Activity Number:
|
242
|
Type:
|
Topic Contributed
|
Date/Time:
|
Tuesday, August 13, 2002 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Section on Physical & Engineering Sciences*
|
Abstract - #300065 |
Title:
|
Experiment Designs for Simultaneous Forward and Inverse Approximations
|
Author(s):
|
Russell Barton*+ and Martin Meckesheimer
|
Affiliation(s):
|
Pennsylvania State University and The Boeing Company/Phantom Works
|
Address:
|
317 Beam Building, University Park, Pennsylvania, 16802, U.S.A.
|
Keywords:
|
Computer Experiments ; Design of Experiments ; Optimal Design
|
Abstract:
|
Computer experiments can be used to construct approximations or 'metamodels' that are fast-running. The rapid response of the metamodel allows model users to explore the response space more thoroughly than would be possible with the original computer model (for example, for optimization and/or insight). For many engineering design problems there are multiple response variables as well as multiple input variables, and inverse maps are possible and of interest. This talk examines some experiment design strategies for simultaneously fitting forward and inverse approximations for deterministic computer models. The experiment design strategy depends on whether both forward and inverse computer models exist. The presentation will focus on strategies for the case in which only the forward computer model exists explicitly.
|