Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Multi-Physics Surrogate Modeling with Dimensionality Reduction

Journal Article · · Transactions of the American Nuclear Society
OSTI ID:22991997
;  [1]; ;  [2]
  1. Purdue University, 400 Central Drive, West Lafayette, IN 47906 (United States)
  2. Idaho National Laboratory: 2525 North Fremont Ave., Idaho Falls, ID 83414 (United States)

This summary provides a new algorithm to construct surrogate models for multi-physics, nonlinear, high dimensional models which typically suffer from the curse of dimensionality. To deal with the high computational cost and difficult error estimation, we employ recent advances in reduced order modeling (ROM) techniques to do an error-preserving reduction for the number of influential model parameters, followed by a conventional surrogate construction process. To implement ROM for a multiphysics model, a range finding algorithm (RFA) is used to identify a linear transformation in which the number of influential parameters can be minimized and the errors resulting from the non-influential parameters can be upper-bounded. The proposed algorithm is applied to the MAMMOTH coupled code system developed at the Idaho National Laboratory. In particular, we investigate the reducibility of Rattlesnake and BISON codes and their multi-physics coupling. The most important result of this work is that the reduction obtained with coupled physics is more significant than that with individual physics models. This summary explores the application of ROM techniques on multi-physics coupling codes system. Concise methodology is provided and selected numerical results are provided to show that coupling system provides additional reduction at interfaces. Future work could be invested to determine the optimum surrogate model format. (authors)

OSTI ID:
22991997
Journal Information:
Transactions of the American Nuclear Society, Journal Name: Transactions of the American Nuclear Society Journal Issue: 1 Vol. 114; ISSN 0003-018X
Country of Publication:
United States
Language:
English

Similar Records

Dimensionality reducibility for multi-physics reduced order modeling
Journal Article · Wed Jul 19 00:00:00 EDT 2017 · Annals of Nuclear Energy (Oxford) · OSTI ID:1400266

ROM-based Subset Selection Algorithm for Efficient Surrogate Modeling
Journal Article · Fri Jul 01 00:00:00 EDT 2016 · Transactions of the American Nuclear Society · OSTI ID:23042942

Surrogate Models for TREAT Transient Calculations
Conference · Sun Jun 11 00:00:00 EDT 2017 · OSTI ID:1635347