On the Consequences of the "No Free Lunch" Theorem for Optimization on the Choice of an Appropriate MDO Architecture

Abstract : Multidisciplinary design optimization (MDO) based on high- delity models is challenging due to the high computational cost of evaluating the objective and constraints. To choose the best MDO architecture, a trial-and-error approach is not possible due to the high cost of the overall optimization and complexity of the implementation. We propose to address this issue by developing a generic methodology that applies to any (potentially expensive) physical problem and generates a scalable approximation that can be quickly computed, for which the input and output dimensions may be set independently. This facilitates evaluation of MDO architectures for the original MDO problem by capturing its structure and behavior. The methodology is applied to two academic MDO test cases: the Supersonic Business Jet problem and the propane combustion problem. Well-known architectures (MDF, IDF and BLISS) are benchmarked on various instances to demonstrate the dependency between the performance of the architecture and the problem dimensions.