STIR: Advances in Stochastic Computational Electromagnetics Modeling Techniques
Abstract
The predictive power of computational electromagnetics (CEM) techniques has improved such that numerical artifacts are no longer the limiting factor in the accuracy of the simulation. Rather, simulation accuracy is limited by the inherent uncertainty in the input parameters. Device performance is subject to statistical uncertainty that is not accounted for by deterministic CEM modeling techniques. We conducted a comprehensive study and review of the state of the art in emerging stochastic CEM techniques for uncertainty quantification (UQ). Our review of existing techniques motivated 1) the creation of a classification framework for UQ techniques in CEM, 2) the analysis of accuracy limitations of a particular class of techniques within this framework Ð namely polynomial basis-function-expansion techniques, 3) the development of a new sensitivity analysis technique based on an electric-field-integral-equation approach, and 4) the development of a new sampling-based UQ method that is suitable for problems with modest to large variability in the CEM simulated field values. Our project culminated with the development of a set of research recommendations that outlines the key challenges and research questions as well as the potential impact of future breakthroughs. Development of robust CEM techniques for quantifying uncertainty in electromagnetic device design would be truly transformative.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 14, 2019
- Source ID
- W911NF1510234
Entities
People
- Susan Hagness
Organizations
- Army Contracting Command
- Defense Advanced Research Projects Agency
- University of Wisconsin–Madison