Statistical-probability simulation of the organoleptic properties of grape wines

Abstract

Approaches to the evaluation of generalized wine quality indices based on a set of the quantitative characteristics of single parameters and their organoleptic rating were studied with the use of statistical-probability simulation methods. A general linear model (multiple linear regression) was constructed to predict degustation evaluation values from the concentrations of volatile substances (acetaldehyde, ethyl acetate, methanol, higher alcohols, acetic acid, and furfural) and wine quality classes (high, medium, low, and adulterated). The wine quality class was identified by discriminant analysis based on the concentrations of the above volatile substances. A program module was developed for the automation of a calculation procedure. The average absolute deviation of predicted values from degustation evaluation data in a test sample was 5.8%.