Species Transport and Reacting Flow

Species conservation

For a mixture with chemical species, NeuralFlow transports independent mass fractions and reconstructs the dependent species from . The conservative species equation is

The last species is dependent so that the composition sum remains exact after admissibility control.

Diffusive species flux

The diffusive mass flux is represented by an effective Fickian coefficient,

For GE turbulence,

The finite-volume face operator uses harmonic interpolation plus a non-orthogonal correction. Species sensible-enthalpy diffusion returns to the mixture energy equation through

Finite-rate chemistry

For reaction , the forward Arrhenius constant is

With molar concentration , the forward rate of progress is

For the current third-body branch, the forward rate is multiplied by

where is the stored third-body efficiency.

For a reversible reaction, the source computes

and subtracts the product-side rate

Species and energy sources

The stoichiometric molar production rate of species is

and the corresponding mass source is

The current split stiff-source path advances the local composition over an interval and constructs the CFD source from the integrated change,

The energy source is reconstructed from species enthalpies,

Cell-integrated contributions are obtained by multiplying these volumetric source rates by the control-volume measure.

Stiff source integration

The chemistry subproblem is a local ODE system for composition and temperature, schematically

A stiff backward-difference integration path advances this system from to . The integrated state change is then converted to the finite-volume source rather than assuming that the beginning-of-step instantaneous rate remains constant over the chemistry interval.

Turbulent finite-rate closure

When turbulent finite-rate chemistry is active with a turbulent flow model, the local chemistry interval is tied to a Kolmogorov-scale time,

In the turbulence-limited finite-rate option, the integrated chemistry contribution is additionally multiplied by a fine-scale reacting-volume fraction. With kinematic viscosity , turbulence time scale , length scale and model constant , this factor limits the volume participating in the reaction and therefore reduces the effective source when turbulent mixing is the controlling process.

The effective chemistry source used by the CFD cell is therefore proportional to .

Mechanism and thermochemical data

Reaction mechanisms, species molecular weights and polynomial thermochemical data are stored in the NeuralFlow mixture/chemical-mechanism model. Standard CHEMKIN-style mechanism and thermodynamic inputs are supported by the supported import path, while the solver consumes the NeuralFlow species/reaction model.

Current coupling limits

The current NeuralFlow model contains laminar finite-rate and turbulent finite-rate volumetric reaction modes. Split volumetric reaction integration is currently disabled when homogeneous VOF is active; nonreacting species transport remains a separate coupled capability.