It is now possible for a designer to explore and then generate a ranged set of system and subsystem top-level specifications for a large scale complex system, especially one in which an engine is either a system or subsystem. A careful evaluation of statistical experimentation and approximation techniques for deterministic computer experiments and methods for Hierarchical Robust Preliminary Design Exploration (HRPDE) has been completed. This evaluation has highlighted the dichotomy between the accuracy of metamodels and computational complexity in simulationbased robust design. To enhance the accuracy of meta-models in the design of complex systems the use of kriging has been investigated and compared with results using response surface methodology. For use of the HRPDE in practice further investigation to deal with this dichotomy is warranted, and an extension of the original proposal has been granted to accomplish these studies.
展开▼