Multi-objective optimization of an industrial isoprene production unit by using genetic algorithm approach

Abstract

The present work deals with the multi-objective optimization of an industrial Isoprene production unit by using Genetic Algorithm (GA). The purpose of this plant is to produce high purity Isoprene for obtaining synthetic and thermoplastic rubber from a C5 cut arising from a pyrolysis gasoline unit. The chemical process consists basically of a dimerization reactor and a separation column train. The Isoprene industrial process is a complex process, and not easy to be solved by commercial simulators mainly due to the lack of thermodynamics properties. A neural network approach has been previously developed in order to model industrial process from historical data (Nascimento et al., 2000; Alves and Nascimento, 2002 and 2004a). In this work, a genetic algorithm (GA)-search, an optimization technique based on principle of natural genetics, is used to perform the optimization procedure. GAs are chosen as an optimization tool because of their successful application in many others engineering and industrial optimization problems (Alves et al., 2004b; Laquerbe et al., 2001; Pibouleau et al., 1999). The reason for a great part of their success is their ability to exploit the information accumulated about an initially unknown search space in order to bias subsequent searches into useful subspaces, i.e., their adaptation. Moreover, GAs use objective functions information and not derivatives or other auxiliary knowledge to perform an effective search for better and better structures. Then, the aim of this paper is to present and discuss the applicability of a genetic algorithm as an alternative procedure for a multi-objective optimization of an industrial process that may be difficult to handle by classical methods (Lim et al., 2001). In this case the optimization of the entire plant involves 21 variables to be optimized. So, in order to decrease the combinatorics of the problem, the global model was divided into three sections and each one was optimized separately, but sequentially, by using the optimal conditions from previous optimization section procedure. For this, a multi-objective genetic algorithm (MOGA) based on a Pareto sort (PS) procedure was implemented to manage this specific problem. The optimization procedure employed in this work does not require necessarily a formal objective function. It should deal with either qualitative set of process constraints and quantitative or economical analysis by using an objective functions that describe the technico-economical goasl. It is important to keep in mind the main objective to be achieved, for example, higher production for a given product specification at lower energy consumption.. A constrained optimization procedure was used to take into account product quality, safe operations conditions, and energy consumption. The GA model developed may find solutions near the global optimum within reasonable time and computational costs.