Computational Fluid Dynamic Simulations for Critical Infrastructure (CFD-SCI)

Project # 24-016 | Year 1 of 3

Sean Breckling^a, John DiBenedetto^b, Clifford Watkins^b, Jorge Reyes^c, James Watts^b

^aNevada National Security Sites (NNSS), ^bSpecial Technologies Lab (STL), ^cVirgina Tech, ^dColorado School of Mines
_{This work was done by Mission Support and Test Services, LLC, under Contract No. DE-NA0003624 with the U.S. Department of Energy, the NNSA Office of Defense Programs, and supported by the Site-Directed Research and Development Program. DOE/NV/03624–1927.}

Abstract

The goal of this project is to develop new capabilities within the complex to address complicated problems requiring expertise in computational fluid dynamics. There are several established topics of interest to date. One focus is the development of a computational model validating a mitigation technique for situations where gasses or contamination cause insufficient breathable air in occupied spaces common in public transportation (Strategic Partnership Projects (SPP) focus). The second topic is concerned with simulating and characterizing air flow in high hazard chemical production facilities well enough such that an accurate digital twin-type model can be deployed. While these simulation and verification use cases will provide much needed understanding of both vulnerable environments and movement of nonproliferation materials into the environment, a more global result will be developing a core Science & Technology capability for airflow simulation as it applies to our sensing needs.

Background

This project brings together previously disconnected teams and skillsets from within the complex. The topics of interest require expertise from chemists, mechanical engineering, computational mathematics, and the physics of fluid dynamics.

Technical Approach

The project will use an iterative approach, combining simulation and limited verification tests. The second objective is to create a group of analysts that can tackle these types of problems as an available team to Nevada National Security Sites testbed stakeholders and others who need better understanding of the environment surrounding Defense Nuclear Nonproliferation (DNN) and Strategic Intelligence Partnership Projects (SIPP) sensing efforts. Important research questions include:

1. Can we simulate confined spaces well enough to devise practical, deployable mitigations to gas threats in public transportation?

Numerical simulations of incompressible fluid flows at these length, temperature, pressure, and velocity scales are a substantial computational challenge. Several well-known tools, e.g., SolidWorks Flow Simulator and COMSOL, have been used successfully to complete such simulations satisfactorily. These methods have been cross validated against numerical simulations from first principles and laboratory experiments.

2. Can we simulate drafty heating, ventilation, and air conditioning (HVAC) systems in production facilities sufficiently such that a full, but computationally-reduced characterization is feasible?

At the prevailing length scales present in production facilities of interest, and near inviscid fluid kinematics, a complete and usefully parameterized computational model is a peta-scale problem. We have the in-house expertise to simulate any particular scenario, e.g., a facility in one solitary configuration. However, given the highly chaotic and nonlinear behavior of fluids, the results of one particular simulation are not broadly useful for our customers.

To develop a well-parameterized “digital twin,” a substantial computational model reduction must be performed. The parameters in question include states of facility gantries (e.g., doors, windows, and HVAC vents), prevailing internal activities (e.g., equipment operation and occupancy), and even the local prevailing weather.

Our proof of concept will be performed in three phases. The first is a completely synthetic fluid model with as few as four parameters (in and out-flow velocity, with sources of heat and cooling). The second is a fully 3D scenario using the same parameter space but is cross-validated against measurements at a low-speed wind tunnel. The third and final phase is a practical characterization and “digital twin” of a simple one-room facility. In Fiscal Year (FY) 2023, “Reduced Order Modelling” (ROM) tools are not widely implemented in popular engineering software. Additionally, while several such tools have been developed in literature, it is not evident that they have been implemented at the scales proposed. Therefore, in the eyes of both SPP and DNN customers, the feasibility of such a capability hinges on the development of these proofs-of-concept.

Results and Technical Accomplishments

We have developed computational tools to simulate near-inviscid flows, at room temperature, at practical length scales. Utilizing finite element solver packages (FreeFem++ and FEniCSx), we now have the in-house ability to solve a host of practical problems.

Simulations of the railcar have been performed successfully. We now have the ability to simulate any practical mitigation strategy, including the “high flow, low speed” approach championed by John DiBenedetto. These simulations are being packaged for a meeting with Los Angeles Metro, in preparation for the 2028 Summer Olympics.

With these tools in place, through extensive simulations, Cliff Watkins and Sean Breckling were able to identify key parameter ranges for data used to train our first and second proof-of-concept ROMs. In FY 2023, one and two-parameter ROMs were implemented successfully. These were promising results, but when the parameter space is so limited, even turbulent flows can be approximated with little difficulty.

Further parameterization requires a substantial computational effort, which began in earnest on the Athena cluster at Los Alamos Operations. The next phase includes the implementation of a continuous data-assimilation technique, to incorporate sensor data common to the facilities of interest.

Conclusions and Path Forward

We are convinced the LESROMs (Large Eddy Simulation – Reduced Order Modelling) are the right approach for our full-scale model. These new techniques couple multiple model reduction methodologies and have shown great promise in literature.

Lessons learned from the first “digital twin” proof-of-concept have guided the construction of our second phase. Incorporating measurement data into the LESROM methodology stands to reduce the number of full numerical simulations required to produce a sufficiently accurate digital twin of the low-speed wind tunnel.

Publications

• Title: A note on the long-time stability of pressure solutions to the 2D Navier Stokes equations
  Journal / Conference: Submitted to the journal of “Applied Mathematics and Computation”
  Year: 2023
  Author(s): Sean Breckling, Joe Fiordilino, Sidney Shields, Jorge Reyes
  

Back to Communications and Computing Index