Reduced Order Hydrodynamic Model for Nuclear Weapons Effects Relevant Simulations
Abstract
DTRA currently needs nuclear weapons effects (NWE) models that fill the gap between very simple, fast-running, low-order models and high-fidelity, slow-running, first-principles models. The fast-running models do not provide enough fidelity and the first principles models run far too slowly when quick turnaround time is required. Recent advances in computing combined with a resurgence of data-driven algorithms (machine learning) have enabled new analysis techniques in many fields of science and engineering. Of particular interest are problems where large datasets are leveraged to reduce computational time/cost of a first-principles model, achieving a model order reduction (MOR). Models that employ these data-driven techniques, often called reduced order models (ROMs) can provide a mathematically sound approach to approximating high-dimensional partial differential equations (PDEs) with low-dimensional models. These data-driven techniques can be applied to hydrodynamic systems of equations, such as the Euler system. The challenge with NWE relevant hydrodynamic simulations is that they often involve travelling waves, moving shocks, or sharp gradients and discontinuities. Although these types of physical features have previously been difficult to reduce with data-driven techniques, development of the Dynamic Mode Decomposition (DMD) algorithm has led to improved capabilities for their reduction [1, 2]. Furthermore, the recent development of Multi-Resolution Dynamic Mode Decomposition (mrDMD) [1] has improved DMD further by enabling the tracking of coherent simulation features with more precision and achieving improved model reduction properties. The goal of this proposal is to develop mrDMD and related methodologies for use with NWE relevant hydrodynamic simulations, and implement these methodologies in a simulation framework. The result will be a reduced order modeling tool for simulating NWE hydrodynamic simulations. Having a reduced order modeling tool would allow an operator to adjust model fidelity to the appropriate level for the given amount of computational resources available in a given time, thus enabling better tactical decision making for Counter Weapons of Mass Destruction (C-WMD) applications.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 16, 2019
- Source ID
- HDTRA11810038
Entities
People
- Weston Lowrie
Organizations
- Applied Research Associates (United States)
- Defense Threat Reduction Agency