Survey of Dynamic Mode Decomposition Methods

Technical Report · Mon Aug 24 04:00:00 EDT 2020

DOI:https://doi.org/10.2172/1657110· OSTI ID:1657110

Hardy, Zachary Kenneth ^[1]

---

1. Los Alamos National Lab. (LANL), Los Alamos, NM (United States); Texas A & M Univ., College Station, TX (United States)

Dynamic mode decomposition (DMD) is a data-driven reduced order modeling (ROM) technique used for dynamic systems. The widely adopted algorithm was first introduced and demonstrated on fluid flow data by Schmid. In recent years, various other fields, such as nuclear engineering, have begun to adopt this method. For example, DMD has been used for estimating α-eigenvalues, as an ROM for pulsed neutron problems, for predicting isotopic composition in burnup calculations, as acceleration techniques for iterative methods, and in capturing dynamic behaviors in molten salt reactor transients. This report seeks to demonstrate the capabilities and limits of the standard DMD algorithm, and identify problem spaces where variants may be better suited. The primary variant this report considers is Multi-Resolution DMD (mrDMD). Because this serves as a survey, synthetically produced data is used in lieu of simulation results. The remainder of this report will go into detail on the DMD theory, outline the standard DMD and mrDMD algorithms, present test cases highlighting the applicability of each, and finally present a discussion on how to determine the best suited algorithm for a given problem. All calculations performed in this report are carried out using the open source DMD library, PyDMD.

• Research Organization: Los Alamos National Laboratory (LANL), Los Alamos, NM (United States)
• Sponsoring Organization: USDOE
• DOE Contract Number: 89233218CNA000001
• OSTI ID: 1657110
• Report Number(s): LA-UR--20-26751
• Country of Publication: United States
• Language: English

Related Subjects

73 NUCLEAR PHYSICS AND RADIATION PHYSICS
97 MATHEMATICS AND COMPUTING
Dynamic Mode Decomposition