Adaptive Search for Optimum in a Problem of Oil Stabilization Process Design

The formulation of a model for an industrial problem in process design leads to an optimization problem with a small, implicitly defined, feasible region, a region which is difficult to identify a priori. The difficulties of obtaining a good solution with conventional optimization methods are discussed. A novel method is proposed, based on the paradigm of evolutionary computing and a two stage search: the first stage aims to find a set of points covering the feasible region and the second stage is a search for the optimum, modelling the evolution of the population, the set of points, found in the first stage. The results of profit optimization for an industrial case study are presented.