Development of High-Fidelity, Machine-Learning Compatible Computational Fluid Dynamics Software Modules based on AMReX

Modeling turbulent flows with complex geometries accurately and cost-effectively is a crucial requirement in various engineering fields. Among the essential steps in any Computational Fluid Dynamics (CFD) simulation, mesh generation and refinement are often the most critical, though resource intensive. Additionally, the complexities involved in modeling the physics of such complex engineering flows pose significant challenges. Moreover, the computational overhead and turnaround time associated with solving high Reynolds number turbulent reacting flows can become formidable obstacles for practical applications. Many of the currently available commercial/academic CFD solvers suffer from non-intuitive interfaces requiring steep learning curve for non-expert users. To address these concerns, there is a pressing need for a cost-effective, easy-to-use, high-order accurate, and high-performance CFD tool. The objective of this SBIR/STTR initiative is to develop an advanced Computational Fluid Dynamics (CFD) software application capable of high-fidelity simulations encompassing multiple physics phenomena and designed for high Reynolds number turbulent compressible flows, which often involve complicated stationary or moving wall boundaries. To achieve this, we plan to leverage the capabilities of two scientific computational codes: AMReX, a scientific library under development at Lawrence Berkeley National Lab (LBL), and our in-house Cartesian Grid Euler (CGE) code. During Phase I of the SBIR/STTR project, we will showcase several of the features mentioned above. Subsequently, in Phase II, we will further enhance the CFD framework and introduce multi-physics models. Importantly, the CFD application's versatility will be underpinned by AMReX's massively parallel and hardwareagnostic framework, enabling deployment on various platforms, including laptops, cloud environments, or HPC clusters. A successful Phase I/Phase II endeavor will result in a versatile computational tool deployable across various computing scales, ranging from peta and exascale systems to small to medium-sized high-performance clusters, laptops, and cloud platforms. The primary outcome of this innovative capability will be a software package strategically designed to harness CDI's established customer base within the wind energy, nuclear engineering and aerospace sectors. Simultaneously, it will substantially enhance CDI's market presence in broader engineering and scientific domains, capitalizing on the expertise and capabilities offered by AMReX and attracting a more diverse user base. Continuum Dynamics, Inc. estimates cumulative sales revenues of $8.0M and additional cumulative software licensing revenues of $6.4M during the first 10 years of commercialization.