Modified Efficient Global Optimization for a Hat-Stiffened Composite Panel with Buckling Constraint

Abstract

The optimization method for composite structural components described herein uses modified efficient global optimization with a multi-objective genetic algorithm and a kriging response surface. For efficient global optimization using kriging, the kriging response surface is used as a representative of the function value. The stochastic distribution of the kriging is used to improve the estimation error of the kriging surrogate model. Using efficient global optimization, a hat-stiffened composite panel was optimized to reduce the weight with the buckling load constraint. The expected improvement was used as a single objective function of a particle swarm optimization. Nevertheless, it is difficult to obtain a feasible solution that satisfies buckling load constraints with the progress of optimization. Using a multi-objective genetic algorithm, we obtain the feasible optimal structure satisfying the constraints. The expected improvement objective function is divided into two objective functions: weight reduction and the uncertainty of satisfaction of the buckling load constraint. Kriging approximation, which is improved with the selected Pareto optimal frontier, reduces the computational cost. Also, a genetic algorithm is used to optimize the stiffened panel configuration. The fractal branch-and-bound method is used for stacking sequence optimizations. This method obtained a feasible optimal structure at a low computational cost.