Constrained evolutionary optimization of a distillation train in chemical engineering

The optimal design and synthesis of distillation systems remains one of the most challenging problems in process engineering. The goal of this paper is to introduce an evolutionary approach for the optimization of the total energy consumption of distillation systems with constraints. Moreover, the contribution of this paper is a novel constraint handling technique that manages design goals as equality constraints, such as the purity and the recovery of the final components. In the literature of these problems prevail the use of inequality constraints; although easy to apply they may lead the search to suboptimal solutions. The case study is a distillation column sequence (DCS) for the separation of four components; this problem is easy to describe yet complex to solve so our approach can show its advantages. The evolutionary algorithm Boltzmann Univariate Marginal Distribution Algorithm, (BUMDA), performs the optimization. AspenONE©software is used for the rigorous evaluation of the fitness function of the population. The results show the efficacy performance of the proposed approach reaching near optimal designs in less than 3000 function evaluations.