Direct search methods for the fast optimisation of batch distillation processes

Abstract

Batch distillation can be conveniently modelled by professional flow-sheet simulators. Optimisation can be performed by coupling an external optimiser to the simulator. The most frequently used method is the genetic algorithm (GA), which, however, requires a high number of simulations to evaluate the objective function. Two direct search methods, the Nelder-Mead simplex and the Box-complex algorithms are applied to reduce the computational intensity of optimisation. Calculations are performed for a case study from the literature where the profit of the regeneration of a multicomponent azeotropic waste solvent mixture was maximised by a GA. The influence of the parameters of the optimisation methods is investigated for each method. The highest profit is reached by the simplex algorithm. Both the simplex and complex algorithms generally outperform GA with a much lower number of simulations. Therefore, direct search methods can be used for fast and efficient optimisation of batch distillation processes.