1st Int Workshop on Surrogate Modelling and Space Mapping for Eng Opt, 11/16{19/00, TDU FIRST-ORDER FRAMEWORKS FOR MANAGING MODELS IN ENGINEERING OPTIMIZATION

Approximation/model management optimization (AMMO) is a rigorous methodology for attaining solutions of highdelity optimization problems with minimal expense in highdelity function and derivative evaluation. First-order AMMO frameworks allow for a wide variety of models and underlying optimization algorithms. Recent demonstrations with aerodynamic optimization achieved three-fold savings in terms of highdelity function and derivative evaluation in the case of variable-resolution models and ve-fold savings in the case of variabledelity physics models. The savings are problem dependent but certain trends are beginning to emerge. We give an overview of the rst-order frameworks, current computational results, and an idea of the scope of the rst-order framework applicability.