Method for constructing Baijiu sensory characteristic and quality expression prediction model

Abstract

The invention discloses a method for constructing a Baijiu sensory characteristic and quality expression prediction model, and belongs to the technical field of Baijiu analysis. The invention discloses a method for constructing a Baijiu sensory characteristic and quality expression prediction model. The method comprises the following steps: obtaining Baijiu sensory characteristic descriptor quantitative data; meanwhile, obtaining content data of 217 flavor substances in the white spirit by utilizing a gas chromatography method and a gas chromatography-mass spectrometry method; on the basis, important related flavor substances corresponding to different sensory characteristics are obtained by analyzing wine sample sensory characteristic descriptor quantitative data and flavor substance content data; furthermore, an important flavor substance corresponding to each sensory feature is taken as an independent variable, sensory feature quantitative data is taken as a dependent variable, and a wine body sensory quantification and quality evaluation prediction model is constructed by applying a regression analysis algorithm, so that evaluation personnel are helped to quickly and visually identify the sensory features of the white wine; the system solves the problem that it is difficult to give consideration to quality evaluation scientificity and efficiency in white spirit enterprise quality control.

Description

Technical Field

[0001] The invention relates to a method for constructing a liquor sensory characteristic and quality expression prediction model, and belongs to the technical field of liquor flavor analysis. Background Art

[0002] As one of the oldest distilled spirits, Baijiu has a variety of complex and unique processes, such as natural inoculation of various microorganisms and long-term solid-state fermentation, which give Baijiu complex sensory and flavor characteristics. The complex sensory characteristics of Baijiu also make it extremely difficult to accurately recognize and precisely control them, which has a serious impact on the stability and even quality improvement of Baijiu.

[0003] The receptor of liquor is human, so the cognitive subject of liquor is also human, but human cognitive behavior is affected by various external environments and internal emotions, which causes fluctuations in the evaluation of liquor quality. Therefore, the introduction of the sensory quantitative description system based on multivariate statistical analysis technology has greatly reduced the uncertainty of individual tasters in the evaluation results and improved the accuracy of liquor evaluation. But at the same time, for liquor production enterprises, although the scientific sensory evaluation system has greatly improved the accuracy of evaluation, the evaluation efficiency is relatively low. Therefore, how to improve the evaluation accuracy while improving the evaluation efficiency is a difficult problem that needs to be solved urgently.

[0004] In order to solve the defects in sensory cognition of liquor, the research on flavor substances of liquor has been paid more and more attention by enterprises and research institutions. The purpose of studying liquor flavor is to clarify the chemical nature of liquor and solve the problem of poor batch stability of product flavor quality caused by the complexity of solid-state brewing process. In recent years, the research technology of liquor flavor has become increasingly advanced, especially instrumental analysis technology and omics technology. The Xu Yan team of Jiangnan University has established a technical method system for the research of Chinese liquor flavor substances based on flavor-oriented technology by applying the academic thought of modern flavor chemistry research combining modern instrumental analysis with sensory analysis. Its research idea comes from "omics technology", such as the very mature "genomics" and "proteomics" technology. Applying this technology to the study of liquor flavor is called "flavoromics". Its non-targeted academic idea is to collect as much chemical information as possible through metabolomics strategy, and then combine a variety of chemometrics technology and modern sensory analysis technology to connect the chemical components in the sample with the senses, so as to mine, identify or predict important flavor compounds that affect flavor perception. Therefore, flavoromics is also called "flavor metabolomics", which is also the main research method of liquor flavor chemistry at present.

[0005] In order to solve the problems of scientificity and low efficiency of sensory cognition of liquor, the existing technology usually constructs a prediction model for liquor quality based on flavor substances as information elements. For example, Chinese patent CN118760987A discloses a method for constructing a liquor quality prediction model, which specifically uses flavor substances in liquor samples of different qualities as information, adjusts the hyperparameters of the XGBoost model based on grid search and cross-validation methods, and finally constructs an XGBoost liquor quality prediction model; Chinese patent CN116805150A discloses a liquor quality prediction method, which specifically preprocesses multivariate data in the liquor fermentation process based on liquor fermentation parameters to generate a multivariate time series in the liquor fermentation process; inputs the multivariate time series obtained in real time into the constructed and trained multivariate data prediction network to obtain the corresponding liquor fermentation parameters. Real-time prediction results of fermentation quality; Chinese patent CN115078573A discloses a method for predicting the quality grade of mellow-type liquor base wine, specifically step (1): using GC-MS to detect mellow-type liquor base wines of different quality grades with known quality grades to obtain fingerprint spectra of volatile flavor components; step (2): constructing a K nearest neighbor model and using the fingerprint spectrum obtained in step (1) to train and optimize the K nearest neighbor model to obtain a preliminary discrimination model; step (3): using the CART algorithm to calculate the fingerprint spectrum obtained in step (1), thereby screening out characteristic flavor substances, and using the obtained characteristic flavor substances to optimize the preliminary discrimination model to obtain a prediction model for predicting the quality grade of mellow-type liquor base wine; step (4): using the prediction model obtained in step (3) to predict the quality grade of mellow-type liquor base wine.

[0006] Although the prior art has disclosed that the use of these prediction methods can solve the problems of scientific cognition and low evaluation efficiency of liquor quality to a certain extent, they only perform discriminant analysis on the quality grade of liquor and cannot further quickly and quantitatively obtain more sensory characteristic information of liquor, and thus cannot design the wine body of diversified sensory characteristic products based on the sensory characteristics of different types of liquor. Summary of the invention

[0007] [Technical issues]

[0008] Liquor sensory perception is very complex, and it is very challenging for liquor tasters to accurately and quickly recognize the sensory characteristics of liquor and make correct qualitative assessments. For liquor production companies, liquor characteristic recognition and quality judgment generally rely on a professional tasting team, but the tasting process has high requirements for the experience and stability of the tasters. The liquor sensory quantitative description system can solve the problem of evaluation accuracy, but it will bring high labor cost investment.

[0009] Therefore, the study of trace complex flavor substances in liquor has become a research focus, in order to unlock the mystery of liquor's sensory properties. However, due to the complexity of liquor's flavor, how to use mature flavor detection technology to quickly and accurately predict the sensory characteristics and quality of liquor, assist in manual evaluation, and improve efficiency remains a problem that needs to be solved urgently.

[0010] [Technical solution]

[0011] To solve the above problems, the purpose of the present invention is to provide a method for constructing a sensory characteristic and quality expression prediction model of liquor using relevant flavor substances; the method takes Luzhou-flavor liquor as an example, firstly establishes a sensory quantitative description system of liquor, obtains typical sensory characteristic quantitative data, and adopts liquid-liquid microextraction (LLME), solid phase microextraction (SPME) and other pre-treatment methods combined with gas chromatography (GC) and gas chromatography-mass spectrometry to quantify a variety of trace flavor substances in liquor, and on this basis, analyzes the correlation between a variety of trace flavor substances and their sensory quantitative data, obtains important related flavor substances with different sensory characteristics, and finally constructs a sensory quantitative and quality evaluation prediction model of liquor using regression analysis algorithm, which assists wine tasters to quickly perform wine tasting and quality identification work, reduces costs and improves efficiency; this is of great value for quickly grasping the sensory characteristics of liquor (semi-finished products) in the actual production process and then classifying and storing them, and designing diversified characteristic products.

[0012] In order to achieve the above purpose, the technical solutions provided are as follows:

[0013] The present invention provides a method for constructing a prediction model for sensory characteristics and quality expression of liquor, the method comprising the following steps:

[0014] (1) Using GBT16861-1997 “Sensory analysis - identification and selection of descriptors for establishing sensory profiles by multivariate analysis methods”, a quantitative sensory description system for liquor was constructed to obtain descriptors of typical sensory characteristics of liquor and their quantitative data as well as quantitative data of overall sensory evaluation of aroma, flavor, and mouthfeel of liquor samples;

[0015] (2) GC-MS / GC-MS were used to determine the contents of 217 flavor compounds in wine samples;

[0016] (3) performing correlation analysis on the quantitative data of the sensory characteristic descriptors obtained in step (1) and the quantitative data of the overall sensory evaluation of the aroma, taste, and mouthfeel of the wine sample and the data of the contents of the 217 flavor substances obtained in step (2) to obtain important related flavor substances corresponding to each sensory characteristic descriptor and the overall sensory evaluation of the aroma, taste, and mouthfeel of the wine sample;

[0017] (4) Using each sensory characteristic descriptor and the quantitative data of the overall sensory evaluation of the aroma, taste, and mouthfeel of the wine sample as the dependent variable and the corresponding important related flavor substances as the independent variable, a regression analysis algorithm was used to construct a prediction model for the sensory quantitative and quality evaluation of liquor.

[0018] In one embodiment, in the process of constructing the quantitative sensory description system of liquor in step (1), the sensory characteristic descriptors are screened according to the usage frequency and evaluation intensity of the descriptors in the statistical evaluation.

[0019] In one embodiment, the typical sensory characteristic descriptors of step (1) are 28; 15 of them are aromas, including aged aroma, cellar aroma, sour aroma, ester aroma, soil aroma, burnt taste, grain aroma, rotten taste, raw taste, mellow aroma, green taste, sweet aroma, aldehyde aroma, aftertaste, and woody aroma; 3 tastes, including sweetness, sourness, and bitterness; 10 taste descriptors, including mellow, soft, long, refreshing, astringent, rough, delicate, dry and spicy, light, and short taste.

[0020] In one embodiment, the quantitative calculation method of typical sensory characteristic descriptors in step (1) is: Among them, taking the cellar aroma of wine samples as an example: A is the sum of the number of times the cellar aroma appears, B is the sum of the intensity of the cellar aroma evaluation, and C is the number of evaluators; the quantitative sensory evaluation of the aroma, taste, and mouthfeel of the wine samples is scored based on the 10-point system.

[0021] In one embodiment, the correlation analysis in step (3) is performed by using SPSS software to perform correlation module analysis, and the Pearson correlation coefficient is combined to determine the important related flavor substances corresponding to each sensory characteristic.

[0022] In one embodiment, in step (4), in the process of using regression analysis algorithm to construct a prediction model for sensory quantitative and quality evaluation of liquor, the model is adjusted by R 2 The values ​​and standard estimated errors are used as optimization targets and are optimized by adjusting the types of input variables (flavor substances).

[0023] In one embodiment, the model adjusted R 2 The value is the largest, the standard estimation error is the smallest, and the overall fitting effect of the model is the best.

[0024] The present invention provides a liquor sensory quantitative and quality evaluation prediction model obtained by the above-mentioned method.

[0025] In one embodiment, the liquor sensory quantitative and quality evaluation prediction model is specifically:

[0026] (1) Aroma:

[0027] Prediction value of cellar aroma = 11.00 + 1.34 × 10 -5 ×Acetal+2.05×10 -4 ×isobutanol+6.12×10 -4 ×Butyl hexanoate+6.83×10 -4 × Isoamyl hexanoate - 1.20 × 10 -4 ×Ethyl valerate;

[0028] Ester aroma prediction value = 4.47 + 3.10 × 10 -3 × Isoamyl acetate - 4.84 × 10 -6 × ethyl acetate + 1.85 × 10 -3 ×Butyl acetate+3.40×10 -6 ×Ethyl hexanoate-1.83×10 -3 × Propyl hexanoate + 2.05 × 10 -4 ×Ethyl heptanoate+4.33×10 -3 × Isobutyraldehyde - 4.84 × 10 -4 ×Isovaleraldehyde-3.47×10 -3 × propionaldehyde + 5.30 × 10 -5 ×isobutanol+5.23×10 -5 × n-propanol -4.20×10 -5 × n-butanol + 1.84 × 10 -4 ×2-pentanone;

[0029] Chen Xiang's predicted value = 14.83 + 3.28 × 10 -6 × Ethyl hexanoate - 5.23 × 10 -5 ×Ethyl valerate+2.30×10 -4 ×Butyl hexanoate-1.13×10 -4 ×Ethyl heptanoate+9.31×10 -5 × acetaldehyde;

[0030] Grain aroma prediction value = 8.47 + 39.1 × 10 -4 ×γ-valerolactone-3.54×10 -4 ×Butyl acetate+1.75×10 -4 × ethyl formate + 2.75 × 10 -3 ×Butyl valerate;

[0031] Prediction value of aroma = -1.96-2.53×10 -4 ×2-pentanol+7.14×10 -5 × Isobutanol - 1.37 × 10 -5 × acetaldehyde + ethyl hexanoate × 2.09 × 10 -6 -2.42×10 -4 × isoamyl acetate;

[0032] Sweet aroma prediction value = -2.19 + 7.66 × 10 -6 × n-butanol -3.64×10 -4 ×Ethyl valerate+4.49×10 -4 ×Butyl acetate-1.46×10 -4 ×Butyl hexanoate+3.19×10 -4 × Isoamyl hexanoate + 1.14 × 10 -4 ×Ethyl heptanoate;

[0033] Aldehyde aroma prediction value = -4.00-6.94×10 -5 ×2-pentanone+5.87×10 -6 × n-propanol -1.19×10 -5 ×Acetal+4.77×10 -5 × acetaldehyde + 1.09 × 10 -4 ×Ethyl formate-8.40×10 -4 ×Butyl acetate-1.99×10 -5 ×Ethyl butyrate+4.49×10 -6 × ethyl acetate;

[0034] Green flavor prediction value = 4.26 + anti, anti-6.07 × 10 -2 ×2.4.-nonadienal+3.27×10 -3 × trans-2-octenal;

[0035] Raw taste prediction value = 3.75 + 5.40 × 10 -3 × trans-2-octenal+4.87×10 -4 × trans-2-decenal;

[0036] Soil fragrance prediction value = 6.04 + 4 - 1.47 × 10 -3 × methylphenol + 1.85 × 10 -2 ×4-ethylphenol+7.78×10 -6 × Butyric acid + 1.00 × 10 -4 ×Isovaleric acid;

[0037] Burnt smell prediction value = 3.49-2.56×10 -3 ×4-methylphenol+3.61×10 -2 ×4-ethylphenol+1.11×10 -2 ×4-ethylguaiacol;

[0038] Sour aroma prediction value = -6.71-3.00×10 -5 ×Butyric acid+2.09×10 -4 × Heptanoic acid +2.82×10 -4 × Caprylic acid +1.16×10-5 × Acetic acid + 2.92 × 10 -4 × Propionic acid;

[0039] Bad taste prediction value = 6.34 + 1.43 × 10 -5 × Butyric acid - 7.97 × 10 -6 × Hexanoic acid + 5.26 × 10 -4 × Caprylic Acid - 1.10×10 -5 ×Acetic acid+2.05×10 -4 × Propionic acid + 3.29 × 10 -6 ×Ethyl lactate;

[0040] Prediction value of costusroot = 0.26 + 4.59 × 10 -4 ×linalool+1.57×10 -2 ×2-Thiophene carboxaldehyde+3.19×10 -4 ×4-ethylguaiacol -2.31×10 -3 × trans-2-nonenal;

[0041] Prediction value of aftertaste = -0.73-3.49×10 -4 ×Valeric acid-2.45×10 -4 × Caprylic acid +3.91×10 -5 ×Butyric acid+9.79×10 -4 × Heptanoic acid + 1.01 × 10 -6 ×Ethyl lactate;

[0042] (2) Taste and texture

[0043] Sweetness prediction value = 75.05-1.46×10 -6 ×Ethyl lactate-2.75×10 -4 × Propionic acid + 3.10 × 10 -6 × ethyl hexanoate + 8.26 × 10 -5 ×Ethyl valerate-4.51×10 -4 ×2-pentanol+6.03×10 -6 × Acetic acid - 1.04 × 10 -3 ×Isoamyl acetate

[0044] Sourness prediction value = 33.45 + 2.52 × 10 -4 × Propionic acid +6.90×10 -4 × Caprylic acid ++6.44×10 -4 × Isoamyl acetate - 4.85 × 10 -4 ×Isobutyraldehyde

[0045] Bitterness prediction value = 0.54 + 1.76 × 10 -2 ×2.3.5.6-Tetramethylpyrazine+4.25×10-7 ×Ethyl lactate+3.98×10 -3 ×2-Propionylfuran+2.31×10 -5 ×Furfural

[0046] Predicted value of body = 44.04-2.88×10 -4 × Propionic acid + 1.30 × 10 -5 × ethyl hexanoate + 4.76 × 10 -4 ×2-pentanol-1.77×10 -4 ×Ethyl heptanoate+6.64×10 -4 ×Butyl Hexanoate

[0047] Predicted value of thinness = 6.53 + 2.79 × 10 -4 × Propionic acid +2.13×10 -2 ×2-Propionylfuran-3.78×10 -4 ×2-pentanol+1.32×10 -4 ×Ethyl valerate

[0048] Fine prediction value = 18.12-3.29×10 -6 ×Ethyl lactate-1.83×10 -4 × Propionic acid - 1.85 × 10 -3 ×Butyl acetate+9.71×10 -5 × Isobutanol - 1.64 × 10 -3 × Isoamyl acetate + 5.80 × 10 -6 × ethyl acetate + 4.89 × 10 -4 ×Isovaleraldehyde

[0049] Rough prediction value = 23.02 + 6.8 × 10 -6 ×Ethyl lactate-4.99×10 -4 ×2-pentanol+8.59×10 -4 × Propyl hexanoate + 1.95 × 10 -3 × Isoamyl acetate - 9.17 × 10 -5 ×Isobutanol

[0050] Soft prediction value = 55.79-6.53×10 -6 × ethyl acetate + 2.68 × 10 -4 ×Ethyl valerate+9.48×10 -4 ×2-pentanol-4.73×10 -4 ×Ethyl heptanoate+1.49×10 -2 ×2,2-diethoxyethanol-1.22×10 -3 ×Isobutyraldehyde

[0051] Prediction value of dryness and spiciness = 1.33 + 1.56 × 10-5 ×Ethyl butyrate+2.26×10 -4 ×2-pentanone-1.72×10 -5 ×Acetal+6.37×10 -5 ×Ethyl Heptanoate

[0052] Long-term prediction value = 32.35-4.76×10 -6 ×Ethyl lactate+6.91×10 -6 × ethyl hexanoate + 5.83 × 10 -4 ×2-pentanol-2.48×10 -4 ×Ethyl valerate-2.75×10 -5 ×Acetal+3.16×10 -4 × Isobutanol - 1.47 × 10 -3 ×Butyl acetate+8.12×10 -5 ×Acetaldehyde

[0053] Short taste prediction value = 6.15 + 2.46 × 10 -6 ×Ethyl lactate+8.11×10 -4 × Furfuryl alcohol - 7.75 × 10 -5 × isobutanol + 1.33 × 10 -3 ×Butyl Acetate

[0054] Prediction value of refreshing and quiet = 50.35-2.54×10 -4 × Propionic acid +7.50×10 -6 × ethyl hexanoate + 2.89 × 10 -3 ×Butyl acetate+2.15×10 -4 ×Ethyl valerate+2.24×10 -5 ×Acetal-3.62×10 -4 ×Ethyl heptanoate-1.20×10 -5 ×Ethyl acetate

[0055] Prediction value of astringency = 12.87 + 7.49 × 10 -6 ×Ethyl lactate+1.54×10 -4 × Propionic acid - 1.05 × 10 -4 ×Ethyl valerate+2.05×10 -4 ×Butyl acetate+1.17×10 -4 ×Ethyl Heptanoate

[0056] (3) Overall evaluation

[0057] Aroma score = 6.49 + 4.12 × 10 -7 × Ethyl hexanoate - 1.47 × 10 -6 ×Acetal+4.92×10 -6× Isobutanol - 9.87 × 10 -6 ×Ethyl heptanoate-1.43×10 -7 × Ethyl acetate - 1.43 × 10 -4 ×Isobutyraldehyde+3.12×10 -6 × acetaldehyde + 1.48 × 10 -5 × ethyl formate + 1.29 × 10 -5 ×Isovaleraldehyde+1.23×10 -5 ×Butyl hexanoate-1.74×10 -7 ×Ethyl lactate-1.19×10 -5 × Propionic acid;

[0058] Taste score = 6.49 + 3.32 × 10 -7 × Ethyl hexanoate - 3.59 × 10 -5 ×Butyl acetate-5.52×10 -7 ×Acetal+6.67×10 -6 × Isobutanol - 7.83 × 10 -5 × Isoamyl acetate - 1.51 × 10 -5 × Caprylic Acid - 5.26 × 10 -6 × Propionic acid - 2.00 × 10 -7 ×Ethyl lactate;

[0059] Overall evaluation = 6.28 + 2.03 × 10 -7 ×Ethyl hexanoate-8.57×10 -5 ×Butyl acetate-6.52×10 -6 ×Ethyl heptanoate+4.54×10 -6 × isobutanol + 3.22 × 10 -5 ×Butyl hexanoate-6.13×10 -7 × Hexanoic acid + 3.52 × 10 -7 × Acetic acid - 5.69 × 10 -6 × Propionic acid - 1.78 × 10 -7 ×Ethyl lactate.

[0060] The present invention also provides the application of the above-mentioned method or the above-mentioned liquor sensory quantification and quality evaluation prediction model in liquor production monitoring.

[0061] Beneficial effects:

[0062] The method for constructing a prediction model for the expression of sensory characteristics and quality of liquor provided by the present invention not only solves the problem that traditional evaluation methods are not concrete and general in expression, but also utilizes mature flavor detection technology combined with multivariate statistical analysis technology to quickly and accurately predict the sensory characteristics and quality of liquor, and assist in manual evaluation, which has broad application prospects for liquor companies to improve production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 A flowchart of the method for constructing a prediction model for sensory characteristics and quality expression of liquor according to the present invention;

[0064] Figure 2 This is a distribution diagram of descriptive words for sensory quantitative analysis of Luzhou-flavor liquor selected in Example 1;

[0065] Figure 3 For Example 4, the sensory quantification and quality evaluation prediction model of Luzhou-flavor liquor is applied to the aroma quantitative prediction graphs of three test liquor samples;

[0066] Figure 4 For Example 4, the sensory quantitative and quality evaluation prediction model of Luzhou-flavor liquor is applied to the quantitative prediction graphs of the taste and mouthfeel of three test liquor samples;

[0067] Figure 5 For Example 4, the sensory quantitative and quality evaluation prediction model of Luzhou-flavor liquor is applied to the overall evaluation quantitative prediction graph of three test liquor samples;

[0068] Figure 6 This is a fitting effect diagram of the prediction model for sensory quantification and quality evaluation of liquor aroma of actual liquor samples in the workshop of Example 5;

[0069] Figure 7 This is a fitting effect diagram of the prediction model for the taste and mouthfeel of actual liquor samples in the workshop of Example 5 and the sensory quantification and quality evaluation of liquor;

[0070] Figure 8 This is a fitting effect diagram of the prediction model for sensory quantification and quality evaluation of liquor based on the actual liquor samples from the workshop in Example 5. DETAILED DESCRIPTION

[0071] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than 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 belong to the scope of protection of the present invention. The following specific implementation methods further describe the present invention.

[0072] Sources of materials involved in the present invention:

[0073] GC 2030plus gas chromatograph (equipped with FID detector, Shimadzu Corporation, Japan), 7890A-5975C gas chromatograph (Agilent Technologies, USA), MPS2XL multifunctional autosampler (Gestel, Germany), 50 / 30μm DVB / CAR / PDMS three-phase extraction head (Supelco, USA); DMA35 / Ex / Petrol handheld density meter (Anton Paar, Austria);

[0074] Liquor samples: 58 Luzhou-flavor liquor samples with sensory differences were selected for the construction of liquor sensory quantitative description system, analysis of important related components of liquor typical sensory characteristics, and construction of liquor sensory characteristics and quality expression prediction model. Among them, 50 samples were used for model construction and 8 samples were used for model verification. 3 liquor samples with sensory differences were selected from the 8 verification samples to demonstrate the actual application effect of the model.

[0075] Professional sensory tasters: 20 professional certified tasters are selected, and a leader of the evaluation team is appointed to identify the tasters’ evaluation motivation and the tasters’ ability to repeat the evaluation.

[0076] Reagents: sodium chloride (analytical grade), ether, ethanol (chromatographic grade), anhydrous sodium sulfate (analytical grade), ultrapure water (18.25 mΩ·cm).

[0077] Example 1 Construction of Liquor Sensory Quantitative Description System

[0078] (1) Screening of descriptors for sensory characteristics of liquor based on M value

[0079] A blind evaluation method was adopted, and 58 wine samples were randomly presented to professional sensory tasters in multiple rounds. The wine glasses used were standard white wine tasting glasses, and about 20 mL of wine was poured into each glass. The product characteristics were evaluated using the sensory quantitative description analysis method described in GBT16861-1997 "Sensory analysis through multivariate analysis methods to identify and select descriptors for establishing sensory profiles". The qualitative expression of characteristics can refer to the white wine sensory characteristic description word library (Table 1), and appropriate words in the table can be selected to describe the sensory characteristics of the wine samples, or other characteristic descriptors not mentioned in the table can be selected for qualitative expression; the quantitative expression of characteristics refers to the white wine sensory analysis scoring scale table (Table 2), and the sensory quantitative description analysis table is used to describe the product characteristics, and the aroma, taste, mouthfeel and comprehensive performance of the wine samples are scored on a 10-point scale;

[0080] After the wine samples were tasted by 20 judges, a total of 121 sensory descriptors were obtained, including 54 olfactory descriptors, 4 taste descriptors, and 34 flavor descriptors. The collected descriptors were preliminarily sorted out, integrating terms with the same meaning and deleting inappropriate terms. Finally, 23 olfactory descriptors, 3 taste descriptors, and 14 flavor descriptors were obtained;

[0081] Table 1. Liquor sensory characteristics description corpus

[0082]

[0083]

[0084] Table 2. Comparison table of sensory intensity and quality evaluation scores of wine samples

[0085]

[0086] In the process of constructing the sensory quantitative description system, the sensory characteristic descriptors were screened according to the usage frequency and evaluation intensity of the descriptors in the statistical evaluation. The contribution index M of each descriptor was calculated by formula (1). The larger the M value, the greater the importance of the descriptor in expressing the sensory characteristics of the product. Generally, descriptors with M values ​​greater than 0.02 and M values ​​of taste and flavor greater than 0.1 were selected. The 40 descriptors selected were screened according to the M value evaluation method, and 6 aroma descriptors such as Dongjiu odor and 4 taste descriptors such as stimulation were eliminated. The descriptors of aftertaste, woody fragrance, dry and spicy, light, short taste and bitter were retained through negotiation, that is, 30 aroma, taste and flavor descriptors were obtained, which were 17, 3 and 10 respectively.

[0087]

[0088] M represents the square root of the product of the frequency F and relative intensity I of each descriptor;

[0089] F represents the percentage of the number of times a descriptive word is actually mentioned to the total number of times the descriptive word may be mentioned;

[0090] I represents the evaluation intensity actually given by the evaluation team to a descriptor and its percentage of the maximum possible evaluation intensity of the descriptor.

[0091] Table 3. Ranking of M values ​​of different sensory characteristics of wine

[0092]

[0093] (2) Rescreening sensory feature descriptors based on multivariate statistical analysis

[0094] Cluster analysis and principal component analysis were used to perform multivariate statistical analysis on the 30 descriptors initially screened above. The results of cluster analysis showed that the aroma of dregs and the taste of dregs, the mellow aroma and the light aroma, the raw taste and the green taste, the sweet aroma, the aldehyde aroma, the woody aroma, the aftertaste, and the pickled vegetable taste all had high similarity. According to the discussion results, the light aroma and the aroma of dregs were eliminated, that is, 15 aromas (aged aroma, cellar aroma, sour aroma, ester aroma, soil aroma, burnt taste, grain aroma, dregs, raw taste, mellow aroma, green taste, sweet aroma, aldehyde aroma, aftertaste, and woody aroma), 3 tastes (sweetness, sourness, and bitterness) and 10 taste descriptors (mellow, soft, long, refreshing, astringent, rough, delicate, dry and spicy, light, and short taste) were retained, for a total of 28 sensory descriptors;

[0095] The principal component analysis of the 28 retained sensory descriptors showed that the cumulative contribution rate of the first three principal components reached more than 80%, indicating that the first three principal components have provided sufficient information of the original data, that is, these 28 sensory descriptors reflect the basic characteristics of the original wine;

[0096] (3) Quantification of sensory characteristics based on the geometric mean of sensory characteristics

[0097] Due to individual differences among tasters, there is uncertainty in the 5-point scale for measuring the sensory characteristics of the 58 original wines mentioned above. Therefore, the sensory characteristics are quantified from the perspective of the frequency and intensity of sensory descriptors, that is, the geometric mean of each sensory characteristic of the wine sample is used to quantify each sensory characteristic (for example, the sum of the number of occurrences of the cellar aroma of a wine sample is A, the sum of its evaluation intensity is B, and the number of tasters is C, then the geometric mean of the cellar aroma sensory characteristic of the wine sample can be expressed as The average score of the aroma, taste and mouthfeel of the wine samples (out of 10 points) can be directly used for the overall sensory quantitative evaluation of the wine. This completes the construction of a quantitative system of 28 sensory descriptors and corresponding aroma, tone and mouthfeel for 58 original wines.

[0098] Example 2 Determination of the content of 217 flavor substances in liquor

[0099] The gas chromatography method is used to determine the contents of 58 flavor substances in liquor, including the following steps:

[0100] (1) Liquor sample pretreatment

[0101] Directly use a 5mL pipette to transfer 5mL of wine sample to a 10mL stoppered graduated colorimetric tube, add 50μL of three internal standard mixed solutions (tert-amyl alcohol, n-amyl acetate, 2-ethyl-n-butyric acid), the final concentrations of the internal standards in the 5mL wine sample system are tert-amyl alcohol 15.98mg / 100mL, n-amyl acetate 17.19mg / 100mL, 2-ethyl-n-butyric acid 18.28mg / 100mL, respectively, and mix thoroughly to form the sample to be tested;

[0102] (2) Determination of the content of 58 flavor substances in liquor samples

[0103] The flavor substance content in the sample to be tested in step (1) was determined by gas chromatography, and the chromatographic conditions were as follows: the chromatographic column was a CP-Wax 57CB capillary chromatographic column, 50m×0.25mm×0.25μm; the carrier gas was nitrogen, the chromatographic column flow rate was 1.0mL / min, the injection port temperature was 270°C, the injection volume was 1μL, the split ratio was 50:1, the hydrogen flow rate was 30mL / min, the air flow rate was 300mL / min, and the detector temperature was 300°C; the temperature program was 35°C for 6min, increased to 60°C at 6°C / min, retained for 3min, then increased to 80°C at 4.5°C / min, retained for 2min, increased to 180°C at 9.5°C / min, retained for 2min, increased to 210°C at 9.5°C / min, and retained for 10min.

[0104] The gas chromatography method is used to determine the content of 159 flavor substances in liquor, including the following steps:

[0105] (1) Pretreatment method for determination of 106 flavor compounds in liquor by liquid-liquid microextraction (LLME) combined with gas chromatography-mass spectrometry

[0106] The wine sample was shaken, 20 mL of the wine sample diluted to 14% voL was taken into a 100 mL sealed sample bottle, NaCl was added to saturation (about 9 g), 50 μL of internal standard mixed solution (2-ethylbutyric acid 2000 mg / L, 2-octanol 1600 mg / L, the final concentration of internal standard in the 20 mL wine sample system was 2-ethylbutyric acid 5 mg / L, 2-octanol 4 mg / L), and then 5 mL of ether was transferred, and 350 r / min was oscillated for 30 min, and extracted 3 times in total. The three organic phases were combined into a 20 mL sealed reagent bottle, 5 g of anhydrous sodium sulfate was added, and dried overnight. The next day, the extract was transferred into a 25 mL KD concentrator tube, concentrated to 0.5 mL by nitrogen blowing, and then transferred to an injection bottle containing a 300 or 500 μL liner tube for GC-MS analysis;

[0107] Pretreatment method for determination of 53 flavor substances in liquor by solid phase microextraction (SPME) combined with gas chromatography-mass spectrometry:

[0108] The sample was shaken, 8 mL of the wine sample was diluted to 5% voL and placed in a 15 mL headspace bottle, 3 g of sodium chloride and 40 μL of internal standard (2-octanol 200 mg / L, the internal standard concentration in the 8 mL wine sample system was 1 mg / L) were added, the headspace bottle cap was quickly tightened, the 50 / 30 μm DVB / CAR / PDMS three-phase extraction head was aged at 260°C for 15 min, the headspace injection heater was kept at 50°C and oscillated at 300 rpm for 5 min, then the extraction head was inserted and continued to oscillate and extract for 45 min, and the injection port was analyzed for 5 min;

[0109] (2) Liquid-liquid microextraction (LLME) and solid phase microextraction (SPME) combined with gas chromatography-mass spectrometry were used to determine the content of flavor compounds in liquor;

[0110] Among them, the chromatographic and mass spectrometry conditions are as follows: the chromatographic column is a DB-WaxUI type inert chromatographic column, 60m×0.25mm×0.25μm; the carrier gas is helium (purity ≥99.999%); the flow rate is 1.0mL / min; the injection port temperature is 230℃; the injection volume is 1μL; no splitting; the initial temperature is 40℃, maintained for 5min, then increased to 100℃ at 3.5℃ / min, maintained for 2min, then increased to 200℃ at 5℃ / min, maintained for 4min, and finally increased to 220℃ at 20℃ / min, maintained for 15min. After the program is completed, it is run at 40℃ for 1min; electron bombardment source (EI); electron bombardment energy is 70ev; quadrupole temperature is 150℃; ion source temperature is 230℃; transmission line temperature is 280℃; selected ion monitoring (SIM).

[0111] Example 3 Analysis of important related flavor substances of typical sensory characteristics of original wine

[0112] The correlation analysis module of SPSS software was used to perform correlation analysis on the 28 sensory descriptors of 58 original liquors constructed in Example 1 and the corresponding quantitative data of aroma, breath, and taste with the content data of 217 flavor substances. The important related flavor substances of 28 typical sensory characteristics and 3 sensory overall scores were determined by combining the Pearson correlation coefficient, and 60 important related flavor substances corresponding to different sensory characteristics were obtained, as shown in Table 4.

[0113] Table 4 Important related flavor substances of typical sensory characteristics of liquor

[0114]

[0115]

[0116] Example 4: Construction of a prediction model for liquor sensory quantification and quality evaluation

[0117] 50 wine samples were used as training samples and 8 wine samples were used as test samples. Based on the contents of 60 important related flavor substances obtained in Example 3 above as independent variables, 28 aroma sensory organs and 3 sensory overall score data as dependent variables, a wine body sensory quantitative prediction and quality evaluation model (flavor substance content unit μg / L) was constructed;

[0118] The constructed model is based on the model adjusted R 2The value and standard estimated error are taken as the optimization target, and the optimization is carried out by adjusting the types of input variables (flavor substances). Table 5 describes the optimization process by taking the cellar aroma prediction model as an example. Model 6 selects 6 out of 10 input variables (isobutyl alcohol, isopentyl hexanoate, acetal, butyl hexanoate, ethyl valerate), and model R 2 The value is the largest, the standard estimation error is the smallest, and the overall model fitting effect is the best, so the cellar fragrance sensory characteristics select model 6 as the final optimization model. The optimization process of other sensory characteristics is the same as that of the cellar fragrance, and will not be repeated one by one;

[0119] Table 5 Optimization of prediction model for sensory quantitative and quality evaluation of liquor

[0120]

[0121] The optimal prediction model finally obtained is as follows:

[0122] (1) Aroma

[0123] Prediction value of cellar aroma = 11.00 + 1.34 × 10 -5 ×Acetal+2.05×10 -4 ×isobutanol+6.12×10 -4 ×Butyl hexanoate+6.83×10 -4 × Isoamyl hexanoate - 1.20 × 10 -4 ×Ethyl valerate;

[0124] Ester aroma prediction value = 4.47 + 3.10 × 10 -3 × Isoamyl acetate - 4.84 × 10 -6 × ethyl acetate + 1.85 × 10 -3 ×Butyl acetate+3.40×10 -6 ×Ethyl hexanoate-1.83×10 -3 × Propyl hexanoate + 2.05 × 10 -4 ×Ethyl heptanoate+4.33×10 -3 × Isobutyraldehyde - 4.84 × 10 -4 ×Isovaleraldehyde-3.47×10 -3 × propionaldehyde + 5.30 × 10 -5 ×isobutanol+5.23×10 -5 × n-propanol -4.20×10 -5 × n-butanol + 1.84 × 10 -4 ×2-pentanone;

[0125] Chen Xiang's predicted value = 14.83 + 3.28 × 10 -6 × Ethyl hexanoate - 5.23 × 10 -5 ×Ethyl valerate+2.30×10 -4 ×Butyl hexanoate-1.13×10-4 ×Ethyl heptanoate+9.31×10 -5 × acetaldehyde;

[0126] Grain aroma prediction value = 8.47 + 39.1 × 10 -4 ×γ-valerolactone-3.54×10 -4 ×Butyl acetate+1.75×10 -4 × ethyl formate + 2.75 × 10 -3 ×Butyl valerate;

[0127] Prediction value of aroma = -1.96-2.53×10 -4 ×2-pentanol+7.14×10 -5 × Isobutanol - 1.37 × 10 -5 × acetaldehyde + ethyl hexanoate × 2.09 × 10 -6 -2.42×10 -4 × isoamyl acetate;

[0128] Sweet aroma prediction value = -2.19 + 7.66 × 10 -6 × n-butanol -3.64×10 -4 ×Ethyl valerate+4.49×10 -4 ×Butyl acetate-1.46×10 -4 ×Butyl hexanoate+3.19×10 -4 × Isoamyl hexanoate + 1.14 × 10 -4 ×Ethyl heptanoate;

[0129] Aldehyde aroma prediction value = -4.00-6.94×10 -5 ×2-pentanone+5.87×10 -6 × n-propanol -1.19×10 -5 ×Acetal+4.77×10 -5 × acetaldehyde + 1.09 × 10 -4 ×Ethyl formate-8.40×10 -4 ×Butyl acetate-1.99×10 -5 ×Ethyl butyrate+4.49×10 -6 × ethyl acetate;

[0130] Green flavor prediction value = 4.26 + anti, anti-6.07 × 10 -2 ×2.4.-nonadienal+3.27×10 -3 × trans-2-octenal;

[0131] Raw taste prediction value = 3.75 + 5.40 × 10 -3 × trans-2-octenal+4.87×10 -4 × trans-2-decenal;

[0132] Soil fragrance prediction value = 6.04 + 4 - 1.47 × 10 -3 × methylphenol + 1.85 × 10 -2 ×4-ethylphenol+7.78×10 -6 × Butyric acid + 1.00 × 10 -4 ×Isovaleric acid;

[0133] Burnt smell prediction value = 3.49-2.56×10 -3 ×4-methylphenol+3.61×10 -2 ×4-ethylphenol+1.11×10 -2 ×4-ethylguaiacol;

[0134] Sour aroma prediction value = -6.71-3.00×10 -5 ×Butyric acid+2.09×10 -4 × Heptanoic acid +2.82×10 -4 × Caprylic acid +1.16×10 -5 × Acetic acid + 2.92 × 10 -4 × Propionic acid;

[0135] Bad taste prediction value = 6.34 + 1.43 × 10 -5 × Butyric acid - 7.97 × 10 -6 × Hexanoic acid + 5.26 × 10 -4 × Caprylic Acid - 1.10×10 -5 ×Acetic acid+2.05×10 -4 × Propionic acid + 3.29 × 10 -6 ×Ethyl lactate;

[0136] Prediction value of costusroot = 0.26 + 4.59 × 10 -4 ×linalool+1.57×10 -2 ×2-Thiophene carboxaldehyde+3.19×10 -4 ×4-ethylguaiacol -2.31×10 -3 × trans-2-nonenal;

[0137] Prediction value of aftertaste = -0.73-3.49×10 -4 ×Valeric acid-2.45×10 -4 × Caprylic acid +3.91×10 -5 ×Butyric acid+9.79×10 -4 × Heptanoic acid + 1.01 × 10 -6 ×Ethyl lactate;

[0138] (2) Taste and texture

[0139] Sweetness prediction value = 75.05-1.46×10 -6×Ethyl lactate-2.75×10 -4 × Propionic acid + 3.10 × 10 -6 × ethyl hexanoate + 8.26 × 10 -5 ×Ethyl valerate-4.51×10 -4 ×2-pentanol+6.03×10 -6 × Acetic acid - 1.04 × 10 -3 ×Isoamyl acetate

[0140] Sourness prediction value = 33.45 + 2.52 × 10 -4 × Propionic acid +6.90×10 -4 × Caprylic acid ++6.44×10 -4 × Isoamyl acetate - 4.85 × 10 -4 ×Isobutyraldehyde

[0141] Bitterness prediction value = 0.54 + 1.76 × 10 -2 ×2.3.5.6-Tetramethylpyrazine+4.25×10 -7 ×Ethyl lactate+3.98×10 -3 ×2-Propionylfuran+2.31×10 -5 ×Furfural

[0142] Predicted value of body = 44.04-2.88×10 -4 × Propionic acid + 1.30 × 10 -5 × ethyl hexanoate + 4.76 × 10 -4 ×2-pentanol-1.77×10 -4 ×Ethyl heptanoate+6.64×10 -4 ×Butyl Hexanoate

[0143] Predicted value of thinness = 6.53 + 2.79 × 10 -4 × Propionic acid +2.13×10 -2 ×2-Propionylfuran-3.78×10 -4 ×2-pentanol+1.32×10 -4 ×Ethyl valerate

[0144] Fine prediction value = 18.12-3.29×10 -6 ×Ethyl lactate-1.83×10 -4 × Propionic acid - 1.85 × 10 -3 ×Butyl acetate+9.71×10 -5 × Isobutanol - 1.64 × 10 -3 × Isoamyl acetate + 5.80 × 10 -6 × ethyl acetate + 4.89 × 10 -4 ×Isovaleraldehyde

[0145] Rough prediction value = 23.02 + 6.8 × 10 -6 ×Ethyl lactate-4.99×10 -4 ×2-pentanol+8.59×10 -4 × Propyl hexanoate + 1.95 × 10 -3 × Isoamyl acetate - 9.17 × 10 -5 ×Isobutanol

[0146] Soft prediction value = 55.79-6.53×10 -6 × ethyl acetate + 2.68 × 10 -4 ×Ethyl valerate+9.48×10 -4 ×2-pentanol-4.73×10 -4 ×Ethyl heptanoate+1.49×10 -2 ×2,2-diethoxyethanol-1.22×10 -3 ×Isobutyraldehyde

[0147] Prediction value of dryness and spiciness = 1.33 + 1.56 × 10 -5 ×Ethyl butyrate+2.26×10 -4 ×2-pentanone-1.72×10 -5 ×Acetal+6.37×10 -5 ×Ethyl Heptanoate

[0148] Long-term prediction value = 32.35-4.76×10 -6 ×Ethyl lactate+6.91×10 -6 × ethyl hexanoate + 5.83 × 10 -4 ×2-pentanol-2.48×10 -4 ×Ethyl valerate-2.75×10 -5 ×Acetal+3.16×10 -4 × Isobutanol - 1.47 × 10 -3 ×Butyl acetate+8.12×10 -5 ×Acetaldehyde

[0149] Short taste prediction value = 6.15 + 2.46 × 10 -6 ×Ethyl lactate+8.11×10 -4 × Furfuryl alcohol - 7.75 × 10 -5 × isobutanol + 1.33 × 10 -3 ×Butyl Acetate

[0150] Prediction value of refreshing and quiet = 50.35-2.54×10 -4 × Propionic acid +7.50×10 -6 × ethyl hexanoate + 2.89 × 10 -3 ×Butyl acetate+2.15×10 -4×Ethyl valerate+2.24×10 -5 ×Acetal-3.62×10 -4 ×Ethyl heptanoate-1.20×10 -5 ×Ethyl acetate

[0151] Prediction value of astringency = 12.87 + 7.49 × 10 -6 ×Ethyl lactate+1.54×10 -4 × Propionic acid - 1.05 × 10 -4 ×Ethyl valerate+2.05×10 -4 ×Butyl acetate+1.17×10 -4 ×Ethyl Heptanoate

[0152] (3) Overall evaluation

[0153] Aroma score = 6.49 + 4.12 × 10 -7 × Ethyl hexanoate - 1.47 × 10 -6 ×Acetal+4.92×10 -6 × Isobutanol - 9.87 × 10 -6 ×Ethyl heptanoate-1.43×10 -7 × Ethyl acetate - 1.43 × 10 -4 ×Isobutyraldehyde+3.12×10 -6 × acetaldehyde + 1.48 × 10 -5 × ethyl formate + 1.29 × 10 -5 ×Isovaleraldehyde+1.23×10 -5 ×Butyl hexanoate-1.74×10 -7 ×Ethyl lactate-1.19×10 -5 × Propionic acid;

[0154] Taste score = 6.49 + 3.32 × 10 -7 × Ethyl hexanoate - 3.59 × 10 -5 ×Butyl acetate-5.52×10 -7 ×Acetal+6.67×10 -6 × Isobutanol - 7.83 × 10 -5 × Isoamyl acetate - 1.51 × 10 -5 × Caprylic Acid - 5.26 × 10 -6 × Propionic acid - 2.00 × 10 -7 ×Ethyl lactate;

[0155] Overall evaluation = 6.28 + 2.03 × 10 -7 ×Ethyl hexanoate-8.57×10 -5 ×Butyl acetate-6.52×10 -6 ×Ethyl heptanoate+4.54×10 -6× isobutanol + 3.22 × 10 -5 ×Butyl hexanoate-6.13×10 -7 × Hexanoic acid + 3.52 × 10 -7 × Acetic acid - 5.69 × 10 -6 × Propionic acid - 1.78 × 10 -7 ×Ethyl lactate;

[0156] This model was used to perform sensory quantitative prediction and verification on 8 test wine samples. The verification results showed that the relative deviation of the model ((predicted value - actual value) / actual value) ranged from 3 to 11%. Figure 4-Figure 6 From the above 8 wine samples, 3 wine samples (wine samples 1, 2 and 3) with large sensory differences were selected to use the model to predict the sensory situation in terms of aroma, taste and overall sensory score.

[0157] Example 5 Practical application effect of liquor sensory quantitative and quality evaluation prediction model

[0158] Three kinds of liquor samples produced in the workshop were taken to quantitatively detect the 60 flavor substances described in Example 3. Then, according to the model constructed in Example 4, the aroma sensory characteristics and quality prediction effects of the three liquor samples were evaluated. The results showed that the relative deviation range of the prediction results of the three sample models was 4.5-15.9%. Figure 7-Figure 8 The figure shows the fitting effect of one of the wine sample aroma, taste and overall score models.

[0159] In summary, the constructed model can better predict the sensory characteristics and flavor quality of different types of wine samples.

[0160] The embodiments provided above are not intended to limit the scope of the present invention, and the steps described are not intended to limit the execution order thereof. Those skilled in the art may make obvious improvements to the present invention in combination with existing common knowledge, which also fall within the scope of protection defined by the claims of the present invention.

Claims

1. A method for constructing a prediction model for sensory characteristics and quality expression of liquor, characterized in that: The method comprises the following steps: (1) Using GBT16861-1997 “Sensory analysis - identification and selection of descriptors for establishing sensory profiles by multivariate analysis methods”, a quantitative sensory description system for liquor was constructed to obtain descriptors of typical sensory characteristics of liquor and their quantitative data as well as quantitative data of overall sensory evaluation of aroma, flavor, and mouthfeel of liquor samples; (2) GC-MS / GC-MS were used to determine the contents of 217 flavor compounds in wine samples; (3) performing correlation analysis on the quantitative data of the sensory characteristic descriptors obtained in step (1) and the quantitative data of the overall sensory evaluation of the aroma, taste, and mouthfeel of the wine sample and the data of the contents of the 217 flavor substances obtained in step (2) to obtain important related flavor substances corresponding to each sensory characteristic descriptor and the overall sensory evaluation of the aroma, taste, and mouthfeel of the wine sample; (4) Using each sensory characteristic descriptor and the quantitative data of the overall sensory evaluation of the aroma, taste, and mouthfeel of the wine sample as the dependent variable and the corresponding important related flavor substances as the independent variable, a regression analysis algorithm was used to construct a prediction model for the sensory quantitative and quality evaluation of liquor.

2. The method according to claim 1, characterized in that In the process of constructing the quantitative sensory description system of liquor in step (1), the sensory characteristic descriptors are screened according to the usage frequency and evaluation intensity of the descriptors in the statistical evaluation.

3. The method according to claim 1, characterized in that There are 28 typical sensory characteristics descriptors in step (1); 15 of them are aroma, including aged aroma, cellar aroma, sour aroma, ester aroma, soil aroma, burnt taste, grain aroma, rotten taste, raw taste, mellow aroma, green taste, sweet aroma, aldehyde aroma, aftertaste, and woody aroma; 3 tastes, including sweetness, sourness, and bitterness; 10 taste descriptors, including mellow, soft, long, refreshing, astringent, rough, delicate, dry and spicy, light, and short taste.

4. The method according to claim 1, characterized in that: The quantitative calculation method of typical sensory characteristic descriptors in step (1) is: Among them, taking the cellar aroma of wine samples as an example: A is the sum of the number of times the cellar aroma appears, B is the sum of the intensity of the cellar aroma evaluation, and C is the number of evaluators; the quantitative sensory evaluation of the aroma, taste, and mouthfeel of the wine samples is scored based on the 10-point system.

5. The method according to claim 1, characterized in that The correlation analysis in step (3) is performed by using SPSS software to perform correlation module analysis, and the Pearson correlation coefficient is combined to determine the important related flavor substances corresponding to each sensory characteristic.

6. The method according to claim 1, characterized in that In the process of using regression analysis algorithm to construct the prediction model of liquor sensory quantitative and quality evaluation, the R 2 The values ​​and standard estimated errors are used as optimization targets and then optimized by adjusting the types of input flavor substances.

7. The method according to claim 6, characterized in that The model adjusted R 2 The value is the largest, the standard estimation error is the smallest, and the overall fitting effect of the model is the best.

8. A sensory quantitative and quality evaluation prediction model for liquor obtained by the method described in any one of claims 1 to 7.

9. The liquor sensory quantitative and quality evaluation prediction model according to claim 8, characterized in that: The liquor sensory quantitative and quality evaluation prediction model is specifically as follows: (1) Aroma: Prediction value of cellar aroma = 11.00 + 1.34 × 10 -5 ×Acetal+2.05×10 -4 ×isobutanol+6.12×10 -4 ×Butyl hexanoate+6.83×10 -4 × Isoamyl hexanoate - 1.20 × 10 -4 ×Ethyl valerate; Ester aroma prediction value = 4.47 + 3.10 × 10 -3 × Isoamyl acetate - 4.84 × 10 -6 × ethyl acetate + 1.85 × 10 -3 ×Butyl acetate+3.40×10 -6 ×Ethyl hexanoate-1.83×10 -3 × Propyl hexanoate + 2.05 × 10 -4 ×Ethyl heptanoate+4.33×10 -3 × Isobutyraldehyde - 4.84 × 10 -4 ×Isovaleraldehyde-3.47×10 -3 × propionaldehyde + 5.30 × 10 -5 ×isobutanol+5.23×10 -5 × n-propanol -4.20×10 -5 × n-butanol + 1.84 × 10 -4 ×2-pentanone; Chen Xiang's predicted value = 14.83 + 3.28 × 10 -6 × Ethyl hexanoate - 5.23 × 10 -5 ×Ethyl valerate+2.30×10 -4 ×Butyl hexanoate-1.13×10 -4 ×Ethyl heptanoate+9.31×10 -5 × acetaldehyde; Grain aroma prediction value = 8.47 + 39.1 × 10 -4 ×γ-valerolactone-3.54×10 -4 ×Butyl acetate+1.75×10 -4 ×Ethyl formate+2.75×10 -3 ×Butyl valerate; Prediction value of aroma = -1.96-2.53×10 -4 ×2-pentanol+7.14×10 -5 × Isobutanol - 1.37 × 10 -5 × acetaldehyde + ethyl hexanoate × 2.09 × 10 -6 -2.42×10 -4 × isoamyl acetate; Sweet aroma prediction value = -2.19 + 7.66 × 10 -6 × n-butanol -3.64×10 -4 ×Ethyl valerate+4.49×10 -4 ×Butyl acetate-1.46×10 -4 ×Butyl hexanoate+3.19×10 -4 × Isoamyl hexanoate + 1.14 × 10 -4 ×Ethyl heptanoate; Aldehyde aroma prediction value = -4.00-6.94×10 -5 ×2-pentanone+5.87×10 -6 × n-propanol -1.19×10 -5 ×Acetal+4.77×10 -5 × acetaldehyde + 1.09 × 10 -4 ×Ethyl formate-8.40×10 -4 ×Butyl acetate-1.99×10 -5 ×Ethyl butyrate+4.49×10 -6 × ethyl acetate; Green flavor prediction value = 4.26 + anti, anti-6.07 × 10 -2 ×2.4.-nonadienal+3.27×10 -3 × trans-2-octenal; Raw taste prediction value = 3.75 + 5.40 × 10 -3 × trans-2-octenal+4.87×10 -4 × trans-2-decenal; Soil fragrance prediction value = 6.04 + 4 - 1.47 × 10 -3 × methylphenol + 1.85 × 10 -2 ×4-ethylphenol+7.78×10 -6 × Butyric acid + 1.00 × 10 -4 ×Isovaleric acid; Burnt smell prediction value = 3.49-2.56×10 -3 ×4-methylphenol+3.61×10 -2 ×4-ethylphenol+1.11×10 -2 ×4-ethylguaiacol; Sour aroma prediction value = -6.71-3.00×10 -5 ×Butyric acid+2.09×10 -4 × Heptanoic acid +2.82×10 -4 × Caprylic acid +1.16×10 -5 × Acetic acid + 2.92 × 10 -4 × Propionic acid; Bad taste prediction value = 6.34 + 1.43 × 10 -5 × Butyric acid - 7.97 × 10 -6 × Hexanoic acid + 5.26 × 10 -4 × Caprylic Acid - 1.10×10 -5 ×Acetic acid+2.05×10 -4 × Propionic acid + 3.29 × 10 -6 ×Ethyl lactate; Prediction value of costusroot = 0.26 + 4.59 × 10 -4 ×linalool+1.57×10 -2 ×2-Thiophene carboxaldehyde+3.19×10 -4 ×4-ethylguaiacol -2.31×10 -3 × trans-2-nonenal; Prediction value of aftertaste = -0.73-3.49×10 -4 ×Valeric acid-2.45×10 -4 × Caprylic acid +3.91×10 -5 ×Butyric acid+9.79×10 -4 × Heptanoic acid + 1.01 × 10 -6 ×Ethyl lactate; (2) Taste and texture Sweetness prediction value = 75.05-1.46×10 -6 ×Ethyl lactate-2.75×10 -4 × Propionic acid + 3.10 × 10 -6 × ethyl hexanoate + 8.26 × 10 -5 ×Ethyl valerate-4.51×10 -4 ×2-pentanol+6.03×10 -6 × Acetic acid - 1.04 × 10 -3 ×Isoamyl acetate Sourness prediction value = 33.45 + 2.52 × 10 -4 × Propionic acid +6.90×10 -4 × Caprylic acid ++6.44×10 -4 × Isoamyl acetate - 4.85 × 10 -4 ×Isobutyraldehyde Bitterness prediction value = 0.54 + 1.76 × 10 -2 ×2.3.5.6-Tetramethylpyrazine+4.25×10 -7 ×Ethyl lactate+3.98×10 -3 ×2-Propionylfuran+2.31×10 -5 ×Furfural Predicted value of body = 44.04-2.88×10 -4 × Propionic acid + 1.30 × 10 -5 × ethyl hexanoate + 4.76 × 10 -4 ×2-pentanol-1.77×10 -4 ×Ethyl heptanoate+6.64×10 -4 ×Butyl Hexanoate Predicted value of thinness = 6.53 + 2.79 × 10 -4 × Propionic acid +2.13×10 -2 ×2-Propionylfuran-3.78×10 -4 ×2-pentanol+1.32×10 -4 ×Ethyl valerate Fine prediction value = 18.12-3.29×10 -6 ×Ethyl lactate-1.83×10 -4 × Propionic acid - 1.85 × 10 -3 ×Butyl acetate+9.71×10 -5 × Isobutanol - 1.64 × 10 -3 × Isoamyl acetate + 5.80 × 10 -6 × ethyl acetate + 4.89 × 10 -4 ×Isovaleraldehyde Rough prediction value = 23.02 + 6.8 × 10 -6 ×Ethyl lactate-4.99×10 -4 ×2-pentanol+8.59×10 -4 × Propyl hexanoate + 1.95 × 10 -3 × Isoamyl acetate - 9.17 × 10 -5 ×Isobutanol Soft prediction value = 55.79-6.53×10 -6 × ethyl acetate + 2.68 × 10 -4 ×Ethyl valerate+9.48×10 -4 ×2-pentanol-4.73×10 -4 ×Ethyl heptanoate+1.49×10 -2 ×2,2-diethoxyethanol-1.22×10 -3 ×Isobutyraldehyde Prediction value of dryness and spiciness = 1.33 + 1.56 × 10 -5 ×Ethyl butyrate+2.26×10 -4 ×2-pentanone-1.72×10 -5 ×Acetal+6.37×10 -5 ×Ethyl Heptanoate Long-term prediction value = 32.35-4.76×10 -6 ×Ethyl lactate+6.91×10 -6 × ethyl hexanoate + 5.83 × 10 -4 ×2-pentanol-2.48×10 -4 ×Ethyl valerate-2.75×10 -5 ×Acetal+3.16×10 -4 × Isobutanol - 1.47 × 10 -3 ×Butyl acetate+8.12×10 -5 ×Acetaldehyde Short taste prediction value = 6.15 + 2.46 × 10 -6 ×Ethyl lactate+8.11×10 -4 × Furfuryl alcohol - 7.75 × 10 -5 × isobutanol + 1.33 × 10 -3 ×Butyl acetate Prediction value of refreshing and quiet = 50.35-2.54×10 -4 × Propionic acid +7.50×10 -6 × ethyl hexanoate + 2.89 × 10 -3 ×Butyl acetate+2.15×10 -4 ×Ethyl valerate+2.24×10 -5 ×Acetal-3.62×10 -4 ×Ethyl heptanoate-1.20×10 -5 ×Ethyl acetate Prediction value of astringency = 12.87 + 7.49 × 10 -6 ×Ethyl lactate+1.54×10 -4 × Propionic acid - 1.05 × 10 -4 ×Ethyl valerate+2.05×10 -4 ×Butyl acetate+1.17×10 -4 ×Ethyl Heptanoate (3) Overall evaluation Aroma score = 6.49 + 4.12 × 10 -7 × Ethyl hexanoate - 1.47 × 10 -6 ×Acetal+4.92×10 -6 × Isobutanol - 9.87 × 10 -6 ×Ethyl heptanoate-1.43×10 -7 × Ethyl acetate - 1.43 × 10 -4 ×isobutyraldehyde+3.12×10 -6 × acetaldehyde + 1.48 × 10 -5 × ethyl formate +1.29×10 -5 ×Isovaleraldehyde+1.23×10 -5 ×Butyl hexanoate-1.74×10 -7 ×Ethyl lactate-1.19×10 -5 × Propionic acid; Taste score = 6.49 + 3.32 × 10 -7 × Ethyl hexanoate - 3.59 × 10 -5 ×Butyl acetate-5.52×10 -7 ×Acetal+6.67×10 -6 × Isobutanol - 7.83 × 10 -5 × Isoamyl acetate - 1.51 × 10 -5 × Caprylic Acid - 5.26 × 10 -6 × Propionic acid - 2.00 × 10 -7 ×Ethyl lactate; Overall evaluation = 6.28 + 2.03 × 10 -7 ×Ethyl hexanoate-8.57×10 -5 ×Butyl acetate-6.52×10 -6 ×Ethyl heptanoate+4.54×10 -6 × isobutanol + 3.22 × 10 -5 ×Butyl hexanoate-6.13×10 -7 × Hexanoic acid + 3.52 × 10 -7 × Acetic acid - 5.69 × 10 -6 × Propionic acid - 1.78 × 10 -7 ×Ethyl lactate.

10. Application of the liquor sensory quantification and quality evaluation prediction model described in any one of claims 8 or 9 in liquor production monitoring.

Images