Flavor Blending Method Based on Linear Hypothesis Algorithm: A Case Study on the Taste of Green Tea

Abstract

Abstract: A linear hypothesis algorithm was developed to calculate the blending proportions of granular green tea sourced from Zhejiang, China. Two similarity coefficients ( [[EQUATION]]and [[EQUATION]]) were derived using the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) to select appropriate raw samples and final blending recipes, respectively. [[EQUATION]] identified the optimal blending proportions through datasets (G1 to G6), comprising quantitative sensory (QDA) and chemical compound analyses, while [[EQUATION]] ranked the top five samples. Integer programming was applied as a model constraint, and the model solution was obtained using the Monte Carlo method. The findings demonstrated that blending proportions based on Dataset-G1 and Dataset-G4 achieved higher [[EQUATION]] values. Sensory evaluation validated that these blended samples closely replicated the target sample's taste profile. The study established precise granularity parameters for tea blending to optimize flavor, providing a model applicable to other tea blending methodologies.

Keywords: Tea blending, Taste, TOPSIS, Factor analysis, Linear hypothesis