A Unified Mathematical and Algorithmic Framework for Managing Multiple Information Sources of Multi-Physics Systems
Abstract
Decision processes for complex, multidisciplinary systems draw on multiple information sources, including multifidelity models, historical data, operational data, experimental data, and expert opinions. In many cases there is not just a set of computational models clearly ranked in terms of fidelity; rather, there are multiple sources of information with different types of distortion of the true system. These sources are not commensurable with a scalar-valued measure of fidelity they tell us different things about the problem, with their collective information being greater than the individual parts. This project comprised an integrated research program leveraging the foundations and methods of information theory, decision theory, and machine learning. These elements were brought together in new ways with multidisciplinary design optimization (MDO), multifidelity modeling, uncertainty quantification, and reduced-order modeling. The specific project goals were to: (1) Develop statistical approaches for defining and quantifying fidelity. (2) Establish decision-theoretic methods for optimally managing sources of uncertain multi-physics information. (3) Create reduced models with goal-driven adaptation to multi-physics interactions and with quantified uncertainty. (4) Formulate an information-theoretic approach for handling multi-physics coupling. (5) Create a scalable framework for solving multi-physics analysis and design problems under uncertainty. All goals were achieved. The MURI project made particularly high-impact contributions in developing the methods and practice of multi-information source optimization, laying the mathematical and algorithmic foundations of multifidelity uncertainty quantification, and advancing design and uncertainty quantification of complex multidisciplinary systems.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 17, 2021
- Accession Number
- AD1146062
Entities
People
- Karen Willcox
Organizations
- Massachusetts Institute of Technology