, rutin, chlorogenic acid, scopoletin and total polyphenols were analyzed by Logarithmetic. 10 After normalization, SPSS19.0 software was used to perform linear regression analysis with the original production line and the new production line and classification (1,2) as dependent variables and 16 chemical composition indicators as independent variables to explore the impact of different indicators on classification. The results are shown in Table 21.

[0141]

[0142] It can be seen from Table 21 that when linear regression analysis was performed with the original production line as Category 1 and the new production line as Category 2, it was found that within the 95% confidence interval, there were significant differences in the eight indicators of reducing sugar, K2O, ash content, pH value, rutin, chlorogenic acid, scopoletin and total polyphenols, and the collinearity tolerance factors were large, which were 0.235, 0.523, 0.617, 0.506, 0.446, 0.318 and 0.371, respectively, indicating that the classification of the original production line and the new production line had an important influence on these eight indicators.

[0143] The results of the correlation analysis between the selected 8 indicators and other indicators are shown in Table 22. Except for the insignificant correlation between reducing sugar and chlorine and chlorogenic acid, there are significant correlations between the remaining indicators; except for the insignificant correlation between K2O and Schmuck value, chlorine and pH value, there are significant correlations between the remaining indicators; except for the insignificant correlation between ash content and chlorine, sugar-alkali ratio and chlorogenic acid, there are significant correlations between the remaining indicators; except for the insignificant correlation between pH value and total nitrogen, K2O, rutin and chlorogenic acid, there are significant correlations between pH value and total nitrogen, K2O, rutin and chlorogenic acid. Except for the correlation with total nitrogen, total sugar, sugar-nitrogen ratio and pH value, rutin has significant correlation with other indicators; except for the correlation with reducing sugar, sugar-nitrogen ratio, ash and pH value, chlorogenic acid has significant correlation with other indicators; except for the correlation with potassium-chlorine ratio, scopoletin has significant correlation with other indicators; except for the correlation with total sugar and pH value, the total amount of polyphenols has significant correlation with other indicators.

[0144] From the correlation analysis, it can be seen that the correlations between the above-mentioned 8 indicators are mostly not significant, while the correlations between them and the remaining indicators are mostly significant. This further verifies that there is no collinearity between the 8 selected indicators, and that the 8 selected indicators can replace most of the information of the remaining indicators, revealing that the screening of the above 8 indicators plays an important role in the classification and differentiation of the original production line and the new production line.

[0145]

[0146] Eight indicators with significant differences were selected as characteristic indicators for further analysis of the differences in chemical composition between the two production lines of Xinyuan.

[0147] Repeat the above steps and use the PLS-DA method (partial least squares discriminant analysis) to analyze the eight main chemical components of the products of the original production line and the new production line B, and analyze the characteristic chemical indicators of the quality differences between the new and original product lines. Fig.29 As shown in the figure, the samples of the two production lines are clearly distinguished, indicating that the products of the two production lines have significant differences in their intrinsic chemical composition. This shows that the selected characteristic indicators can accurately characterize the differences of the new original line products. The above characteristic indicator screening method is reasonable, and the selected characteristic indicators can be used for the difference analysis of the new original line.

[0148] like Fig.30 , 31 As shown in the figure, the cluster centers of the two production lines are far apart, proving that the Xinyuan production line has significant differences in intrinsic chemical composition. Different chemical composition indicators are distributed near the two cluster centers. The indicator closer to the cluster center indicates that the indicator has a greater contribution to the classification.

[0149] from Fig.32 As shown, the cumulative contribution rate of the first two principal components to the new original line classification reaches 81.75%, which can explain most of the information of the model. Through the contribution of the first two principal component indicators, the contribution rate of each chemical indicator to the new original line classification is calculated. The calculation results are shown in Table 23.

[0150]

[0151] It can be seen from the table that 5 chemical components such as reducing sugar have important contributions to the new thread, and 3 chemical components such as rutin have important contributions to the original thread.

[0152] cb, SPSS19.0 software was used to measure the indicators of 35 volatile components of the selected cigarette samples, and variance analysis, correlation analysis and discriminant analysis were performed.

[0153] A statistical analysis was conducted on the chemical composition indicators of the original production line and the new production line. The results are shown in Tables 24 and 25 below.

[0154]

[0155]

[0156] It can be seen from Table 25 that among the 28 chemical component indicators tested, the average content of propylene glycol is the highest, 405.49, and its standard deviation is the highest, 157.84; the average content of dihydroactinol is the lowest, 0.68, and the standard deviation of geranyl acetone is the lowest, 0.12; from the coefficient of variation RSD values ​​of each intrinsic chemical component indicator of the original production line and the new production line in Table 19, it can be seen that in the original production line, the coefficient of variation of propylene glycol, furfuryl alcohol, butyrolactone and triglyceride is relatively large, and in the new production line, the coefficient of variation of pyridine, 3-cyclopentene-1,4-dione, butyrolactone, 5-methylfurfural, phenylacetaldehyde, 1-(1H-pyrrol-2-yl)ethanone, 5-hydroxymethylfurfural, nicotine, β-dihydrodamascone, dihydroactinol, megatrienone B, megatrienone D and neophytadiene is relatively large, indicating that the above indicators are quite different between the original production line and the new production line.

[0157] There are certain differences in the average content and standard deviation of each volatile component between the original production line and the new production line, indicating that the volatile components are significantly different between different production lines.

[0158] The PLS-DA method (partial least squares discriminant analysis) was used to analyze the characteristic volatile components of the products of the original production line and the new production line B, and to analyze the characteristic volatile component indicators of the quality differences between the new and original product lines. Fig.35 As shown, the samples from the two production lines are clearly distinguished, indicating that the products from the two production lines have significant differences in characteristic volatile components.

[0159] like Fig.36 , 37 As shown in the figure, the cluster centers of the two production lines are far apart, proving that the Xinyuan production line has significant differences in characteristic volatile components. Different characteristic volatile component indicators are distributed near the two cluster centers. The indicator closer to the cluster center indicates that the indicator has a greater contribution to the classification.

[0160] from Fig.38 As shown in Table 26, the cumulative contribution rate of the first four principal components to the new original line classification reaches 87.75%, which can explain most of the information of the model. Through the contribution of the first four principal component indicators, the contribution rate of each chemical indicator to the new original line classification is calculated. The calculation results are shown in Table 26.

[0161]

[0162] It can be seen from the table that 14 volatile components such as furfuryl alcohol have a certain contribution to the original line B. 13 volatile components such as propylene glycol have a certain contribution to the new line B.

[0163] From the above analysis results, it can be seen that all 27 volatile indicators detected have more or less contributed to the classification of the new original line. However, due to the large number of volatile components detected, the complexity and difficulty of the analysis are increased, and it is difficult to find the differences in the volatile components of the new original line products. Therefore, it is necessary to screen the various volatile component indicators obtained by analysis, and then select the volatile characteristic indicators that can characterize the differences of the new original line B products.

[0164] The 28 chemical composition indicators of the original production line and the new production line were calculated by Log 10 After normalization, SPSS19.0 software was used to perform one-way ANOVA on the treated chemical indicators according to the independent sample T test method. The results are shown in Table 27.

[0165]

[0166] As shown in Table 27, since the number of samples of the original production line (51) and the new production line (401) are different, the variance of the original production line and the new production line is different. The single sample T test shows that under the condition of 95% test confidence interval, there is no significant difference between the original production line and the new production line in butyrolactone, menthol, naphthalene, nicotine and dihydroactin, that is, the significant P values ​​are 0.121, 0.414, 1.000, 0.728 and 0.216 respectively. Therefore, the above five indicators have no important role in the difference analysis between the original production line and the new production line, and the remaining indicators have important roles in the difference between the original production line and the new production line.

[0167] The 23 chemical composition indicators are divided according to different functional group properties, see Table 28.

[0168]

[0169] A correlation analysis was conducted on 23 indicators that play an important role in the difference analysis between the original production line and the new production line. The results are shown in Table 29.

[0170]

[0171]

[0172] Chemical components with correlation greater than 0.50 are included in Table 30.

[0173]

[0174] As can be seen from Table 30, there is a certain correlation between the 23 chemical components. The statistical analysis of the chemical components with strong correlation shows that propylene glycol (V1) has a very significant positive correlation with triglyceride (V11); furfural (V4) has a very significant positive correlation with 1-(H-pyrrol-2-yl)ethanone (V7); furfuryl alcohol (V5) has a very significant positive correlation with benzaldehyde (V9), 1-(1H-pyrrol-2-yl)ethanone (V13), 2,3-dihydro-3,5-dihydroxy-6-methyl-4H-pyran-4-one (V14), 5-hydroxymethylfurfural (V19) and solanone (V22); 3-cyclopentene-1,4-dione (V6) has a very significant positive correlation with 1-(3-furanyl)-ethanone (V7), 5-hydroxymethylfurfural (V10), phenylacetaldehyde (V12), and trans-thin There was a very significant negative correlation between furfural (V16) and β-dihydrodamascone (V23); there was a very significant positive correlation between 1-(H-pyrrol-2-yl)ethanone (V7) and β-dihydrodamascone (V23); there was a very significant positive correlation between 5-hydroxymethylfurfural (V10) and phenylacetaldehyde (V12), and a very significant negative correlation with trans-menthone (V16); there was a very significant positive correlation between phenylacetaldehyde (V12) and neophytadiene (V28), and a very significant negative correlation with trans-menthone (V16); there was a very significant positive correlation between 1-(1H-pyrrol-2-yl)ethanone (V13) and 2,3-dihydro-3,5-dihydroxy-6-methyl-4H-pyran-4-one (V14) and 5-hydroxymethylfurfural (V19); there was a very significant positive correlation between 2,3 -Dihydro-3,5-dihydroxy-6-methyl-4H-pyran-4-one (V14) had a significant positive correlation with 5-hydroxymethylfurfural (V19), nicotine (V20), solanone (V22), geranylacetone (V24), dihydroactinol (V25), megastigmatrienone B (V26), and megastigmatrienone D (V27); menthone (V15) had an extremely significant positive correlation with trans-menthone (V16) and menthol (V17); trans-menthone (V16) had an extremely significant positive correlation with menthol (V17) and β-dihydrodamascone (V23); 5-hydroxymethylfurfural (V19) had an extremely significant positive correlation with solanone (V22); nicotine (V20) and dihydroactinol There was an extremely significant positive correlation between solanone (V22) and β-dihydrodamascone (V23) and geranylacetone (V24); geranylacetone (V24) had a significant positive correlation with dihydroactinol (V25), megastigmatrienone B (V26), and megastigmatrienone D (V27); there was an extremely significant positive correlation between dihydroactinol (V25) and megastigmatrienone B (V26), megastigmatrienone D (V27) and neophytadiene (V28); there was an extremely significant positive correlation between megastigmatrienone B (V26), megastigmatrienone D (V27) and neophytadiene (V28).

[0175] Eliminating the above indicators that have no significant contribution to the B classification of cigarettes from new and original production lines: the five chemical components of butyrolactone, menthol, naphthalene, nicotine and dihydroactinol, and combining the indicators of the same type in Table 3-20, the chemical components with strong correlation can replace each other to a certain extent due to their similar effects on cigarettes. Trans-menthone can be used to replace menthone; megastigmatrienone B can replace megastigmatrienone D; furfural can replace 1-(H-pyrrol-2-yl)ethanone; 5-hydroxymethylfurfural can replace phenylacetaldehyde and 2,3-dihydro-3,5-dihydroxy-6-methyl-4H-pyran-4-one.

[0176] Eighteen indicators with significant differences were selected as characteristic indicators: propylene glycol, pyridine, 2,3-butanediol, furfural, furfuryl alcohol, 3-cyclopentene-1,4-dione, 1-(3-furanyl)-ethanone, benzaldehyde, 5-methylfurfural, benzyl alcohol, trans-menthone, 5-hydroxymethylfurfural, triglyceride, solanone, β-dihydrodamascone, geranylacetone, megastigmatrienone D and neophytadiene. They were used for further analysis of the differences in volatile components of the two production lines of Xinyuan.

[0177] Repeat the above steps and use the PLS-DA method (partial least squares discriminant analysis) to analyze the 18 volatile components of the products of the original production line and the new production line B, and analyze the characteristic chemical indicators of the quality differences between the two product lines.

[0178] like Fig.39 As shown in the figure, the samples of the two production lines are clearly distinguished, indicating that the products of the two production lines have significant differences in volatile components. This shows that the selected characteristic indicators can accurately characterize the differences of the new original line products. The above characteristic indicator screening method is reasonable, and the selected characteristic indicators can be used for the difference analysis of the new original line.

[0179] like Fig.40 , 41 As shown in the figure, the class centers of the two production lines are far apart, proving that the volatile components of the Xinyuan production line are significantly different. Different volatile component indicators are distributed near the two class centers. The indicator closer to the class center indicates that the indicator has a greater contribution to the classification.

[0180] from Fig.42 As shown in Table 31, the cumulative contribution rate of the first two principal components to the new original line classification reaches 93.16%, which can explain most of the information of the model. Through the contribution of the first two principal component indicators, the contribution rate of each volatile component indicator to the new original line classification is calculated. The calculation results are shown in Table 31.

[0181]

[0182] It can be seen from Table 31 that 12 volatile components such as furfural and propylene glycol have a certain contribution to the new line B. 6 volatile components such as furfuryl alcohol have a certain contribution to the original line B.

[0183] In summary, the overall information of near-infrared spectroscopy was used to conduct a difference analysis on the new original line B products through principal component analysis combined with Mahalanobis distance (PCA-MD) analysis method. The results showed that there was a significant difference in intrinsic quality between the new production line and the original production line B products.

[0184] The overall information of chromatographic fingerprint was used to analyze the difference of the new original line B products through principal component analysis combined with Mahalanobis distance (PCA-MD) analysis method. The results showed that there were significant differences in the intrinsic quality between the new production line and the original production line B products.

[0185] Both spectral fingerprints and chromatographic fingerprints can be used to evaluate the overall differentiation of the new original line B products.

[0186] Using chemometrics software such as SPSS 19.0 and SIMCA 11.5, partial least squares-discriminant analysis (PLS-DA), regression analysis, correlation analysis, and variance analysis methods, characteristic index screening and difference analysis were performed on the main chemical composition indicators and volatile components, and characteristic indicators that play an important role in the difference evaluation of the new original production line were screened out. The verification results show that the screened characteristic indicators can be applied to the difference evaluation of the new original line products.

[0187] The scope of protection claimed by the present invention is not limited to the above specific embodiments. For those skilled in the art, the present invention may have various deformations and changes. Any modifications, improvements and equivalent substitutions made within the concept and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for evaluating the quality differences of cigarette products based on multidimensional spatial chemical fingerprints, characterized in that: The following steps are involved: a. Establish near infrared discrimination and classification model Collect the near-infrared diffuse reflectance spectrum data of the samples, and establish a near-infrared discrimination and classification model after optimization processing; Take the control sample to establish the near-infrared correction model; import the collected sample into the near-infrared correction model of the control sample for calculation, predict and analyze the chemical composition content of the sample and the control sample; classify the samples according to different batches, apply the principal component analysis method, draw the spectral classification effect diagram, calculate the distribution of samples in each category and draw the principal component score diagram, b. Establishing chromatographic discrimination and classification model HS-GC-MS was used to collect the gas chromatography fingerprint data of the samples, and the gas chromatography discrimination and classification model was established after optimization processing; c. Index detection The selected cigarette samples were tested for conventional chemical indices, polyphenol content, ash content and pH value, and variance analysis, correlation analysis and discriminant analysis were performed; The indicators of the selected cigarette samples were tested for volatile components, and variance analysis, correlation analysis and discriminant analysis were performed.

2. The method for evaluating the quality difference of cigarette products based on multidimensional chemical fingerprints according to claim 1, characterized in that: Step c includes the following steps: The models of steps a and b are subjected to pattern recognition, component difference analysis, impact analysis on PLS-DA separation contribution, component data normalization, classification of different types of cigarette samples, independent sample t-test or linear regression analysis, selection of chemical indicators with significant differences, correlation analysis, replacement of chemical components with smaller significant differences, PLS-DA analysis, and characteristic indicators to characterize cigarette differences.

3. The method for evaluating the quality difference of cigarette products based on multidimensional spatial chemical fingerprints according to claim 2, characterized in that: In step c, the influence of PLS-DA separation contribution is analyzed, PLS-DA score diagram and class center, chemical index factor distribution diagram and score loading diagram are drawn, and the characteristic chemical indicators of the quality differences of the two product lines of Xinyuan are analyzed; the differences in the intrinsic chemical composition of the products of the two production lines are compared, and the number of principal components of the PLS-DA class model of the main chemical composition indicators is calculated based on the indicators closer to the class center, and the contribution rate of each indicator to the classification of the Xinyuan line is calculated, and then the indicators that can characterize the differences between the two product lines are selected.

4. The method for evaluating the quality difference of cigarette products based on multidimensional spatial chemical fingerprints according to claim 1, characterized in that: In step b, collecting sample gas chromatography fingerprint data includes the following steps: Take 100 cigarettes and mix them thoroughly. After balancing at constant temperature and humidity for 48 hours, store them for later use. Weigh the sample into a headspace bottle, add 1 μL of 25 mg / mL naphthaleneethanol solution as an internal standard, place the headspace bottle in a headspace sampler, and perform automatic sampling for GC / MS analysis. A three-factor three-level orthogonal experiment was conducted on the headspace heating temperature, headspace equilibration time and sample amount to determine the optimal headspace heating temperature, optimal headspace equilibration time and optimal sample amount; repeated measurements were performed under the optimal conditions, and the relative standard deviations of the six measurements of 31 components were calculated.

5. The method for evaluating the quality difference of cigarette products based on multidimensional chemical fingerprints according to claim 4, characterized in that: In step b, the optimal conditions for setting the headspace sampler are: sample amount 0.6 g, headspace heating temperature 130° C., and heating time 30 min.

6. The method for evaluating the quality difference of cigarette products based on multidimensional spatial chemical fingerprints according to claim 1, characterized in that: The optimization processing in steps a and b includes signal noise reduction, outlier detection, spectrum preprocessing, modeling spectral region selection and determination of the maximum number of principal components.

7. The method for evaluating the quality difference of cigarette products based on multidimensional chemical fingerprints according to claim 1, characterized in that: In step a, collecting sample near-infrared diffuse reflectance spectrum data includes the following steps: The prepared sample was crushed, passed through a 60-mesh sieve, and placed in a sealed bag for later use. The spectrometer was preheated at room temperature for 2 hours. The sample was placed in a 5 cm quartz cup with more than half the capacity, gently pressed to be flat, and placed on a rotating table to collect the near-infrared diffuse reflectance spectrum data of the sample.

Images