open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

I am currently working on Viollet validation case availiable in Telemac3D v7p2r0. I tried to change the vertical turbulence model to mixing length model (keeping the k-epsilon model as horizontal turbulence model), however I received an error message : Program received signal SIGSEGV : Segmentation fault - invalid memory reference.
Could you, please, help me to understand what's behind all of this?

Best regards,

Sophie

Hello Sophie

I'm not able to consult my source files but I think that when using k-eps model, you can not use another turbulence model for the vertical direction. K-eps, generally uses both horizontal and vertical velocity compennent in order to compute kinetic energy, diffusion and source terms (production, destruction and diffusions).

with my best regards

Riadh ATA

Hello Riadh,

Thanks for your reply.

It is something new on Version v7p2r0 : K-epsilon model turbulence model can now be used independently on the horizontal and on the vertical.

It succeeded to use turbulent constant viscosity model on the horizontal and k-epsilon model on the vertical; but the opposite gives an error message.
As I told you in my first message, the use of k-epsilon model on the horizontal and the mixing length model on the vertical is not working either, we have the same error message : Program received signal SIGSEGV : Segmentation fault - invalid memory reference.

Could you/someone help me?

Best regards,

Sophie

Hi
Try to add DEBUGGER=1 to see in which subroutine the crash occurs...

Regards

Hello,

I have added DEBUGGER = 1, it seems that the crash occurs in PREADV subroutine.
Please find in attachment the file ".sortie".

Best regards,

Sophie

Sophie

Please send me your case at my mail.

regards
Riadh

Dear Riadh,

Please find in attachment Viollet validation case.

Best regards,

Sophie

Hello Sophie

Actually, we found an effective bug with these new options. There are so many options when we cross all the choices of turnulence models in horizontal and vertical directions. It seems that we did not consider for several combinations.

My colleague in charge of Telemac-3D will send you a tentative fix and the bug will be fixed on the next stabilized release of V7P2r1 and of course on the trunk.

with my best regards

Riadh ATA

Hello Riadh,

Thank you for your prompt reply.

Best regards,

Sophie

Hello,

Please find enclosed a tentative fix that solves the segmentation fault. There were some subroutines that had not been updated to allow k-epsilon on a direction and not on the other. Files are coming from the trunk and have to be adapted for version 7.2. You then have to re-compile the system.

You also have an additional single subroutine cstkeps.f to be replaced in your FORTRAN FILE.

Anyway, physical results are not very good with k-epsilon on the horizontal and mixing length on the vertical for this test case and with this tentative fixe. I am not sure that theoretically, this combo must lead to good results. A few bugs may be still present, but I will not be able to work on this before next Thursday, sorry.

Hope this helps,

Chi-Tuan