CN-121980420-A - White spirit grade identification method, controller and equipment

CN121980420ACN 121980420 ACN121980420 ACN 121980420ACN-121980420-A

Abstract

The application discloses a white spirit grade identification method, a controller and equipment, wherein the white spirit grade identification method comprises the steps of inputting 17 key features of white spirit to be identified into a final identification model, outputting the grade of the white spirit to be identified by the final identification model, wherein the final identification model is obtained by carrying out parameter optimization on an initial identification model based on the 17 key features, the 17 key features are features which are extracted from 36 basic features and 5 derivative features of the white spirit and are more than a correlation threshold value in correlation with the white spirit grade based on correlation analysis, an arrangement importance method and a recursive feature elimination method, and the 5 derivative features are features which are generated based on the 36 basic features and are used for representing the organic acid overall content, acidity balance, wine stability, esterification potential and metal element balance of the white spirit. The white spirit grade identification method disclosed by the application can improve the objectivity and accuracy of white spirit grade identification.

Inventors

Zhu Xiajian
TIAN DONGWEI
YAN PENGCHENG
LI TINGXIA
WANG JINLONG
LU LUNWEI
LU HU

Assignees

贵州习酒股份有限公司

Dates

Publication Date: 20260505
Application Date: 20260126

Claims (10)

1. A white spirit grade identification method is characterized by comprising the following steps: inputting 17 key characteristics of white spirit to be identified into a final identification model; The final recognition model is obtained by carrying out parameter optimization on an initial recognition model based on the 17 key features, wherein the recognition model comprises a Support Vector Machine (SVM) model, a multi-layer perceptron (MLP) model, a K-nearest KNN model and a Decision Tree (DT) model, the 17 key features are features which are extracted from 36 basic features and 5 derivative features of white spirit and have association with the white spirit grade larger than an association threshold value based on correlation analysis, an arrangement importance method and a recursive feature elimination method, and the 5 derivative features are features which are generated based on the 36 basic features and are used for representing the overall content of organic acid, acidity balance, wine stability, esterification potential and metal element balance of the white spirit; The final recognition model outputs the grade of the white spirit to be recognized; The 36 basic features include acetic acid concentration, lactic acid concentration, 12 organic acid concentration, and 22 metal element concentrations, the 12 organic acids including fumaric acid, tartaric acid, glyceric acid, malic acid, maleic acid, citric acid, succinic acid, 2-hydroxybutyric acid, 2-methylsuccinic acid, 2-hydroxy-3-methylbutyric acid, 3-phenyllactic acid, and azelaic acid, the 22 metal elements including Na, mg, al, K, ca, ti, V, cr, mn, fe, co, ni, cu, zn, ga, as, se, sr, cd, sb, ba and Pb, the 5 derivative features including total organic acid concentration, acidity balance feature, total metal element concentration, esterification feature, and metal element balance feature, and the 17 key features including Na, al, K, ca, co, ti, lactic acid, acetic acid, tartaric acid, malic acid, glyceric acid, maleic acid, 2-methylsuccinic acid, 2-hydroxy-3-methylbutyric acid, and 3-phenyllactic acid concentration, total organic acid concentration, and metal element balance feature.
2. The method according to claim 1, wherein the 5 derived features are features generated based on the 36 basic features for characterizing the organic acid overall content, acidity balance, wine body stability, metal element balance and esterification potential of the white wine, comprising: The total organic acid concentration used for representing the integral content of organic acids of the white spirit is generated according to the sum of the concentration of the acetic acid, the concentration of the lactic acid and the concentration of the 12 organic acids; according to the ratio of the concentration of the acetic acid to the concentration of the lactic acid, the generated acidity balance characteristic is used for representing the acidity balance of the white spirit; According to the sum of the concentrations of the 22 metal elements, the generated total metal element concentration used for representing the stability of the white spirit body; according to the sum of the concentrations of Na, mg, al, K, ca and Fe, the esterification characteristic used for representing the esterification potential of the white spirit is generated, wherein the larger the esterification characteristic is, the larger the esterification potential of the white spirit is; and generating the metal element balance characteristic for representing the metal element balance of the white spirit according to the ratio of the sum of the Na concentration and the K concentration to the sum of the Ca concentration and the Mg concentration.
3. The method according to claim 2, wherein the 17 key features are features extracted from 36 basic features and 5 derivative features of white spirit that have a correlation with white spirit level greater than a correlation threshold using the initial recognition model based on correlation analysis, a rank importance method, and a recursive feature elimination method, comprising: The method comprises the steps of obtaining white spirit samples with the grades being A, B, C, D from high to low, wherein the number of the samples of each grade is the same; inputting 36 basic features and 5 derivative features of each sample in the training set; Based on the correlation analysis, a first feature subset is obtained, wherein the first feature subset comprises features, of the basic features and the derivative features, of which the correlation coefficient with the white spirit level is larger than a correlation threshold value; acquiring a second feature subset, a third feature subset, a fourth feature subset and a fifth feature subset based on the arrangement importance method and the initial recognition model, wherein the second to fifth feature subsets are features with importance larger than an importance threshold value in the basic features and the derivative features in the corresponding models respectively; Acquiring a sixth feature subset, a seventh feature subset, an eighth feature subset and a ninth feature subset based on a recursive feature elimination method and the initial recognition model, wherein the sixth to ninth feature subsets are features with influence performance larger than an influence threshold value in the basic features and the derivative features respectively in the corresponding models; Extracting the 17 key features with occurrence frequencies greater than a frequency threshold from the first to ninth feature subsets.
4. A method according to claim 3, wherein the final recognition model is obtained by parameter optimization of an initial recognition model based on the 17 key features, comprising: After the 17 key features are extracted, inputting the 17 key features of each sample in the training set into the initial recognition model, and acquiring a training recognition result by adopting five-fold cross validation; Under the condition that the training recognition result does not meet a result threshold, adjusting parameters of the initial recognition model until the training recognition result meets the result threshold, and determining the final recognition model; Inputting 17 key features of each sample in the test set into the final recognition model to obtain a test recognition result; And under the condition that the test identification result meets the result threshold value, acquiring the final identification model.
5. The method of claim 4, wherein in the event that the test recognition result does not meet the result threshold, continuing to perform the adjustment action until both the training recognition result and the test recognition result meet the result threshold, obtaining the final recognition model.
6. The method of any one of claims 3-5, further comprising, after extracting the 17 key features: Inputting 17 key features of each sample in the training set into the initial recognition model to obtain a training recognition result; Inputting 36 basic features of each sample in the training set into the initial recognition model to obtain a first recognition result; Inputting 36 basic features and 5 derivative features of each sample in the training set into the initial recognition model to obtain a second recognition result; Inputting the concentration of 22 metal elements of each sample in the training set into the initial recognition model to obtain a third recognition result; inputting the concentration of acetic acid, the concentration of lactic acid and the concentration of 12 organic acids of each sample in the training set into the initial recognition model to obtain a fourth recognition result; and comparing the training recognition result with the first to fourth recognition results, wherein the training recognition result is optimal.
7. The method of claim 4 or 5, further comprising, after extracting the 17 key features: Inputting 17 key features of each sample in the training set into the initial recognition model, and acquiring a training recognition result by adopting five-fold cross validation; Inputting features included in a first feature subset of each sample in the training set into the initial recognition model to obtain a fifth recognition result, wherein the first feature subset comprises the concentrations of Al, co, ti, na, V, lactic acid, acetic acid, maleic acid, citric acid, 2-methylsuccinic acid, 2-hydroxybutyric acid and azelaic acid, the total organic acid concentration and the metal element balance feature; And comparing the training recognition result with the fifth recognition result, wherein the training recognition result is optimal.
8. A controller for performing the spirit level recognition method of any one of claims 1-7.
9. A spirit level recognition device comprising a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the spirit level recognition method of any one of claims 1-7.
10. A computer readable storage medium, characterized in that a computer program is stored thereon, wherein the computer program, when executed by a processor, implements the spirit level recognition method of any one of claims 1-7.

Description

White spirit grade identification method, controller and equipment Technical Field The application relates to the technical field of white spirit, in particular to a white spirit grade identification method, a controller and equipment. Background The white spirit is divided into different grades according to the quality, and the grade of the white spirit is clear, so that the diversified requirements of different consumer groups can be met, on the one hand, the quality of the white spirit can be monitored by staff, the stability of the white spirit production process can be improved, and the like. The existing white spirit grade identification mainly depends on subjective evaluation of people, so that the white spirit grade identification result has the problems of strong subjectivity, poor objectivity, non-uniform standard and the like, and the accuracy of the white spirit grade identification result is reduced. Disclosure of Invention In view of the above, the application provides a white spirit grade identification method, a controller and a device, which can improve the objectivity and accuracy of white spirit grade identification. In order to solve the problems, the technical scheme provided by the application is as follows: The application provides a white spirit grade identification method, which comprises the steps of inputting 17 key characteristics of white spirit to be identified into a final identification model; the final recognition model is obtained by carrying out parameter optimization on an initial recognition model based on the 17 key features, wherein the recognition model comprises a support vector machine SVM model, a multi-layer perceptron MLP model, a K-nearest KNN model and a decision tree DT model, the 17 key features are features which are extracted from 36 basic features and 5 derivative features of white spirit and are more than an associated threshold value in correlation based on correlation analysis, an arrangement importance method and a recursive feature elimination method, the 5 derivative features are features which are generated based on the 36 basic features and are used for representing the overall content of organic acid, acidity balance, wine stability, esterification potential and metal element balance of white spirit, the final recognition model outputs the grade of white spirit to be recognized, the 36 basic features comprise concentration of acetic acid, concentration of lactic acid, concentration of 12 organic acids and concentration of 22 metal elements, the 12 organic acids comprise fumaric acid, tartaric acid, glyceric acid, malic acid, maleic acid, citric acid, succinic acid, 2-hydroxybutyric acid, 2-methyl-2-hydroxy succinic acid, 2-hydroxy-3-methyl butyric acid, esterification potential and metal element balance, the 12-phenyllactic acid and the 12 total concentration of 25 metal elements comprise concentration of metal elements, the total concentration of 25 metal elements and the total concentration of the metal elements to be recognized, the 17 key characteristics comprise Na, al, K, ca, co, ti, lactic acid, acetic acid, tartaric acid, malic acid, glyceric acid, maleic acid, 2-methylsuccinic acid, 2-hydroxy-3-methylbutyric acid and 3-phenyllactic acid concentration, total organic acid concentration and metal element balance characteristics. The 5 derivative features are features which are generated based on the 36 basic features and are used for representing the integral content of organic acid, acidity balance, wine stability, metal element balance and esterification potential of the white wine, and comprise the steps of generating the total organic acid concentration which is used for representing the integral content of the organic acid of the white wine according to the sum of the acetic acid concentration, the lactic acid concentration and the 12 organic acid concentrations, generating the acidity balance feature which is used for representing the acidity balance of the white wine according to the ratio of the acetic acid concentration to the lactic acid concentration, generating the total metal element concentration which is used for representing the wine body stability according to the sum of the 22 metal element concentrations, generating the esterification feature which is used for representing the esterification potential of the white wine according to the sum of the Na, the Mg, the Al, the K, the Ca and the Fe, generating the esterification potential of the white wine according to the ratio of the sum of the Na concentration and the K concentration to the sum of the Ca concentration to the ratio of the Ca concentration to the balance of the metal element concentration. The 17 key features are features which are extracted from 36 basic features and 5 derivative features of white spirit and have correlation with the white spirit grade greater than a correlation threshold by using the initial recognition model, and comprise the steps of obtainin