Direct and indirect design prediction in genetic algorithm for inverse design problems

An improved genetic algorithm is proposed and tested for five different test cases: surface fittings of a wing and geographical terrain, an inverse design of an airfoil and wing shapes at subsonic flow, and an inverse design of an airfoil shape at transonic flow. The new algorithm emphasizes the use of both direct and indirect design predictions based on local surrogate models in genetic algorithm structure. Local response surface approximations are constructed by using neural networks. For all the demonstration problems considered herein, remarkable reductions in the computational times have been accomplished.