Robust Design Optimization for Product Platform Design

Georgia Tech Research Corporation, Atlanta GA

Investigators

Abstract

This grant provides funding to develop a systematic yet flexible approach to build a series of metamodels for robust design space exploration and optimization of large systems along a design timeline. Six statistical metamodeling approaches will be evaluated, namely, response surface modeling, kriging, multivariate adaptive regression splines, classification and regression trees, artificial neural networks and wavelets. Both classical and space-filling experimental designs will be tested with the most promising statistical metamodeling approaches and verified for use along a design time-line. The design of families of general aviation aircraft (to verify product platform development), and the design of vehicle body structures (to verify the metamodeling and robust design methodology) are used to evaluate the efficacy of the method. This work, if successful, will provide the proof-of-concept for the development of a formal, generic, mathematically rigorous method for use in rapidly exploring concepts in the early stages of project initiation. The outcome is anticipated to have an impact both on the engineering design community and the statistical community bringing both communities closer to addressing real problems with tools anchored in a rigorous and mathematically correct foundation. In addition, the successful use of this method by manufacturing enterprises will allow them to rapidly and cost-effectively design product platforms thereby enhancing their competitiveness.