Evaluation of genetic algorithms and simulated annealing for phase equilibrium and stability problems

Phase equilibrium calculations require global minimization of free energy, and phase stability analysis too often involves global minimization of tang ent plane distance function (TPDF). In this study, two stochastic global op timization techniques, namely, genetic algorithm (GA) and simulated anneali ng (SA) are evaluated and compared for phase equilibrium and stability prob lems. Typical examples and different thermodynamic models are considered. T he results show that GA is generally more efficient and reliable than SA fo r phase equilibrium calculations. Both GA and SA exhibited poor reliability for locating the global minimum of free energy function for some complex p hase equilibrium systems. For these problems, a hybrid GA incorporating SA for individual learning, is proposed and its improved capability is shown. The results on phase stability problems show that GA is able to locate the global minimum of TPDF with 100% reliability in all the examples tried. It is also found to be very efficient compared to other global techniques repo rted in the literature. (C) 2001 Elsevier Science B.V. All rights reserved.