An adaptive response surface method and Gaussian global-best harmony search algorithm for optimization of aircraft stiffened panels

Abstract

For the engineering optimization problems characterized by high computational cost and multiple local optima, the combination use of surrogate model and meta-heuristic algorithm is increasingly popular. In this paper, a bi-loop optimization framework of stiffened panels is proposed to search the global optimum, which includes an adaptive response surface method (ARSM) loop and a Gaussian global-best harmony search (GGHS) loop, aiming to improve the global optimization capacity in an efficient manner. The ARSM loop involves an inner loop to select the data set for training the response surface function, where the spherical data set can provide an accurate prediction of the optimum condition compared to the original RSM. Typical aircraft stiffened panels in NASA are employed to demonstrate the effectiveness of the proposed framework. Results demonstrate that the proposed ARSM has higher prediction accuracy of weight and buckling load compared to the traditional RSM, and the optimum design of GGHS combined with ARSM has the lowest relative error by comparison with different harmony search algorithms. Finally, a fast prediction model of aircraft stiffened panel is developed, which can provide approximated optimum conditions without detailed optimization process.