Framework for Multidisciplinary Design Based on Response-Surface Approximations

Abstract

A framework has been proposed to allow for the multidisciplinary design of coupled, nonhierarchic systems. This approach is based on the ability to decompose a model-based analysis of a coupled system into subspaces or contributing disciplines. These subspaces are defined in terms of the design variables that they can influence and the information they contribute to the characterization of the complete system. The subspace coupling is based on the information they exchange. By using a set of response surface approximations, experts responsible for a particular subspace can make design decisions with the goals of improving the complete system merit and satisfying system constraints. Because design variables can be shared between subspaces, coordination of the subspace design decisions is achieved by the solution of a fully approximate optimization problem involving the complete set of system design variables. The implementation of this framework using two flight vehicle concept design problems is presented.