Optimisation of a batch extractive distillation process with hybrid methods

Hégely, László (2026) Optimisation of a batch extractive distillation process with hybrid methods. SEPARATION AND PURIFICATION TECHNOLOGY, 383. No.-136197. ISSN 1383-5866

Text BED opt cikk published.pdf - Published Version Restricted to Registered users only Download (6MB) | Request a copy

Abstract

Batch distillation processes are usually optimised by evolutionary algorithms, such as the genetic algorithm (GA). These are capable of finding the global optimum but can be very time-consuming. Direct search methods, such as the Nelder-Mead simplex and Box-complex algorithms, are faster but can converge to local optima. This work proposes hybrid optimisation methods as combinations of a global and a local method. The global methods considered are GA and Latin hypercube sampling (LHS) as a random search method. The best result of the global method is an initial point of the local method, the simplex or the complex algorithm. The hybrid methods are applied to minimise the profit of methanol recovery from a waste solvent mixture by batch extractive distillation. A large number of optimisations are performed with different numbers of generations (GA) and sample points (LHS). The highest profit is reached by the GA + C (GA followed by the complex algorithm) method, which gives consistently good results. GA + S (GA followed by the simplex algorithm) is faster than GA + C but shows a higher variation among the profit values obtained. The recommended number of GA generations is 5–15 for GA + S and 50 for GA-C, although in the latter case, good results can be obtained even with a single GA generation. The hybrid methods GA + S, GA + C and LHS + C (LHS followed by the complex algorithm) consistently outperform GA itself, making them promising candidates for the efficient optimisation of batch distillation processes.

Actions (login required)

View Item