ANN-based modeling of moisture sorption in date powder: effects of particle size and storage conditions on thermodynamic behavior and functional properties
摘要
This study investigated the effects of particle size (180, 250, and 355 μm), temperature (15, 25, and 35 °C), and relative humidity (10–90%) on the adsorption isotherms, thermodynamic behavior, and functional stability of Sukkari date powder. It addresses a key gap in the literature by systematically evaluating particle size effects on sorption–thermodynamic responses, an area scarcely examined and rarely modeled in tandem with artificial neural networks (ANNs). Five empirical isotherm models and multiple ANN architectures were evaluated for predicting equilibrium moisture content (EMC). Physical properties (color, density, flowability, and hydration) were measured at room temperature across particle sizes. Adsorption followed Type III behavior; temperature decreased EMC at water activity below 0.5, while particle size had no significant effect on EMC. Particle size markedly altered functionality: larger particles darkened color (L*: 52.79 to 48.33), increased bulk density (0.558 to 0.685 g/cm³), improved flowability (Hausner ratio: 1.22 to 1.18), increased water absorption (155% to 203%), and decreased solubility (78% to 65%); initial moisture rose slightly with size (5.37% to 5.55%). BET and Oswin best described the experimental isotherms, whereas the ANN achieved superior predictive accuracy. The Oswin model suggested maintaining moisture contents below 3% (dry basis) at a water activity of 0.3 to ensure microbiological safety and stability. Thermodynamically, increasing EMC reduced water-binding energy (Qₛₜ) and spontaneity (ΔG), while entropic penalties (ΔS) increased. By isolating particle-size effects and integrating ANN with classical models, this work provides novel, actionable targets for drying, packaging, and storage to extend shelf life and improve the processability of date powder.