Rapid Parametric Sensitivity Analysis for Plume-Surface Interaction Simulations

NASA modeling and simulation activities are mandated to provide uncertainty characterization, quantification (UQ), and propagation for all of their simulation tools and results. In the Phase I of this project, CFD Research addressed this need by implementing two approaches for sensitivity analysis into the Gas Granular Flow Solver, Loci/GGFS, used by NASA for prediction of Lunar and Martian Plume-Surface Interaction (PSI) effects such as dust lofting, obscuration, debris transport, and surface cratering. The first method, the intrusive methodology Forward Automatic Differentiation (FAD), enables run-time sensitivity analysis and propagation of the underlying sub-model uncertainties through a simulation in a minimal number of runs. The second method was the nonintrusive Sensitivity Quantification for Uncertainty Analysis Toolkit (SQUAT). Both approaches quantified sensitivities in a PSI validation problem. In Phase II, CFD Research will mature both methodologies. The efficiency and applicability of FAD will be improved for a broad class of problems in Loci/GGFS and other Loci solvers including Loci/CHEM, which is used for a variety of applications by NASA. SQUAT will be extended to work with all Loci-based solvers. Both uncertainty analysis methods can also be adapted for implementation or integration with other CFD solvers to enable critically needed UQ and sensitivity analysis for a wide range of NASA and non-NASA applications. At the end of this project, a full suite of UQ tools will be available to the analyst for sensitivity analysis, allowing identification of dominant sub-model contributors of uncertainty, guide improvements, and provide a rapid propagation of critical uncertainties to the simulation output metrics. The resulting tools will be delivered to NASA for ready application to analysis of Lunar and Martian landers, including the Human Lander System, to aid in quantifying and propagating uncertainties in current simulations. NASA modeling and simulation standards mandate providing uncertainty quantification (UQ) through characterization and propagation of uncertainties present in underlying submodels. Current simulation tools supporting HLS lack efficient means for rapid,  practical uncertainty assessment of simulations that rely on physics submodels with limited fidelity and considerable uncertainties. The Phase II project provides significant improvements to the Forward Automatic Differentiation (FAD) sensitivity method in Loci/CHEM and Loci/GGFS. This enables rapid uncertainty analysis for a broader class of problems. When used in conjunction with the non-intrusive Sensitivity Quantification for Uncertainty Analysis Toolkit (SQUAT), our approach enables identification of the dominant uncertainty contributors through efficient means. Our approach will 1) enable run-time sensitivity analysis with FAD, 2) identification of dominant submodel contributors with FAD and SQUAT, 3) provide rapid propagation of critical uncertainties, and 4) provide UQ application support for NASA on production level problems.  Phase II objectives are to significantly improve the implementation of FAD and SQUAT workflow for Loci/GGFS and Loci/CHEM. Using both methods, the entire range of parameter space and problem size for complex simulations supporting human landing systems will enable propagation of uncertainties through both solvers. The Phase II work plan includes: 1.    Implement identified improvements to the underlying models, flux Jacobians, and flux limiters currently in Loci/CHEM and Loci/GGFS  for propagating FAD-enabled sensitivity derivatives.  2.    Implement specific improvements to the iterative convergence performance of Loci/CHEM and Loci/GGFS. 3.    Develop and implement a linearized approach for solving for FAD sensitivities for Loci/CHEM and Loci/GGFS.  4.    Implement enhancements to I/O APIs in Loci/CHEM and Loci/GGFS for FAD-enabled computations of sensitivity derivatives. 5.    Extend SQUAT method for usage with all Loci-based solvers and to better enable integration with other CFD solvers and physics-based tools. 6.    Deploy both sensitivity approaches on NASA HPC systems and work with NASA personnel to apply them to production level cases. Deliverables include: 1) Improved FAD capability in Loci/CHEM and Loci/GGFS 2) Improved SQUAT capability for all Loci-based solvers,  and 3) Production support for UQ tools for NASA analysist.  

Immediate NASA applications include the improvements across the Loci-based solver family for a broad range of numerical simulations, especially in determining uncertainties present in models used therein. This work can be extended to other solvers with similar benefits. Identification and understanding of model uncertainties will have a direct impact on missions requiring propulsive landing and take-off, such as the Commercial Lunar Payload Services landers, for the Human Lander System, and future Martian robotic and human landers. Uncertainties persist in a wide range of non-NASA sand and dust related military and civilian applications such as rotorcraft brownout, engine dust ingestion, and obscuring the warfighter. In addition, multiphase flows occur in many applications in chemical, and fossil-energy conversion industries where accurate physics modeling plays a huge role in the flow behavior of real particulate systems.

This is early/mid-stage (TRL 3) — the most realistic path in is NASA SBIR/STTR, which funds small businesses and research institutions to develop technology aligned with NASA's needs (equity-free, phased funding). Check whether a current SBIR/STTR solicitation topic overlaps with this project's technology area, or contact the project directly (above) to ask.

None of these are guaranteed paths for this specific project — TechPort itself doesn't have an "apply" button. Reaching out to the contact(s) above with a specific question is usually the fastest way to find out what's actually open.