HYBRID TWO-LEVEL GENETIC OPTIMIZATION ALGORITHM WITH DIFFERENT FIDELITY MODELS FOR AERODYNAMIC DESIGN PROBLEMS
I. L. Chernyshev · 2002
A new hybrid genetic algorithm for solving optimization problems in multidimensional space has been developed. The idea behind the method is to simultaneously use the numerical models of different fidelity. The bulk of computations is based on low-fidelity and low time-consuming model, while the small amount of computations are based on a high-fidelity model for refining the behavior of the objective function. As a result, the computational accuracy of the developed algorithm corresponds to higher accuracy model, and total computational time is reduced by several times as compared to a conventional genetic algorithm. Basic features of the developed algorithm are described and several examples of applying this algorithm to transonic aircraft aerodynamic configuration design are presented in the paper.