Optimal synthesis of thermally coupled distillation sequences using a novel MILP approach

This paper introduces a novel MILP approach for the design of distillation columns sequences of zeotropic mixtures explicitly including from conventional to fully thermally coupled sequences and divided wall columns with a single wall. The model is based on the use of two superstructure levels. In the upper level a superstructure that includes all the basic sequences of separation tasks and all the possibilities of heat exchange (condensers and reboilers). In the lower level an extended tree that explicitly includes all the alternatives for a given separation task (different thermal states and compositions of the feed to a given separation task). In that way, it is possible to a priori optimize all the possible separation tasks involved in the superstructure. A set of logical relationships relates the feasible sequences (upper level) with the optimized tasks in the extended tree resulting in a MILP to select the optimal sequence. Performance of the model in terms of robustness and computational time is illustrated with several examples.