Design of optimal wine distillation recipes using multi-criteria decision-making techniques


In the wines and spirits industry, distillation is commonly used to balance the productivity and distinct flavor and ensure high-quality and toxic-free products. Generally, distillation involves five main operating variables: heart, head, and tail cut volumes, as well as heating power and reflux rate to track the alcohol strength of distillate and achieve the desirable ethanol recovery. In most industries, the operating variables are defined by testing and smelling methods conducted by experienced operators. However, this method is expensive and time-consuming, and designing spirit distillation recipes based on the distiller’s decision is also challenging.

These challenges can be addressed by using model-based approaches and standard optimization techniques to design optimal distillation recipes. This approach has enabled the design of several spirit distillation models, based on a set of algebraic and non-differential equations, for different distillation columns. Additionally, since the distillation of spirits involves simultaneous optimization of many conflicting objectives, the trade-off in meeting the objectives can be addressed by designing the recipes within the multi-objective optimization (MOO) framework.

The MOO-based solution consists of optimal points and optimal decision variables known as Parent front and Pareto set, respectively. The multi-objective evolutionary algorithms (MOEA) use the Pareto dominance concepts to produce various optimal points in one optimization run. Research revealed that this problem can be solved by considering optimal operating conditions and design variables based on multi criteria decision making (MCDM) and MOEA techniques, which are yet to be applied to design optimal recipes for spirit distillation.

On this account, Dr. Ricardo Luna from Viña Concha y Toro, Professor Francisco López from Universitat Rovira i Virgili and Professor José R. Pérez-Correa from Pontificia Universidad Católica de Chile solved the multi-objective dynamic optimization problem by using MCDM techniques to design recipes for wine distillations. Specifically, the approach was based on the Thompson Sampling Efficient Multi-Objective (TS-EMO) algorithm. The work is currently published in the journal, Computers and Chemical Engineering.

In their approach, eight distillation objectives, more than in the previous studies, were considered. The objectives included enhancing efficiency in energy consumption, productivity in ethanol recovery, and others associated with food safety and sensory quality. On the other hand, the decision variables included heart-cut volume, head-cut volume, and heating power path. Furthermore, seven MCDM methods were utilized to establish the best optimal distillation recipes based on five different priorities. Finally, the selected solutions were compared with traditional approaches to solve the multi-objective optimization problem.

Results showed that the initial eight objectives were reduced to four independent objectives via correlation analysis of the Pareto front solutions. Each of the MCDM algorithms produced varied solutions when the objectives were all given equal priority. In contrast, when some objectives were given higher priority, the MCDM algorithms selected similar solutions in most cases. Among the algorithms, VIKOR’s recipe was characterized by medium energy consumption and high ethanol recovery, making it the best for producing young spirits.. The authors also realized that a balance performance was achieved when the ethanol recovery was assigned the highest priority, recovering 85% of the ethanol, making this recipe appropriate for producing aged spirits.

In summary, the design of optimal wine distillation recipes based on MCDM and multi-objective optimization techniques was reported. Bayesian algorithms like TS-EMO proved effective for solving the complex and challenging MOO dynamic distillation problem comprising eight objectives. For instance, one optimization run involving TS-EMO produced several Pareto solutions compared to the W-method, which achieved one solution per run. Furthermore, the presented optimization approach applies to any spirit alembic distillations. In a statement to Advances in Engineering, first author Dr. Ricardo Luna said the presented methodology could effectively define recipes for the optimal operation of different distillation processes using different equipment and is applicable to any spirit or fruit distillation process.


Luna, R., López, F., & Pérez-Correa, J. (2021). Design of optimal wine distillation recipes using multi-criteria decision-making techniquesComputers & Chemical Engineering, 145, 107194.

Go To Computers & Chemical Engineering

Check Also

A simple approach for adding thixotropy to an elasto-visco-plastic rheological model to facilitate structural interrogation of human blood - Advances in Engineering

A simple approach for adding thixotropy to an elasto-visco-plastic rheological model to facilitate structural interrogation of human blood