Accelerating Design of ML and AI Experiments in Scientific Simulation

Across industry and scientific fields, Artificial Intelligence (AI) and Machine Learning (ML) have ushered in a sea change in how engineers and scientists interact with their designs and tools. This is felt especially in the field of computational science, where simulations sit on the same silicon as these “thinking” algorithms. In the field of computational fluid dynamics, current Reynolds Averaged Navier Stokes (RANS) closure models typically cannot properly model all regimes of a given flow state. However, it has now been shown that by incorporating high fidelity simulation data into the closure model through machine learning, these closure models can be significantly improved. Furthermore, data-driven models have shown promise of some extrapolatory power, meaning that improvement can potentially be achieved even outside the narrow configuration of their training dataset. That being said, significantly more work is required to understand the interplay between the data-driven model, the simulation, and expectations of robustness and accuracy as the model's application is expanded into increasingly diverse scenarios. This problem is characteristic of many challenges faced by those looking to adopt data-driven techniques into their solvers. As such, in this Phase II effort, we are proposing to significantly advance our in situ AI toolkit demonstrated in Phase I. We plan to enable powerful new capabilities for Catalyst to serve AI and ML technologies more efficiently. Further, we will conduct exciting research into adjoint-driven model sensitivity analysis, which will provide powerful insights into how a data-driven model impacts a simulation state, and how it can be improved given quantities of interest directly relevant to the simulation.

Accurately modeling fluid flow directly supports NASA’s Aeronautics, Space Technology, and Exploration, and is a key concern of the Aeronautics Research Mission Directorate. In order to target this directorate, we are directly employing the NASA FUN3D computational fluid dynamics solver. We not only seek to use NASA’s solver for demonstrations, but to expand the solver’s capabilities with the in situ visualization and analysis platform Catalyst, which will provide a direct connection to modern machine learning capabilities. Specifically, we propose to study Reynold’s Averaged Navier Stokes (RANS) data driven closure models, which pose a fascinating challenge for machine learning approaches, as well as show significant promise to improve the model accuracy of RANS simulations, employed every day by NASA scientists and engineers. Machine learning has demonstrated a sweeping impact across many sectors. Our proposed effort targets a need in aerospace engineering - with impact across other branches of engineering - for more robust and more accurate turbulence models. Design of ever more sophisticated aerodynamic components in wings and aircraft engines for instance undergoes significant iteration. By enabling technologies for users to incorporate high fidelity simulation data into the analysis of many design configurations, the user can be more confident in their results, and iterate more quickly and efficiently. In our Phase II effort, we are targeting the NASA FUN3D solver, we are directly investing in a wide reaching technology utilized from aerospace to wind energy, from military to even the trucking industry. This is an ideal platform for our exploration into adjoint-driven model improvement, thanks to FUN3D’s robust discrete adjoint solver. Adjoints are widely used for complex analyses such as shape optimization or material property tuning. By employing this technology for improved data-driven modeling, along with the many technology advancements we are proposing, we can reduce the barrier for entry to adoption of data-driven models, and empower users to explore what is possible with machine learning enhanced physics.

This is a mature technology (TRL 7+) — the realistic path in is usually NASA's Technology Transfer Program: licensing an existing NASA patent, or a Space Act Agreement to use NASA facilities/expertise directly. NASA also runs a startup licensing program with no upfront fee for companies formed to commercialize a specific NASA technology.

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.