Global Optimization Algorithms in Multidisciplinary Design Optimization

While Multidisciplinay Design Optimization (MDO) literature focuses mainly on the development of different formulations, through the manipulation of design variables, less attention is generally devoted to the combination of specific MDO formulations with existing nonlinear optimization algorithms. In this paper, the focus is on the application of a Global Optimization (GO) algorithm to an MDO problem. We first introduce and describe some MDO approaches from the literature. Then, we consider our MDO formulation where we deal with the GO boxconstrained problem min f(x), f:Rn in R a<x<b We assume that the solution of the latter problem requires the use of a derivative-free methods since the derivatives of f(x) are unavailable and/or the function must be treated as a 'black-box'. Within this framework we study some globally convergent modifications of the evolutionary Particle Swarm Optimization (PSO) algorithm, suitably adapted for boxconstrained optimization. Finally, we report our numerical experience. Preliminary results are provided for a simple hydroelastic problem. Two different numerical tools are involved: a fluid dynamic solver, to simulate the flow around hydrofoils traveling in proximity of the air-water interface, and a simplified torsion-flexional wing model.