Hello everybody,

I have an issue concerning the QUICK-discretization scheme. I have performed two runs on a pretty coarse grid (about 600k element, 100k nodes, no special boundary layer resolution (that is pretty hight y+) and no densities in the wake regions) that was generated with ICEMCFD. I conducted two runs with the blend factor at sbf=0.75 and sbf=1 using the first run as initial solution for the second run. After that, I have tried to use the QUICK-scheme with the sbf=1-solution as the initial guess. With that configuration the solver crashed immediately, saying:

ERROR #004100018 has occurred in subroutine FINMES. Message: Fatal overflow in linear solver.

After that, I tried without an initial solution. This time the solver crashed after the first outer loop, saying:

A wall has been placed at portion(s) of an OUTLET boundary condition (at 59.7% of the faces, 59.7% of the area) to prevent fluid from flowing into the domain. The boundary condition name is: OUT. The fluid name is: Air Ideal Gas. If this situation persists, consider switching to an Opening type boundary condition instead.

****** Notice ****** The maximum Mach number currently is 5.930E+03. If this message persists and you experience robustness problems,consider setting the expert parameter 'max continuity loops' to 2.

ERROR #004100018 has occurred in subroutine FINMES. Message: Fatal overflow in linear solver.

Did anyone ever have the same or similar problem an might know what the problem is?

Regards and thank you Andreas

Hi,

A big problem with QUICK is that it is not very numerically stable, and I think that is exactly what you are experiencing. This is especially important for sonic flows - QUICK is terrible for sonic flows. In my experience QUICK is only good for low Mach number flows with very even grid spacing (or preferably constant grid spacing).

I strongly recommend you use the High resolution differencing scheme as it is much more robust for sonic flows. There is a loss of formal accuracy to second order in general and first order at strong gradients but QUICK will have major boundedness and other stability problems so is unlikely to be more accurate.

Glenn Horrocks

Hey Glenn,

thanks for your advice. One further issue with my flow - actually it is not sonic at all! I have a frees stream velocity of about 36 m/sec and maximum velocities of about 45 m/sec around my article. So the CFX-warning for my maximum Mach number confuses me even more. So I suppose that something is going totally wrong with the QUICK scheme... However I have no idea what it could be.

Regards

Hi,

If the flow is not high speed but it is giving warning messages about supersonic flows then something is seriously wrong. Either your simulation is diverging in a big way (go to high resolution differencing, it is much more stable) or your simulation setup is wrong.

Glenn Horrocks

[Reza_computational]

Hello everybody
I have encountered the error (some faces have been replaced with walls to prevent fluid flowing out or into the boundary) once i was solving flow in centrifugal compressor with cfx and QUICK scheme...

[ghorrocks]

It is not an error. It is a warning.

It just means that the inlet or outlet has reverse flow and it has stopped the reverse flow using artificial walls. This can occur when you use QUICK because it is allows resolution for finer features than the default scheme.

[feizaghaee]

how can i use QUICK scheme?

[ghorrocks]

You have to select it using CEL. Set up a run using some other differencign scheme and replace the entry with QUICK.

[feizaghaee]

where can i read description of setting up QUICK scheme

[ghorrocks]

Not sure it is documented. To set it up you need to edit the CCL. Find the Spatial differencing key and change whatever is there (probably upwind or hybrid by default) and in a text editor change it to "QUICK". Reload the CCL and you will be using QUICK.

[feizaghaee]

thanks in advance

[baratian]

Hi every body
I have a problem with QUICK scheme
In a 2d pure convection for 2 cells when i apply the scheme ij cell property phi eliminates and only neighbors remain! So how can i calculate the ij cell ? Just like the example in versteeg imean the dotted cells in pic attached
Thanks in advance

A11-ConvertImage.png

[ghorrocks]

Are you asking how does the QUICK differencing scheme work? If so then look in any CFD textbook, Versteeg covers this I am sure. If this is not your question then please restate your question more clearly.