Digital Evaluation Of Tea Infusion Quality Based On Stoichiometry

Tea is a unique flavor beverage.Tea quality has a decisive role in the taste.However,with a wide range of tea categories and different characteristics,it is difficult to correctly identify the quality of tea.Traditionally,tea quality evaluation relies on human sensory evaluation and chemical composition detection.The former requires strict requirements on reviewers,environmental conditions,equipment and procedures,and the later needs more instruments,reagents,manpower and time.They are accurate and authoritative,but time-consuming and labor-intensive,which cannot meet the requirements of quality detection nowdays.It is urgent to find a simpler,faster and more accurate evaluation method.Tea infusion can be drunk directly,which is the most important part of tea and the way for effective ingredients to enter human body.Thus,this study took tea infusion as research object.Black tea and green tea were selected as samples.Characteristics signals of tea infusion were collected by electronic tongue,color difference,ultraviolet spectrum and image recognition techniques,combining with different analysis methods to digitally evaluated the quality of tea infusion.The main research contents and results were as follows:(1)The sensory evaluation score was closely related to tea grade.The content of chemical components varied greatly among different kinds of tea.And the content of chemical components of different grades of tea varied from tea to tea.There were correlations between sensory evaluation score and the content of some chemical components.(2)The sensor response values of different grades of Dianhong Congou black tea and Kemmun black tea infusion were collected separately using an electronic tongue and combined with analysis methods to differentiate the types and grades of tea.The results showed that principal component analysis and discriminant analysis could simultaneously identify the types and grades of black tea,and the discriminant analysis was more effective.In the discriminant analysis scatter plot,tea samples were arranged regularly according to grades without overlapping.The accuracy of discriminant analysis cross validation was 98.20%.The accuracy of BP neural network for black tea grade recognition was 95.00%.(3)The color difference values and ultraviolet spectrum values of different grades of Huangshan Maofeng tea infusion were collected to distinguish the grades by principal component analysis.The fusion data of ultraviolet spectrum and color difference were more effective in differentiating grades than the data of color difference or ultraviolet spectrum alone.In the scatter plot,the grades of fusion data tea samples decreased in the counterclockwise direction.And the tea samples of the same grade were more aggregated.The random forest model was chosen,and the grade classification and verification accuracy were both 1(100%)by inputting fusion data.(4)The electronic tongue response values of tea samples were collected and support vector machine models of the chemical component content and electron tongue response values were developed.The optimal penalty parameter c of the model was in the range of 4-11.3137,and the kernel parameter g was in the range of 0.7071-4.A method for rapid determination of amino acid content in tea using colorimetric capsule image recognition was proposed for the first time.The results showed that the accuracy of the amino acids standard concentration prediction model was 0.88.The coefficient of determination(R~2)between the true value of chemical component detection and the model predicted value of amino acid content of black tea and green tea samples was0.8069,the root mean square error(RMSE)was 0.3709,and the root mean square error of prediction(RMSEP)was 0.0566.