Dynamic modeling and simulation of an alembic pear wine distillation

• Published in: Food and bioproducts processing 2013-10, Vol.91 (4), p.447-456
• Main Authors: Sacher, Johannes, García-Llobodanin, Laura, López, Francisco, Segura, Hugo, Pérez-Correa, José Ricardo
• Format: Article
• Language: English
• Subjects: acetaldehyde; Alcoholic beverages; Aromas; condensation; Congeners; distillates; distillation; distilled spirits; Dynamic modeling; dynamic models; Dynamical systems; Dynamics; Ethanol; Ethyl alcohol; gas chromatography; heat transfer; Mathematical models; methanol; odor compounds; Pear spirit; Pears; phenylethyl alcohol; Pot still; prediction; Recovery; simulation models; Vitaceae; Volatile composition; wines
• ISSN: 0960-3085; 1744-3571
• DOI: 10.1016/j.fbp.2013.04.001

•The traditional alembic has been modeled as a two equilibrium stage system.•The liquid phase can be treated as a quasi-binary mixture.•Thermodynamic properties can be calculated independently of congener concentrations.•Activity coefficients can be predicted for most congeners with the UNIFAC method.•Ethanol and 15 congeners in the distillate for a pear spirit were well predicted. In this research a phase equilibrium model to predict the recovery of several aroma compounds in spirits distilled in a traditional Charentais alembic has been developed. Due to the congeners’ quasi-infinite dilution, the properties of the mixture are assumed to be dependent on the ethanol mole fraction only. Thus, the system is treated as a quasi-binary mixture; the ethanol recovery, the dynamics of the condensation and the heat transfer processes are modeled independently of the congener's concentration. The model was calibrated with laboratory data obtained during the elaboration of pear distillates. The concentration of ethanol and 15 congeners in the distillate were measured with gas chromatography during the experiments. Fitting just two parameters, one for heating power and one for heat loss, was sufficient to reproduce the experimental ethanol recovery curves. This allowed an accurate prediction of most of the congener's recovery curves. The prediction of the concentration of five congeners, including acetaldehyde and methanol, are not significantly different from measured values. The prediction of the concentration of six congeners was biased but with relatively small errors (ranging between 8% and 25%). Significantly large prediction biases (more than 30%) were observed for ethyl hexanoate, ethyl decanoate and phenethyl alcohol.