Solvers¶

Arxyne dispatches simulations through its solver registry (defined in arxyne_platform/core/solver.py). Each solver runs in a Docker container with a JSON config in / JSON results out interface.

v0.1.0 Stability¶

Tier	Meaning
Required	Must be present — core experience depends on it
Recommended	Useful, not essential for first run

Solver Licenses¶

Each bundled solver carries its own license. Arxyne distributes solver Docker images separately from the platform (BSL 1.1). Solvers are not covered by the Arxyne platform license — you must comply with each solver's terms.

Solver	License	Notes
OpenFOAM	GPLv3	Included, open source
XLB	Apache 2.0	Included, open source
Chrono	BSD-3	Included, open source
DoMINO	NGC required	NVIDIA NIM, user provides NGC API key

Solver Registry¶

arxyne --solvers

Domain	Solver	Version	Image	Stability
`aero`	OpenFOAM	v2312	`arxynehq/solver-openfoam`	Required
`aero`	XLB (GPU LBM)	main	`arxynehq/solver-xlb`	Recommended
`aero`	NVIDIA DoMINO NIM	2.1.0	`nvcr.io/nim/nvidia/domino-automotive-aero`	Recommended
`dynamics`	Project Chrono	9.0.1	`arxynehq/solver-chrono`	Recommended

Planned domains (not yet implemented): structural FEA, thermal CHT, crash, NVH. These will ship in future releases.

OpenFOAM (Aerodynamics)¶

Domain: aero KPIs: cd, cl, cm_pitch Method: RANS (simpleFoam), steady-state

Dispatch¶

arxyne --product ahmed --aero
arxyne --product ahmed --aero --solver openfoam

XLB (Aerodynamics — GPU LBM)¶

Domain: aero KPIs: cd, cl Method: Lattice Boltzmann (KBC), unsteady

GPU-native LBM solver. Known limitations: incompressible only (no transonic), no CmPitch output (moment computation not implemented). Half-models (ONERA M6) use an EquilibriumBC symmetry approximation. Cl under-predicts vs OpenFOAM due to stationary ground (no moving ground BC on uniform grids at automotive Re).

Dispatch¶

arxyne --product ahmed --aero --solver xlb     # local GPU

NVIDIA DoMINO NIM (Aerodynamics — AI Surrogate)¶

Domain: aero Stability: Recommended KPIs: cd, cl, surface Cp, wall shear Method: Neural operator inference

arxyne --product drivaer-notchback --aero --solver domino

Requires NVIDIA GPU (>= 8 GB VRAM) and NGC API key. Returns results in seconds instead of minutes — useful for design-space sweeps.

Project Chrono (Vehicle Dynamics)¶

Domain: dynamics Stability: Recommended KPIs: lateral_g, yaw_rate_dps, roll_angle_deg, speed_kmh Method: Multi-body dynamics (Project Chrono 9.0.1)

Dispatch¶

arxyne --product drivaer-notchback --dynamics

Runs all four ISO manoeuvres (step steer, lane change, steady-state cornering, cruise). Each manoeuvre writes a USD sublayer with time-sampled chassis transforms and KPIs.

Validation Data¶

Known solver-vs-experiment comparisons for the bundled reference products:

Product	Solver	KPI	Simulated	Experimental	Error
Ahmed 25°	OpenFOAM	Cd	0.278	0.295	5.8%
Ahmed 25°	XLB	Cd	0.278	0.295	5.8%
DrivAer notchback	OpenFOAM	Cd	0.275	0.260	5.8%
DrivAer notchback	XLB	Cd	0.270	0.260	3.8%