| Complex systems design poses an interesting as well as demanding information management problem for system level integration and design. The high interconnectivity of disciplines combined with the specific knowledge and expertise in each of these calls for a system level view that is broad, as in spanning across all disciplines, while at the same time detailed enough to do the disciplinary knowledge justice. The treatment of this requires highly evolved information management and decision approaches, which result in design methodologies that can handle this high degree of complexity.; The solution is to create models within the design process, which predict meaningful metrics representative of the various disciplinary analyses that can be quickly evaluated and thus serve in system level decision making and optimization. Such models approximate the physics-based analysis codes used in each of the disciplines and are called metamodels since effectively, they model the (physics-based) models on which the disciplinary analysis codes are based.; The thesis formulates a new metamodel implementation process to be used in complex systems design, utilizing a Gaussian Process prediction method. It is based on a Bayesian probability and inference approach and as such returns a variance prediction along with the most likely value, thus giving an estimate also for the confidence in the prediction. Within this thesis, the applicability and appropriateness at the theoretical as well as practical level are investigated, and proof-of-concept implementations at the disciplinary and system levels are provided. |
