MULTI-OBJECTIVE DIFFERENTIAL EVOLUTION: MODIFICATIONS AND APPLICATIONS TO CHEMICAL PROCESSES

Abstract

Industrial problems are complex in nature, and often have multiple conflicting objectives. Multi-objective optimization (MOO) helps engineers in exploring trade-offs among different objectives. There are several stochastic MOO techniques in practice, but they are not robust and efficient on application problems. This study improves multi-objective differential evolution (MODE) in key aspects, such as search termination based on the improvement in non-dominated solutions obtained with generations, better exploration of search space using taboo list, and constraints handling by dynamically relaxing them. The improved MODE (I-MODE) algorithm has been tested on many benchmark functions and then applied to solve chemical engineering application problems. The developed search termination criterion has also been tested in conjunction with other algorithms. Bio-ethanol process, bio-diesel plant and membrane distillation system are modeled and then optimized for multiple objectives.