Preliminary computer-aided design of powdered soft beverages integrating heuristic knowledge

This study aimed to develop a computer-aided tool for the preliminary design of juice-type powdered soft beverages (PSBs). The tool integrated predictive models of bulk properties, expert knowledge, and product specifications into a mathematical model optimized for formulation cost. A comprehensive literature review enabled to identify key ingredients-sweeteners, acidulants, anticaking agents, and carriers-and their influence on PSB performance. It also established formulation rules and quality factors, linking them to the physicochemical and bulk properties of the mixture. Given the lack of existing models for powder formulation, experiments were conducted to correlate flow factor and cohesion of PSBs with their composition. Linear polynomials with respect to the concentration of key ingredients were derived, and an exponential equation was fitted to relate these two properties. Additionally, a simple mixing rule was validated experimentally to estimate particle size distribution (PSD) based on ingredient composition. A formulation model was then developed, incorporating predictive equations for sweetness intensity, acidity, flavor content, PSD, fines content, and flow factor, along with various constraints. The model was optimized using MILP (Mixed-Integer Linear Programming) to generate potential PSB formulations that met manufacturing and consumer performance requirements at minimum cost. Four promising blends were identified and prototyped at laboratory scale. The prototypes performed as predicted, and after expert evaluation, they were found to resemble commercial products, requiring only minor adjustments to meet desired specifications.