Method for distinguishing flavor type grade of Maotai-flavor liquor based on infrared spectrum

Abstract

The invention relates to the field of identification of flavor grades of white spirit, and particularly discloses a method for identifying flavor grades of Maotai-flavor white spirit based on infrared spectroscopy, which comprises the following steps: S1, collecting base liquor samples of Maotai-flavor white spirit with different flavor grades, and carrying out sensory evaluation on the base liquor samples of the white spirit; s2, carrying out chemical value determination on 58 substances in the baijiu base liquor sample by using a gas chromatograph; s3, carrying out spectrum collection on the base liquor sample of the white spirit by using a Fourier transform infrared spectrometer; s4, establishing a base liquor aroma type grade qualitative judgment model based on the infrared spectrum; s5, using SIMCA software to establish a partial least square discriminant analysis classification model to discriminate the aroma type grade of the base liquor; and S6, carrying out Fourier transform infrared spectrum collection on the base liquor to be graded, and carrying out rapid flavor type grade judgment on the base liquor to be graded by utilizing the models established in the step S4 and the step S5. According to the method, the prediction model is established by combining infrared spectrum with an advanced mathematical analysis method, and online real-time detection of the sample can be realized.

Description

Technical Field

[0001] The invention patent relates to the technical field of liquor flavor grade identification, specifically, to a method for identifying the flavor grade of sauce-flavor liquor based on infrared spectroscopy. Background Art

[0002] Daqu sauce-flavor liquor is made through a one-year production cycle, two feedings, nine distillations, eight fermentations, seven wine extractions, long-term storage, and careful blending. Among them, the seven wine extraction rounds respectively obtain raw sand wine, rough sand wine, big return wine (three to five rounds), small return wine, and dry dregs wine with different flavor characteristics. The wines of each round are divided into three typical flavors according to different styles: sauce-flavored, mellow and sweet, and bottom-flavored. The base wine of each flavor can be divided into superior, first-level, second-level, and third-level according to different quality. Careful blending means blending and seasoning base wines of different rounds, flavors, and grades according to product design plans, according to certain processes and proportions, to make the wine fuller and more typical in style. In this process, the company's identification basis for base wines of various flavors and grades, base wine quality management, and the professional skills of sommeliers are the key to determining the stability and quality of the final finished wine, which is directly related to the rise and fall of the company. However, the main aroma components of Maotai-flavor liquor are not as clear as those of other types of liquor, resulting in the lack of an authoritative indicator for the content of the main components in the classification and grading of Maotai-flavor liquor, and it is almost entirely dependent on manual sensory evaluation. Although sensory reviewers are professionally trained, manual sensory evaluation is easily interfered by external factors such as reviewers' experience, mental state, physical condition, and environment, making the evaluation results uncertain. There is also a serious demand contradiction due to the lack of Maotai-flavor liquor evaluation talents in Guizhou and the uneven skill levels. At the same time, the lack of classification and grading standards and the lack of information data management have led to limited product design and insufficient personalized production of products, which have become technical bottlenecks that have affected the transformation and development of the entire Maotai-flavor liquor industry.

[0003] To this end, the inventors of the present application propose a method for distinguishing the flavor grade of Maotai-flavor liquor based on infrared spectroscopy to solve the above-mentioned problems in the prior art. Summary of the invention

[0004] The purpose of the present invention is to solve the problems of the prior art raised in the above background technology and to provide a method for distinguishing the flavor grade of Maotai-flavor liquor based on infrared spectroscopy.

[0005] The above-mentioned purpose of the present invention is achieved like this:

[0006] A method for distinguishing the flavor grade of Maotai-flavor liquor based on infrared spectroscopy comprises the following steps:

[0007] S1. Collect samples of sauce-flavor liquor base wines of different flavor levels and conduct sensory evaluation on the liquor base wine samples;

[0008] S2. Use gas chromatograph to measure the chemical values ​​of 58 substances in the liquor base wine sample;

[0009] S3, using Fourier transform infrared spectrometer to collect spectrum of liquor base wine sample;

[0010] S4. Establishing a qualitative discrimination model for base wine aroma grade based on infrared spectrum;

[0011] S5. Use SIMCA software to establish a partial least squares discriminant analysis (PLS-DA) classification model to discriminate the base wine flavor grade;

[0012] S6. Collect Fourier transform infrared spectrum of the base wine to be graded, and use the model established in step S4 and step S5 to quickly determine the flavor grade of the base wine to be graded.

[0013] Furthermore, in step S1, the workshop team conducts a preliminary wine grade classification of the white wine base wine after picking the wine, and then hands it over to the enterprise, project team and expert sensory reviewers for sensory review of the white wine base wine samples based on the aroma, taste, body sensation and aroma left in the empty cup.

[0014] Furthermore, in step S2, an Agilent gas chromatograph is used to determine the chemical values ​​of 58 flavor substances; the chromatographic conditions are: an HP-INNOWAX19091N-213* chromatographic column is used. The injection port temperature is 240°C, the detector temperature is 240°C, the temperature is kept constant at 32°C for 10 minutes, the temperature is increased to 50°C at 7.15°C / min, the temperature is increased to 180°C at 8°C / min and maintained for 4 minutes, the temperature is increased to 210°C at 30°C / min and maintained for 5 minutes, the column flow rate is 0.80, split injection is used, the split ratio is 45:1, and the injection volume is 1 μL.

[0015] Furthermore, the infrared spectra in step S3 and step S6 are collected by using a Thermo Fisher near-infrared spectrometer to collect spectra of the wine samples, with a scanning range of 4000-10000cm-1, 64 scans, a resolution of 16cm-1, and each sample is collected once, and the spectrum is collected with air as the background.

[0016] Furthermore, the specific steps of establishing the base wine flavor grade qualitative discrimination model based on infrared spectrum in step S4 are:

[0017] First, select Discriminant analysis under the "Description" tab and use the Mahalanobis distance to classify the spectra;

[0018] Then enter the "Pathlength" tab, select the optical path correction method, and select Constant—no correction, that is, the optical path is constant;

[0019] Enter the "Classes" tab and enter the name of each category;

[0020] Go to the "Standards" tab, click the "open standards" button to add spectra, and select the class corresponding to each spectrum in the class column;

[0021] Enter the "Spectra" tab, set the spectrum preprocessing method, and select first derivative (to eliminate baseline drift);

[0022] Enter the "Regions" tab and select the modeling range of the spectrum;

[0023] Click the "Edit Regions" button to enter the band selection interface. After setting the band, the discrimination model is established.

[0024] Furthermore, the specific steps of step S4 are:

[0025] (1) First, data preprocessing is performed to standardize or normalize the 1*40000 spectral data so that its distribution range is [0, 1] or [-1, 1];

[0026] (2) Neural network design:

[0027] Input layer: receives 1*40000 spectral data;

[0028] Convolutional layer 1: uses 32 5-point filters followed by a ReLU activation function;

[0029] Pooling layer 1: Use maximum pooling of size 2;

[0030] Convolutional layer 2: uses 64 5-point filters followed by a ReLU activation function;

[0031] Pooling layer 2: Use maximum pooling of size 2;

[0032] Convolutional layer 3: uses 128 5-point filters followed by a ReLU activation function;

[0033] Global maximum pooling layer: extracts the maximum response of each filter and compresses the data dimension to 1*128;

[0034] Fully connected layer: output feature data with a length of 100;

[0035] (3) Feature extraction: input the training set and test set into the designed neural network to extract 1*100 feature data;

[0036] (4) Hierarchical SVM model design:

[0037] The first-level classifier: Use SVM to distinguish between "Jiangxiang Premium" and other grades of liquor;

[0038] Second-level classifier: For data that are not classified as "superior sauce-flavored sauce", SVM is used to distinguish "superior mellow and sweet sauce" from the following grades;

[0039] The third-level classifier: for the data with the second-level classification result of "mellow and sweet level 1", SVM is used to distinguish "mellow and sweet level 1" from "mellow and sweet level 2";

[0040] The fourth-level classifier: for the data whose second-level classification result is "mellow and sweet level 2", SVM is used to distinguish "mellow and sweet level 2" from "mellow and sweet level 3";

[0041] (5) Model training:

[0042] Use the training set data to train the neural network and extract features;

[0043] Use the extracted features to train a hierarchical SVM model;

[0044] (6) Model evaluation:

[0045] Use the validation set to evaluate the performance of the neural network and SVM models;

[0046] Adjust network structure and parameters to optimize performance;

[0047] (7) Prediction and classification:

[0048] Preprocess and extract features from the spectral data of new samples.

[0049] Furthermore, in step S5, the method of using the SIMCA software Auntie Jin's base wine grade qualitative discrimination model is:

[0050] Using the chemical values ​​of the 58 substances measured in step S2, use the SIMCA software to first import the chemical values ​​of the 58 substances. First, import the data, select File, then select New regular project, import the data, and click Finish;

[0051] Then establish the PLS-DA classification model, select New modelPLS-DA, select Workset, enter Set class, select Autofit, select Permutation N, and select Scores to establish the discriminant model.

[0052] Furthermore, in step S6, the base wine to be graded is quickly graded, including qualitative identification of the base wine grade, specifically:

[0053] First, the spectral information of the base wine sample to be graded is processed, and the spectral features are extracted and classified by SVM to obtain the identification results;

[0054] According to the steps of establishing the infrared spectrum qualitative model, the identification result is obtained;

[0055] Using SIMCA software, the partial least squares discriminant analysis (PLS-DA) classification and discrimination model was established to obtain the identification results.

[0056] Furthermore, the flavor grade of the base wine to be classified and graded is finally determined based on the qualitative identification results of the base wine grade using infrared spectroscopy and SIMCA software and the corresponding sensory evaluation personnel.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] 1. In the method of the present invention, the sample does not need to be pre-treated and no chemical reagents are used, which reduces the detection cost and does not pollute the environment.

[0059] 2. In the method of the present invention, the amount of sample required for detection is small, and the detection time of each sample is about 30 seconds, which can truly realize rapid, efficient and non-destructive detection.

[0060] 3. Compared with manual evaluation, the method of the present invention greatly reduces human errors and subjective factors in evaluation, and improves the accuracy and consistency of identification.

[0061] 4. The model constructed in the present invention has high accuracy and provides important support for the quality and evaluation of liquor during the production process.

[0062] 5. The method of the present invention establishes a method for quickly distinguishing the flavor grade of the base wine of Maotai-flavor liquor, overcoming the defect that the existing public technology can only identify the grade.

[0063] 6. The method of the present invention establishes a more scientific and accurate base wine flavor grade identification model of infrared spectrum + SIMCA software. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 is a flow chart of a method in an embodiment of the present invention;

[0065] Figure 2 is an infrared scanning spectrum of a single sample in an embodiment of the present invention;

[0066] Figure 3 It is a scatter plot of the partial least squares model in the embodiment of the present invention. DETAILED DESCRIPTION

[0067] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0068] The implementation of the present invention is described in detail below in conjunction with specific embodiments. For those skilled in the art, the specific meanings of the above terms can be understood according to specific circumstances.

[0069] The following are preferred embodiments of the present invention:

[0070] The embodiment of the present invention provides a method for distinguishing the flavor grade of Maotai-flavor liquor based on infrared spectroscopy. Figure 1 The flowchart of the method of the present invention specifically comprises the following steps:

[0071] Step 1: Collect samples of sauce-flavor liquor base wine of different flavor grades, and conduct sensory evaluation on the liquor base wine samples;

[0072] Step 2: Use gas chromatograph to measure the chemical values ​​of 58 substances in the liquor base wine sample;

[0073] Step 3, using a Fourier transform infrared spectrometer to collect spectra of the liquor base wine sample;

[0074] Step 4: Establish a qualitative discrimination model for base wine aroma grade based on infrared spectroscopy;

[0075] Step 5: Use SIMCA software to establish a partial least squares discriminant analysis (PLS-DA) classification model to discriminate the base wine flavor grade;

[0076] Step 6: Collect Fourier transform infrared spectrum of the base wine to be graded, and use the model established in steps 4 and 5 to quickly determine the flavor grade of the base wine to be graded.

[0077] Furthermore, in step one, the workshop team will conduct a preliminary classification of the base liquor into different grades after picking the liquor, and then hand it over to the enterprise, project team and expert sensory reviewers for sensory review of the base liquor samples based on the aroma, taste, body sensation and aroma left in the empty cup.

[0078] In this embodiment, in step 2, an Agilent gas chromatograph is used to measure the chemical values ​​of 58 flavor substances; the chromatographic conditions are: using an HP-INNOWAX19091N-213* chromatographic column. The injection port temperature is 240°C, the detector temperature is 240°C, the temperature is kept constant at 32°C for 10 minutes, the temperature is increased to 50°C at 7.15°C / min, the temperature is increased to 180°C at 8°C / min and maintained for 4 minutes, the temperature is increased to 210°C at 30°C / min and maintained for 5 minutes, the column flow rate is 0.80, split injection is used, the split ratio is 45:1, and the injection volume is 1 μL.

[0079] The infrared spectra in step 3 and step 6 are collected by using a Thermo Fisher near-infrared spectrometer to collect spectra of the wine samples (such as Figure 2 As shown), the scanning range was 4000-10000cm-1, the number of scans was 64 times, the resolution was 16cm-1, each sample was collected once, and the spectrum was collected with air as the background.

[0080] In this embodiment, the specific steps of establishing a qualitative discrimination model for base wine flavor grade based on infrared spectrum in step 4 are:

[0081] First, select Discriminant analysis under the "Description" tab and use the Mahalanobis distance to classify the spectra;

[0082] Then enter the "Pathlength" tab, select the optical path correction method, and select Constant—no correction, that is, the optical path is constant;

[0083] Enter the "Classes" tab and enter the name of each category;

[0084] Go to the "Standards" tab, click the "open standards" button to add spectra, and select the class corresponding to each spectrum in the class column;

[0085] Enter the "Spectra" tab, set the spectrum preprocessing method, and select first derivative (to eliminate baseline drift);

[0086] Enter the "Regions" tab and select the modeling range of the spectrum;

[0087] Click the "Edit Regions" button to enter the band selection interface. After setting the band, the discrimination model is established.

[0088] The specific steps of establishing the base wine aroma grade qualitative discrimination model based on infrared spectrum in the above step 4 are:

[0089] (1) First, data preprocessing is performed to standardize or normalize the 1*40000 spectral data so that its distribution range is [0, 1] or [-1, 1];

[0090] (2) Neural network design:

[0091] Input layer: receives 1*40000 spectral data;

[0092] Convolutional layer 1: uses 32 5-point filters followed by a ReLU activation function;

[0093] Pooling layer 1: Use maximum pooling of size 2;

[0094] Convolutional layer 2: uses 64 5-point filters followed by a ReLU activation function;

[0095] Pooling layer 2: Use maximum pooling of size 2;

[0096] Convolutional layer 3: uses 128 5-point filters followed by a ReLU activation function;

[0097] Global maximum pooling layer: extracts the maximum response of each filter and compresses the data dimension to 1*128;

[0098] Fully connected layer: output feature data with a length of 100;

[0099] (3) Feature extraction: input the training set and test set into the designed neural network to extract 1*100 feature data;

[0100] (4) Hierarchical SVM model design:

[0101] The first-level classifier: Use SVM to distinguish between "Jiangxiang Premium" and other grades of liquor;

[0102] Second-level classifier: For data that are not classified as "superior sauce-flavored sauce", SVM is used to distinguish "superior mellow and sweet sauce" from the following grades;

[0103] The third-level classifier: for the data with the second-level classification result of "mellow and sweet level 1", SVM is used to distinguish "mellow and sweet level 1" from "mellow and sweet level 2";

[0104] The fourth-level classifier: for the data whose second-level classification result is "mellow and sweet level 2", SVM is used to distinguish "mellow and sweet level 2" from "mellow and sweet level 3";

[0105] (5) Model training:

[0106] Use the training set data to train the neural network and extract features;

[0107] Use the extracted features to train a hierarchical SVM model;

[0108] (6) Model evaluation:

[0109] Use the validation set to evaluate the performance of the neural network and SVM models;

[0110] Adjust network structure and parameters to optimize performance;

[0111] (7) Prediction and classification:

[0112] Preprocess and extract features from the spectral data of new samples.

[0113] In this embodiment, the method of using the qualitative discrimination model of Auntie Jin's base wine grade using SIMCA software is as follows:

[0114] Using the chemical values ​​of the 58 substances measured in step 2, use the SIMCA software to first import the chemical values ​​of the 58 substances. First, import the data, select File, then select New regular project, import the data, and click Finish;

[0115] Then establish the PLS-DA classification model, select New modelPLS-DA, select Workset, enter Set class, select Autofit, select Permutation N, and select Scores to establish the discriminant model.

[0116] In this embodiment, the rapid identification and grading of the base wine to be graded includes qualitative identification of the base wine grade, specifically:

[0117] First, the spectral information of the base wine sample to be graded is processed, and the spectral features are extracted and classified by SVM to obtain the identification results;

[0118] According to the steps of establishing the infrared spectrum qualitative model, the identification result is obtained;

[0119] Using SIMCA software, the partial least squares discriminant analysis (PLS-DA) classification and discrimination model was established to obtain the identification results.

[0120] In the above method of the present embodiment, the flavor grade of the base wine to be classified and graded is finally determined by using the infrared spectrum and the qualitative identification results of the base wine grade by SIMCA software and the corresponding sensory evaluation personnel.

[0121] The following is the four-round data statistics of 58 flavor components in the embodiment of the present invention:

[0122]

[0123]

[0124]

[0125]

[0126]

[0127]

[0128]

[0129]

[0130]

[0131]

[0132] The difficulty and significance of the technical problem solved by the method of the present invention are:

[0133] In recent years, infrared spectroscopy has the advantages of good analysis reproducibility, fast sample detection, no need for chemicals, no sample loss and high accuracy. It has been widely used in the fields of food detection and agricultural product quality and safety, and can be combined with mathematical analysis methods to achieve rapid classification of samples. In the method of the present invention, infrared spectroscopy is combined with advanced mathematical analysis methods to establish a prediction model, which can achieve online real-time detection of samples; in the technical scheme of the present invention: first evaluation is adopted, spectrum and gas phase are collected, the spectrum is directly modeled by infrared and the spectrum is processed, and other software is used for classification and grading; the gas phase data is modeled and classified by SIMCA software; an identification model for aroma grade is established (aroma: sauce-flavored, mellow and sweet, grade: excellent grade one, two and three).

[0134] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for distinguishing the flavor grade of Maotai-flavor liquor based on infrared spectroscopy, characterized in that: The following steps are involved: S1. Collect samples of sauce-flavor liquor base wines of different flavor levels and conduct sensory evaluation on the liquor base wine samples; S2. Use gas chromatograph to measure the chemical values ​​of 58 substances in the liquor base wine sample; S3, using Fourier transform infrared spectrometer to collect spectrum of liquor base wine sample; S4. Establishing a qualitative discrimination model for base wine aroma grade based on infrared spectrum; S5. Use SIMCA software to establish a partial least squares discriminant analysis (PLS-DA) classification model to discriminate the base wine flavor grade; S6. Collect Fourier transform infrared spectrum of the base wine to be graded, and use the model established in step S4 and step S5 to quickly determine the flavor grade of the base wine to be graded.

2. The method for distinguishing the flavor grade of Maotai-flavor liquor based on infrared spectroscopy according to claim 1, characterized in that: In step S1, the workshop team will conduct a preliminary wine grade classification on the liquor base wine after picking the wine, and then hand it over to the enterprise, project team and expert sensory reviewers for sensory review of the liquor base wine samples based on the aroma, taste, body feeling and empty cup aroma.

3. The method for distinguishing the flavor grade of Maotai-flavor liquor based on infrared spectroscopy according to claim 1, characterized in that: In step S2, an Agilent gas chromatograph is used to determine the chemical values ​​of 58 flavor substances; the chromatographic conditions are: an HP-INNOWAX19091N-213* chromatographic column is used. The injection port temperature is 240°C, the detector temperature is 240°C, the temperature is kept constant at 32°C for 10 minutes, the temperature is increased to 50°C at 7.15°C / min, the temperature is increased to 180°C at 8°C / min and maintained for 4 minutes, the temperature is increased to 210°C at 30°C / min and maintained for 5 minutes, the column flow rate is 0.80, split injection is used, the split ratio is 45:1, and the injection volume is 1 μL.

4. The method for distinguishing the flavor grade of Maotai-flavor liquor based on infrared spectroscopy according to claim 1, characterized in that: The infrared spectra in step S3 and step S6 are collected by using a Thermo Fisher near-infrared spectrometer to collect spectra of the wine samples, with a scanning range of 4000-10000cm-1, 64 scans, a resolution of 16cm-1, and one acquisition for each sample, and the spectrum is collected with air as the background.

5. The method for distinguishing the flavor grade of Maotai-flavor liquor based on infrared spectroscopy according to claim 4 is characterized in that: The specific steps of establishing the base wine flavor grade qualitative discrimination model based on infrared spectrum in step S4 are: First, select Discriminant analysis under the "Description" tab and use the Mahalanobis distance to classify the spectra; Then enter the "Pathlength" tab, select the optical path correction method, and select Constant—no correction, that is, the optical path is constant; Enter the "Classes" tab and enter the name of each category; Go to the "Standards" tab, click the "open standards" button to add spectra, and select the class corresponding to each spectrum in the class column; Enter the "Spectra" tab, set the spectrum preprocessing method, and select firstderivative—first-order derivative spectrum (to eliminate baseline drift); Enter the "Regions" tab and select the modeling range of the spectrum; Click the "Edit Regions" button to enter the band selection interface. After setting the band, the discrimination model will be established.

6. The method for distinguishing the flavor grade of Maotai-flavor liquor based on infrared spectroscopy according to claim 5 is characterized in that: The specific steps of step S4 are: (1) First, data preprocessing is performed to standardize or normalize the 1*40000 spectral data so that its distribution range is [0, 1] or [-1, 1]; (2) Neural network design: Input layer: receives 1*40000 spectral data; Convolutional layer 1: uses 32 5-point filters followed by a ReLU activation function; Pooling layer 1: Use maximum pooling of size 2; Convolutional layer 2: uses 64 5-point filters followed by a ReLU activation function; Pooling layer 2: Use maximum pooling of size 2; Convolutional layer 3: uses 128 5-point filters followed by a ReLU activation function; Global maximum pooling layer: extracts the maximum response of each filter and compresses the data dimension to 1*128; Fully connected layer: output feature data with a length of 100; (3) Feature extraction: input the training set and test set into the designed neural network to extract 1*100 feature data; (4) Hierarchical SVM model design: The first-level classifier: Use SVM to distinguish "Jiangxiang Premium" from other grades of liquor; Second-level classifier: For data that are not classified as "Jiangxiang Premium" in the first-level classification, SVM is used to distinguish "Pure and Sweet Premium" from the following grades; The third-level classifier: for the data whose second-level classification result is "mellow and sweet level 1", SVM is used to distinguish "mellow and sweet level 1" from "mellow and sweet level 2"; The fourth-level classifier: for the data whose second-level classification result is "mellow and sweet level 2", SVM is used to distinguish "mellow and sweet level 2" from "mellow and sweet level 3"; (5) Model training: Use the training set data to train the neural network and extract features; Use the extracted features to train a hierarchical SVM model; (6) Model evaluation: Use the validation set to evaluate the performance of the neural network and SVM models; Adjust network structure and parameters to optimize performance; (7) Prediction and classification: Preprocess and extract features from the spectral data of new samples.

7. The method for distinguishing the flavor grade of Maotai-flavor liquor based on infrared spectroscopy according to claim 1, characterized in that: The method of using the qualitative discrimination model of Auntie Jin's base wine grade using the SIMCA software in step S5 is: Using the chemical values ​​of the 58 substances measured in step S2, use the SIMCA software to first import the chemical values ​​of the 58 substances. First, import the data, select File, then select New regular project, import the data, and click Finish; Then establish the PLS-DA classification model, select New modelPLS-DA, select Workset, enter Set class, select Autofit, select PermutationN, and select Scores to establish the discriminant model.

8. The method for distinguishing the flavor grade of Maotai-flavor liquor based on infrared spectroscopy according to claim 7, characterized in that: In step S6, the base wine to be graded is quickly graded, including qualitative identification of the base wine grade, specifically: First, the spectral information of the base wine sample to be graded is processed, and the spectral features are extracted and classified by SVM to obtain the identification results; According to the steps of establishing the infrared spectrum qualitative model, the identification result is obtained; Using SIMCA software, the partial least squares discriminant analysis (PLS-DA) classification and discrimination model was established to obtain the identification results.

9. The method for distinguishing the flavor grade of Maotai-flavor liquor based on infrared spectroscopy according to claim 8, characterized in that: The aroma grade of the base wine to be classified and graded is finally determined based on the qualitative identification results of the base wine grade using infrared spectroscopy and SIMCA software as well as the corresponding sensory evaluation personnel.

Images