Multiobjective Optimization of Industrial Processes Using Elitist Multiobjective Differential Evolution (Elitist-MODE)

Abstract

Multiobjective optimization (MOO) refers to solving more than one-objective optimization problems simultaneously to attain a set of solutions, i.e., the Pareto front. In this study, an elitist strategy of multiobjective differential evolution (Elitist-MODE) algorithm is used to find the Pareto optimal solutions for a set of multiobjective test problems [POL (unconstrained), CONSTR-EX (constrained), and TNK (constrained)] and two-industrial engineering process problems, namely, styrene reactor and polyethylene terephthalate reactor (PET). A detailed analysis is reported for the test problems and real-world industrial problems, and the obtained Pareto fronts are compared with the Pareto front obtained using MODE and MODE III algorithms. Simultaneous maximization of selectivity and yield (for the styrene reactor) and simultaneous minimization of acid- and vinyl-end group concentrations (for the PET reactor) are considered. A detailed analysis of obtained Pareto fronts with respect to the set of decision variables for the individual industrial study is reported. It is observed that the improved strategies of MODE algorithms in general converge to the true Pareto front both for the test problems (constrained/unconstrained) and complex industrial problems. It is also suitable for solving the problems having disconnected Pareto front.

Keywords:

• Differential evolution
• Elitist-MODE
• Genetic algorithms
• Industrial processes
• MODE
• MOOP

Notes

No Pareto front resulted.