CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   Temperature depending density (

Birnensuppe June 23, 2016 06:20

Temperature depending density
I am struggling at the moment on the implementation of temperature depended density. I created a user function with 100 K temperature steps. (My temperature range is from 300 K to 30 000K).
The problem is, that the solver has fatal linear overflow after 5 iterations.
I also tried to implement a function which matches the density, but with 10 K temperature steps, this also creates a fatal linear overflow.
Then I modified the function so that for a temperature below 1000 K the density is constant (I know this is not physical valid, it was just a test), the simulations runs fine until the temperature exceeds 1000 K.

ERROR #004100018 has occurred in subroutine FINMES.
Message: Fatal overflow in linear solver.

Some also faced this problem:

Rearranging the table doesnít works for me and also my mathematical function doesnít work.
I you need further information about my setup, please ask.
Thank in advance for your response
Best Greetings

Lance June 23, 2016 06:27

um.... 30 000 K is pretty much, having in mind that the surface temperature of the sun is roughly 5000 K. What is your application? What physics are you trying to model? What are the actual values of the density of your fluid above 1000 K? Are you sure the continuum assumption is valid?
It seems that CFX is having a hard time dealing with your density function, and you probably need to model whatever you are modeling in another way.

cfdgremlin June 23, 2016 06:52

You need to be very careful when developing your own functions/expressions for material properties, such as density. Strictly speaking, they need to be thermodynamically consistent in order for them to work properly and produce correct solutions.

My advice would be to, if possible, use one of the existing material models to define your fluid, as these are tried and tested. If none of these fit your requirement, and you need to develop your own expressions, then there is some documentation at the end of the section on Material Properties which discusses the criteria needed for thermodynamic consistency.

Birnensuppe June 23, 2016 07:30

@Lance: Yes I know. I am simulation eletric arcs with special look at temperatur radiation. I hope the maximum temperature is around 15 000 K, but the density function is implemented in a range from 300K to 30 000K.
I take the density values from Gleizes' publication:

For the testfile I set the density below 1000 K to 3.4692E-01 [kg/m^3] (This is the density at 1000K) after that I followed publication. Are you interessted that I post all missing 29 000 values?
For 10 000 K the density is 1.7218E-02 [kg/m^3] for example.

Yes, I am sure the continuum assumption is valid. Thank you for this thaught.

@cfdgremlin: Thank you for your advice, since I take the values from an trustworthy publication, I hope that the values are correct.

To specify, I startet with ideal gas provided by CFX, then I add temperatur depended conductivity, dynamic viscosity, specific heat capacity and thermal conductivity, it's all working fine, but not the density.

In the solver theory guide it is often mentioned, that the density can be variable. Is it possible that CFX calculated the density by itself? (I tried to find it, but wasn't sucessful)

Kind regards

ghorrocks June 23, 2016 08:18

The original question is an FAQ:

And as the others have stated it is almost certainly linked to your extreme conditions and/or the material properties linked to it.

Note that beyond a few thousand Kelvin most materials are plasma and CFX does not have a plasma model. CFX's capabilities in the plasma regime are quite limited. So CFX may well be incapable of modelling what you are intending.

Birnensuppe June 24, 2016 04:17

Thank you for your reply ghorrocks.
My question is, whether it is possible to implement a temperature dependent density.
The fact, that the solver runs fine with a constant density at a value for 300 K or for 15 000 K and with a modified density which is constant. As soon as the setup exceeds the temperature range where the density is hold constant by the modified function, the solver is diverging. The difference of the value for the density from 1000 K to 1100 K is 3.4692E-01 - 3.1538E-01 = 0.3154E-01 [kg/m^3], which shouldn't be such a big step, which could disturb the solver. This matter astonish me.
In addition I don't think CFX should have a problem to simulate the setup with a temperature around 300 K. I hope you agree with me in this point. That makes me think that either I made a basic mistake or CFX has a basic failure. (I hope it's me ;) )

ghorrocks June 24, 2016 07:35

Yes, you can define a temperature dependant density.

Are you using ideal gas or some other compressible EOS? If not, why not? And if not, what EOS are you using?

Birnensuppe June 24, 2016 09:47

2 Attachment(s)
I am using ideal gas EOS with the option value, because I couldn't find the necessary material parameters for ideal gas EOS, which are accurate for my temperature range.
If I change the EOS to ideal gas, CFX calculates the density with the ideal gas law, but that not exact enough for my temperature range. There I want to implement an density, which is more accurate for my setu
I have attached my material setup for the density.

All times are GMT -4. The time now is 08:20.