Computational Intelligence for Process-optimization Software

Abstract

This work describes a general algorithm for a cooperative hyper-heuristics that enables the optimization of systems of nonlinear algebraic equations with algebraic constraints. The hyper-heuristics comprises the following agents: Genetic Algorithms, Simulated Annealing and Particle Swarm Optimization. Information exchanges take place effectively among them since the immediate incorporation of solution candidates speeds up the search. Algorithmic performance is illustrated with general test models, most of them corresponding to process systems that have currently been employed in PSE. When running in parallel, numerical results demonstrate that the collaborative hybrid structure with embedded intelligent learning contributes to improve results in terms of effectiveness and accuracy. The combination of several heuristic optimization approaches into a hyper-heuristics provides enhanced benefits over traditional strategies since this method helps to find proper comprehensive solutions, also contributing to achieve and accelerate convergence.