[oscol227]

Hi!

I get the Floating point exception: Zero divide error in the CFX-solver.

I have implemented functions from data-sheets that describes material data (density, specific heat capacity, dynamic viscosity and thermal conductivity) based on the pressure an temperature which is then used in the materials-tab for a new defined material. However when i start the solver it stucks at the first iteration loop for about 5 minutes and then crashes.

I have tried different meshes, inital conditions, different number of maximum points in the materials-tab and extrapolation of temp and pressure. Nothing helps!

When i change the material to ideal gas air it works just fine so it must be the new material i have described that is causing the problem. If I evaluate the functions and plot for different ranges i cannot see anything that seems wrong.

I hope someone here can help me
Regards
Oscar

[Opaque]

For a start, try using one of the "data sheets" at a time to isolate which is giving you trouble.

Not sure what equation of state you are trying to model, but you should be very aware that inputs for density and specific heat capacity are not independent and must satisfy specific thermodynamic relationships (see ANSYS CFX documentation for details); otherwise, the material database will not be correct if it can be computed at all.

Hope the above helps,

[oscol227]

Thanks for the answer!

Have located the density to be the problem. If I do not use the data for the density the solver does not crash.

If I plot the data I have for density for different pressures and temperatures I do however not see anything wrong with the plot. It looks smooth and fine, it does not contain so much data points for certain pressures and temperatures which maybe the problem. Does anyone know who CFX interpolates the data? Can the few number of data points at a certain range be a problem? Note that the density only have minor changes at those locations.

[ghorrocks]

CFX uses a table to interpolate the physical properties. You can control the settings of this interpolation table in the materials tab.

Be aware that in early stages of the simulation you can have weird values of flow parameters which lead to wild physical properties. If the physical properties model goes unstable during this time the simulation can crash, even though the final converged result is fine. If this is causing problems then you need to use a better initial condition or "bend" your material model to eliminate the instability during convergence.

[oscol227]

Okey, the strange thing is that I start from initial values that are from an earlier solution which are close to what i expect the results to look like. The earlier solution is however obtain when using a function for the density instead of the data sheet, but it is not as accurate as the data sheet. I have also tried to modified the interpolation settings in the materials-tab but changing ranges and/or number of points does not help.

[ghorrocks]

I recommend you do a simulation where you write out a res file on the time step before it crashes. You probably have the density going bezerk somewhere, so you need to find out where and how.

[oscol227]

Forgot to mentioned it is a steady state simulation I am doing.

The problem is that it crashes for the first iteration so there is no results to investigate and the inital start is not containing any wierd values, as mentioned earlier it is started from an earlier solution case that is very close to the expected results.

[ghorrocks]

In that case the hints listed in this FAQ are applicable: http://www.cfd-online.com/Wiki/Ansys...do_about_it.3F

And be extra-especially specially careful that the density function you have defined is well behaved in the range of densities you expect to see.