Microencapsulation of flavored liquor by spray drying: Optimization using response surface methodology and genetic algorithm-based support vector regression

Abstract

Listen

Translate article

Flavored liquor microcapsules were prepared by spray drying with baijiu as the main core material, and gelatin, β-cyclodextrin, and maltodextrin as the composite wall material. The experiments were first conducted in a lab-scale spray dryer with ethanol retention as the objective function, and feed solid content, inlet air temperature, and feed rate as the influencing factors. The operating parameters were optimized using the commonly used response surface methodology (RSM) and a machine learning method of genetic algorithm-based support vector regression (GA-SVR). Subsequently, the pilot experiments were conducted in a pilot-scale spray dryer based on the optimized parameters. The qualities of products from the two dryers were evaluated. The results indicated that the ethanol retentions for the lab-scale and pilot-scale dryers were 47.78% and 46.07%, respectively. Compared with the RSM, the GA-SVR opted for a strategy of simultaneously increasing both inlet air temperature and feed rate while maintaining almost the same ethanol retention. The optimal operating parameters were determined as 35.7 wt% of feed solid content, 110 °C of inlet air temperature, and 9.5 mL·min−1 of feed rate for the lab-scale dryer and 28 mL·min−1 for the pilot-scale dryer. The microcapsules prepared by the two dryers exhibited a spherical shape with fine texture, and had a certain degree of heat resistance. The pilot-scale dried products showed slightly better flowability and lower residual water content. This work is aimed to provide a practical guidance for the machine learning method in optimizing the spray-dried microencapsulation.