Computational intelligence based frameworks for engineering optimization

The modern engineering design optimization process often replaces laboratory experiments with computer simulations. This a setup formulates an optimization problem of a black-box function, namely, which has no analytic expression and which is computationally expensive to evaluate. This has motivated the development of computational-intelligence based frameworks, as often they can perform well in challenging settings. However, a main bottle neck in their implementation is the limited number of function evaluations. To this end, metamodels are used to approximate the expensive simulation and to obtain predicted objective values at a lower computational cost. While a variety of metamodels have been proposed, the optimal type is typically problem-dependant and unknown prior to the optimization search. To address this issue, this chapter describes framework which employs multiple metamodels concurrently, thereby benefiting from the different approximations. A detailed performance analysis based on an engineering problem shows the merit of the proposed framework.