Surrogate Model Management in Genetic Algorithms with Fuzzy Controllers

Surrogate modeling techniques are of particular interest for engineering design when high-fidelity, thus expensive analysis codes are used. They provide sufficiently accurate solutions by using numeric approximation models. Recently, surrogates have been employed adding engineering and expert knowledge to improve the accuracy and the convergence of the algorithm. This paper proposes a granular-surrogate model, which in turns provides a structure to extract and represent some knowledge with fuzzy logic. The extracted rule-based understanding of the granule's activity allows us to design two fuzzy controllers to manage the parameters update, providing a self-adaptive granular surrogate model according to the characteristics of the function handled. With this proposal, we are changing from a data-driven surrogate to a knowledge-based one, showing the effectiveness of the algorithm in standard benchmarks.