[bungusbeefcake]

Hey guys, I've been running this problem for weeks now without any luck in solving it! Basically I'm running a sim on a SCRamjet/waverider geometry. I've been trying to run it at Mach 7. Whenever I run the sim with Mach 7(2090 m/s velocity in inlet and 2090,0,0 initial conditions) I get the following error:
A floating point exception has occurred: floating point exception [Overflow]. The specific cause cannot be identified. Please refer to the troubleshooting section of the User's Guide.
Context: star.kwturb.KwTurbSolver
The error occurs before the 1st iteration!
I can mesh the geometry no problem and I'm running the sim with the following physics: All Y+ wall treatment, coupled energy, coupled flow, gas - air, gradients, ideal gas, k-omega turbulence, RANS, SST (menter) k-Omega, steady, 3D, transition boundary distance, turbulence suppression and turbulent.
I have tried remeshing in various configurations using volumetric controls both isotropic and anisotropic in case this was an issue, changing the number of prism layers and thickness, changing the courant number. I'm confused as to where to go as I've done the exact same conditions on a different geometry and that had no issues.
I have tried to start the sim at 20m/s and ramp up to 2000m/s and although I don't automatically get the error, the velocity in the fluid remains at approx 20m/s after a number of iterations, which to me makes no sense!
I can give anyone who wants it the sim, meshed or not if they would like to have a look?
Would anyone have any suggestions?
Thanks for any help or guidance as its killing me to just keep running into this wall!!

Regards,

Bungus

[fluid23]

It sounds like maybe your ramp is too quick? Over how many iterations did you ramp up from 20 to 2000 m/s? It takes time for changes in boundary condition to propogate through your domain. You might also try ramping your courrant/CFL number.

Getting to Mach 7 will take some time.

[fluid23]

Can you post a shot of your mesh?

[bungusbeefcake]

Originally Posted by MBdonCFD

Can you post a shot of your mesh?

This is one of my mesh's, I have many different ones, some more coarse, some finer, some with larger prism layers. The ramp is probably happening too fast, I only tried it a few times and only 20 or so iterations between them.
The Mach 7 worked from the initial on a previous model, thats what really confuses me! Let me know what you think.

[fluid23]

Try to ramp over like 400 iterations or so and allow the changes to propagate through your domain.

As for why one model works and another doesn't, its hard to say. Are these the same mesh settings and different geometry, different mesh and same geoemtry, or different geometry and different mesh?

[fluid23]

Also, it looks like maybe you would benefit from mesh enhancement around your shock locations and in the wake, but it's hard to tell for sure from these shots. Look at the help documentation for wake refinement and field function mesh refinement.

[bungusbeefcake]

I'll try to do it over 400 iterations. The model is much different than the original one so I have a different mesh and different model, however all the mesh parameters were scaled to the new model based on length scale, but I don't know if this changes anything or not. I might try to scale the model down to the original size and see if it works that way.

Have you any other thoughts? Any experience with high mach numbers? and if so then what were the physics you would suggest?
I was thinking about the wake refinement but thats why I tried to use the volumetric controls.

[fluid23]

I wouldn't scale your model because that changes your primary Reynolds number. How much larger is the newm model? If you are wanting to make a good comparison I would keep all the same settings as the original file unless it will double your cell count or something.

I don't have much experience with high Mach, I do helicopters and rotors. There we typically 'top out' at less than 200 kts. However, I think your physics selection is appropriate. 9 times out of 10 divergence is an issue of mesh quality. If you have access to the Steve Portal, log in and go to the macro hut. There you will find a handy little java macro called Mesh Quality Analyzer. Download and install if you are able and then run it. Don't worry about the residuals to start, just do mesh quality. That should help you locate any problem cells. If you still see divergence after fixing those, then run the residual analysis too. This will highlight all the cells with the top 5%, 10% (or whatever you choose) residual values.

[bungusbeefcake]

I've had a look through the simulation. I can get the simulation to run now, but the mesh needs to be approximately 20 times smaller. This being the case, the only way to technically run the sim is to scale it. I know about the issues with the Reynolds number but at least I'll have an order of magnitude approximation so its not so bad.

There is a few other possible confusing issues which I have but I'm altering mesh's to determine the sensitivity of the mesh on the sim.

[fluid23]

That seems extreme. What was your cell count before and what are you thinking it needs to be now? I think scaling will cause far more than Re issues for you. Plus, it doesn't really change the relative size of the cells. Smaller geometry will mean smaller cells, not fewer cells. Flip side, larger geometry will require larger cells, not more cells. There may be some slight variation, but it would be more a product of your mesher than anything else.

[meatty]

Switch to Real Gas Redlich-Kwong and use the segregated solver. It is much more stable in this very high speed regime.