Prediction Uncertainties beyond the Range of Experience: A Case Study in Inertial Confinement Fusion Implosion Experiments
Scientists often predict physical outcomes, e.g., experimental results, with the assistance of computer codes that, at their best, only coarsely approximate reality. Coarse predictions are challenging in critical part due to the multitude of seemingly arbitrary yet consequential decisions that must be made such as choice of relevant data, calibration of code parameters, and construction of empirical discrepancy forms. In this paper, we present a case study in the context of inertial confinement fusion (ICF) implosion experiments where extrapolative predictions are needed with quantified uncertainties. The purpose of this case study is to reflect relevant statistical methods, as applied to ICF model fitting and prediction, to document the numerous decisions that must be made in the prediction pipeline, to extend a complex example in extrapolation to the uncertainty quantification (UQ) community, and to reflect on the challenges we encountered supporting extrapolations with imperfect models and thereby recommend several future research directions. We conclude with a discussion about the UQ community's role in less than ideal predictive scenarios like our ICF exercise.
- Research Organization:
- Los Alamos National Lab. (LANL), Los Alamos, NM (United States)
- Sponsoring Organization:
- USDOE
- Grant/Contract Number:
- 89233218CNA000001
- OSTI ID:
- 1524384
- Report Number(s):
- LA-UR-17-30694
- Journal Information:
- SIAM/ASA Journal on Uncertainty Quantification, Vol. 7, Issue 2; ISSN 2166-2525
- Publisher:
- SIAMCopyright Statement
- Country of Publication:
- United States
- Language:
- English
Web of Science
Analysis of NIF scaling using physics informed machine learning
|
journal | January 2020 |
Similar Records
A New Appraisal- Lessons from the History of Efforts to Value Green and High-Performance Home Attributes in the United States
DiaMonD: An Integrated Multifaceted Approach to Mathematics at the Interfaces of Data, Models, and Decisions