The Fidelity LES solver is the industry’s first high-fidelity computational fluid dynamics (CFD) solver that expands the practical application of large eddy simulations (LES) to a broad range of engineering applications. Designed to tackle the toughest fluid dynamics challenges, it accurately predicts traditionally complex problems for CFD in aeroacoustics, aerodynamics, combustion, heat transfer, and multiphase. Fidelity LES software introduces a paradigm shift to the industry with the ability to leverage both computer processing units (CPUs) and graphical processing units (GPUs), reducing the turnaround time for LES simulations from days to hours. The solver has been optimized to consume as little memory as possible and scales linearly to hundreds of GPUs across dozens of nodes.

Global Uniqueness

Given the same surface and generating site distribution, the mesh is a unique outcome regardless of partitioning

Inherent Quality for Scale-resolving

Produces grids with inherent quality (face normals aligned with displacement vector, volumes deform continuously under motion) and easily supports regular isotropic topologies in large parts of the volume

Body-fitted at boundaries

No cut-cells or immersed boundaries; a fully body-fitted approach that also support non-convex boundary cells

Scalable for HPC

Algorithmically lends itself to massive parallelism, is designed to leverage the same compute resources as the flow solver (large CPU counts), resulting in less movement of large data files

Fast Mesh Design

Because of Euclidean distance properties of Voronoi diagrams, mesh slices can be previewed very quickly and with less compute, enabling fast design and iteration of meshes

Length Scale Independence

Volumetric and surface feature length scales are decoupled, enabling robust mesh generation for complex geometries and length scale disparities

KEEP Schemes

Use of Kinetic Energy and Entropy Preserving schemes to provide a stable and low-dissipation methodology suitable for a range of multi-physics simulation types

Skew-symmetric Operators

Use of skew-symmetric operators allows for provable stability, enabling between 2nd-4th order accuracy in different regions of the mesh

Scalable for HPC

Designed to be massively parallel on HPC architectures. The GPU-native version of the code (supporting both AMD and Nvidia hardware) is also massively parallel and can leverage cutting edge hardware for cost and time-to-result reduction

Improved Accuracy & Performance vs DDES

Because of the wall modeling paradigm and DDES sensitivity to shielding functions, in many instances our WM-LES has demonstrated improved performance and accuracy over hybrid RANS/LES methods

Wall-modeled Approach

Resolving (with the mesh) only the outer boundary layer structures and leveraging wall-models allows tractable mesh sizes for industrially-relevant simulations

Time-dependent Simulation

Certain quantities of interest, for example acoustics, require a time-dependent approach to accurately characterize the flow features of interest. Our software includes both the solver as well as additional time-dependent post-processing utilities to facilitate extracting actionable information from results

High-speed Flows

Compressible formulation has demonstrated ability to characterize flowfields ranging from subsonic, transonic, supersonic, and even hypersonic regimes.

Aeroacoustics

LES has proven extremely valuable in the simulation of aeroacoustics, and CharLES has been leveraged in both high- and low-speed flows to understand acoustics. Additional post-processing of noise propagation using the Ffowcs Williams-Hawking (FW-H) approach

Leveraging HPC Resources

Successful scaling of flow solvers on CPU resources to 105-106 concurrent ranks
Transitioned code base to leverage GPU acceleration due to cost efficiencies (~10X cheaper on $-basis)

• Voronoi Meshing: Fast, scalable, and robust mesh generation for complex geometries

• LES Solver: Advanced numerical methods and extreme scalability for accurate LES

• GPU Acceleration: Solver advancements leveraging GPU computing for massive LES simulations

• High-Speed Simulations: Accurate simulation of complex scenarios in hours, not days

• Analysis Tools: Robust post-processing capabilities for large datasets

• Dimensional Reduction: Efficient tools to handle and analyze high-fidelity, time-dependent data

• Acoustic Predictions: Efficient implementation of Ffowcs Williams-Hawkings formulation for far-field acoustics

• Quantitative Imaging: Enriches PNGs with custom metadata for rapid computation and data analysis

• Modal Decomposition: Various methods for both image and full-field data