CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   CFX (https://www.cfd-online.com/Forums/cfx/)
-   -   Large Eddy Simulation with too high Mach number (https://www.cfd-online.com/Forums/cfx/82198-large-eddy-simulation-too-high-mach-number.html)

 Roland R November 19, 2010 08:56

Large Eddy Simulation with too high Mach number

Hello,

I would like to calculate a Large Eddy Simulation to the investigation of an aero acoustics problem in high speed compressible flow. The LES calculation was initialized from a converged transient SST result. The geometry is quite complicated, I am fully aware that the mesh should be more finer but I have very short time.

I applied tetra mesh, its quality is about 0.3-0.5 . The mesh is fine in the narrow cross sections but it’s coarse in the zones which is not investigated.

The time step is 1e-6 s. Based on the convergence history the residuals is OK, but the Mach number is unrealistic:

COEFFICIENT LOOP ITERATION = 15 CPU SECONDS = 2.018E+05
----------------------------------------------------------------------
| Equation | Rate | RMS Res | Max Res | Linear Solution
+----------------------+------+---------+---------+------------------
| U-Mom | 0.58 | 2.5E-05 | 5.9E-03 | 7.5E-04 OK
| V-Mom | 0.80 | 2.8E-05 | 1.3E-02 | 8.2E-04 OK
| W-Mom | 0.83 | 5.8E-05 | 1.6E-02 | 1.2E-03 OK
| P-Mass | 0.89 | 9.2E-06 | 3.8E-03 | 5.0 5.1E-04 OK
+----------------------+------+---------+---------+------------------
| H-Energy | 0.84 | 3.1E-05 | 1.2E-02 | 5.8 3.5E-08 OK
+----------------------+------+---------+---------+------------------
+--------------------------------------------------------------------
| Notice: The maximum Mach number is 4.458E+01.
+--------------------------------------------------------------------

I thought that this numerical error was caused by the mesh but based on the result the mesh quality is OK in the zone of the high Mach number. The error can be detected in various locations but it always is in one node. (not in a large area). This notice can be seen after every time step then the solver stops with „overflow”.

Could anybody help me to solve this problem?

Roland

 joey2007 November 19, 2010 12:54

Guess there is something basic wrong. Check your setup.

BTW: The resolveable vortices size depend on your cell size. IMHO you have to be fine anywhere or us DES/SAS

 ghorrocks November 19, 2010 17:58

Use the post processor to find where the area of rapid flow is, and whether it is a problem.

 Roland R November 25, 2010 08:37

Quote:
 Originally Posted by ghorrocks (Post 284173) Use the post processor to find where the area of rapid flow is, and whether it is a problem.
OK, I have found some critical nodes where the Mach number is too high, but I don't understand the cause of this large numerical error. In these nodes the quality of cells is acceptable...although the size of the cells is large. Can the large element size cause so high Mach number in the case of a Large Eddy Simulation?

By the way, In my opinion the convergence history is qiute unusual. The notice (high Much number) was present for 470 timesteps, but the convergence was stabil. Finally in one iteration the calculation has stoped.

TIME STEP = 474 SIMULATION TIME = 4.7400E-05 CPU SECONDS = 6.306E+05
----------------------------------------------------------------------
COEFFICIENT LOOP ITERATION = 1 CPU SECONDS = 6.306E+05
----------------------------------------------------------------------
| Equation | Rate | RMS Res | Max Res | Linear Solution |
+----------------------+------+---------+---------+------------------+
| U-Mom |10.36 | 2.7E-03 | 1.7E+00 | 8.5E-02 OK|
| V-Mom | 4.02 | 9.0E-04 | 5.8E-01 | 4.7E-02 OK|
| W-Mom | 1.16 | 3.8E-04 | 1.7E-01 | 7.4E-02 OK|
| P-Mass | 7.51 | 5.0E-05 | 1.2E-02 | 5.8 3.3E-04 OK|
+----------------------+------+---------+---------+------------------+
| H-Energy | 0.59 | 1.2E-03 | 6.9E-01 | 6.8 1.1E-07 OK|
+----------------------+------+---------+---------+------------------+
----------------------------------------------------------------------
COEFFICIENT LOOP ITERATION = 2 CPU SECONDS = 6.307E+05
----------------------------------------------------------------------
| Equation | Rate | RMS Res | Max Res | Linear Solution |
+----------------------+------+---------+---------+------------------+
| U-Mom | 1.45 | 3.8E-03 | 2.5E+00 | 7.2E-03 OK|
| V-Mom | 1.14 | 1.0E-03 | 6.7E-01 | 1.4E-02 OK|
| W-Mom | 3.84 | 1.5E-03 | 9.6E-01 | 1.7E-02 OK|
| P-Mass | 0.80 | 4.0E-05 | 1.1E-02 | 5.8 2.3E-04 OK|
+----------------------+------+---------+---------+------------------+
| H-Energy | 1.00 | 1.2E-03 | 9.2E-01 | 6.8 1.2E-07 OK|
+----------------------+------+---------+---------+------------------+
----------------------------------------------------------------------
COEFFICIENT LOOP ITERATION = 3 CPU SECONDS = 6.308E+05
----------------------------------------------------------------------
| Equation | Rate | RMS Res | Max Res | Linear Solution |
+----------------------+------+---------+---------+------------------+
| U-Mom | 0.00 | 8.3E-17 | 4.9E-14 | 9.8E+11 * |
| V-Mom | 0.00 | 3.8E-17 | 2.3E-14 | 3.0E+12 * |
| W-Mom | 0.00 | 1.1E-16 | 7.4E-14 | 8.8E+11 * |
| P-Mass | 0.00 | 0.0E+00 | 3.6E-28 | 10.5 6.0E+08 F |

+----------------------+------+---------+---------+------------------

...how can it happened? What can cause this sudden "shock" in my calculation?

Thanks
Roland

 ghorrocks November 25, 2010 17:43

No, not an unusual convergence history at all. Just a simple divergence. You need to stabilise it numerically.

Try:
* Double precision number
* Tighter convergence
* Improving mesh quality
* Smaller timesteps
* lower order discretisation (not a good idea with LES but at least it will probably converge)
* Read the documentation on "obtaining convergence"

 All times are GMT -4. The time now is 12:26.