[Gokkul]

Hello everyone

I am doing steady state 3D simulation for turbine. Before assembling the whole turbine, I tried simulating only the first stator. Everything was fine and I was able to get results. But when I am trying after assembling 2 stage turbine, I am getting errors like "Floating point exception: Overflow" or "Fatal Overflow of linear solver" before 10th iterations. So I tried simulating the first rotor alone, I am getting the same error. The mesh analysis has no error and even in the CFX solver, under mesh statistics, everything says OK. I tried using different mesh from coarse to very fine mesh. I am using SST for turbulence model. I can see that only the rotor has the error. It would be great if someone could assist me in this situation. If needed I can share my mesh file or CFX file.

Thanks in advance!

[Gert-Jan]

For these kind of questions, we kindly refer to the following link:

https://www.cfd-online.com/Wiki/Ansys_FAQ#CFX-5

[Gokkul]

Hello Gert

Thanks for the response. I will look into the link.
Also, with the same rotor geometry and mesh, when I tried Air at 25C as fluid the simulation ran without any errors. The error pops up when I use Air as Ideal Gas. That's why it was very confusing and don't know where exactly the problem is!

[Gert-Jan]

The source of error can be anywhere. Changing from constant properties to properies that depend on Temperature and Pressure is a huge step. Even since I don't know what Energy model you use (None, Thermal, Total)

I would suggest to start gently with low rotor speed and ramp up slowly and increase other complexity step by step. Don't try to hit the top in one go.

[Opaque]

As Gert said, the error can be anywhere, so it is best to start slow and with controlled variables/degrees of freedom.

Pick a mesh first (regardless of accuracy), set up the physics, and try to obtain some results in it. Practice different settings to understand how they influence the path of convergence (read the modeling guidelines for help).

Attempting to change multiple parameters w/o knowing what they will achieve will not get you anywhere.

For steady-state, I will select the AutoTimescale option unless you understand how to use and compute an appropriate Physical Timescale. Read the documentation about what the diagnostics in the output file mean. Those numbers are not written to show progress, but how it is progressing and indicate the level of difficulty encountered when solving your problem.

[Gokkul]

Hello Gert and Opaque

Thanks a lot for your valuable comments. I will try to implement it in my simulation!

Also, I created backup file before the error and when I was post-processing the results, the magnitude of velocity is too high (1.9e3).

[ghorrocks]

This is typical in a diverging simulation. Look at the FAQ quoted by Gert-Jan for general tips on what to look at. Some of the other FAQs (eg the one on "Floating point error") are also relevant - you don't quite have a floating point error yet, but you are approaching it if it is diverging.