In industry, distillation is usually the preferred process to separate multicomponent mixtures. A given multicomponent distillation process can be carried out by various configurations (sequences) that differ significantly in energy consumption. So it is important to be able to generate a complete search space of sequences that are attractive for a given application, and evaluate energy consumption of each. This will enable us to choose a sequence that consumes least energy.

Earlier attempts to generate a search space for zeotropic (ideal to near ideal) distillation were either incomplete, or included a large number of configurations which were never optimal. Our prior work in this area describes a method to generate a search space which is complete in terms of useful sequences and at the same time, does not include the sub-optimal configurations. Because the sub-optimal configurations are not included, the size of the search space is significantly reduced. This method uses binary integer variables associated with edges in a supernetwork. However, though the search space is smaller, it still grows exponentially as the number of components in the feed increases. Hence, a computationally efficient method to generate the search space was desired.

The goal of this work is to describe a novel computationally efficient formulation. The new formulation uses binary integer variables in a node-based network. Representing the decision variables as a matrix easily captures the physical reality of the problem. This matrix representation also helps us to easily automate the method. The new formulation can also be extended to include thermally coupled configurations in the search space almost trivially. This approach is very fast computationally, and the resulting search space can be combined with any desired objective function evaluator to give the truly optimum distillation sequence for the application on hand.