Abstract:

We examine the mathematical statement of multidisciplinary optimization (MDO) problems or, more specifically, the formulation of MDO problems as optimization problems, and the consequences of problem formulation for the practical solution of the resulting computational problem by optimization algorithms. Because of the complexity and expense of the constituent analyses, most efforts in dealing with systematic MDO problem formulation focus on methods that aim at affording the user the maximum disciplinary autonomy. We examine some notions of autonomy and consider broad classes of MDO problem formulations in light of disciplinary autonomy, as well as the effects of the techniques used for attaining autonomy via distributing the disciplinary subproblems. We also comment on the promise of MDO versus the state of the art and the open questions.