Jingyang Fuzhuan tea year discrimination method

Abstract

The invention discloses a Jingyang Fuzhuan tea year distinguishing method, and belongs to the technical field of tea year distinguishing. Fuzhuan tea samples in different storage years are measured and analyzed by adopting a headspace solid-phase micro-extraction gas chromatograph-mass spectrometer, a Fuzhuan tea storage year predicting model is constructed according to the characteristics of volatile organic compounds and by integrating a machine learning method, and the Jingyang Fuzhuan tea year distinguishing method is used for distinguishing the Jingyang Fuzhuan tea years. And judging the storage year of the Fuzhuan tea through the prediction model. The Jingyang Fuzhuan tea year distinguishing method is simple in treatment step, convenient to operate and suitable for large-scale sample treatment and screening, and the headspace solid-phase microextraction gas chromatograph-mass spectrometer technology is stable and mature; a linear regression model and a support vector machine classification model are established by using the content values of the 10 compounds, and the specific storage year of the Fuzhuan tea can be predicted by combining an error reference value, so that the operation is simple, convenient and objective, the interference of human factors is eliminated, and the accuracy of identifying the storage year of the Fuzhuan tea can be remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of tea vintage identification technology, and in particular to a method for determining the vintage of Jingyang Fu tea. Background Technology

[0002] Fu brick tea is a type of post-fermented dark tea originating from China. It is also known as Fu brick tea and is made from raw dark tea leaves through a series of unique processing techniques. Tea aroma is one of the main sensory characteristics reflecting tea quality. Identifying the key aroma compounds in different aroma types of tea will enrich the basic theory of tea aroma chemistry and provide a scientific reference for the identification and control of tea aroma quality.

[0003] Currently, reports on the impact of storage time on tea flavor are lacking regarding Fu brick tea, an ancient fully fermented tea. Existing research primarily focuses on samples of Liubao tea, Pu'er tea, green tea, and white tea, lacking systematic tracking and data support for the dynamic evolution of Fu brick tea's characteristic flavor during long-term storage. Furthermore, how to accurately predict storage years based on these chemical fingerprints remains a blank. Summary of the Invention

[0004] The purpose of this invention is to provide a method for determining the age of Jingyang Fu tea. The processing steps are simple and convenient, suitable for the processing and screening of large-scale samples. The headspace solid phase microextraction gas chromatography-mass spectrometry (HS-SPME-MS) technology is stable and mature. The headspace solid phase microextraction process only requires less than 2g of tea sample and 3-10mL of purified water, avoiding tea sample waste. By establishing a linear regression model and a support vector machine classification model using the content values ​​of 10 compounds, and combining them with error reference values, the specific storage year of Fu tea can be predicted. The operation is simple and objective, eliminates human interference, and can significantly improve the accuracy of identifying the storage year of Fu tea.

[0005] To achieve the above objectives, the present invention provides a method for determining the age of Jingyang Fu tea, comprising the following steps: Step 1: Collect and pre-treat samples of Jingyang Fu tea from different storage years, and determine the volatile organic compounds in the samples; Step 2: Based on orthogonal partial least squares regression and support vector machine, key volatile organic compounds with VIP values ​​> 1 were screened out from the volatile organic compounds in Step 1; Step 3: Construct an annual prediction model based on the key volatile organic compounds obtained in Step 2; Step 4: Use the year prediction model obtained in Step 3 to determine the year of Fu brick tea samples with unknown storage years.

[0006] Preferably, in step one, the pretreatment method is as follows: weigh the tea powder of Jingyang Fu tea sample and place it in a 20mL headspace vial, add sodium chloride, then add 2-octanol internal standard solution, add boiling water to brew, shake and mix thoroughly to perform headspace solid phase microextraction.

[0007] Preferably, in step one, the amount of tea powder of Jingyang Fucha sample weighed is 0.5-2g, the amount of sodium chloride added is 0.5-2g, the amount of 2-octanol internal standard solution used is 5-10μm, the concentration is 1mg / mL, and the amount of boiling water used is 3-10mL.

[0008] Preferably, in step one, the headspace solid-phase microextraction conditions are as follows: the extraction head aging temperature is 250-300℃, the aging time is 3-10 min, the heating chamber temperature is 80-100℃, the heating time is 30-60 min, the suction time is 30-60 min, and the desorption is performed at 240-250℃ for 3-10 min.

[0009] Preferably, in step one, the GC parameters are as follows: HP-5MS column, 30m×0.25mm×μm; carrier gas He, purity >99.999%, flow rate 1.8mL / min; splitless injection mode; injection port temperature 240℃; temperature program is: initial temperature 40℃ held for 3-10min, increased to 170℃ at 2℃ / min held for 3-10min, increased to 240℃ at 8℃ / min held for 3-10min; MS conditions were: electron ion source temperature 200℃, EI ionization energy 70eV; mass scan range m / z 30-500; and interface temperature 220℃.

[0010] Preferably, in step two, when screening key volatile organic compounds, qualitative analysis is performed by searching the NIST spectral library and calculating the retention index, and quantitative analysis is performed by using the internal standard method. In the quantitative analysis, 2-octanol is used as an internal standard to perform relative quantification of the volatile compound content. Orthogonal partial least squares discriminant analysis is performed using SIMCA 14.1 software to screen out key volatile organic compounds with VIP values ​​> 1 as variables in the sample data.

[0011] Preferably, in step three, the top 10 key volatile organic compounds with VIP values ​​> 1 are used as modeling variables to obtain the prediction model, as shown in equation (I): Equation (Ⅰ); Wherein, R1 is the content of methyl salicylate, R2 is the content of 3-hexen-1-ol, R3 is the content of (1R,2R,3S,5R)-(-)-2,3-pinenediol, R4 is the content of 6-methyl-5-hepten-2-one, R5 is the content of phenethyl alcohol, R6 is the content of α-terpineol, R7 is the content of (Z)-3,7-dimethyl-2,6-octadien-1-ol, R8 is the content of benzyl alcohol, R9 is the content of 1-octen-3-ol, and R10 is the content of trans-β-ionone.

[0012] Therefore, the present invention adopts the above-mentioned method for determining the age of Jingyang Fu tea. The processing steps are simple and convenient, and it is suitable for the processing and screening of large-scale samples. Moreover, the headspace solid phase microextraction gas chromatography-mass spectrometry (HS-SPME-MS) technology is stable and mature. The headspace solid phase microextraction process only requires less than 2g of tea sample and 3-10mL of purified water, avoiding tea sample waste. By using the content values ​​of 10 compounds to establish a linear regression model and a support vector machine classification model, combined with the error reference value, the specific storage year of Fu tea can be predicted. The operation is simple and objective, eliminates the interference of human factors, and can significantly improve the accuracy of identifying the storage year of Fu tea.

[0013] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0014] Figure 1 This is a heat map of volatile organic compounds during the fermentation process of this invention; Figure 2 This is a scatter plot of the predicted and actual storage years using leave-one-out cross-validation and external validation in the PLS regression model of this invention. Figure 3 This is a performance comparison chart of the prediction models constructed in this invention; Figure 4 This is the classification result of the confusion matrix of this invention; Figure 5 This invention presents multi-category ROC curves and micro-average curves for tea with different storage years and AUC values. Detailed Implementation

[0015] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0016] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0017] Example 1 This invention provides a method for determining the age of Jingyang Fu brick tea, the specific steps of which are as follows: 1. The headspace sampling method for the samples was as follows: A whole piece of Jingyang Fu brick tea (all tea samples were finished Jingyang Fu brick tea from the Pudao brand, from 2016, 2017, 2018, 2019, 2020, 2021, 2022, 2023, and 2024) was taken out. Samples were evenly taken from the four corners and the center of the tea brick using the five-point sampling method. The samples were then evenly ground and passed through an 80-mesh sieve to obtain tea powder. 1g of tea powder was weighed and placed in a 20mL headspace vial. 1g of sodium chloride was added, along with 5μL of 1mg / mL 2-octanol internal standard solution. 5mL of boiling water was added, and the mixture was thoroughly shaken and mixed.

[0018] 2. The experimental instruments used were a headspace solid-phase microextraction (HS-MS) gas chromatography-mass spectrometry (GC-MS) system (5977B-7890B-sniffer 9100 instrument, Agilent Technologies, USA), a chromatographic column (HP-5MS, Agilent Technologies, USA), and a three-phase extraction head (CAR×DVB×PDMS, 50 / 30μm, Sigma-Aldrich, Shanghai, China). The headspace solid-phase microextraction conditions were as follows: aging temperature of the extraction head (CAR×DVB×PDMS, 50 / 30μm) was 270℃ for 5 min; the heating oven temperature was 80℃ for 60 min; the suction time was 45 min; and the analysis was performed at 240℃ for 5 min. GC parameters were as follows: HP-5MS column (30m × 0.25mm × μm); carrier gas He (purity > 99.999%); flow rate 1.8mL / min; splitless injection mode; injector temperature 240℃; temperature program: initial temperature 40℃, hold for 4 min, increase to 170℃ at 2℃ / min, hold for 5 min, increase to 240℃ at 8℃ / min, hold for 2 min. MS conditions were as follows: electron ion source temperature 200℃, EI ionization energy 70eV; mass scan range m / z 30-500; adapter temperature 220℃.

[0019] 3. Qualitative analysis was performed using NIST spectral library search combined with retention index calculation. All experiments were repeated three times, and data are expressed as mean ± standard deviation. For quantitative analysis, the internal standard method was used, with 2-octanol as the internal standard for relative quantification of volatile compound content. Hierarchical clustering was performed using Python (version 3.9.13), and cluster heatmaps were generated; see [link to Python documentation]. Figure 1 Cluster analysis was performed on metabolite data from 2016 to 2024 (9 time points), resulting in a heatmap containing 5 clusters (C1-C5). Partial least second length discriminant analysis was performed using SIMCA 14.1 software to screen key volatile organic compounds with VIP values ​​> 1 as variables in the sample data, as shown in Table 1.

[0020] Table 1. Values ​​of VIP>1 for volatile compounds during fermentation.

[0021] 4. Based on the original sample data, the content values ​​of the top 10 volatile compounds with VIP > 1 were taken as characteristic variables, including R1: Methyl salicylate, R2: 3-Hexen-1-ol, R3: (1R,2R,3S,5R)-(-)-2,3-Pinanediol, R4: 5-Hepten-2-one, 6-methyl-, R5: Phenylethyl Alcohol, R6: α-Terpineol, R7: 2,6-Octadien-1-ol, 3,7-dimethyl-,(Z)-, R8: Benzyl The data were preprocessed using pandas and numpy, with the following parameters: alcohol (benzyl alcohol), R9: 1-Octen-3-ol, and R10: trans-β-Ionone. Tea samples were divided into training and test sets in a 7:3 ratio, with a fixed random seed used to ensure reproducibility. In the modeling phase, pipeline techniques were employed to integrate data standardization and PLS regression, using five principal components to extract feature information.

[0022] The validation process employs a dual strategy: First, LOOCV cross-validation is performed on 18 samples from the training set, with one sample used for validation at a time while the rest are used for training, thus preventing data leakage. Then, external validation is performed on 9 samples from the independent test set to comprehensively evaluate the model's generalization ability. After model training, the model formula is as follows: ; Wherein, R1 is the content of methyl salicylate, R2 is the content of 3-hexen-1-ol, R3 is the content of (1R,2R,3S,5R)-(-)-2,3-pinenediol, R4 is the content of 6-methyl-5-hepten-2-one, R5 is the content of phenethyl alcohol, R6 is the content of α-terpineol, R7 is the content of (Z)-3,7-dimethyl-2,6-octadien-1-ol, R8 is the content of benzyl alcohol, R9 is the content of 1-octen-3-ol, and R10 is the content of trans-β-ionone.

[0023] The results analysis phase provides multi-dimensional evaluation indicators, including R. 2 The coefficient of determination and root mean square error of RMSE are calculated, and a prediction scatter plot is generated simultaneously. See [link / reference]. Figure 2 The R of the LOOCV training set 2=0.815, RMSE=1.23. Most of the 18 samples (blue dots) on the training set are closely distributed around the ideal black line, with only a few points deviating significantly. This indicates that the model fits the training data well and effectively captures the linear relationship between features and storage years. The R-value on the external test set is... 2 =0.767, RMSE=0.91. The distribution of the 9 samples (red triangles) on the test set also closely follows the black ideal line with no significant deviation, indicating that the model has excellent generalization ability. Specific information on the actual and predicted storage years of Jingyang Fucha samples in the training and test sets is shown in Table 2. The prediction error for the vast majority of samples is within ±2 years, with only a few long-term stored samples showing slightly larger errors, which do not exceed the range of their respective categories.

[0024] Table 2. Predicted and actual storage times of Jingyang Fucha samples in the training and test sets.

[0025] 5. Using the same data preprocessing method as in step 4, classify the tea samples by year into three levels (short-term storage, medium-term storage, and long-term storage). StandardScaler is used to standardize the original features using Z-scores. Then, a dual feature selection strategy using F-test and random forest is employed to screen key features (R5 and R9) and generate an enhanced feature set. In the modeling phase, a multi-model comparison framework is constructed, including SVM (radial basis function kernel and linear kernel), logistic regression, random forest, and Gaussian Naive Bayes. See [link to relevant documentation]. Figure 3 Among them, SVM (linear kernel) and logistic regression models performed best. Hyperparameter grid search was performed to optimize the models, and the performance of all models was tested on the original feature set and the enhanced feature set respectively. Finally, the SVM (linear kernel) model was selected as the best model.

[0026] Model training and validation employed a 7:3 hierarchical splitting strategy, using macro-average AUC and accuracy as core evaluation metrics. Comprehensive performance analysis was conducted using confusion matrix and multi-class ROC curves. (See attached...) Figure 4 and Figure 5 Samples stored for 1-3 years and 7-9 years were accurately classified. Only one sample stored for 4-6 years was misclassified as 1-3 years. Eight out of nine samples were correctly classified, and the accuracy of the confusion matrix was 88.9%. The model performed well overall, with a micro-average AUC of 94.4% on the ROC curve. It could perfectly classify samples stored for a long time. In samples stored for a short and medium period, the model misclassified some samples, but the micro-average AUC was still high, which reflects the excellent performance of the model. The results of both tests jointly verified the stability of the model performance.

[0027] Therefore, the present invention adopts the above-mentioned method for determining the age of Jingyang Fu tea. The processing steps are simple and convenient, and it is suitable for the processing and screening of large-scale samples. Moreover, the headspace solid phase microextraction gas chromatography-mass spectrometry (HS-SPME-MS) technology is stable and mature. The headspace solid phase microextraction process only requires less than 2g of tea sample and 3-10mL of purified water, avoiding tea sample waste. By using the content values ​​of 10 compounds to establish a linear regression model and a support vector machine classification model, combined with the error reference value, the specific storage year of Fu tea can be predicted. The operation is simple and objective, eliminates the interference of human factors, and can significantly improve the accuracy of identifying the storage year of Fu tea.

[0028] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for determining the age of Jingyang Fuzhuan tea, characterized in that: Includes the following steps: Step 1: Collect and pre-treat samples of Jingyang Fu tea from different storage years, and determine the volatile organic compounds in the samples; Step 2: Based on orthogonal partial least squares regression and support vector machine, key volatile organic compounds with VIP values ​​> 1 were screened out from the volatile organic compounds in Step 1; Step 3: Construct an annual prediction model based on the key volatile organic compounds obtained in Step 2; Step 4: Use the year prediction model obtained in Step 3 to determine the year of Fu brick tea samples with unknown storage years.

2. The method for determining the age of Jingyang Fuzhuan tea according to claim 1, characterized in that: In step one, the pretreatment method is as follows: weigh the tea powder of Jingyang Fu tea sample and place it in a 20mL headspace vial, add sodium chloride, then add 2-octanol internal standard solution, add boiling water to brew, shake and mix thoroughly to perform headspace solid phase microextraction.

3. The method for determining the age of Jingyang Fuzhuan tea according to claim 1, characterized in that: In step one, the amount of tea powder for Jingyang Fucha sample is 0.5-2g, the amount of sodium chloride added is 0.5-2g, the amount of 2-octanol internal standard solution used is 5-10μm with a concentration of 1mg / mL, and the amount of boiling water used is 3-10mL.

4. The method for determining the age of Jingyang Fuzhuan tea according to claim 1, characterized in that: In step one, the headspace solid-phase microextraction conditions are as follows: the extraction head aging temperature is 250-300℃, the aging time is 3-10 min, the heating chamber temperature is 80-100℃, the heating time is 30-60 min, the suction time is 30-60 min, and the desorption is performed at 240-250℃ for 3-10 min.

5. The method for determining the age of Jingyang Fuzhuan tea according to claim 1, characterized in that: In step one, the GC parameters were as follows: HP-5MS column, 30m × 0.25mm × μm; carrier gas He, purity > 99.999%, flow rate 1.8mL / min; splitless injection mode; injection port temperature 240℃; temperature program: initial temperature 40℃ held for 3-10min, increased to 170℃ at 2℃ / min held for 3-10min, increased to 240℃ at 8℃ / min held for 3-10min. MS conditions were: electron ion source temperature 200℃, EI ionization energy 70eV; mass scan range m / z 30-500; and interface temperature 220℃.

6. The method for determining the age of Jingyang Fuzhuan tea according to claim 1, characterized in that: In step two, qualitative analysis was performed by searching the NIST spectral library and calculating the retention index when screening key volatile organic compounds. Quantitative analysis was performed using the internal standard method, with 2-octanol as the internal standard to perform relative quantification of the volatile compound content. Orthogonal partial least squares discriminant analysis was performed using SIMCA 14.1 software to screen key volatile organic compounds with VIP values ​​> 1 as variables in the sample data.

7. The method for determining the age of Jingyang Fuzhuan tea according to claim 1, characterized in that: In step three, the top 10 key volatile organic compounds with VIP values ​​> 1 are used as modeling variables to obtain the prediction model, as shown in equation (I): Equation (Ⅰ); Wherein, R1 is the content of methyl salicylate, R2 is the content of 3-hexen-1-ol, R3 is the content of (1R,2R,3S,5R)-(-)-2,3-pinenediol, R4 is the content of 6-methyl-5-hepten-2-one, R5 is the content of phenethyl alcohol, R6 is the content of α-terpineol, R7 is the content of (Z)-3,7-dimethyl-2,6-octadien-1-ol, R8 is the content of benzyl alcohol, R9 is the content of 1-octen-3-ol, and R10 is the content of trans-β-ionone.

Images