Activity title | Multi-Fidelity Methods for Multidisciplinary Design Optimization |
Activity Reference | AVT-385 |
Panel | AVT |
Security Classification | NATO UNCLASSIFIED |
Status | Active |
Activity type | RLS |
Start date | 2023-01-01T00:00:00Z |
End date | 2025-06-30T00:00:00Z |
Keywords | Multidisciplinary Design Optimization, MultiFidelity Methods, Surrogate Modeling |
Background | Multidisciplinary Design Optimization (MDO) approaches have proliferated for the design of vehicles operating in all domains to meet ever-increasing user requirements. Such requirements may drive systems into states involving physics costly to model and simulate. Consequently, multi-fidelity (MF) methods have rapidly emerged to broadly use traditional analysis tools while sparsely leveraging higher fidelity and higher cost approaches to capture challenging physics in an affordable manner. Many different MF strategies have appeared, warranting classification, explanation, and comparative study of strengths and weaknesses. |
Objectives | The objectives of the Lecture Series are: (1) to overview MF methods and the application of MF methods to relevant design problems; (2) to overview existing methods for surrogate modeling; (3) to study the mathematical character of existing and emerging MF methods; (4) to examine the application of MF methods to a variety of benchmark problems; and (5) to explore research topics in MF methods and enabling technologies. Lecture series notes are expected to be of lasting and broad utility. |
Topics | • MF method nomenclature and principles, MF method performance criteria
• Survey of MF strategies and enabling tools
• Examples drawn from selected analytic functions and literature cases
• Surrogate modeling fundamentals, including Gaussian process approaches, neural nets
• Adaptive sampling and active learning approaches for design space exploration and optimization; Bayesian optimization
• MF surrogate methods for cost-sensitive design studies
• Comparative study of MF methods applied to analytic functions
• Vehicle benchmark design problems
• Comparative study of MF methods applied to vehicle benchmark design problems
• Computational frameworks for MF design
• MF methods for Uncertainty Quantification and design under uncertainty
• Research in MF methods
|
| Contact Panel Office |
