Prediction method and equipment for year of Maotai-flavor liquor, medium and product

Abstract

The invention discloses a method, equipment, medium and product for predicting the year of Maotai-flavor liquor, and relates to the technical field of chromatography and spectrum, chemometrics and artificial intelligence analys.The method comprises the steps that a chromatographic separation technology is utilized, chromatographic data and spectrum data of a Maotai-flavor liquor sample to be detected are collected, the chromatographic data are preprocessed, and the year of the Maotai-flavor liquor sample to be detected is obtained; determining the preprocessed data; decomposing and processing the three-dimensional fluorescence spectrum data based on a high-order cyclic factor algorithm, and determining a decomposition result; taking the processed data as four trained Bayesian network models, and outputting four year prediction results of the Maotai-flavor liquor; and taking the median of the four year prediction results of the Maotai-flavor liquor as the final year prediction result of the Maotai-flavor liquor sample to be detected. According to the method, the year prediction accuracy and prediction efficiency of the Maotai-flavor liquor can be improved.

Description

Technical Field

[0001] This application relates to the fields of chromatography and spectroscopy, chemometrics, and artificial intelligence analysis technology, and in particular to a method, equipment, medium, and product for predicting the vintage of Maotai-flavor liquor. Background Art

[0002] High-quality Maotai-flavor baijiu possesses a pleasant, aged aroma, a characteristic solely created by time. The "vintage" of Maotai-flavor baijiu refers to its aging period. Starting from the first distillation of the first batch of liquor after the Double Ninth Festival, a one-year brewing cycle is followed by seven distillations of base liquor from the fermented mash. After at least three years of aging, blending and flavoring are performed. Once the quality is deemed acceptable, it is aged for another six months before packaging and release. In total, the process from distillation to release requires at least five years. The freshly distilled base liquor has a strong and pungent aroma and flavor, which becomes mellow and smooth after long-term aging, with the Maotai-flavor aroma becoming more prominent. The longer the aging, the smoother the body and the more elegant the aroma of the Maotai-flavor baijiu.

[0003] Currently, the more scientific methods for identifying the age of Maotai-flavor liquor mainly include: intelligent sensory analysis, which uses artificial intelligence such as electronic noses or electronic tongues to collect fingerprint spectra of Maotai-flavor liquor, extract characteristic peaks of the liquor sample for data analysis. However, the flavor of Maotai-flavor liquor is complex and variable, and the differences between Maotai-flavor liquors of different years are subtle, requiring extremely high precision in the analytical model. Therefore, determining the age of Maotai-flavor liquor requires a high-precision model; gas chromatography or liquid chromatography analysis, which analyzes the aroma substances in the liquor and establishes a database for identification or prediction. When using this method to determine the age of Maotai-flavor liquor, although it can accurately analyze the chemical components in the liquor, it is difficult to directly and accurately reflect the year information; mass spectrometry, which quantitatively analyzes the different spectra formed by the mass-charge ratio arrangement of charged particles. This method can be used in combination with other technologies, such as gas chromatography-mass spectrometry fingerprinting. Mass spectrometry has advantages in determining the age of Maotai-flavor liquor in terms of volatile component analysis, but the chemical component database for determining the age of Maotai-flavor liquor is not yet perfect. The current vintage prediction results for Maotai-flavor liquor show low accuracy, which directly restricts the feasibility and practicality of vintage identification for Maotai-flavor liquor, posing numerous challenges and difficulties. Summary of the Invention

[0004] The purpose of this application is to provide a method, equipment, medium, and product for predicting the vintage of Maotai-flavor liquor, which can solve the problems of inaccurate and inefficient vintage prediction in related technologies.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] Firstly, this application provides a method for predicting the vintage of Maotai-flavor liquor, including:

[0007] Chromatographic and spectral data of the Maotai-flavor liquor sample to be tested were collected; the spectral data included three-dimensional fluorescence spectral data; the chromatographic data were preprocessed to determine the preprocessed data; the chromatographic data included gas chromatography data, gas chromatography-mass spectrometry data, and ion chromatography data; the three-dimensional fluorescence spectral data were processed based on a high-order cyclic factor decomposition algorithm to determine the decomposition result; the preprocessed data and the decomposition result were used as processed data, and four Maotai-flavor liquor year prediction results were output based on the processed data and four trained Bayesian network models; the four trained Bayesian network models were constructed based on gas chromatography data, gas chromatography-mass spectrometry data, ion chromatography data, and three-dimensional fluorescence spectral data, respectively; the median of the four Maotai-flavor liquor year prediction results was used as the final year prediction result for the Maotai-flavor liquor sample to be tested.

[0008] Secondly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for predicting the vintage of Maotai-flavor liquor as described above.

[0009] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for predicting the vintage of Maotai-flavor liquor as described above.

[0010] Fourthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the method for predicting the vintage of Maotai-flavor liquor as described above.

[0011] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0012] This application provides a method, equipment, medium, and product for predicting the vintage of Maotai-flavor liquor. By preprocessing gas chromatography, gas chromatography-mass spectrometry, and ion chromatography data of Maotai-flavor liquor samples, the accuracy and consistency of the preprocessed data are ensured. Furthermore, three-dimensional fluorescence detection technology is used to determine unique three-dimensional fluorescence spectral data for Maotai-flavor liquor from different years. A high-order cyclic factor decomposition algorithm is used to factorize the three-dimensional fluorescence spectral data. Compared with traditional methods, the high-order cyclic factor decomposition algorithm exhibits higher robustness and reproducibility when handling complex samples containing unknown and interfering components, effectively avoiding the occurrence of imaginary solutions, thus providing a solid foundation for the accuracy of Maotai-flavor liquor vintage prediction. Finally, this application utilizes four trained Bayesian network models, based on real Maotai-flavor liquor samples from different years, to achieve a fast and efficient workflow from data processing to result output, accurately predicting the vintage of Maotai-flavor liquor. This integrated technical system not only improves the scientific rigor and reliability of vintage identification but also enhances the efficiency of Maotai-flavor liquor vintage detection and prediction. Attached Figure Description

[0013] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0014] Figure 1 This is a flowchart illustrating a method for predicting the vintage of a type of Maotai-flavor liquor provided in an embodiment of this application.

[0015] Figure 2 A flowchart outlining the entire process of identifying the vintage of Maotai-flavor liquor;

[0016] Figure 3 The image shows the gas chromatographic data of a sample of Maotai-flavor liquor.

[0017] Figure 4 The images show the chromatograms of a sauce-flavored baijiu sample before and after gas chromatography data preprocessing; among them, Figure 4 (a) in the figure is the chromatogram of gas chromatography data before preprocessing; Figure 4 (b) in the figure is the chromatogram after gas chromatography data preprocessing;

[0018] Figure 5 The GC-MS total ion current TIC chromatogram of a Maotai-flavor liquor sample;

[0019] Figure 6 The images show the chromatograms of a sauce-aroma baijiu sample before and after gas chromatography-mass spectrometry data preprocessing; among them... Figure 6(a) in the figure is the spectrum before gas chromatography-mass spectrometry data preprocessing; Figure 6 (b) in the figure is the spectrum after gas chromatography-mass spectrometry data preprocessing;

[0020] Figure 7 The ion chromatography data of the Maotai-flavor liquor sample are shown.

[0021] Figure 8 The images show the chromatograms of a Maotai-flavor liquor sample before and after ion chromatography data preprocessing; among them, Figure 8 (a) in the figure is the chromatogram before ion chromatography data preprocessing; Figure 8 (b) in the figure is the chromatogram after preprocessing of ion chromatography data;

[0022] Figure 9 The images show the three-dimensional fluorescence spectra of a Maotai-flavor liquor sample before and after preprocessing; among them, Figure 9 (a) in the image is the spectrum of the three-dimensional fluorescence spectroscopy data before preprocessing; Figure 9 (b) in the figure is the spectrum after preprocessing of the three-dimensional fluorescence spectroscopy data;

[0023] Figure 10 The image shows the three-dimensional fluorescence spectrum of a sample of Maotai-flavor liquor; among which, Figure 10 (a) in the figure is a schematic diagram of the three-dimensional fluorescence spectrum of the Maotai-flavor liquor sample; Figure 10 (b) in the figure is a schematic diagram of the three-dimensional fluorescence factor decomposition of the Maotai-flavor liquor sample;

[0024] Figure 11 The three-dimensional fluorescence spectral data of the Maotai-flavor liquor sample are factor decomposition component spectra; among them, Figure 11 (a) in the figure is the excitation spectrum of each component; Figure 11 (b) in the figure shows the emission spectra of each component; Figure 11 (c) in the figure represents the concentration response signal of each component; Figure 11 (d) in the graph represents the trend of residual changes;

[0025] Figure 12 The three-dimensional fluorescence spectral data of Maotai-flavor liquor samples from different years are shown in the excitation-emission (EX / EM) spectra of each component after factorization; among them, Figure 12 (a) in the figure is the excitation-emission (EX / EM) spectrum corresponding to component 1; where Figure 12 (b) in the figure is the excitation-emission (EX / EM) spectrum corresponding to component 2; where Figure 12 (c) in the figure is the excitation-emission (EX / EM) spectrum corresponding to component 3; Figure 12 (d) in the figure is the excitation-emission (EX / EM) spectrum corresponding to component 4; Figure 12(e) in the figure is the excitation-emission (EX / EM) spectrum corresponding to component 5; Figure 12 (f) in the figure is the excitation-emission (EX / EM) spectrum corresponding to component 6. Detailed Implementation

[0026] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0028] Example 1:

[0029] Maotai-flavor baijiu (using Daqu) uses glutinous sorghum as its core ingredient and high-temperature Daqu as its saccharification and fermentation agent. It undergoes multiple processes including solid-state fermentation, distillation, aging, blending, packaging, and storage. No substances not produced during the fermentation process are added. Its unique "12987" brewing process—a one-year production cycle, two rounds of feeding, nine rounds of steaming, eight rounds of fermentation, and seven rounds of distillation—creates its distinctive flavor characteristics. Among various aged baijiu, Maotai-flavor aged baijiu holds an important position in the high-end market due to its superior quality and is highly sought after by consumers, especially aged Maotai-flavor baijiu. The aging and storage time determines the maturity of Maotai-flavor liquor. When assessing the age (vintage), three stages should be considered: first, the storage time in containers (earthenware jars) after the base liquor from the newly distilled batches is stored; second, the aging time after blending and flavoring the base liquor from the aging batches (generally calculated as the weighted average of the main base liquors used); and third, the storage time in the original packaging bottles (jars) after the finished liquor is bottled. With the continued rise in market demand for aged Maotai-flavor liquor (aged Maotai-flavor liquor), the importance of vintage identification is increasingly prominent. Accurate vintage identification can effectively protect consumers' legitimate rights and interests, preventing them from purchasing counterfeit vintage liquor at high prices due to information asymmetry; it also helps to regulate market order, promote the healthy and orderly development of the Maotai-flavor liquor industry, and allow consumers to more truly appreciate its profound quality and cultural value.

[0030] In the field of vintage (age) identification and prediction of Maotai-flavor liquor, various data analysis techniques are commonly used. Among them, Partial Least Squares Discriminant Analysis (PLS-DA) is widely applied to the processing of chemical composition data for Maotai-flavor liquor of known vintages. This method first automatically fits and generates a Principal Component Analysis (PCA) score map, filters information based on validity indicators, and then constructs a PLS-DA model to achieve accurate vintage identification of Maotai-flavor liquor. Principal Component Analysis can reduce the dimensionality of various data in the liquor sample, deeply analyze the distribution characteristics of liquor samples from different years in the principal component space, and can be used to distinguish key features of different years. Random Forest algorithm uses a large amount of feature data such as the rich flavor substance content and physicochemical indicators of the liquor sample to build a prediction model. Support Vector Machine (SVM) can handle high-dimensional data and complex nonlinear relationships, and can efficiently classify and predict liquor sample data from different years, demonstrating significant value and potential in this field.

[0031] The Bayesian algorithm employed in this application analyzes existing data on the components of baijiu and past experience in identifying the vintage of Maotai-flavor baijiu, making the vintage identification more reasonable. The algorithm's dynamic updating capability allows it to update the posterior probability distribution in a timely manner based on Bayes' theorem as new baijiu sample data is collected, ensuring the identification results constantly reflect changes in the actual situation and guaranteeing accuracy. Furthermore, its probabilistic presentation of results perfectly aligns with the inherent uncertainty in identifying the vintage of Maotai-flavor baijiu, providing a more comprehensive reflection of the probabilities of each year and facilitating users' judgment based on probability. By combining these advantages with other technologies, not only can intelligent data mining and processing be achieved, but the efficiency of Maotai-flavor baijiu vintage identification can also be balanced with improved accuracy, enabling precise vintage identification of Maotai-flavor baijiu and providing a solid technical foundation for verifying the authenticity of Maotai-flavor baijiu vintages.

[0032] like Figure 1 and Figure 2 As shown, this application provides a method for predicting the vintage of Maotai-flavor liquor, including steps 101-105.

[0033] Step 101: The sample of Maotai-flavor liquor to be tested is processed using three-dimensional fluorescence detection technology to determine the spectral data; the spectral data includes the three-dimensional fluorescence spectral data.

[0034] Among them, three-dimensional fluorescence detection technology was used to process the sample of the Maotai-flavor liquor to be tested and to determine the spectral data;

[0035] Step 102: Preprocess the chromatographic data to determine the preprocessed data; the chromatographic data includes gas chromatography data, gas chromatography-mass spectrometry data, and ion chromatography data.

[0036] Specifically, the sample of the Maotai-flavor liquor to be tested is processed using gas chromatography data detection technology to determine gas chromatography data; the sample of the Maotai-flavor liquor to be tested is processed using gas chromatography-mass spectrometry (GC-MS) detection technology to determine GC-MS data; and the sample of the Maotai-flavor liquor to be tested is processed using ion chromatography data detection technology to determine ion chromatography data.

[0037] Step 103: Based on the high-order cyclic factor decomposition algorithm, process the three-dimensional fluorescence spectral data to determine the decomposition results; based on internal standard correction, process the gas chromatography data, gas chromatography-mass spectrometry data, and ion chromatography data to determine the processed data results.

[0038] In some embodiments, before step 103, the method further includes: background subtraction of the three-dimensional fluorescence spectral data.

[0039] Step 104: Use the preprocessed data and the decomposition results as processing data, and output four prediction results for the vintage of Maotai-flavor liquor based on the processing data and four trained Bayesian network models; the four trained Bayesian network models are respectively constructed based on historical gas chromatography data, historical gas chromatography-mass spectrometry data, historical ion chromatography data and historical three-dimensional fluorescence spectroscopy data.

[0040] Step 105: Take the median of the four predicted vintages of the Maotai-flavor liquor as the final vintage prediction result of the Maotai-flavor liquor sample to be tested.

[0041] Specifically, gas chromatography (GC), gas chromatography-mass spectrometry (GC-MS), ion chromatography (IC), and three-dimensional fluorescence spectroscopy (EEMs) combined with Bayesian algorithms are used to analyze volatile aroma components, organic acids, anions, and characteristic fluorescent compounds in baijiu, thereby identifying the vintage of sauce-flavored baijiu.

[0042] By collecting gas chromatography, gas chromatography-mass spectrometry, and ion chromatography data of Maotai-flavor liquor from different years, meticulous data preprocessing was performed using Modellab software. This included peak alignment, retention time correction, and baseline correction. This process denoised the original signals, eliminating instrument electronic noise and environmental interference, resulting in clearer chromatographic peaks. Simultaneously, retention times were calibrated to prevent variations in retention time from affecting component identification, ensuring comparability between different batches of samples and guaranteeing data accuracy and consistency. Furthermore, advanced chemometric machine learning algorithms based on Bayesian networks were combined to construct an artificial intelligence model capable of accurately identifying the vintage of Maotai-flavor liquor. In addition, this application employed three-dimensional fluorescence detection technology to create unique three-dimensional fluorescence spectra for Maotai-flavor liquor from different years. The HILOFAC algorithm was used for spectral factorization. Compared to traditional methods, the HILOFAC algorithm demonstrated higher robustness and reproducibility when handling complex samples containing unknown and interfering components, effectively avoiding imaginary solutions and providing a solid foundation for the vintage identification of Maotai-flavor liquor.

[0043] This application's solution not only integrates advanced analytical techniques and artificial intelligence machine learning models to construct an automated and intelligent model training and prediction mechanism, but also, relying on real sample datasets, achieves a rapid and efficient process from data collection to result output, accurately predicting the vintage of Maotai-flavor liquor. This integrated technical system not only improves the scientific rigor and reliability of vintage identification, overcoming the technical challenge of rapid vintage determination for Maotai-flavor liquor, but also brings a brand-new intelligent quality evaluation tool to the Maotai-flavor liquor industry and even the entire food testing field, providing an intelligent solution with broad application value and market prospects for the testing field. With its accuracy, efficiency, objectivity, intelligence, and standardization, this technical system provides strong technical support for the vintage identification of Maotai-flavor liquor, laying the foundation for the standardized development of the industry and the healthy and orderly progress of the market. At the same time, it provides consumers with reliable quality assurance and scientific purchasing guidance, enhancing consumer confidence and promoting the continued prosperity of the market.

[0044] In some embodiments, step 101 specifically includes: performing chromatographic peak alignment and retention time correction processing on the chromatographic data based on the multiple internal standard method to determine the retention time; converting the retention time into a retention index, and determining the corrected gas chromatographic data, corrected gas chromatographic-mass spectrometry data, and corrected ion chromatographic data based on the retention index; and using the corrected gas chromatographic data, the corrected gas chromatographic-mass spectrometry data, and the corrected ion chromatographic data as the preprocessed data.

[0045] A Bayesian network model was established based on gas chromatography data, gas chromatography-mass spectrometry data, ion chromatography data, and three-dimensional fluorescence spectroscopy data, combined with chemometrics and machine learning algorithms, to identify the vintage of Maotai-flavor liquor. Specifically, this included:

[0046] 1. Data collection, the process is as follows.

[0047] (1) Gas chromatographic data: The parameters of the gas chromatograph are set according to the characteristics of the sauce-flavored liquor sample. The sample signal is recorded by the detector and the gas chromatographic data is collected.

[0048] (2) Gas Chromatography-Mass Spectrometry Data: Based on the characteristics of the sauce-flavored liquor sample, the parameters of the gas chromatography-mass spectrometry data exchanger are set. The sample is ionized with the help of the ion source, and ions with different mass-to-charge ratios are separated by the mass analyzer. The mass spectrometry signal of the sample is accurately recorded by the detector, and gas chromatography-mass spectrometry data is collected.

[0049] (3) Ion chromatography data: Based on the characteristics of the sauce-flavored liquor sample, adjust the parameters of the ion chromatography data instrument, use a high-sensitivity detector to accurately capture and record the sample signal, and collect ion chromatography data.

[0050] (4) Three-dimensional fluorescence spectral data: Take a sample of Maotai-flavor liquor to make a mother liquor. Based on the actual detection spectrum, select a suitable dilution concentration, appropriate excitation wavelength and emission wavelength range and bandwidth, and scan to collect three-dimensional fluorescence spectral data.

[0051] 2. The process of chromatographic data preprocessing is as follows.

[0052] Data preprocessing is performed in each solution, including baseline correction and internal standard correction. When performing internal standard correction, an internal standard with similar properties to the target analyte, good separation performance, and known concentration is selected. The internal standard method formula is used to calculate the correction, thereby accurately correcting the chromatographic data to obtain preprocessed data and improving analytical accuracy.

[0053] (1) Import multichromatographic data.

[0054] Gas chromatography data, gas chromatography-mass spectrometry data, and ion chromatography data were converted into CDF standard text format, imported into ModelLab Chroman software, and then a separate solution was created for each data type within the software.

[0055] (2) Internal standard correction.

[0056] By performing specific operations on the independent variable (i.e., the characteristic compound), such as using filtering and smoothing methods, the peak area is normalized to eliminate errors caused by fluctuations in instrument status.

[0057] 3. Three-dimensional fluorescence spectroscopy data processing.

[0058] (1) Data preprocessing.

[0059] The obtained three-dimensional fluorescence data was converted into a standard CSV file and imported into ModelLab Specman 3D software. Background subtraction was performed on the three-dimensional fluorescence spectral data to eliminate first- and second-order Rayleigh and Raman scattering.

[0060] (2) Correction and analysis of the three-dimensional fluorescence spectrum using the Hierarchical and Local Factor Analysis Combined (HILOFAC) algorithm. The three-dimensional fluorescence spectrum is three-dimensional fluorescence spectral data.

[0061] a. The trilinear component model or higher-order linear component model used for higher-order correction is as follows.

[0062] Three-dimensional fluorescence spectroscopy data exhibits high-dimensional collinearity. It consists of multiple samples, each described by an I×J×K data array. The high-dimensional matrix composed of all its components can be represented in the following form, i.e., a trilinear component model or a higher-order linear component model.

[0063]

[0064] In three-dimensional fluorescence spectroscopy data, x ijk Let be a matrix element in X(I×J×K), corresponding to the fluorescence intensity at k for the i-th excitation wavelength and the j-th emission wavelength; N is the number of columns in the loading matrix, i.e., the number of factors; a in b jn c kn These are elements in the excitation matrix A, emission matrix B, and relative concentration matrix C, respectively; e ijk Let X be the element of a high-dimensional residual array E, with a size of I×J×K. This high-order linear component model includes three parts: known (target) analytes, unknown (or uncorrected) analytes, and background interference (noise). Here, X(I×J×K) represents the fluorescence intensity at K for the (I-th excitation wavelength and the J-th emission wavelength), and n represents the number of analytes.

[0065] b. HILOFAC algorithm modeling.

[0066] To fully utilize the excitation matrix A, emission matrix B, and relative concentration matrix C for accurate factorization, the linear decomposition calculation formula of the above trilinear component model using Singular Value Decomposition (SVD) is as follows.

[0067] FM(:,:,n)=A(:,n)*C(k,n)*B(:,n) T

[0068] a (i) T =((C + X i.. B). / (B T B)+(B + X i.. C). / (C T C))

[0069] c (k) T = (B + X ..k A). / (A T A)+(A + X ..k B). / (B T B))

[0070] b (j) T =(A + X .j. C). / (C T C)+(C + X .j. A). / (A T A))

[0071] In the formula, FM k In the (I×J×N) three-dimensional matrix, k represents a sample, referring to the index of all rows in the current matrix, meaning all rows participate in the calculation. Unlike previous parallel factorization algorithms, to avoid the appearance of imaginary solutions and sensitivity to the number of factors, this algorithm introduces tensor singular value decomposition to perform high-order cyclic decomposition on the trilinear data, making a... (i) b (j) c (k) The loss function can be minimized through iterations. Generally, after a finite number of iterations (usually <10), the decomposition result satisfying the condition for the residuals can be obtained. The decomposition results of matrices A, B, and C after factorization using the above formula can be directly used in subsequent machine learning processes, such as qualitative discriminant and quantitative regression analysis. Here, FM(:,;,n) is used in the formula to calculate the equation related to the excitation matrix A, emission matrix B, and relative concentration matrix C. The excitation matrix A is related to parameters obtained by a specific formula. The relative concentration matrix C is obtained by calculating the relevant parameters using a specific formula. The parameters related to the emission matrix B are obtained by calculating using a specific formula.

[0072] In some embodiments, step 104 specifically includes: analyzing the corrected gas chromatography data, the corrected gas chromatography-mass spectrometry data, the corrected ion chromatography data, and the three-dimensional fluorescence spectral decomposition results respectively to determine the characteristic compounds corresponding to the sample of Maotai-flavor liquor to be tested; the characteristic compounds characterize the relationship between different types of compounds in Maotai-flavor liquor; the characteristic compounds include a first characteristic compound, a second characteristic compound, a third characteristic compound, and a fourth characteristic compound; the first characteristic compound, the second characteristic compound, the third characteristic compound, and the fourth characteristic compound are used as four types of samples; the four types of samples are respectively input into the corresponding trained Bayesian network models, and the probabilities of Maotai-flavor liquor corresponding to each year for each type of sample are output; the year corresponding to the highest probability among the probabilities of Maotai-flavor liquor in each year for each type of sample is selected as the predicted year of Maotai-flavor liquor for each type of sample; the predicted year of Maotai-flavor liquor for the four types of samples is used as the four predicted year of Maotai-flavor liquor.

[0073] In some embodiments, the four types of samples are input into the corresponding trained Bayesian network models, and the probability of each type of sample corresponding to each year of Maotai-flavor liquor is output. Specifically, this includes: for any type of sample, calculating the prior probability of that type of sample in each year of Maotai-flavor liquor; evaluating the probability of each compound in that type of sample according to a Gaussian normal distribution; calculating the conditional probability of that type of sample in each year of Maotai-flavor liquor based on the probability of each compound in that type of sample; determining the posterior probability of each year of Maotai-flavor liquor based on the conditional probability and the prior probability; and using the posterior probability of each year of Maotai-flavor liquor as the probability of that type of sample corresponding to each year of Maotai-flavor liquor.

[0074] In some embodiments, steps 201-203 are included before step 104.

[0075] Step 201: Use the historical chromatographic data and historical spectral data of Maotai-flavor liquor samples from different years as four types of historical samples to construct four Bayesian network models.

[0076] Step 202: Based on the four types of historical samples and the four Bayesian network models, determine the predicted year values ​​for the four types of Maotai-flavor liquor.

[0077] Step 203: Based on the predicted vintage values ​​of the four types of Maotai-flavor liquor and the K-fold cross-validation method, determine four trained Bayesian network models.

[0078] In some embodiments, step 203 specifically includes: for any type of Maotai-flavor liquor's predicted year, using K-fold cross-validation, evaluating the predicted year value for that type of Maotai-flavor liquor to determine the mean and variance of the predicted year value; training a Bayesian network model corresponding to the predicted year value of the type of Maotai-flavor liquor with the objective of minimizing the mean and variance of the predicted year value of the type of Maotai-flavor liquor, determining the target Bayesian network model corresponding to the predicted year value of the type of Maotai-flavor liquor; obtaining the target predicted year values ​​for the four types of Maotai-flavor liquor output by the four target Bayesian network models, and using the median of the target predicted year values ​​for the four types of Maotai-flavor liquor as... The target prediction result of the vintage of Maotai-flavor liquor; determine whether the difference between the target prediction result of the vintage of Maotai-flavor liquor and the actual vintage of Maotai-flavor liquor is less than or equal to a preset value, and determine the first judgment result; if the first judgment result is yes, determine the four target Bayesian network models as four trained Bayesian network models; if the first judgment result is no, with the goal of minimizing the deviation between the target prediction result of the vintage of Maotai-flavor liquor and the actual vintage of Maotai-flavor liquor, and based on the K-fold cross-validation method, train the four target Bayesian network models, and determine the four trained target Bayesian network models as the four trained Bayesian network models.

[0079] In practical applications, the preprocessed gas chromatography data, gas chromatography-mass spectrometry data, ion chromatography data, and three-dimensional fluorescence spectroscopy data are divided into training, validation, and test sets, respectively. A Bayesian network chemometrics algorithm is used for model training.

[0080] The core of the Bayesian network model is Bayes' theorem, which describes the probability of a hypothesis being true given certain evidence. According to Bayes' theorem, the classification C of a Maotai-flavor liquor sample can be predicted by analyzing the content X of its compounds. k .

[0081] Given: X = {x1, x2, x3, ..., x} m} and C k ={c1,c2,c3,……c k}

[0082] Where X represents the characteristic compound (i.e., the independent variable), which includes multiple different types of compounds; m represents different compounds; and k represents the year of production for Maotai-flavor liquor. The calculation process involves the following steps:

[0083] (a) First, calculate the prior probability P(C) of each type of Maotai-flavor liquor. k Assume there are M samples of Maotai-flavor liquor, where M k Each sample belongs to category C. k The prior probability of Maotai-flavor liquor is:

[0084] P(C k ) = M k / M

[0085] (II) Calculate the conditional probability P(X / C) of Maotai-flavor liquor under each category. k According to Naive Bayes' theorem, C k The conditional probability of compound X in soy sauce-flavored baijiu under the given conditions is:

[0086]

[0087] (iii) Evaluate the probability of each compound based on the Gaussian normal distribution.

[0088]

[0089] Where, x i It is the content of compound i, μ ki It is the mean, σ ki It is the standard deviation.

[0090] (iv) Calculate the posterior probability by combining the prior probability and the conditional probability, which is the probability of the content of the soy sauce aroma type of liquor compound X the classification of soy sauce aroma type of liquor.

[0091] P(C k |X)=P(C k )·P(X|C k )

[0092] (v) Finally, the year category corresponding to the maximum posterior probability is the classification of Maotai-flavor liquor.

[0093]

[0094] Calculate the probability P(X|C) of X under each category. k The probability of finding the category with the highest probability is the year of the sauce-flavored baijiu.

[0095] The above process uses preprocessed chromatographic data and preprocessed three-dimensional fluorescence spectra as four types of samples to train a Bayesian network model, establishing the dependencies and conditional probability distributions between compounds. The Bayesian network model is evaluated based on the prediction deviation of the Maotai-flavor liquor sample year. Further, based on the model evaluation results, key parameters that significantly affect the performance of the Maotai-flavor liquor year prediction model are identified, and the optimal parameter settings are found. This mainly involves estimating the mean and variance. Specifically, in the Bayesian network model, K-fold cross-validation is used to evaluate the model's accuracy and parameter reliability. Based on the mean and variance of the predicted Maotai-flavor liquor year for each sample type determined after evaluation, the Bayesian network model with the best performance is selected as the final Maotai-flavor liquor year prediction model. A new dataset is prepared in the test set for predicting Maotai-flavor liquor years. The four Bayesian network models in this application provide four independent liquor year prediction results. Outliers (pred) are removed using the following formula, and the median is calculated as the final predicted year for Maotai-flavor liquor.

[0096]

[0097] Here, Q1 and Q3 represent the 1st and 3rd quartiles, respectively, and 1.5IQR represents 1.5 times the interquartile range. The interquartile range (IQR) is the difference between Q3 and Q1, representing the middle 50% range of the data. IQR is commonly used to measure the dispersion of data and to detect outliers.

[0098] This application has the following beneficial effects:

[0099] (1) The three-dimensional fluorescence detection method used in this application has the advantages of simple sample preprocessing, fast detection speed, high sensitivity and low cost, and can realize the rapid detection of three-dimensional fluorescence data of sauce-flavored liquor.

[0100] (2) This application realizes a method for identifying the age of Maotai-flavor liquor by three-dimensional fluorescence detection based on the HILOFAC algorithm with strong anti-interference ability. It can extract the pure signal of the three-dimensional fluorescence spectrum in the presence of Rayleigh scattering, peak overlap, multiple impurity peaks and other unknown interferences, simplifying the preprocessing steps, saving experimental time and cost, and realizing real-time, non-destructive and green analysis.

[0101] (3) This application realizes fingerprint spectrum of sauce-flavored liquor of different years based on gas chromatography data, gas chromatography-mass spectrometry data, ion chromatography data and three-dimensional fluorescence technology, and realizes the detection and prediction technology for sauce-flavored liquor of different years.

[0102] (4) This application realizes the construction of a smart rapid detection model for the age of Maotai-flavor liquor based on the combination of gas chromatography data detection technology, ion chromatography data detection technology, gas chromatography-mass spectrometry data technology, three-dimensional fluorescence spectrum technology and Bayesian network algorithm. This reduces the interference of environmental factors such as background, reduces the complexity of the model, improves the detection and identification accuracy, and realizes the rapid identification of the age of Maotai-flavor liquor.

[0103] (5) This application realizes the automation and standardization of the three-dimensional fluorescence spectrum data analysis and modeling process based on the rapid detection of the age of Maotai-flavor liquor. Through the designed data preprocessing, model training, model optimization and sample prediction methods, efficient processing of Maotai-flavor liquor age data is achieved, which significantly improves the accuracy of the age identification of Maotai-flavor liquor on the existing basis.

[0104] (6) Innovatively, this application uses robust statistical methods to calculate the IQR and median of the vintage prediction results of sauce-flavored liquor obtained based on four different instrumental analysis methods, thereby eliminating the influence of errors between different analysis methods and obtaining reliable statistical results.

[0105] Example 2:

[0106] I. Gas chromatography data combined with chemometrics for the identification of vintage in Maotai-flavor liquor.

[0107] Sample solution preparation: Take 990 μL of the wine sample to be tested in a 2.0 mL injection bottle, add 10 μL of the three-component internal standard solution (2.0% v / v tert-amyl alcohol, n-amyl acetate, 2-ethylbutyric acid 50% ethanol solution), cap and seal, shake and mix well, let stand to stabilize and equilibrate, and then perform GC analysis.

[0108] Gas chromatograph settings: GC injector temperature 250℃; FID detector temperature 250℃; carrier gas control mode linear velocity, carrier gas linear velocity 12.6 cm / L from 0.0 min to 5.0 min (20.8 cm / L from 5.0 min to 85.0 min); split injection, split ratio 30:1; injection volume 1.0 μL; capillary column DB-WaxUI (60 m × 0.25 mm × 0.25 μm); column oven temperature program: 30℃ (hold for 12 min), increase to 75℃ at 5.0℃ / min (hold for 1.0 min), then increase to 100℃ at 3.0℃ / min (hold for 2.0 min), then increase to 160℃ at 2.0℃ / min, then increase to 180℃ at 5.0℃ / min, then increase to 240℃ at 10.0℃ / min (hold for 13.0 min).

[0109] The data was imported in batches into ModelLab chemometrics modeling software, preprocessed, and then used for modeling and analysis.

[0110] Reference Figure 3 and Figure 4 It was observed that the retention times of each compound drifted, and the spectra corrected for retention times yielded "aligned" spectra, which ensured the accuracy of the analytical results.

[0111] Tert-amyl alcohol (internal standard), n-amyl acetate (internal standard), and 2-ethylbutyric acid (internal standard) were selected as internal standards. The retention time was converted into the retention index by the multiple internal standard method, and the retention time of other compounds contained in Maotai-flavor liquor was corrected to make the retention time more accurate and reduce the error caused by inaccurate measurement.

[0112] The acquired data was divided into a test set and a validation set. The test set contained 120 data samples, while the validation set consisted of 12 data samples. A Naive Bayes algorithm was used to construct a model for predicting the vintage of Maotai-flavor liquor. To evaluate the model's performance, K-fold cross-validation was used to analyze the results. The model's quality was evaluated based on accuracy, precision, and recall (all should reach above 95%), and the modeling parameters were adjusted to avoid overfitting.

[0113] The dataset of unknown years was used to predict the year of Maotai-flavor liquor. The year was predicted by gas chromatographic data of collected Maotai-flavor liquor samples, and the results are shown in Table 1.

[0114] By combining gas chromatography data with a Bayesian network model, the vintage of Maotai-flavor liquor can be identified and predicted. The deviation range for vintage prediction of Maotai-flavor liquor is 0 to 1 year.

[0115] Table 1. GC Prediction Results of Vintage-Specific Maotai-flavor Baijiu

[0116]

[0117]

[0118] II. Gas chromatography-mass spectrometry data combined with chemometrics for the identification of vintage in Maotai-flavor liquor.

[0119] Sample solution preparation: Weigh 1.8 g of chromatographically pure NaCl (baked at high temperature) into a 20 mL headspace vial, transfer 4.0 mL of ultrapure water and 1.0 mL of the wine sample to be tested, and accurately add 2.5 μL of a three-component internal standard solution (10 μg / mL 4-bromofluorobenzene, 1,2-dichlorobenzene-[D4], acenaphthene-[D10]) in methanol solution, seal the vial, vortex and mix well, and then run the solution on a GCMS instrument for analysis.

[0120] Solid-phase microextraction settings for AOC6000SPMEArrow injector: SmartSPMEArrow extraction head DVB / CarbonWR / PDMS 1.10mm, aging temperature 260℃, aging time (before / after extraction) 3min, equilibration time 5min, extraction temperature 50℃, extraction time 15min, desorption time at the injection port 2min.

[0121] Gas chromatography front-end settings: GC injection port temperature 250℃; constant linear velocity control mode, carrier gas linear velocity 25.5cm / s; split injection split ratio 5:1; injection volume 1.0μL; capillary column SH-PolarWax (60m×0.25mm×0.25μm); column oven temperature program: 40℃ (hold for 5 min), then increase to 250℃ at 3.0℃ / min (hold for 15 min).

[0122] MS mass spectrometry electron impact ion source (EI) settings: electron capability 70 Ev, ion source temperature 250℃, interface temperature 250℃, Q3Scan scan range 35 m / z~400 m / z.

[0123] The data was imported in batches into ModelLab chemometrics modeling software, preprocessed, and then used for modeling and analysis.

[0124] Reference Figure 5 and Figure 6 Retention time correction and peak alignment were performed on the gas chromatography-mass spectrometry (GC-MS) data. Then, the mass spectrometry data was deconvolved, with an ion peak mass-to-charge ratio (MTR) matching tolerance of 0.8 Da. In mass spectrometry, MTR measures the mass-to-charge ratio of an ion in Da. The peak intensity threshold was set to 1000; the peak detection slope was set to 200; the XIC (Extracted Ion Chromatogram) smoothing level was set to 5 (moderate); and the peak signal-to-noise ratio factor was set to 5. Using these deconvolution parameters, potential co-elution peaks in the total ion current data were effectively separated and analyzed.

[0125] Samples of Maotai-flavor liquor from different years were collected and preprocessed before mass spectrometry analysis to obtain their mass spectrometry data. The acquired data were divided into a test set and a validation set. The test set contained 120 data samples, and the validation set consisted of 12 data samples. A Bayesian algorithm was used to construct a prediction model, and K-fold cross-validation was used to cross-validate the model. The quality of the model was evaluated and optimized based on accuracy, precision, and recall (all should reach above 95%), ultimately establishing a data model for liquor vintages.

[0126] Data sets of unknown years were used to predict the year of Maotai-flavor liquor. The year was predicted by gas chromatography-mass spectrometry data of collected Maotai-flavor liquor samples, and the results are shown in Table 2.

[0127] The vintage of Maotai-flavor liquor was predicted using gas chromatography-mass spectrometry data combined with a Bayesian neural network model. The predicted vintage for this Maotai-flavor liquor had a margin of error of 0–1 year.

[0128] Table 2. GC-MS Prediction Results of Vintage-Specific Maotai-flavor Baijiu

[0129] sample Actual vintage of sauce-flavored baijiu GC-MS data prediction year Deviation / Year 1 3 3 0.0 2 4 4 0.0 3 7 8 1.0 4 9 9 0.0 5 12 12 0.0 6 13 13 0.0 7 15 15 0.0 8 16 16 0.0 9 17 17 0.0 10 18 18 0.0 11 20 21 1.0 12 24 24 0.0

[0130] III. Ion chromatography data combined with chemometrics are used for the vintage identification of Maotai-flavor liquor.

[0131] Sample solution preparation: Accurately transfer 9.8 mL of secondary deionized water into a 20 mL glass test tube using a 10.0 mL dispenser, accurately add 200 μL of the wine sample to be tested, and then accurately add 20 μL of internal standard solution (2‰ v / v 2-ethylbutyric acid 50% ethanol solution). Seal the tube with a sealing film, shake well, and after stabilization and equilibrium, filter the solution through a 0.22 μm needle filter into the injection tube. Then, run the solution on an IC for analysis.

[0132] High-performance ion chromatography (HPLC) setup: Dionex IonPac AS11-HC4 × 250 mm anion column; EGC 500 KOH eluent generator; AERS 4 mm self-circulating suppressor with a suppression current of 104 mA; flow rate: multi-step gradient type, 0–40 min flow rate 1.0 mL / min, 40–47 min flow rate 0.8 mL / min, 47–50 min flow rate 1.0 mL / min; injection volume 25.0 μL, dilution factor 50.0; detector: conductivity detector, detection cell temperature 35 °C.

[0133] Import the data in batches into ModelLab chemometrics modeling software, and perform modeling analysis after data preprocessing.

[0134] Reference Figure 7 and Figure 8 If chromatographic peaks are not aligned, the ion chromatography data will seriously affect the data quality. Therefore, retention time correction must be performed on the chromatogram.

[0135] The acquired data set consists of 120 data samples in the test set and 12 data samples in the validation set. A Bayesian algorithm was used to construct a vintage prediction model for Maotai-flavor liquor. K-fold cross-validation was used. The quality of the model was evaluated by accuracy, precision, and recall (all exceeding 95%), and parameters were adjusted to avoid overfitting.

[0136] Based on data from different years of Maotai-flavor liquor, a Bayesian network algorithm model was used to predict the year of Maotai-flavor liquor. The prediction results for the Maotai-flavor liquor samples are shown in Table 3. The prediction deviation range for the samples is 0 to 1 year, with a minimum deviation of 0 years.

[0137] Table 3. Prediction Results of IC for Vintage Maotai-flavor Baijiu

[0138] sample Actual vintage of sauce-flavored baijiu IC Data Forecast Year Deviation / Year 1 3 3 0.0 2 4 4 0.0 3 7 7 0.0 4 9 9 0.0 5 12 12 0.0 6 13 12 1.0 7 15 15 0.0 8 16 17 1.0 9 17 17 0.0 10 18 18 0.0 11 20 20 0.0 12 24 23 1.0

[0139] IV. Three-dimensional fluorescence combined with chemometrics for the identification of vintage in Maotai-flavor liquor.

[0140] Sample solution preparation: Weigh the sample to be tested into a volumetric flask, add methanol to the mark, and centrifuge ultrasonically to obtain the stock solution for each sample. Dilute the stock solution with methanol separately before testing. Select the most suitable dilution concentration for the reference standard and sample based on the actual spectral response for detection.

[0141] Three-dimensional fluorescence spectrometer settings: excitation wavelength λex: 200~600nm, bandwidth 5nm; emission wavelength λem: 200~600nm, bandwidth 5nm; photomultiplier diode (PMT) voltage: 600V; cuvette size 1cm, sample temperature 25℃.

[0142] Import the data in batches into ModelLab chemometrics modeling software, and perform modeling analysis after data preprocessing.

[0143] Reference Figure 9 and Figure 10 Background subtraction is performed on the three-dimensional fluorescence spectral data. This can be achieved through scattering subtraction (Rayleigh scattering) and interpolation algorithms (piecewise Hemite spline interpolation).

[0144] Reference Figure 11 and Figure 12 Factor decomposition was performed on the three-dimensional fluorescence spectral data.

[0145] The test set contains 120 data samples, and the validation set contains 12 data samples. A Bayesian algorithm is used to construct a vintage prediction model for Maotai-flavor liquor, namely a Bayesian network model. K-fold cross-validation is employed, and the model quality is evaluated through accuracy, precision, and recall (all should reach above 95%). Parameters are adjusted to prevent overfitting.

[0146] The dataset of unknown years was used to predict the year of Maotai-flavor liquor. The three-dimensional fluorescence data of the collected Maotai-flavor liquor samples were used to predict the year, and the results are shown in Table 4.

[0147] This study uses three-dimensional fluorescence data combined with a Bayesian neural network model to predict the vintage of Maotai-flavor liquor. The predicted vintage for Maotai-flavor liquor has a margin of error of 0–1 year.

[0148] Table 4. EEMs Prediction Results of Maotai-flavor Baijiu by Year

[0149] sample Actual vintage of sauce-flavored baijiu EEMs ​​data forecast year Deviation / Year 1 3 3 0.0 2 4 4 0.0 3 7 7 0.0 4 9 8 1.0 5 12 12 0.0 6 13 13 0.0 7 15 15 0.0 8 16 16 0.0 9 17 17 0.0 10 18 18 0.0 11 20 21 1.0 12 24 24 0.0

[0150] A Bayesian network model was used to predict the year of Maotai-flavor liquor samples. The results are shown in Table 5. The year prediction was performed using a Bayesian network algorithm based on gas chromatography data, gas chromatography-mass spectrometry data, ion chromatography data, and three-dimensional fluorescence data. The error range for the predicted year of Maotai-flavor liquor was 0–1 year, with a maximum deviation of 1 year. The median of the year prediction results from the four methods was calculated to overcome the insufficient sensitivity and accuracy of single detection methods, thus obtaining objective year prediction results. The maximum deviation of the final corrected year prediction was 0.5 years.

[0151] Table 5. Comprehensive Judgment Analysis of Vintage Prediction Results for Maotai-flavor Baijiu

[0152]

[0153] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the methods described above.

[0154] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the methods described above.

[0155] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the methods described above.

Claims

1. A method for predicting the vintage of a sauce-flavored baijiu, characterized in that, include: Chromatographic and spectral data of the Maotai-flavor liquor sample to be tested were collected; the spectral data included three-dimensional fluorescence spectral data. The chromatographic data is preprocessed to determine the preprocessed data; the chromatographic data includes gas chromatographic data, gas chromatographic-mass spectrometry data, and ion chromatographic data. The three-dimensional fluorescence spectral data are processed based on a high-order cyclic factor decomposition algorithm to determine the decomposition results. The preprocessed data and the decomposition results are used as processing data, and based on the processing data and four trained Bayesian network models, four prediction results for the vintage of Maotai-flavor liquor are output; the four trained Bayesian network models are respectively constructed based on gas chromatography data, gas chromatography-mass spectrometry data, ion chromatography data and three-dimensional fluorescence spectroscopy data; The median of the four predicted vintages of the Maotai-flavor liquor was taken as the final vintage prediction result for the Maotai-flavor liquor sample to be tested.

2. The method for predicting the vintage of Maotai-flavor liquor according to claim 1, characterized in that, The chromatographic data is preprocessed to determine the preprocessed data, specifically including: Based on the multiple internal standard method, the chromatographic data were subjected to peak alignment and retention time correction to determine the retention time; The retention time is converted into a retention index, and based on the retention index, the corrected gas chromatography data, corrected gas chromatography-mass spectrometry data, and corrected ion chromatography data are determined. The corrected gas chromatography data, the corrected gas chromatography-mass spectrometry data, and the corrected ion chromatography data are used as preprocessed data.

3. The method for predicting the vintage of Maotai-flavor liquor according to claim 2, characterized in that, The preprocessed data and the decomposition results are used as processing data. Based on the processing data and four trained Bayesian network models, four prediction results for the vintage of Maotai-flavor liquor are output, specifically including: The corrected gas chromatography data, the corrected gas chromatography-mass spectrometry data, the corrected ion chromatography data, and the decomposition results are analyzed respectively to determine the characteristic compounds corresponding to the tested Maotai-flavor liquor sample; the characteristic compounds characterize the relationship between different types of compounds in Maotai-flavor liquor; the characteristic compounds include a first characteristic compound, a second characteristic compound, a third characteristic compound, and a fourth characteristic compound; The first characteristic compound, the second characteristic compound, the third characteristic compound, and the fourth characteristic compound are regarded as four types of samples; The four types of samples are input into the corresponding trained Bayesian network models, and the probability of each type of data sample corresponding to each year of Maotai-flavor liquor is output. The year with the highest probability among the probabilities of Maotai-flavor liquor in each year of each sample category is selected as the predicted year of Maotai-flavor liquor for each sample category. The predicted year of Maotai-flavor liquor corresponding to the four types of samples is used as the predicted year of Maotai-flavor liquor.

4. The method for predicting the vintage of Maotai-flavor liquor according to claim 3, characterized in that, The four types of samples are input into the corresponding trained Bayesian network models, and the output is the probability of each type of sample corresponding to a certain year of Maotai-flavor liquor, specifically including: For any type of sample, calculate the prior probability of that type of sample in each year of Maotai-flavor liquor; The probability of each compound in this type of sample is assessed based on the Gaussian normal distribution; Based on the probability of each compound in this type of sample, calculate the conditional probability of this type of sample in each year of Maotai-flavor liquor. Based on the conditional probability and the prior probability, determine the posterior probability of each year of Maotai-flavor liquor; The posterior probability of each year of Maotai-flavor liquor is used as the probability of Maotai-flavor liquor corresponding to each year for that type of sample.

5. The method for predicting the vintage of Maotai-flavor liquor according to claim 1, characterized in that, Before outputting the four predicted vintages of Maotai-flavor liquor based on the processed data and four trained Bayesian network models, the following steps are also included: Gas chromatography data, gas chromatography-mass spectrometry data, ion chromatography data, and three-dimensional fluorescence spectroscopy data of Maotai-flavor liquor samples from different years were used as four types of samples to construct four Bayesian network models. Based on the four types of historical samples and the four Bayesian network models, the predicted values ​​of the four types of Maotai-flavor liquor years are determined; Based on the predicted vintage values ​​of four types of Maotai-flavor liquor and the K-fold cross-validation method, four trained Bayesian network models were determined.

6. The method for predicting the vintage of Maotai-flavor liquor according to claim 5, characterized in that, Based on the predicted vintage values ​​of four types of Maotai-flavor liquor and the K-fold cross-validation method, four pre-trained Bayesian network models were determined, specifically including: For any type of Maotai-flavor liquor, the predicted year value is evaluated using the K-fold cross-validation method, and the mean and variance of the predicted year value are determined. With the goal of minimizing the mean and variance of the predicted year values ​​of the soy sauce-flavored liquor, a Bayesian network model corresponding to the predicted year values ​​of the soy sauce-flavored liquor is trained, and the target Bayesian network model corresponding to the predicted year values ​​of the soy sauce-flavored liquor is determined. Obtain the target predicted values ​​of four types of Maotai-flavor liquor years from the output of four target Bayesian network models, and take the median of the target predicted values ​​of the four types of Maotai-flavor liquor years as the target prediction result of Maotai-flavor liquor years; Determine whether the difference between the target predicted year of the sauce-flavored liquor and the actual year of the sauce-flavored liquor is less than or equal to a preset value, and determine the first judgment result; If the first judgment result is yes, then the four target Bayesian network models are determined to be four trained Bayesian network models; If the first judgment result is negative, the objective is to minimize the deviation between the target prediction result of the sauce-flavored liquor year and the actual year of the sauce-flavored liquor. Based on the K-fold cross-validation method, four target Bayesian network models are trained, and the four trained target Bayesian network models are determined to be the four trained Bayesian network models.

7. The method for predicting the vintage of Maotai-flavor liquor according to claim 1, characterized in that, Before determining the decomposition results, the process based on the high-order cyclic factor decomposition algorithm for the three-dimensional fluorescence spectral data further includes: Background subtraction is performed on the three-dimensional fluorescence spectral data.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the method for predicting the vintage of Maotai-flavor liquor according to any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the method for predicting the vintage of Maotai-flavor liquor as described in any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the method for predicting the vintage of Maotai-flavor liquor as described in any one of claims 1-7.

Images