code_saturne User's Forum

I had just installed and run a simulation with ver. 2.0.3 a few days before 2.1 was released.

On 2.0.3 the simulations ends well with nice convergence. SEEMINGLY so does the 2.1 version, but upon post-processing, strange localised hot-spots appear on the outlet cells.

I've attached explanatory screenshots and 'listing' files.

When run with a passive thermal scalar 2.0.3 converges well, with 2.1 TempC ends at e+20.
 
Any ideas?

Thanks in advance
 
Claus

Listing file for 2.1

Hello Claus,

Comparing the log files side by side, they don't seem to be for the same configuration (slightly different number of cells, different volumes, different Gy gravity component, and no thermal variable with 2.1, so I assume you posted the wrong log file for 2.1).

When comparing results from 2.0 and 2.1, if you set the same scale in ParaView for the temperature variable, do you have significant difference only at the outlet (in which case it could be a gradient reconstruction/extrapolation related issue, which would disappear with a layer of extruded faces at the outlet, though this is not a satisfactory solution), or are the results also significantly different elsewhere ?

If your mesh is not too large, and not confidential, I can take a look at the test case if you post it here; otherwise, I believe this forum's functionalities allow sending directly to another user (not sure if this is reserved for moderators, or allowed to all), or you could send the case to us through the saturne-support at edf.fr e-mail (I can access the forum from outside EDF, while I can only access the support mail account from EDF offices, so I prefer the forum).

Best regards,

  Yvan

Yes, you are right about the small discrepancies between the log files: The log file from 2.0 is created with a mesh which has a small anomaly at the outlet due to the 'viscous layers' function in Salomé - this is NOT the source of the error. I then recreated the geometry to get rid of this anomaly and refine the mesh just a bit - more homogeneous sized cells etc.  For the 2.1 log file I switched off the thermal scalar to see if that was the source of the hotspots - it was not.

From the the xml files I now see that in 2.0 I mistakenly used 982 for G and 9.82 in 2.1 - that accounts for the pressure being off by an order of magnitude in post-pro, but it shouldn't account for the hotspots?

I have compared slices of the velocity profile (with Integrate Variables amongst others) and they are nearly identical in all of the pipe, however, in 2.1 the hotspots appear on the outlet for pressure, temp and velocity.

Pressure relaxation is for 2.0: 0.9  and 2.1: 0.7

Im wondering if I can think of any more useful information...

I've attached xml file for 2.0 and 2.1 and the mesh - If run on a 2.0.3 installation it should converge fine, if run on a 2.1 installation, hotspots should appear.

Thanks in advance for helping.

Regards,

Claus

Hello Claus,

Actually, your results from 2.0 seem to be also wrong. For example the min/max values for the last iteration are :
   ** INFORMATION ON VARIABLES
---------------------------------------------------------------
   Variable      Min. value    Max. value   Min clip   Max clip
---------------------------------------------------------------
v  Pressure    -0.25958E-01   0.16842E+00         --         --
v  VelocityX   -0.19024E-01   0.14981E-02         --         --
v  VelocityY   -0.14191E-01   0.37793E-02         --         --
v  VelocityZ   -0.63386E-02   0.65977E-02         --         --
v  TempC       -0.11086E+02   0.21352E+03          0          0
v  total_pressu-0.70710E+06   0.17488E+06          0          0
---------------------------------------------------------------
 
The min temperature is negative and the max value is upper than 100°c.

Your mesh have bad quality criteria (see attached). The ideal "Weigh" is 0.5 (see "ipond" in the user guide p:32 for more details) and the ideal "non ortho" (represents the angle IFJ) is 0° (the worst being 90°).

I think the mesh quality is the reason of the bad results (for 2.1 and 2.0 versions).

Best regards,

JF