Solving phase equilibrium problems by means of avoidance-based multiobjectivization

Abstract

Phase-equilibrium problems are good examples for real-world engineering optimization problems with a certain characteristic. Despite their low dimensionality, finding the desired optima is difficult as their basins of attraction are small and surrounded by the much larger basin of the global optimum, which unfortunately resembles a physically impossible and therefore unwanted solution. We tackle such problems by means of a multi-objectivization-assisted multimodal optimization algorithm which explicitly uses problem knowledge concerning where the sought solutions are not in order to find the desired ones. The method is successfully applied to three phase equilibrium problems and shall be suitable also for tackling difficult multimodal optimization problems from other domains.