A hybrid global optimization scheme that combines deterministic and stochastic techniques for process design

Abstract

This article aims to develop a hybrid global optimization scheme for the optimal design of chemical processes. With the combination of stochastic techniques and a deterministic local search, the proposed optimization methodology integrates a feasible point strategy, chaotic dynamics, and the information theory (IT). Based on the deterministic local search using a modified feasible point strategy, the chosen trial points are easily refined to fall in the feasible region that has better objective function values. The IT with a defined information free energy is then used to prevent being trapped by a local optimum and to search for the global solution. Instead of using traditional random procedures for stochastic searching, in this article, we apply chaotic dynamics, which have the properties of randomization, ergodicity, and regularity, for the next point generation in order to effectively achieve the global solution. The effectiveness and applicability of the proposed global optimization scheme are demonstrated through two typical chemical process design problems. To provide a guideline for parameter selection, the effects of the algorithm parameters on global optimization performance are investigated systematically. Extensive comparisons with comparative optimization algorithms reveal that the proposed hybrid global optimization scheme is more efficient and superior in locating the global solution for process design problems.