NeuralFlowTheory & User Reference Manual
Governing Carrier-Flow Equations
LIKUA HomeManual Home

2 Governing Carrier-Flow Equations

2.1 Control-volume form

Let be a finite control volume with boundary and outward area vector . The compressible carrier equations solved by NeuralFlow are written in conservative balance form. For compactness, define where is density, is absolute velocity, and is specific total energy. Turbulence and species equations extend this carrier vector when active.

The integral balance is (2.1)

2.2 Continuity

The mass equation is or, for one control volume, (2.2)

The constant-density artificial-compressibility regime does not use Eq. (2.2) as a physical density evolution equation. Its pressure/continuity equation is given in Chapter 4.

2.3 Momentum

For a Newtonian fluid, (2.3) with viscous stress (2.4) Here is the molecular viscosity for laminar flow and contains the modeled turbulent contribution when the active turbulence closure requires it. Equation (2.4) is kept in tensor form in this guide; replacing it by a Laplacian is only valid under additional assumptions on viscosity and divergence and is not used as a general identity.

The control-volume momentum flux through a face is therefore In NeuralFlow the inviscid part is replaced by a numerical face flux, while viscous terms are evaluated from reconstructed gradients and effective transport properties.

2.4 Total energy

The compressible total-energy equation is (2.5) where Fourier heat conduction is For a reacting or multi-species formulation, the energy flux may additionally contain enthalpy carried by species diffusion. NeuralFlow constructs these contributions in the same coupled residual so that their state dependence can enter the AD Jacobian.

Using total enthalpy the inviscid energy flux is . This is the form used naturally by AUSM-family fluxes.

2.5 Species transport

For chemical species, NeuralFlow may solve independent mass-fraction equations and close the final fraction by (2.6) For an independent species , (2.7) with diffusive mass flux and source . Passive-species mode sets the chemical source to zero. The representation preserves the algebraic sum of mass fractions but requires diffusion and thermochemistry terms to remain consistent with the dependent species.

2.6 Turbulence extension

When GE – RANS is enabled, transport equations for turbulent kinetic energy and specific dissipation rate are added to the coupled block. Turbulent stress and effective transport properties then modify Eqs. (2.3) and (2.5). NeuralFlow also includes the turbulent kinetic-energy contribution in the carrier total enthalpy used by the GE-RANS flux path. The detailed closure is given in the GE-RANS chapter.

2.7 Finite-volume residual

For a cell , the spatial residual used throughout the remainder of this guide is defined as (2.8) The sign convention used by the current NeuralFlow matrix assembly is (2.9) so the Newton-like correction is . This convention is developed in detail in Chapter 3.