New Concepts in Computer Simulation-Surrogates
Abstract
Although the advent of fast and inexpensive parallel computers has rendered numerous previously intractable calculations feasible, many numerical simulations remain too resource intensive to be directly inserted into engineering optimization efforts. An alternative to direct insertion considers models for computational systems: the expensive simulation is evoked only to construct and validate a simplified input output model; this simplified input output model then serves as a simulation surrogate in subsequent engineering optimization studies. As compared to the direct insertion approach, surrogates offer more complete, efficient, and robust optimization, greater accommodation of prior information, broad applicability to families of objective functions, and a more interactive, flexible design environment. In this project, a validated surrogate methodology was developed which permits economical and reliable integration of large-scale, possibly noisy numerical simulations into engineering design and optimization studies. The project included both the formulation, analysis, and implementation of construction-validations algorithms, and the application of the resulting techniques to problems of scientific and engineering interest. The sample application are intended to be sufficiently simple to permit numerical experiment, yet sufficiently complex to properly stimulate, constrain, and demonstrate the proposed methodology.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 03, 1997
- Accession Number
- ADA335679
Entities
People
- Anthony T. Patera
Organizations
- Massachusetts Institute of Technology