Predicting the composition of sucrose–ethanol–water mixtures and alcoholic beverages using density, refractive index, and temperature data

Abstract

Accurate alcohol content measurements are a critical regulatory requirement in the alcoholic beverage industry, where bottles labeled with the percentage of alcohol by volume must be within ± 0.3 percentage points of the actual value. While large-scale producers use expensive analysis equipment, smaller breweries and distilleries often rely on manual hydrometer readings, which are prone to error and inefficiency. This project explores a cost-effective alternative to estimate beverage compositions, by using widely available in-line instruments such as Coriolis meters and refractometers. The primary aim is to predict the composition of an unknown solution in terms of Brix for sugar content and percentage alcohol by weight (ABW%) for alcohol content, using the physical properties of temperature, density, and refractive index of the solution. Solutions with known compositions were prepared in the lab, and their corresponding physical measurements were recorded across a wide range of mixture compositions and temperatures. These data were used to train regression models that learned the relationship between physical properties and solution composition. Given the temperature, density, and refractive index of an unknown solution, the models could then predict the mixture composition. The Brix model achieved high accuracy, with most predictions within ±1 Brix. Alcohol predictions were less precise, typically within ± 2 ABV percentage points. The models were validated using both lab-made mixtures and commercially produced beverages. The models performed comparably on both, showing strong generalization. The results also suggest that prediction accuracy would be improved by collecting more measurements. With additional parameters, this approach could be extended beyond three-component mixtures to more complex formulations, and beyond alcoholic beverages to applications in industries such as chemical processing, cosmetics, and food quality monitoring.