CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   Does the Kinematic Diffusivity work well in CFX 10 (

Michael Skoblin February 27, 2007 08:41

Does the Kinematic Diffusivity work well in CFX 10
Hello to everybody. I am trying to model a steady-state premixed methane-air flame at 1 atm using CHEMKIN's chemistry. My approach assumes the Variable Composition Mixture in CFX-10 and building the source terms in energy and species continuity equations. In addition, I need to take into account species diffusion, so that I set the Kinematic Diffusivity for every species (D~10^-5 m^2 s^-1). I implement both sources and diffusion coefficients via FORTRAN routines. The problem I faced with is that I got VERY high peak flame temperatures, 3000 K instead of ~2200 K expected. Of course, first of all I have checked my own routines, and found nothing wrong. As the energy and species sources can be easily checked, the only thing in which I can suspect a bug is the diffusion. Does anybody have any experience in modeling a gaseous mixture with diffusion? Does the Kinematic Diffusivity work well? With best regards, Michael.

opaque February 27, 2007 09:57

Re: Does the Kinematic Diffusivity work well in CF
Dear Michael,

Are you aware that ANSYS CFX includes the enthalpy of formation as part of the static enthalpy? Therefore, you got to be really careful if you plan to introduce sources in either the energy or species equations.

Did you create your own version of the species, or modify the reference enthalpy/entropy values to account for the fact that you are introducing reaction sources ?

Is your flow turbulent or laminar? I imagine must be laminar, otherwise, the kinematic diffusivity is negligible in the main flow. Also, when you input the kinematic diffusivity you must have realized that CFX uses species diffusion into the bulk; therefore the model is quite different that multicomponent diffusion "ala" Stefan-Maxwell as CHEMKIN tries to model with the mixture rule approximation..

Just food for thought,


Michael Skoblin February 27, 2007 14:50

Re: Does the Kinematic Diffusivity work well in CF
Thank you very much, Opaque, for your quick response.

Indeed, I was unaware of that the enthalpy of a species includes the formation enthalpy. It seems to me now that I add the energy twice. At first changing the composition and then introducing energy to the energy equation as the integral of Rho*Cp*dT. (It is Chemkin that makes a "chemistry step" with initial conditions from the current CFX solution field). A good food for thought, indeed.

I use the CFX inherent species library. And, frankly speaking, I do not understand now (maybe, because of the late time) why I have to modify the reference values of the species's enthalpy/entropy to take into account composition changes... I have to think on this.

The flow is laminar. "Diffusion into the bulk" means that the diffusive species is considered as a minor fraction? So that ANSYS CFX does not care that the summary diffusive flow should not carry mass?

Thank you very much again. Michael.

opaque February 27, 2007 15:45

Re: Does the Kinematic Diffusivity work well in CF
Dear Michael,

When doing composition changes, all the species must a have a consistent reference state (whichever the reference state is). If you use the species library from CFX, then you do not need to the reference state nor add any reaction sources to the energy equation.

What you do with the species equations depends on how you set them up.. Are the reactions rate computed by CFX, or by other code?


Michael Skoblin February 28, 2007 05:30

Re: Does the Kinematic Diffusivity work well in CF
Dear Opaque, my approach, in shorts, is as follows:

I do NOT use any CFX reactions. At every false time iteration, when mass fractions are to be solved, ANSYS CFX calls my routine. The latter takes the current solution fields (mass fractions of species, temperature and pressure) and uses them as the initial conditions for Chemkin chemistry code. The species in Chemkin and ANSYS are the same (same NASA coefficients for Cp and other thermodynamic properties). As I use a Physical Time Step option, I can have a false timestep dt at any iteration and can ask Chemkin to modify the current composition. Within Chemkin I can compute the partial density change dRhok = integral of Rho*dYk, where Yk is the mass fraction of the k-th species. Then dRhok/dt is used as a source term in the species continuity equations. Earlier, however, I added a similar term to the energy equation, but now I understand that it was the mistake, which led me to overheating. Started yesterday evening, my corrected simulation shows today pretty low peak temperatures of the methane-air flame - 2100 K. Thanks, Michael.

All times are GMT -4. The time now is 09:33.