Method for evaluating quality grade of base liquor of soft-taste liquor

Abstract

The invention discloses a method for evaluating the quality grade of soft-taste white spirit base liquor, and particularly relates to the technical field of white spirit base liquor quality detection. Volatile and non-volatile components in baijiu base liquor are accurately separated through a multi-dimensional chromatography technology, flavor substances are extracted, target flavor components are quantitatively analyzed in combination with a multivariable correction model, the ratio of the flavor components is calculated, and the flavor compounds of the baijiu base liquor are detected through a high-sensitivity sensor. Analyzing the detection data through an artificial intelligence model, training flavor feature distribution of different quality grades, calculating similarity scores of new samples and standard samples, and automatically determining the quality grade of the base liquor in combination with clustering analysis; according to the method, the problem that a traditional detection means is greatly influenced by the matrix effect is solved, precision, intelligence and standardization of quality evaluation of the baijiu base liquor are achieved, and a scientific basis is provided for quality control, product grading and stability monitoring in the baijiu production process.

Description

Technical Field

[0001] The invention relates to the technical field of liquor base wine quality detection, and in particular to a method for evaluating the quality grade of mellow liquor base wine. Background Art

[0002] The quality grade assessment of the base liquor of mellow and soft liquor refers to a comprehensive analysis of the sensory characteristics, chemical composition, process stability and storage potential of the base liquor, and its classification according to the standard system. The purpose of the assessment is to ensure that the base liquor meets the style requirements of the target product and provide a basis for subsequent blending and quality control.

[0003] The prior art has the following deficiencies: Liquor contains complex organic components, including alcohols, esters, phenols, aldehydes, etc. The matrix components may interfere with the detection of target substances, affect the chromatographic separation effect, and even cause quantitative errors. In addition, if the matrix effect is not fully considered, it may lead to deviations in key indicators such as alcohol content and acid ester ratio, affecting quality assessment. Summary of the invention

[0004] The purpose of the present invention is to provide a method for evaluating the quality grade of a mellow-type liquor base wine, so as to solve the shortcomings in the background technology.

[0005] In order to achieve the above object, the present invention provides the following technical solution: a method for evaluating the quality grade of a mellow-flavored liquor base wine, comprising: Multidimensional chromatography analysis technology is used to separate the volatile and non-volatile components in the liquor base and extract the target flavor substances; Combined with the multivariate calibration model, the target flavor substances were quantitatively analyzed, the flavor component ratios were calculated, and the matrix differences between different samples were adjusted; A high-sensitivity sensor is used to detect the flavor compounds in the adjusted liquor base wine, and an artificial intelligence model is used to analyze the detection data. The flavor characteristic distribution of liquor base wines with known different quality grades is trained, and then the similarity score between the new sample and the standard sample is calculated. The quality grade of the base wine is determined by combining cluster analysis.

[0006] Preferably, the multidimensional chromatography analysis technology includes two-dimensional gas chromatography-mass spectrometry and two-dimensional liquid chromatography, which are used to separate volatile and non-volatile flavor components in the liquor base to improve the resolution and qualitative accuracy of the target flavor substances.

[0007] Preferably, the multivariate correction model uses partial least squares regression to construct a mathematical model between the concentration of target flavor substances and matrix parameters, and the model calculation expression is: ;in: represents the concentration of flavor substances after correction for matrix effects. represents the uncorrected concentration of the target flavor substance, represents the key variables affecting the matrix effect, is the regression coefficient, which is obtained by fitting the training data and represents the correction effect of each matrix parameter on the concentration of the target flavor substance.

[0008] Preferably, the flavor component ratio includes an acid ester ratio and an ester alcohol ratio, wherein the acid ester ratio is used to characterize the balance between the sour aroma and mellowness of the liquor base wine, and the ester alcohol ratio is used to measure the richness of the wine aroma, and the flavor characteristics of the base wine are evaluated based on the ratio calculation results.

[0009] Preferably, the acid ester ratio It reflects the balance between the sour aroma and mellowness of liquor and is defined as: ;in: represents the sum of the corrected concentrations of various organic acids, It represents the sum of the corrected concentrations of each ester; Ester alcohol ratio It measures the aromatic intensity of liquor and is defined as: ;in: is the sum of the corrected concentrations of ester compounds, It is the sum of the concentrations after correction for higher alcohols.

[0010] Preferably, the high-sensitivity sensor includes a surface enhanced Raman spectroscopy sensor, a gas chromatography-ion mobility spectrometry, a nano gas sensor or an electronic nose.

[0011] Preferably, the flavor compound data detected by the high-sensitivity sensor is expressed as: ; Where: X represents the flavor feature vector of a base wine sample, including the concentrations or sensor response values ​​of n flavor compounds, is the measured value of the nth flavor compound; For multiple base wine samples, the data collected by the sensor forms a matrix: ; Where: m is the number of samples, represents the concentration of the nth flavor compound of the mth sample; Training classification model f(X): Q=f(X); where: Q is the base wine quality grade, select base wine sample data Xtrain of known quality grade as the training set, calculate the feature vector Z, and train the classification model f(X) to learn the flavor distribution of different quality grades; Use the K-Means clustering algorithm to automatically divide the base wine samples into K levels: ; Where: J is the objective function, which represents the total distance from all samples to the cluster center, For sample Whether it belongs to category k, K is the number of clustering categories, Represents the cluster center of the k-th level base wine sample.

[0012] Preferably, the similarity score between the new sample and the standard sample is calculated, and the new sample The similarity calculation method between the sample sets of each quality level is: Calculate new samples With each quality grade center The similarity between , the expression is: ;in: The center point of the k-th base wine quality grade, that is, the mean vector, is represented. The similarity between the obtained new sample and the center of each quality grade is compared with the similarity threshold. If the similarity between the obtained new sample and the center of each quality grade is greater than or equal to the similarity threshold, it means that the new sample has a flavor characteristic similar to the quality grade and is classified into this quality grade. If the similarity between the obtained new sample and the center of each quality grade is less than the similarity threshold, it means that the new sample does not belong to any known quality grade, and K-Means clustering or DBSCAN is used for unsupervised analysis.

[0013] In the above technical solution, the technical effects and advantages provided by the present invention are: 1. The present invention constructs an accurate and intelligent quality grading system. The volatile and non-volatile components in the liquor base wine are separated at high resolution by multidimensional chromatography technology, and the matrix effect is corrected by partial least squares regression to calculate the ratios of flavor components such as acid ester ratio and ester alcohol ratio to improve the comparability between samples. Subsequently, high-sensitivity sensors such as SERS, GC-IMS, nano gas sensors, and electronic noses are used to detect the flavor compounds in the liquor base wine to obtain refined flavor data. Finally, the base wine samples are classified and trained in combination with machine learning models such as support vector machines, random forests, and deep neural networks, and the similarity scores between new samples and standard samples are calculated, and K-Means clustering or DBSCAN algorithms are used to complete intelligent grading and anomaly detection.

[0014] 2. The present invention significantly improves the accuracy, consistency and automation of the quality assessment of liquor base wine by integrating standardized data analysis with intelligent algorithms. Compared with the traditional manual sensory evaluation method, this method can objectively and efficiently screen the quality grade of base wine, reduce human errors, and improve the stability of the production process and product consistency. In addition, the method can also automatically identify new quality grades or abnormal samples, which helps liquor companies optimize product flavor management and improve quality control capabilities, and provides a scientific basis for the liquor quality standardization and intelligent evaluation system, promoting the modernization and upgrading of the liquor industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0016] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] For examples, see Figure 1 As shown, the method for evaluating the quality grade of the mellow and soft liquor base wine described in this embodiment includes: Multidimensional chromatography analysis technology is used to separate the volatile and non-volatile components in the liquor base and extract the target flavor substances; Combined with the multivariate calibration model, the target flavor substances were quantitatively analyzed, the flavor component ratios were calculated, and the matrix differences between different samples were adjusted; A high-sensitivity sensor is used to detect the flavor compounds in the adjusted liquor base wine, and an artificial intelligence model is used to analyze the detection data. The flavor characteristic distribution of liquor base wines with known different quality grades is trained, and then the similarity score between the new sample and the standard sample is calculated. The quality grade of the base wine is determined by combining cluster analysis.

[0019] Take samples of the mellow liquor base to be tested, ensure that there is no pollution during the sampling process, and store them in sealed brown glass bottles away from light to prevent oxidation and flavor loss. For different batches of base liquor, use a uniform sampling method to ensure the representativeness of the data.

[0020] If the sample alcohol content is too high (e.g. > 60% vol), distilled water or ethanol water solution (the ratio is the same as that of liquor) can be used for dilution to avoid high concentration affecting the chromatographic separation effect. The sample is filtered using a 0.22 μm microporous membrane to remove particulate matter and prevent instrument contamination.

[0021] Enrichment of volatile components: Solid phase microextraction (SPME) or purge and trap (P&T) technology is used to extract the sample at a constant temperature for 10-30 minutes to enrich trace volatile compounds.

[0022] Non-volatile component extraction: Use liquid-liquid extraction (LLE) or solid phase extraction (SPE) methods to separate non-volatile components in the liquor base, such as phenols, terpenes and other key flavor substances.

[0023] Analysis of volatile flavor components using gas chromatography-mass spectrometry (GC-MS): Chromatographic column selection: Use a nonpolar column (DB-5MS, 30m×0.25mm×0.25μm) or a polar column (DB-WAX, 30m×0.25mm×0.25μm) to optimize the separation of target compounds.

[0024] Chromatographic conditions: The injection port temperature was set at 250°C, and the splitless mode was used to improve the detection sensitivity; carrier gas: 99.999% pure helium, flow rate 1.0 mL / min; temperature program: initial temperature 40°C, maintained for 3 min; temperature increased to 200°C at 5°C / min, maintained for 5 min; temperature increased to 280°C at 10°C / min, maintained for 10 min. Mass spectrometry detection: electron impact ionization source (EI, 70 eV) was used, with a scanning range of m / z 30-500; compound identification was performed by retention index (RI) and standard spectral libraries (such as NIST, Wiley).

[0025] Liquid chromatography-mass spectrometry (LC-MS) was used to analyze non-volatile flavor components: Column selection: A C18 reverse phase column (2.1×100mm, 1.7μm) was used to separate polar and semi-polar flavor components. Mobile phase: Phase A: 0.1% formic acid in water; Phase B: acetonitrile or methanol (containing 0.1% formic acid). Gradient elution: 0-2 min: 5% B, isocratic elution; 2-10 min: 5%-95% B, increasing gradient; 10-15 min: 95% B, isocratic elution; 15-17 min: 95%-5% B, decreasing gradient. Mass spectrometry detection: Ion source: ESI (electrospray ionization) mode; Positive and negative ion scanning mode: m / z 50-1000; Combined with standard control and secondary mass spectrometry (MS / MS) fragment analysis for qualitative and quantitative analysis.

[0026] Extraction of target flavor substances: By comparing with standard substances (such as ethyl acetate, ethyl lactate, ethyl caproate, eugenol, etc.), select the compounds that contribute most to the flavor of the base liquor; calculate the relative content (normalized peak area) and flavor activity value (OAV = C / OT) to screen the key flavor compounds; use statistical analysis (such as principal component analysis (PCA), linear discriminant analysis (LDA)) to screen the flavor components that contribute most to the quality grade of the base liquor.

[0027] Record the flavor composition, relative content, characteristic peaks and retention time of different batches of base wine; establish a base wine quality assessment model by combining machine learning algorithms (such as support vector machine SVM, random forest RF); set the flavor fingerprint characteristics of base wines of different quality levels (such as high-end, mid-end, and low-end) for subsequent quality assessment.

[0028] In the quality assessment of liquor base wine, direct quantitative analysis of target flavor substances may lead to deviations due to differences in matrix components (such as water content, alcohol concentration, non-volatile substances, etc.) of different batches of samples. Therefore, a multivariate correction model is introduced to adjust the content of target flavor substances so that the test results of different batches are comparable.

[0029] Actual concentration of target flavor substances Affected by the response signal S of the detection system and the matrix effect M, it can be expressed as: ;in: is the true concentration of the target flavor substance (mg / L), S is the chromatographic peak area or mass spectrometry ion intensity of the target compound, k is the instrument correction factor, which is determined by the standard curve (i.e., the concentration conversion value corresponding to the unit signal, mg / L / unit signal); M is the matrix effect, including the effects of alcohol content, solubility, and other coexisting components on the target compound. Since the matrix effect M varies from batch to batch, a multivariate calibration model is required to compensate.

[0030] Partial least squares regression was used to construct a mathematical model between the concentration of flavor components and matrix parameters: ;in: represents the concentration of flavor substances after correction for matrix effects (mg / L); represents the uncorrected concentration of target flavor substances (mg / L); Indicates key variables that influence matrix effects, such as alcohol content , total acid Total esters , Higher alcohol content wait; is the regression coefficient, which is obtained by fitting the training data and represents the correction effect of each matrix parameter on the concentration of the target flavor substance. The PLSR model minimizes the prediction error and makes the corrected concentration Closer to the true value, improving the comparability of samples from different batches.

[0031] The flavor component ratio can reflect the flavor balance of liquor, specifically including the acid ester ratio AER and the ester alcohol ratio EAR.

[0032] The acid-acid ratio reflects the balance between the sour aroma and mellowness of liquor and is defined as: ;in: It represents the sum of the corrected concentrations of various organic acids (such as acetic acid, lactic acid, caproic acid, etc.) (mg / L); It represents the sum of the corrected concentrations of various esters (such as ethyl acetate, ethyl lactate, ethyl caproate, etc.) (mg / L).

[0033] When the AER is high, the liquor may be too sour and astringent; when the AER is low, the liquor may not be full-bodied enough. The appropriate acid-ester ratio contributes to the harmonious taste of mellow liquor.

[0034] The ester-to-alcohol ratio measures the aromatic intensity of the liquor and is defined as: ;in: is the sum of the corrected concentrations of ester compounds (mg / L), It is the sum of the corrected concentrations of higher alcohols (such as isopentanol, n-butanol, n-pentanol, etc.) (mg / L). A higher EAR indicates that the wine has a more prominent aroma, while a lower EAR may result in a lighter flavor or a rough taste.

[0035] According to the characteristics of flavor compounds (volatility, polarity, functional groups, etc.), choose the appropriate high-sensitivity sensor technology, including: Surface-enhanced Raman spectroscopy is suitable for low-concentration alcohols, phenols, terpenes and other key flavor components of liquor. Raman signals are enhanced by precious metal substrates such as nano-gold / silver to improve detection sensitivity, and LOD can reach ppb (10⁻ 9 ).

[0036] Gas chromatography-ion mobility spectrometry is suitable for high-sensitivity detection of volatile flavor substances such as esters and aldehydes; combined with GC pre-separation and IMS rapid detection, it improves the qualitative ability in complex matrix environments.

[0037] Nano gas sensors are suitable for detecting aldehydes (furfural, nonanal, etc.) and alcohols (n-butanol, isoamyl alcohol, etc.), with LOD reaching ppt (10⁻¹²). Surface modification technology (such as functionalized graphene and metal oxide nanoparticles) is used to enhance the selective response to the target.

[0038] The electronic nose, composed of multiple gas sensors, can detect a variety of volatile flavor substances at the same time and perform pattern recognition. It is suitable for large-scale quality grading scenarios and can be combined with principal component analysis (PCA) to achieve automatic classification.

[0039] The flavor compound data detected by high-sensitivity sensors (such as SERS, GC-IMS, nano gas sensors) are expressed as: ; Where: X represents the flavor feature vector of a base wine sample, including the concentrations or sensor response values ​​of n flavor compounds, is the measured value of the nth flavor compound (mg / L or sensor signal intensity).

[0040] For multiple base wine samples, the data collected by the sensor forms a matrix: ; Where: m is the number of samples, represents the concentration of the nth flavor compound of the mth sample.

[0041] Use support vector machine (SVM), random forest (RF) or deep neural network (DNN) to train the classification model f(X): Q=f(X); where: Q is the quality grade of the base wine (such as 1 = low grade, 2 = medium grade, 3 = high grade).

[0042] The base wine sample data Xtrain with known quality grades is selected as the training set, the feature vector Z is calculated, and the classification model f(X) is trained to learn the flavor distribution of different quality grades.

[0043] Use the K-Means clustering algorithm to automatically divide the base wine samples into K levels: ; Where: J is the objective function, which represents the total distance from all samples to the cluster center, For sample Whether it belongs to category k (if yes, then 1, otherwise 0), K is the number of cluster categories (such as 3 represents low, medium, and high levels); Represents the cluster center of the k-th level base wine sample, and uses K-Means for unsupervised clustering to obtain the category label.

[0044] Calculate the similarity score between the new sample and the standard sample. The similarity calculation method between the new sample Xnew and the sample set Xclassk of each quality level is: Calculate new samples With each quality grade center The similarity between , the expression is: ;in: represents the center point of the k-th base wine quality grade, that is, the mean vector, The value range is [−1, 1]. The closer the value is to 1, the higher the flavor similarity.

[0045] The similarity between the acquired new sample and the center of each quality grade is compared with the similarity threshold. If the similarity between the acquired new sample and the center of each quality grade is greater than or equal to the similarity threshold, it means that the new sample has a flavor characteristic that is sufficiently similar to the quality grade, and it is classified into the quality grade. The new sample has a high consistency with the standard sample set of the quality grade, indicating that the flavor of the base wine meets the typical characteristics of a certain quality grade. If the similarity between the new sample and the center of each quality grade is less than the similarity threshold, it means that the new sample does not belong to any known quality grade, that is, the flavor characteristics of the sample are significantly different from all known quality grades; it may be a new quality grade, and the classification system needs to be redefined or expanded; it may be an abnormal sample (such as abnormal production process, lack of flavor, etc.), which requires further testing. If the sample data volume is large, unsupervised clustering methods such as K-Means or DBSCAN can be used to determine whether there is a new quality grade; if it is an isolated abnormal sample, further analysis can be performed to determine whether it is a detection error or data anomaly (such as sensor drift, base wine contamination, etc.).

[0046] In the present invention, firstly, multidimensional chromatography technology (GC×GC-MS, 2D-LC, etc.) is used to accurately separate the volatile and non-volatile components in the base liquor of liquor, and extract key flavor substances; then, the target flavor substances are quantitatively analyzed in combination with a multivariate correction model, and the flavor component ratio (such as acid ester ratio, ester alcohol ratio) is calculated to eliminate the matrix effect between different samples and improve data consistency. Then, a high-sensitivity sensor (SERS, GC-IMS, nano gas sensor, etc.) is used to detect the adjusted flavor compounds and collect more detailed flavor fingerprint information. Finally, the detection data is trained and pattern recognized by an artificial intelligence model (SVM, DNN, K-Means clustering), the similarity score between the new sample and the standard sample is calculated, and the quality grade of the base liquor is automatically determined in combination with cluster analysis. This method breaks through the subjective limitations of traditional manual sensory evaluation, realizes the precision, intelligence and standardization of the quality evaluation of the base liquor of liquor, and can be widely used in quality control and graded management in the production process of liquor.

[0047] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0048] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0049] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there may be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the objects associated with each other are in an "or" relationship, but it may also indicate an "and / or" relationship, which can be understood by referring to the context. A person of ordinary skill in the art can appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this article can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0050] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for evaluating the quality grade of a mellow-flavored liquor base wine, characterized in that: include: Multidimensional chromatography analysis technology is used to separate the volatile and non-volatile components in the liquor base and extract the target flavor substances; Combined with the multivariate calibration model, the target flavor substances were quantitatively analyzed, the flavor component ratios were calculated, and the matrix differences between different samples were adjusted; A high-sensitivity sensor is used to detect the flavor compounds in the adjusted liquor base wine, and an artificial intelligence model is used to analyze the detection data. The flavor characteristic distribution of liquor base wines with known different quality grades is trained, and then the similarity score between the new sample and the standard sample is calculated. The quality grade of the base wine is determined by combining cluster analysis.

2. The method for evaluating the quality grade of a mellow and soft liquor base wine according to claim 1, characterized in that: The multidimensional chromatography analysis technology includes two-dimensional gas chromatography-mass spectrometry and two-dimensional liquid chromatography, which is used to separate volatile and non-volatile flavor components in the liquor base to improve the resolution and qualitative accuracy of the target flavor substances.

3. The method for evaluating the quality grade of a mellow and soft liquor base wine according to claim 2, characterized in that: The multivariate correction model uses partial least squares regression to construct a mathematical model between the concentration of target flavor substances and matrix parameters. The model calculation expression is: ;in: represents the concentration of flavor substances after correction for matrix effects. represents the uncorrected concentration of the target flavor substance, represents the key variables affecting the matrix effect, is the regression coefficient, which is obtained by fitting the training data and represents the correction effect of each matrix parameter on the concentration of the target flavor substance.

4. The method for evaluating the quality grade of a mellow and soft liquor base wine according to claim 3, characterized in that: The flavor component ratio includes the acid ester ratio and the ester alcohol ratio, wherein the acid ester ratio is used to characterize the balance between the sour aroma and the mellowness of the liquor base wine, and the ester alcohol ratio is used to measure the richness of the wine aroma, and the flavor characteristics of the base wine are evaluated based on the ratio calculation results.

5. The method for evaluating the quality grade of a mellow and soft liquor base wine according to claim 4, characterized in that: Acid ester ratio It reflects the balance between the sour aroma and mellowness of liquor and is defined as: ;in: represents the sum of the corrected concentrations of various organic acids, It represents the sum of the corrected concentrations of each ester; Ester alcohol ratio It measures the aromatic intensity of liquor and is defined as: ;in: is the sum of the corrected concentrations of ester compounds, It is the sum of the concentrations after correction for higher alcohols.

6. The method for evaluating the quality grade of a mellow and soft liquor base wine according to claim 5, characterized in that: The high-sensitivity sensor includes a surface enhanced Raman spectroscopy sensor, a gas chromatography-ion mobility spectrometry, a nano gas sensitive sensor or an electronic nose.

7. The method for evaluating the quality grade of a mellow and soft liquor base wine according to claim 6, characterized in that: The flavor compound data detected by the high-sensitivity sensor is expressed as: ; Where: X represents the flavor feature vector of a base wine sample, including the concentrations or sensor response values ​​of n flavor compounds, is the measured value of the nth flavor compound; For multiple base wine samples, the data collected by the sensor forms a matrix: ; Where: m is the number of samples, represents the concentration of the nth flavor compound of the mth sample; Training classification model f(X): Q=f(X); where: Q is the base wine quality grade, select base wine sample data Xtrain of known quality grade as the training set, calculate the feature vector Z, and train the classification model f(X) to learn the flavor distribution of different quality grades; Use the K-Means clustering algorithm to automatically divide the base wine samples into K levels: ; Where: J is the objective function, which represents the total distance from all samples to the cluster center, For sample Whether it belongs to category k, K is the number of clustering categories, Represents the cluster center of the k-th level base wine sample.

8. The method for evaluating the quality grade of a mellow and soft liquor base wine according to claim 7, characterized in that: Calculate the similarity score between the new sample and the standard sample. The similarity calculation method between the sample sets of each quality level is: Calculate new samples With each quality grade center The similarity between , the expression is: ;in: The center point of the k-th base wine quality grade, that is, the mean vector, is represented. The similarity between the obtained new sample and the center of each quality grade is compared with the similarity threshold. If the similarity between the obtained new sample and the center of each quality grade is greater than or equal to the similarity threshold, it means that the new sample has a flavor characteristic similar to the quality grade and is classified into this quality grade. If the similarity between the obtained new sample and the center of each quality grade is less than the similarity threshold, it means that the new sample does not belong to any known quality grade, and K-Means clustering or DBSCAN is used for unsupervised analysis.

Images