Hybrid approach for multiobjective optimization and its application to process engineering problems

Abstract

Stochastic search methods can thoroughly explore the search space for the global optimum, but once the search approaches the optimum, they become inefficient to find the optimal solution precisely. Conversely, deterministic local methods are more efficient in finding the optimum precisely, particularly if a good initial estimate is given. In this work, the stochastic and local search methods are combined together for improving the search efficiency without losing reliability for solving multiobjective optimization problems. An improved multiobjective differential evolution (I-MODE) is used for the global search; it includes a termination criterion to decide the switching from the stochastic search to the local search. The normalized normal constraint method is employed to refine the non-dominated solutions obtained by the I-MODE search. The proposed hybrid approach is tested on several unconstrained and constrained biobjective optimization problems. It is then applied to the alkylation process, three-stage fermentation process integrated with extraction, and three-stage fermentation process integrated with pervaporation.