Applied Food Research (Jun 2021)
TeaPot: A chemometric tool for tea blend recipe estimation
Abstract
Retail teas are blends of different tea grades in different proportions with varying quality and characteristics. Since there is not any available alternative, the ratio of stock teas in a blend is determined by expert tasters and sometimes supported by expensive chemical analyses. Instead of this relatively subjective expert dependant method, in this study, we present a novel and very economic approach to determine tea blending proportions. The approach is based on chemometric analysis of the data produced by spectral scanning of simple tea infusions (only prepared using boiling water as extraction solvent). For this purpose, UV–Vis spectra of 21 different stock tea samples were obtained and the data was processed using Principal Component Analysis (PCA). The blending proportions to produce a target tea mixture were calculated by a custom-developed algorithm, named TeaPot, with an easy-to-use graphical interface. In brief, a similar blend to the target is calculated from inversely proportioned Euclidean distances of tea samples’ PCA scores. These PCA scores were weighted with the associative variance explanation proportion of principal components. According to the validation results, over 90% similarity to the target is achieved from TeaPot's estimations. This proposed methodology and algorithm provide a simpler, faster, and affordable solution for tea manufacturers to avoid the effects of seasonal changes, process variables, and species-dependant differences in rental teas. As a result, TeaPot makes it possible to ensure product quality with very little expense and requiring no expertise.
