Chromatographic difference peak identification method and essence and flavor quality evaluation method

Abstract

The application relates to the technical field of chromatographic difference peak identification methods for essence and flavor quality evaluation, and discloses a chromatographic difference peak identification method for essence and flavor quality evaluation, which comprises the following steps: determining a first matrix based on characteristic peak data of each sample in a first sample set, and determining a second matrix based on characteristic peak data of each sample in a second sample set; performing principal component analysis based on the first matrix to obtain a first principal component feature matrix and a second principal component feature matrix; performing bidirectional interaction attention calculation based on the first principal component feature matrix and the second principal component feature matrix to respectively obtain a first fusion feature matrix and a second fusion feature matrix; determining a difference parameter of each characteristic peak based on the first fusion feature matrix, the second fusion feature matrix, the first matrix and the second matrix, and determining a chromatographic difference peak based on the difference parameters. The application also discloses an essence and flavor quality evaluation method.

Description

Technical Field

[0001] This application relates to the field of fragrance quality testing technology, specifically to a chromatographic difference peak identification method and a fragrance quality evaluation method for evaluating the quality of fragrances and flavors. Background Technology

[0002] In modern production, flavorings and fragrances are widely used in food, tobacco, daily chemicals and other fields. Their quality directly affects the quality and reputation of the end products. Therefore, it is crucial to do a good job in the quality control of flavorings and fragrances.

[0003] Gas chromatography-mass spectrometry (GC-MS) is a core method for analyzing fragrance and flavor components and achieving quality control. It can accurately detect various volatile components in fragrances and flavors, helping to differentiate product categories and identify genuine products, thus ensuring batch-to-batch consistency. However, existing GC-MS technology generates massive amounts of complex chromatographic data containing numerous peaks, making data analysis extremely difficult and hindering rapid product quality analysis. Summary of the Invention

[0004] This application provides a chromatographic difference peak identification method for flavor and fragrance quality evaluation and a flavor and fragrance quality evaluation method thereof. This method can screen out characteristic peaks that are more important for quality identification as the basis for product quality evaluation, thereby greatly reducing the amount of data to be analyzed and providing convenient conditions for rapid product quality analysis.

[0005] In a first aspect, this application provides a method for identifying chromatographic difference peaks for evaluating the quality of fragrances and flavorings, including:

[0006] A first matrix is ​​determined based on the characteristic peak data of each sample in the first sample set, and a second matrix is ​​determined based on the characteristic peak data of each sample in the second sample set, wherein the element in the i-th row and k-th column of the first matrix is ​​related to the peak area of ​​the k-th characteristic peak of the i-th sample in the first sample set, and the element in the j-th row and k-th column of the second matrix is ​​related to the peak area of ​​the k-th characteristic peak of the j-th sample in the second sample set.

[0007] Principal component analysis is performed based on the first matrix to obtain the first principal component feature matrix and the second principal component feature matrix.

[0008] Bidirectional interactive attention is calculated based on the first principal component feature matrix and the second principal component feature matrix to obtain the first fusion feature matrix and the second fusion feature matrix respectively.

[0009] The difference parameters of each characteristic peak are determined based on the first fusion feature matrix, the second fusion feature matrix, the first matrix, and the second matrix, and the chromatographic difference peaks are determined based on each difference parameter.

[0010] In some implementations of this application, bidirectional interactive attention calculation is performed based on the first principal component feature matrix and the second principal component feature matrix to obtain the first fusion feature matrix and the second fusion feature matrix, respectively, including:

[0011] Using the first principal component feature matrix as the query and the second principal component feature matrix as the key and value, obtain the first attention score matrix;

[0012] The first attention score matrix is ​​normalized using softmax to obtain the first attention weight matrix.

[0013] The first fused feature matrix is ​​obtained by weighting and fusing the second principal component feature matrix based on the first attention weight matrix.

[0014] Using the second principal component feature matrix as the query and the first principal component feature matrix as the key and value, the second attention score matrix is ​​obtained.

[0015] The second attention score matrix is ​​normalized using softmax to obtain the second attention weight matrix.

[0016] The first principal component feature matrix is ​​weighted and fused based on the second attention weight matrix to obtain the second fused feature matrix.

[0017] In some implementations of this application, the difference parameters of each feature peak are determined based on the first fusion feature matrix, the second fusion feature matrix, the first matrix, and the second matrix, including:

[0018] Based on the first attention weight matrix and the second attention weight matrix, obtain the first fusion weight vector corresponding to each sample in the first sample set and the second fusion weight vector corresponding to each sample in the second sample set.

[0019] The second matrix is ​​weighted based on the first fusion weight vector corresponding to each sample to obtain the second weighted feature vector corresponding to each sample in the second sample set, and the first peak parameter vector is obtained based on the first matrix and each second weighted feature vector.

[0020] The first matrix is ​​weighted based on the second fusion weight vector corresponding to each sample to obtain the first weighted feature vector corresponding to each sample in the first sample set, and the second peak parameter vector is obtained based on the second matrix and each first weighted feature vector.

[0021] Based on the first peak parameter vector and the second peak parameter vector, the difference parameters of each characteristic peak are obtained; the first peak parameter vector and the second peak parameter vector have the same dimension, which is equal to the number of characteristic peaks.

[0022] In some implementations of this application, based on the first attention weight matrix and the second attention weight matrix, the first fusion weight vector corresponding to each sample in the first sample set and the second fusion weight vector corresponding to each sample in the second sample set are obtained, including:

[0023] Construct a first reference weight vector and a first correction weight vector for each sample in the first sample set. The first reference weight vector corresponding to the i-th sample in the first sample set is constructed based on the i-th row element of the first attention weight matrix, and the first correction weight vector corresponding to the i-th sample in the first sample set is constructed based on the i-th column element of the second attention weight matrix.

[0024] The first fusion weight vector is determined based on the average value of the first reference weight vector and the first correction weight vector corresponding to each sample in the first sample set.

[0025] Construct a second reference weight vector and a second correction weight vector for each sample in the second sample set. The second reference weight vector corresponding to the j-th sample in the second sample set is constructed based on the j-th row element of the second attention weight matrix, and the second correction weight vector corresponding to the j-th sample in the second sample set is constructed based on the j-th column element of the first attention weight matrix.

[0026] The second fusion weight vector is determined based on the average value of the second reference weight vector and the second correction weight vector corresponding to each sample in the second sample set.

[0027] In some implementations of this application, obtaining the first peak parameter vector based on the first matrix and each second weighted feature vector includes: for the i-th row of the first matrix, obtaining the difference vector corresponding to the sample based on the absolute difference between each element of the row and each element of the second weighted feature vector corresponding to the i-th sample in the second sample set; traversing all rows of the first matrix to obtain each difference vector corresponding to all samples in the first sample set; summing each difference vector to determine the first peak parameter vector;

[0028] The second peak parameter vector is obtained based on the second matrix and each first weighted feature vector, including: for the j-th row of the second matrix, based on the absolute difference between each element of the row and each element of the first weighted feature vector corresponding to the j-th sample in the first sample set, the difference vector corresponding to the sample is obtained; all rows of the second matrix are traversed to obtain each difference vector corresponding to all samples in the second sample set; and the difference vectors are summed to determine the second peak parameter vector.

[0029] In some implementations of this application, the difference parameters of each characteristic peak are obtained based on the first peak parameter vector and the second peak parameter vector, including:

[0030] The first peak parameter vector and the second peak parameter vector are summed and divided by the total number of samples in the first and second sample sets. Then, the resulting vector is normalized. The k-th element of the normalized vector is the difference parameter of the k-th characteristic peak.

[0031] In some implementations of this application, principal component analysis is performed based on the first matrix and the second matrix to obtain the first principal component feature matrix and the second principal component feature matrix, including:

[0032] Perform principal component analysis on the first matrix to determine the score matrix, principal component coefficient matrix, and number of principal components D;

[0033] The first principal component feature matrix is ​​determined based on the first D columns of the score matrix;

[0034] The second principal component feature matrix is ​​determined based on the first D columns and the second matrix of the principal component coefficient matrix.

[0035] In some implementations of this application, the first principal component feature matrix is ​​a matrix composed of the first D columns of the score matrix, and the second principal component feature matrix is ​​a product matrix of the second matrix and the first D columns of the principal component coefficient matrix.

[0036] In some implementations of this application, determining a first matrix based on the characteristic peak data of each sample in a first sample set, and determining a second matrix based on the characteristic peak data of each sample in a second sample set, includes:

[0037] Obtain gas chromatography-mass spectra of fragrance samples, and determine characteristic peaks that reflect volatile components based on the spectra;

[0038] Obtain the characteristic peak data of each sample in the first sample set, construct the first intermediate matrix, normalize the first intermediate matrix, and obtain the first matrix;

[0039] Obtain the characteristic peak data of each sample in the second sample set, construct the second intermediate matrix, normalize the second intermediate matrix, and obtain the second matrix.

[0040] Secondly, this application also provides a method for evaluating the quality of flavorings and fragrances, including:

[0041] Obtain the gas chromatography-mass spectra of the sample to be tested, and obtain the data of each characteristic peak based on the spectra;

[0042] The chromatographic difference peaks of the sample to be tested are determined based on the chromatographic difference peak identification method as described in the first aspect;

[0043] For each characteristic peak data of the sample to be tested, the data related to each chromatographic difference peak are analyzed to evaluate the quality of the sample to be tested.

[0044] This application has the following beneficial effects:

[0045] This application employs a bidirectional interactive attention mechanism, simultaneously calculating the attention between different types of samples, aggregating bidirectional information to achieve a comprehensive assessment of chromatographic peak differences, thereby screening out characteristic peaks that are more favorable for quality identification as the basis for product quality evaluation, thus reducing the amount of data to be analyzed and providing convenient conditions for achieving rapid product quality analysis. Attached Figure Description

[0046] Figure 1 The GC-MS total ion chromatograms of fragrance and flavor samples A and B provided in one embodiment of this application are shown.

[0047] Figure 2 The figure shows characteristic peak curves of a certain fragrance and flavor sample A and B based on chemical substances provided in one embodiment of this application;

[0048] Figure 3 A distribution diagram of chromatographic peak differences corresponding to the chemical substances provided in one embodiment of this application is shown. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0050] GC-MS generates massive amounts of complex chromatographic data containing numerous chromatographic peaks, which require analysis using chemometric methods. Principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) are two commonly used methods, but both methods present significant challenges in data analysis.

[0051] To address the aforementioned issues, this application provides a chromatographic peak identification method for flavor and fragrance quality evaluation. The method described in this application is a peak screening method based on PCA and interactive attention. PCA is used as an auxiliary dimensionality reduction and noise reduction technique, while an interactive attention mechanism between two sets of samples is introduced to adaptively learn the correlation strength between samples. Combined with weighted difference, this method achieves accurate screening of peak differences, thereby effectively overcoming the limitations of existing technologies and meeting the needs of practical chromatographic analysis.

[0052] The chromatographic peak identification method for evaluating the quality of flavors and fragrances provided in this application includes the following steps:

[0053] Step S1: Determine a first matrix based on the characteristic peak data of each sample in the first sample set, and determine a second matrix based on the characteristic peak data of each sample in the second sample set, wherein the element in the i-th row and k-th column of the first matrix is ​​related to the peak area of ​​the k-th characteristic peak of the i-th sample in the first sample set, and the element in the j-th row and k-th column of the second matrix is ​​related to the peak area of ​​the k-th characteristic peak of the j-th sample in the second sample set.

[0054] Specifically, step S1 may include the following steps:

[0055] S11: Obtain the GC-MS spectrum of the fragrance and flavor sample, and combine it with mass spectrometry analysis to find the characteristic peaks that reflect volatile components (including sugars, amino acids, organic acids, polyphenols, alkaloids, etc.) and their corresponding peak areas, obtain the characteristic peak curve of each sample, and obtain the first curve of the sample.

[0056] S12: Arrange the first curves of each flavor and fragrance A sample (i.e., each first sample in the first sample set) to construct the first intermediate matrix of the first sample set; arrange the first curves of each flavor and fragrance B sample (i.e., each second sample in the second sample set) to construct the second intermediate matrix of the second sample set; wherein the element in the i or j-th row and k-th column of each intermediate matrix represents the peak area of ​​the k-th chromatographic peak of the i or j-th sample in the sample set.

[0057] S13: Calculate the mean and standard deviation of the peak area of ​​each chromatographic peak in the first intermediate matrix, and perform Pareto normalization on the first and second intermediate matrices respectively to obtain the first matrix X. A Second matrix X B .

[0058] Specifically, let the total number of samples in the first sample set be m, the total number of samples in the second sample set be n, and the number of characteristic peaks be l. Then the first matrix X A Let X be an m x l matrix. B It is an n-row, l-column matrix.

[0059] Step S13 can eliminate the influence of the dimensions of each sample data, thereby reducing the interference of baseline drift and detection error on subsequent analysis and improving the comparability of data.

[0060] Step S2: Perform principal component analysis based on the first matrix to obtain the first principal component feature matrix and the second principal component feature matrix.

[0061] Principal component analysis can effectively reduce redundant information in chromatographic peaks of high-dimensional GC-MS data, retain the core features of the data, thereby reducing the complexity of subsequent attention calculations and improving computational efficiency.

[0062] Specifically, step S2 includes:

[0063] Step S21: For the first matrix X A Principal component analysis was performed to obtain the score matrix T, the principal component coefficient matrix P, the variance explained rate, and the number of principal components D. The first matrix X... A The score matrix T and principal component coefficient matrix P, and the number of principal components D can be screened by "cumulative variance explained ≥ 95.0%".

[0064] Step S22: Determine the first principal component feature matrix F based on the first D columns of the score matrix T. A .

[0065] Specifically, the first principal component feature matrix F A This is a matrix consisting of the first D columns of the score matrix T:

[0066]

[0067] Step S23: Determine the second principal component feature matrix based on the first D columns of the principal component coefficient matrix and the second matrix.

[0068] Step S23 is to use the first matrix X A The first D columns of the principal component coefficient matrix P are used to construct the second matrix X. B Projecting onto the same principal component space yields the second principal component eigenma matrix F of the second matrix. B :

[0069]

[0070] Step S3: Perform bidirectional interactive attention calculation based on the first principal component feature matrix and the second principal component feature matrix to obtain the first fusion feature matrix and the second fusion feature matrix respectively.

[0071] Specifically, step S3 involves performing bidirectional interactive attention calculation on the first and second sample sets to determine the attention weights between the samples, thereby capturing the correlation information of different samples and providing a core basis for subsequent chromatographic peak weight calculation. This is divided into two directions: A→B (i.e., first sample set → second sample set) and B→A (second sample set → first sample set).

[0072] Step S3 specifically includes:

[0073] Step S31: Using the first principal component feature matrix F A For the query, the second principal component feature matrix F B Given keys and values, obtain the first attention score matrix S. AB :

[0074]

[0075] Specifically, the first attention score matrix S AB It is an m x n matrix, where m is the number of samples in the first sample set and n is the number of samples in the second sample set.

[0076] Step S32: For the first attention score matrix S AB After normalization (using the softmax function), the first attention weight matrix α in the A→B direction is obtained. AB The first attention weight matrix α AB The element in the i-th row and j-th column satisfies the following relation:

[0077]

[0078] Among them, s ij S represents the first attention score matrix. AB The element in the i-th row and j-th column, the first attention weight matrix α AB It is also an m-row n-column matrix.

[0079] Step S33: Based on the first attention weight matrix α AB For the second principal component characteristic matrix F B Perform weighted fusion to obtain the first fusion feature matrix F AB :

[0080]

[0081] Steps S31 to S33 above are for attention calculation in the A→B direction, and steps S34 to S36 below are for attention calculation in the B→A direction.

[0082] Step S34: Using the second principal component feature matrix F B For the query, the first principal component feature matrix F A Given keys and values, obtain the second attention score matrix S. BA :

[0083]

[0084] Specifically, the second attention score matrix S BA It is an n-row m-column matrix.

[0085] Step S35: For the second attention score matrix S BA Perform softmax normalization to obtain the second attention weight matrix β BA The second attention weight matrix β BA The element in the j-th row and i-th column satisfies the following relation:

[0086]

[0087] Among them, s ji S represents the second attention score matrix. BA The element in the j-th row and i-th column, the second attention weight matrix α BA It is also an n-row m-column matrix.

[0088] Step S36: Based on the second attention weight matrix β BA For the first principal component characteristic matrix F A Perform weighted fusion to obtain the second fusion feature matrix F BA :

[0089]

[0090] Step S4: Based on the first fusion feature matrix F AB The second fusion feature matrix F BA The first matrix X A Second matrix X B Determine the difference parameters of each characteristic peak, and determine the chromatographic difference peaks based on each difference parameter.

[0091] Step S4 involves mapping the attention information to the original peak space, that is, mapping the information corresponding to the bidirectional attention weight back to the chromatographic peak space of the first curve, calculating the difference of each chromatographic peak, and realizing the quantification of peak differences.

[0092] Specifically, step S4 includes:

[0093] Step S41: Based on the first attention weight matrix α AB Second attention weight matrix β BA Obtain the first fusion weight vector corresponding to each sample in the first sample set and the second fusion weight vector corresponding to each sample in the second sample set.

[0094] Specifically, this step can be broken down into the following steps:

[0095] S411: Construct the first reference weight vector and the first correction weight vector corresponding to each sample in the first sample set, wherein the first reference weight vector corresponding to the i-th sample in the first sample set is based on the first attention weight matrix α. AB The i-th row of elements constructs α AB (i, :), the first sample set of the first sample set i The first correction weight vector corresponding to each sample is based on the second attention weight matrix β. BA The i-th column elements constitute β BA ( : , i).

[0096] Step S412: Based on the average value of the first reference weight vector and the first correction weight vector corresponding to each sample in the first sample set, determine the corresponding first fusion weight vector. Let the first fusion weight vector corresponding to the i-th sample in the first sample set be denoted as α. i .

[0097] Step S413: Construct the second reference weight vector and the second correction weight vector corresponding to each sample in the second sample set, wherein the second reference weight vector corresponding to the j-th sample in the second sample set is constructed based on the j-th row element of the second attention weight matrix, i.e., β. BA (j, : ), the second correction weight vector corresponding to the j-th sample in the second sample set is constructed based on the j-th column element of the first attention weight matrix, i.e., α AB ( : , j).

[0098] Step S414: Based on the average of the second reference weight vector and the second correction weight vector corresponding to each sample in the second sample set, determine the corresponding second fusion weight vector. Let the second fusion weight vector corresponding to the j-th sample in the second sample set be denoted as β. j .

[0099] Step S42: Calculate the weighted average of the second matrix based on the first fusion weight vector corresponding to each sample to obtain the second weighted feature vector corresponding to each sample in the second sample set, and obtain the first peak parameter vector based on the first matrix and each second weighted feature vector.

[0100] Specifically, for the i-th row of the first matrix, the difference vector corresponding to the sample is obtained based on the absolute difference between each element of the row and each element of the second weighted feature vector corresponding to the i-th sample in the second sample set; all rows of the first matrix are traversed to obtain the difference vectors corresponding to all samples in the first sample set; the difference vectors are summed to determine the first peak parameter vector.

[0101] Specifically, let the first fusion weight vector be denoted as α. i The matrix X of the sample to be tested (i.e., the second sample) B By performing weighting, we obtain the second weighted feature vector WX corresponding to the i-th sample in the second sample set. B,i Calculate the difference between the i-th sample in the first sample set and the weighted second sample, and obtain the first peak parameter vector of the i-th first sample by taking the absolute value. Iteratively calculate the parameter vector corresponding to each sample in the first sample set, i.e., calculate i from 1 to m, sequentially obtaining the peak difference parameters, and then sum them to obtain the first peak parameter vector p. AB :

[0102]

[0103]

[0104] Step S43: Perform weighted calculation on the first matrix based on the second fusion weight vector corresponding to each sample in the second sample set to obtain the first weighted feature vector corresponding to each sample in the first sample set, and obtain the second peak parameter vector based on the second matrix and each first weighted feature vector.

[0105] Specifically, for the j-th row in the second matrix, the difference vector corresponding to the sample is obtained based on the absolute difference between each element of the row and each element of the first weighted feature vector corresponding to the j-th sample in the first sample set; all rows of the second matrix are traversed to obtain the difference vectors corresponding to all samples in the second sample set; the difference vectors are summed to determine the second peak parameter vector.

[0106] Specifically, the second fusion weight vector is denoted as β. j Matrix X of the reference sample (i.e., the first sample) A By performing weighting, we obtain the first weighted feature vector WX corresponding to the j-th sample in the first sample set. A,j Calculate the difference between the j-th sample in the second sample set and the weighted first sample, and obtain the second peak parameter vector of the j-th second sample by taking the absolute value. Iteratively calculate the parameter vector corresponding to each sample in the second sample set, i.e., calculate j from 1 to n, sequentially obtaining the peak difference parameters, and then sum them to obtain the second peak parameter vector p. BA :

[0107]

[0108]

[0109] Step S44: Based on the first peak parameter vector and the second peak parameter vector, obtain the difference parameters of each characteristic peak.

[0110] The first peak parameter vector and the second peak parameter vector have the same dimension, both equal to the number of characteristic peaks. Step S44 specifically includes: summing the first peak parameter vector and the second peak parameter vector, and then dividing by the total number of samples in the first and second sample sets (i.e., the sum of the number of samples m in the first sample set and the number of samples n in the second sample set), to obtain:

[0111]

[0112] The obtained vector p is normalized, and the k-th element of the normalized vector is the difference parameter dif_k of the k-th characteristic peak.

[0113]

[0114] Where p(k) represents the k-th component element of vector p.

[0115] A higher value for the difference parameter of a characteristic peak indicates a more significant difference in the peak's characteristics across different sample classifications. In other words, this characteristic peak is more beneficial for quality control of fragrances and flavorings. Based on the actual sample testing needs, the characteristic peaks can be arranged from largest to smallest value, and a certain number of peaks at the top can be selected as the difference characteristic peaks required for evaluation.

[0116] This application employs a bidirectional interactive attention mechanism, simultaneously calculating the attention between different categories of samples. By aggregating bidirectional information, it achieves a comprehensive assessment of chromatographic peak differences, addressing the limitations and poor robustness of unidirectional attention assessment. Furthermore, this application establishes a complete workflow of "Principal Component Analysis (PCA) dimensionality reduction - bidirectional attention calculation - original peak space mapping." After simplifying high-dimensional GC-MS data, attention information is mapped back to the original chromatographic peak space through attention weight fusion and difference accumulation, achieving peak difference quantification. In addition, this application achieves symmetrical accumulation in both A-to-B and B-to-A directions, making the selection of difference peaks more comprehensive and ensuring the accuracy of the selection.

[0117] On the other hand, some embodiments of this application also provide a method for evaluating the quality of fragrances and flavorings, including:

[0118] Obtain the gas chromatography-mass spectra of the sample to be tested, and obtain the data of each characteristic peak based on the spectra;

[0119] The chromatographic difference peaks of the sample to be tested are determined based on the chromatographic difference peak identification method in the above embodiments.

[0120] For each characteristic peak data of the sample to be tested, the data related to each chromatographic difference peak are analyzed to evaluate the quality of the sample.

[0121] The following are examples of testing tobacco flavoring samples using the method described in this application.

[0122] Example: Calculation of chromatographic peak differences in fragrance and flavor samples

[0123] (1) Materials, reagents and instruments

[0124] Samples of the same type of tobacco flavoring and fragrance from different sources: 28 samples from source A (reference samples) and 27 samples from source B (samples to be tested). Methanol (chromatographic grade): Tianjin Kemei Chemical Reagent Co., Ltd.; phenethyl acetate (≥99%, Beijing Bailingwei Technology Co., Ltd.); n-Heptadecane (≥99%, Sigma-Aldrich, USA).

[0125] TriplusRSH-Trace1310-ISQLT gas chromatograph-mass spectrometer (Thermo Fisher Scientific, USA); analytical balance AG104 (sensitivity: 0.1 mg, Mettler Toledo, Switzerland); 100 μL pipette (Eppendorf, USA).

[0126] (2) Preparation of standard solutions

[0127] Weigh 0.05 g of phenethyl acetate and 0.015 g of n-heptadecane internal standard sample into a 50 mL volumetric flask, then add methanol to make up to the volume to prepare a mixed internal standard solution, wherein the concentrations of phenethyl acetate and n-heptadecane are 1 mg / mL and 3 mg / mL, respectively.

[0128] (3) Instrument operating conditions

[0129] Static headspace analyzer: Sample equilibrium temperature: 140 ℃; Sample ring temperature: 170 ℃; Transfer line temperature: 180 ℃; Sample equilibrium time: 20 min; Sample vial pressurization pressure: 130 kPa; Pressurization time: 0.20 min; Gas filling time: 0.5 min; Sample ring equilibrium time: 0.2 min; Sample injection time: 0.5 min.

[0130] Gas chromatography conditions: Column: DB-624 capillary column (60 m * 0.32 mm * 1.8 μm); Injector temperature: 180 ℃; Carrier gas: High-purity helium (99.999%); Flow rate: 2.0 mL / min (constant flow mode); Split ratio: 20:1; Temperature program: 80 ℃, hold for 1 min, increase to 210 ℃ at a rate of 4 ℃ / min, then increase to 240 ℃ at a rate of 20 ℃ / min and hold for 10 min.

[0131] Mass spectrometry conditions: transfer line temperature: 240 °C; ionization method: electron impact source (EI); ionization energy: 70 eV; ion source temperature: 280 °C; solvent delay time: 6 min; scan mode: full scan mode (SCAN) + selected ion mode (SIM), SCAN mass range: 33 amu ~ 450 amu.

[0132] (4) Sample pretreatment

[0133] Weigh 0.05 g of sample, accurate to 0.0001 g, place the sample in a 20 mL headspace vial, add 100 μL of internal standard solution, immediately cap and seal, and analyze according to the above conditions.

[0134] (4) Results Analysis

[0135] First, the GC-MS data of the fragrance and flavor samples were analyzed sequentially. Figure 1 Qualitative and quantitative analyses (characteristic peak qualitative and peak area quantitative) were performed on the sample (as shown), and characteristic peak curves containing retention time and relative peak area (compound peak area divided by internal standard area) were obtained for each sample. Figure 2 (As shown). Analysis yielded material information and peak areas for 48 chromatographic peaks.

[0136] Following the steps described above, after Pareto normalization of the chromatographic peak curves, principal component analysis was performed on the sample from source A, and then the difference parameters for each chromatographic peak were calculated. Figure 3 As shown in the figure, the chromatographic peaks that cause significant differences between reference sample A and test sample B are: peak 40 (anesinol), peak 45 (eugenol), peak 11 (myrcene), peak 17 (γ-terpinene), peak 23 (linalool), peak 10 (juniperene), peak 16 ((+)-limonene), peak 8 (α-pinene), peak 41 (p-methoxybenzaldehyde), and peak 39 (geraniol). Therefore, the content of these substances can be used to control the quality of fragrances and flavorings.

[0137] This application method is based on a two-way interactive attention mechanism, which integrates the two-way attention information of samples A and B. The weight calculation is accurately mapped to the original peak space, directly corresponding to each chromatographic peak detected by GC-MS. It has stronger interpretability and clarifies "which peaks are the core of distinguishing samples". It provides direct quantitative basis for the identification of authenticity of fragrances and flavors, batch stability control and component traceability, and has more prominent practicality.

[0138] The specific embodiments described above illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Although the description of this application is presented in conjunction with some embodiments, this does not mean that the features of this application are limited to this embodiment. On the contrary, the purpose of describing the application in conjunction with embodiments is to cover other options or modifications that may be derived based on the claims of this application. This application may also be implemented without using these details. Furthermore, to avoid confusion or obscuring the focus of this application, some specific details have been omitted in the description. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other.

[0139] In the embodiments of this application, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first," "second," or "third" may explicitly or implicitly include one or more of that feature.

[0140] In the embodiments of this application, "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0141] In the description of this application, it should be noted that the terms "upper", "lower", "top", "bottom", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0142] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "set," "install," "connect," and "fit" should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0143] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for identifying chromatographic difference peaks for evaluating the quality of fragrances and flavorings, characterized in that, include: A first matrix is ​​determined based on the characteristic peak data of each sample in the first sample set, and a second matrix is ​​determined based on the characteristic peak data of each sample in the second sample set, wherein the element in the i-th row and k-th column of the first matrix is ​​related to the peak area of ​​the k-th characteristic peak of the i-th sample in the first sample set, and the element in the j-th row and k-th column of the second matrix is ​​related to the peak area of ​​the k-th characteristic peak of the j-th sample in the second sample set. Principal component analysis is performed based on the first matrix to obtain the first principal component feature matrix and the second principal component feature matrix. Bidirectional interactive attention calculation is performed based on the first principal component feature matrix and the second principal component feature matrix to obtain the first fusion feature matrix and the second fusion feature matrix respectively; The difference parameters of each characteristic peak are determined based on the first fusion feature matrix, the second fusion feature matrix, the first matrix, and the second matrix, and the chromatographic difference peaks are determined based on each difference parameter.

2. The method for identifying chromatographic difference peaks for evaluating the quality of fragrances and flavors according to claim 1, characterized in that, Based on the first principal component feature matrix and the second principal component feature matrix, bidirectional interactive attention calculation is performed to obtain the first fusion feature matrix and the second fusion feature matrix, respectively, including: Using the first principal component feature matrix as the query and the second principal component feature matrix as the key and value, obtain the first attention score matrix; The first attention score matrix is ​​normalized using softmax to obtain the first attention weight matrix; The second principal component feature matrix is ​​weighted and fused based on the first attention weight matrix to obtain the first fused feature matrix; Using the second principal component feature matrix as the query and the first principal component feature matrix as the key and value, obtain the second attention score matrix; The second attention score matrix is ​​normalized using softmax to obtain the second attention weight matrix; The first principal component feature matrix is ​​weighted and fused based on the second attention weight matrix to obtain the second fused feature matrix.

3. The method for identifying chromatographic difference peaks for evaluating the quality of fragrances and flavorings according to claim 2, characterized in that, Based on the first fused feature matrix, the second fused feature matrix, the first matrix, and the second matrix, the difference parameters of each feature peak are determined, including: Based on the first attention weight matrix and the second attention weight matrix, obtain the first fusion weight vector corresponding to each sample in the first sample set and the second fusion weight vector corresponding to each sample in the second sample set; The second matrix is ​​weighted based on the first fusion weight vector corresponding to each sample to obtain the second weighted feature vector corresponding to each sample in the second sample set, and the first peak parameter vector is obtained based on the first matrix and each second weighted feature vector. The first matrix is ​​weighted based on the second fusion weight vector corresponding to each sample to obtain the first weighted feature vector corresponding to each sample in the first sample set, and the second peak parameter vector is obtained based on the second matrix and each of the first weighted feature vectors. Based on the first peak parameter vector and the second peak parameter vector, the difference parameters of each characteristic peak are obtained; wherein the first peak parameter vector and the second peak parameter vector have the same dimension, which is equal to the number of characteristic peaks.

4. The method for identifying chromatographic difference peaks for evaluating the quality of fragrances and flavors according to claim 3, characterized in that, Based on the first attention weight matrix and the second attention weight matrix, obtain the first fusion weight vector corresponding to each sample in the first sample set and the second fusion weight vector corresponding to each sample in the second sample set, including: Construct a first reference weight vector and a first correction weight vector corresponding to each sample in the first sample set, wherein the first reference weight vector corresponding to the i-th sample in the first sample set is constructed based on the i-th row element of the first attention weight matrix, and the first correction weight vector corresponding to the i-th sample in the first sample set is constructed based on the i-th column element of the second attention weight matrix; Based on the average value of the first reference weight vector and the first correction weight vector corresponding to each sample in the first sample set, the first fusion weight vector corresponding to it is determined. Construct a second reference weight vector and a second correction weight vector corresponding to each sample in the second sample set, wherein the second reference weight vector corresponding to the j-th sample in the second sample set is constructed based on the j-th row element of the second attention weight matrix, and the second correction weight vector corresponding to the j-th sample in the second sample set is constructed based on the j-th column element of the first attention weight matrix; The second fusion weight vector is determined based on the average value of the second reference weight vector and the second correction weight vector corresponding to each sample in the second sample set.

5. The method for identifying chromatographic difference peaks for evaluating the quality of fragrances and flavors according to claim 3, characterized in that, Obtaining the first peak parameter vector based on the first matrix and each of the second weighted feature vectors includes: for the i-th row of the first matrix, obtaining the difference vector corresponding to the sample based on the absolute difference between each element of the row and each element of the second weighted feature vector corresponding to the i-th sample in the second sample set; traversing all rows of the first matrix to obtain each of the difference vectors corresponding to all samples in the first sample set; summing the difference vectors to determine the first peak parameter vector. Obtaining the second peak parameter vector based on the second matrix and each of the first weighted feature vectors includes: for the j-th row of the second matrix, obtaining the difference vector corresponding to the sample based on the absolute difference between each element of the row and each element of the first weighted feature vector corresponding to the j-th sample in the first sample set; traversing all rows of the second matrix to obtain each of the difference vectors corresponding to all samples in the second sample set; summing each of the difference vectors to determine the second peak parameter vector.

6. The method for identifying chromatographic difference peaks for evaluating the quality of fragrances and flavors according to claim 3, characterized in that, Based on the first peak parameter vector and the second peak parameter vector, the difference parameters of each characteristic peak are obtained, including: The first peak parameter vector and the second peak parameter vector are summed and divided by the total number of samples in the first and second sample sets. Then, the resulting vector is normalized. The k-th element of the normalized vector is the difference parameter of the k-th characteristic peak.

7. The method for identifying chromatographic difference peaks for evaluating the quality of fragrances and flavors according to claim 1, characterized in that, Principal component analysis is performed based on the first matrix to obtain the first principal component feature matrix and the second principal component feature matrix, including: Perform principal component analysis on the first matrix to determine the score matrix, principal component coefficient matrix, and number of principal components D; The first principal component feature matrix is ​​determined based on the first D columns of the score matrix; The second principal component feature matrix is ​​determined based on the first D columns of the principal component coefficient matrix and the second matrix.

8. The method for identifying chromatographic difference peaks for evaluating the quality of fragrances and flavors according to claim 7, characterized in that, The first principal component feature matrix is ​​the matrix formed by the first D columns of the score matrix, and the second principal component feature matrix is ​​the product matrix of the second matrix and the matrix formed by the first D columns of the principal component coefficient matrix.

9. The method for identifying chromatographic difference peaks for evaluating the quality of fragrances and flavors according to claim 1, characterized in that, Determining a first matrix based on the characteristic peak data of each sample in a first sample set, and determining a second matrix based on the characteristic peak data of each sample in a second sample set, including: Obtain the gas chromatography-mass spectra of the fragrance sample, and determine the characteristic peaks that reflect the volatile components based on the spectra; Obtain the characteristic peak data of each sample in the first sample set, construct a first intermediate matrix, normalize the first intermediate matrix, and obtain the first matrix; Obtain the characteristic peak data of each sample in the second sample set, construct a second intermediate matrix, normalize the second intermediate matrix, and obtain the second matrix.

10. A method for evaluating the quality of flavorings and fragrances, characterized in that, include: Obtain the gas chromatography-mass spectrometry (GC-MS) spectrum of the sample to be tested, and obtain the data of each characteristic peak based on the spectrum; Determining chromatographic difference peaks based on the method according to any one of claims 1-9; For each characteristic peak data of the sample to be tested, the data related to each chromatographic difference peak are analyzed to evaluate the quality of the sample to be tested.

Images