Ongoing Research Pursuits

Modeling, Control, and Prediction of Plasma-Coupled Combustion Transients in High-Speed Flow Turbulence

Large-scale numerical simulations are starting to enable the prediction of multi-physics phenomena. Their interactions are usually so complex that current theory and models have a limited predictive capacity. Thus, the Center for Exascale Simulation of Plasma-Coupled Combustion (XPACC), a Multidisciplinary Simulation Center within the Predictive Science Academic Alliance Program II (PSAAP II), is developing high-fidelity numerical simulations in conjunction with exascale-aimed computer science tools and uncertainty quantification (UQ) analyses, such as adjoint-based sensitivity, to provide a route to prediction. Specifically, our prediction is the transient ignition kernel (TIK) characteristics in plasma-coupled, high-speed flow turbulence applications (like those anticipated in scramjet designs). These experiments provide the TIK prediction target, which is only revealed after the simulation prediction is complete.

XPACC is co-directed by Prof. Jonathan Freund (MechSE and AE, University of Illinois at Urbana-Champaign). Simulation results presented here were produced by D. Buchta.

Identify Mechanisms of Pressure Radiation from High-speed Turbulence and Pathways to Quiet

Adjoint-Based Sensitivity Analysis

The utility of large-scale simulations for engineering is limited, especially as the simulation cost and number of uncertain input parameters increases. Put simply, the cost of a brute force design optimization (perturbing inputs to locate the maximum value of the output) would scale with the number of parameters. Instead, leveraging the fact that any set of parameters must also solve the initial boundary value problem, the solution to the corresponding adjoint system provides the local gradient, independent of the number of inputs. The adjoint-based gradient is used to automatically optimize (minimize or maximize) the quantity of interest (QoI). For engineering purposes, typical QoI involves the aeroacoustic intensity, such as those flows near military aircraft, or the amount of consumed reactants indicating a successful ignition (e.g. combustion applications such as scramjets). For uncertainty quantification, the adjoint-based sensitivity provides a useful measure for dimension reduction near the nominal inputs. Ongoing research uses this gradient to improve models by automatically reducing differences between simulation and experimental data.

Data Processing to Improve Models in Numerical Simulations

Experimental data is essential to model validation and improvement. Usually, the data represents a quantity not mathematically equivalent to the simulated quantities. One such example comes from schlieren photography. Schlieren images recorded by a camera are plane-integrated light intensity values (corresponding to a color map used for visualization). The light intensity is proportional to the density gradient magnitude. Gradients in density refract light at different angles. For simulations of the Navier-Stokes equations, the solution requires additional post-processing of the density field, for example. Most often, there is no tractable way to convert the real-valued simulation data (density, momentum, energy, etc.) to experimental (red, blue, green, alpha) pixel images in a meaningful way. Therefore, we are examining the dynamics of experimental video data (i.e., space-time correlations, energetic modes through proper orthogonal decomposition, etc.) to facilitate a comparison to simulation dynamics. The comparison is thus a sketch of the experimental and simulation databases, which are often very large, to inform missing or unknown physics (such as inflow boundary conditions) and to further improve designs of such configurations. Open-source online video databases, such as those at NASA, are being used to test these concepts. 

Simulations of Ionian Plumes

Volcanic plumes on Io, the innermost of the four Galilean moons of the planet Jupiter, form through gas dynamics interactions with a gravity field. We used continuum simulations to investigate these mechanisms, with a focus on linking unknown source conditions to far‐field observations such as plume height, width, and (unexpectedly) multiple deposition rings. Solutions from axisymmetric simulations suggest that concentric-like rings can form through spectacular fluid mechanics. Under the effect of Io's gravity, the gas falls around the main expansion zone, re-compressing near the surface with subsequent expansions and compressions as the gas flows radially from the source. Multiple radial compression zones are one possible mechanism forming rings of material deposition. 

Previous Research Pursuits

Development of Advanced Probabilistic Risk Assessment Models to Predict Human Health Consequences

While working in industry, our group developed risk assessment models related to a comprehensive range of chemical, biological, radiological, and nuclear defense (CBRNE) scenarios, often using the Chemical Terrorism Risk Assessment (CTRA), a quantitative risk assessment (QRA) approach to estimate the risk among chemical terrorism attack scenarios and assist in prioritizing mitigation strategies.

• Applying the CTRA for Food Safety and Defense. Current and potential application of both the CTRA Food Consequence Model and CTRA risk results were presented to illustrate a number of model outputs, including the impact of recall and various chemical parameters. These results provide an improved understanding and visualization of the timing of the food contamination event and the windows for mitigation at various stages.
• Modeling Retailer and Consumer Behavior in a Food Contamination Event. A mathematical model describing consumer and retailer behavior during food contamination was developed as part of the CTRA. This tool can be utilized to prioritize investments in mitigating consequence during a food contamination event, provide insight into relative benefits of enhancing investigations in order to identify recalls, and improve risk communication of recalls.
• Foodborne Contamination Consequence Modeling. Intentional foodborne contaminations have been attempted in the United States and abroad. We developed a foodborne contamination consequence model that estimates the human health consequences of an intentional food contamination scenario. 
• Chemical Supply Chain Incident Model for Human Health Consequence Estimates. In order to quantify risk, human health consequences were estimated from releases and subsequent dispersion of toxic industrial chemicals from targets, such as chemical production facilities, bulk highway transport vehicles, rail cars, barges, and pipelines. An outdoor inhalation consequence model that incorporates key aspects of source terms, plume dispersion, meteorology, population modeling, and release location was developed. The general framework allows for many future applications, such as estimating the risk along an entire rail-line or reduction in risk caused by new policy restricting tanker trucks carrying hazardous materials around specific urban areas.