Abstract
This article addresses the experimental evaluation and mathematical modeling of water sorption isotherms of dried specialty coffee beans that have been processed by wet and semi-dry post-harvest methods. The study analyzed different water activities ranging from 0.1 to 0.85 and temperatures of 25, 35, and 45°C. The experimental isotherms were obtained using the dynamic dew point method (DDI). To model water sorption isotherms, 11 empirical models, 4 machine learning, and the Guggenheim-Anderson-de Boer (GAB) equation were used. The experimental data were randomly split into 75% for model training and 25% for validation. The experimental results show a type II water sorption isotherm and a significant temperature influence. Additionally, the sorption shape of the isotherms suggests that the mucilaginous coating, which covers the beans obtained by the semi-dry method, modifies the curves and plays a protective role against water sorption. The SVM model was the most accurate predictor to describe the upward sigmoidal type II sorption trend. The impact of temperature and water activity, as well as the post-harvest method, on the equilibrium moisture content (ERM = 0.21% and R2ad = 99.8%) suggests that it could be a valuable tool for predicting and optimizing storage conditions for both types of specialty coffee beans.