CMPSTheory & Implementation Manual
Source-Aligned CMPS Capability Map
LIKUA HomeManual Home

Source-Aligned CMPS Capability Map

Authority. This map is based on the uploaded CMPS source archive dated 13 August 2026. It describes capabilities visible in the numerical implementation and distinguishes active production paths from explicit compatibility locks.

SubsystemRepresented capabilityManual location
Carrier flowCompressible and constant-density artificial-compressibility formulations; steady and transient coupled implicit solutionChapters 2–6; Time Integration
Pseudo-time accelerationLocal CFL spectral stepping, stretched-grid aspect-ratio acceleration, time-step smoothing and CFL ramp/controlTime Integration
Convective fluxesAUSM family and HLLC for compressible carrier flow; AUSM+up artificial-compressibility fluxChapter 6
ReconstructionFirst order, second order, bounded central differencing; Green–Gauss and least-square gradients; multiple slope/flux limitersChapter 5
Low-speed conditioningSmith-type preconditioning with steady/unsteady reference-velocity control and isentropic-Mach accelerationChapter 5; VOF Model
VOFHomogeneous fully coupled N-phase VOF, Upwind/HRIC/CICSAM, mixed EOS, bounded simplex updates, MPI consistency, restart validationVOF Model; VOF Numerics
Dilute dispersed phaseSeparate phase momentum/thermal equations, drag/heat transfer, interfacial-area transport, breakup/coalescence closuresChapters 7–8
SpeciesN-minus-one species transport, diffusion and enthalpy diffusionChapter 2; Reacting Flow
ChemistryLaminar and turbulent finite-rate volumetric reaction models with stiff local source integrationReacting Flow
ThermodynamicsIdeal-gas, stiffened-gas and constant-density fluid EOS; polynomial species thermodynamics; mixture propertiesChapter 12; EOS and Caloric Models; VOF Model
TurbulenceGE k–omega family, wall treatments, low-Re/production modifications, curvature correction and SAS optionChapters 14–17
Solid thermalSteady/transient conduction, anisotropic conductivity, fluid–solid thermal couplingSolid Energy; Chapter 11
CharringBoundary-attached one-dimensional charring-material responseChapter 11
Boundary conditionsWall, stagnation/mass-flow/velocity inlets, pressure outlet, far field, symmetry, axis, fluid–solid and propellant surfaceBoundary Conditions
Linear solversBlock Krylov, block ILU/Jacobi, AMG and AMG-preconditioned flexible Krylov configurationsChapter 9
Geometric/agglomeration multigridCorrection, FAS, FMG; V/W/F cycles; multiple transfer operators and hierarchy-quality controlsChapter 18
Parallel executionMPI domain decomposition, interface/halo exchange, CPU cluster and NVIDIA CUDA GPU-cluster executionChapter 13
CPU algebraIntel MKL-backed CPU numerical pathChapters 9 and 13
GPU algebraNVIDIA CUDA accelerator path and multi-GPU executionChapter 13
Post-processingPoint probes, surface monitors, mass flow, forces, moments, pressure/temperature averages and aerodynamic coefficientsMonitoring and Integral Reports
Coordinate systemsPlanar and axisymmetric domains; static and rotating cell-zone reference framesReference Frames
NeuralFlow learned solverSolver-trained recurrent GNO path driven by CMPS finite-volume residual and Jacobian feedbackNeuralFlow Learned Solver
Transport propertiesConstant/polynomial Cp; constant/polynomial/Sutherland/Gupta viscosity; constant/polynomial/kinetic/Gupta conductivityTransport Properties
Body forcesCarrier and dispersed gravity; rotating-zone momentum sourceReference Frames
Dispersed fluxesStandard donor/receiver, Rusanov, AUSM and HLLC regularized particle fluxesChapter 8
Monitoring/reportsPoint probes, surface fields, flux reports, forces, moments, center of pressure, drag/lift coefficientsMonitoring and Integral Reports

Surface integrals and monitoring

Surface reports, flux integrals, forces, moments, coefficients and point probes are summarized in Solution Monitoring and Integral Reports.

Coordinate and reference-frame support

CMPS supports planar and axisymmetric domains and static/rotating cell-zone reference frames. The implemented absolute-velocity rotating-zone formulation, grid-relative carrier fluxes, gravity and current compatibility restrictions are documented in Coordinate Systems, Body Forces and Rotating Reference Frames.

How to read this manual

“Implemented” means a source path exists and is wired into the solver configuration. It does not imply that every combination of models is valid. Compatibility gates documented in this manual are part of the numerical definition: a rejected combination must not be interpreted as supported merely because both individual modules exist.

NeuralFlow solver-trained GNO source audit

NeuralFlow is documented as the CMPS-trained recurrent local Graph Neural Operator: it learns the iterative correction process from intermediate solver states, finite-volume mesh and boundary information, conservation imbalance, and exact differentiated residual feedback. The source audit then follows the directed graph, boundary schema, shared recurrent stage, correction map, rollout, training anchor, transpose-Jacobian physics gradient, and implemented physical metric families.

GE-RANS source audit

Chapter 14 was rebuilt from the active turbulence source. The previous placeholder production/compressibility expression was removed and replaced by the implemented Kato-Launder option, dilatation term, production limiter, low-Reynolds coefficients, curvature corrections, axisymmetric terms and viscous-work energy contribution. Chapter 15 now gives the actual scale-adaptive source and velocity-Laplacian length scale.