Very high aspect ratio
Hi,
I used Ansys meshing to create a mesh with 10E-6 m first boundary layer height. There are few elements with crazy aspect ratio, around 2000000.....What can I do with this? and unbelieveable I did't get any warning. I used this mesh to run the simulation with double precision and no divergence happened, the result oscillated because of large seperation of the flow. Can i trust this? What can happen with such high aspect ratio? Thank you Kind Regards, |
10^-6 m thickness of first Bundary layer? What do you simulate actually?
Let's say the number of Prism Layer is usually between 10 - 15 and the total height of BL depends on the flow behaviour and your expectation. But usually is between cm and m. Check the physics and think about whether you really need such tiny thin BL. In your case, if the thickness is 10^-6 (Height )and cell size (Length) is in cm, of course you will get a cray aspect ratio. |
Quote:
I do need the 10e-6 thickness to get y+ around 1, since I have really high pressure and mach number. Actually I am using mm as the cell size, but there is still some crazy elements. I could not get even smaller cell size due to the enormous number of elements already. And I have 30 layers to get a relative smooth transition from the BL to the normal flow. |
Do you simulate supersonic flow?
|
Quote:
|
Can you just use wall functions?
Stu |
Quote:
Instead of creating a very fine mesh in BL, you can use the wall function. Because the Velocity Profile in BL is predictable and therefore is approximated using Wall Function. Since you want to capture more details in the BL, you can use the SST Turbulence Model and Automatic Wall Function. |
Actually there are 2 methods to solve the flow behaviour near the wall:
1: Wall Funtion Methode For wall function: y+ < 300 is ok delta_y = L*y+ * (74)^2*Re^(-13/14), using this formular to estimate the first node spacing delta_y, if the flow passes over a plate 2: Low-Reynold-Number Methode For Low-Reynold-Number Methode: y+ < 2 |
Also, aspect ratio isn't everything... You can still have a high aspect ratio (though yours is very very high) and have a happy solver if your angles, skewness, etc. are all still good. This is usually fine with boundary elements...
|
Slightly off topic, but can someone advise if multigridding speed up the rate of convergence for very fine prism layers? Because the layer is so thin, the velocity gradient normal to the surface would be very small between cells (i.e. low frequency), and therefore would take a lot of iterations without multigridding. I was just wondering out of curiosity...
|
Quote:
|
Quote:
|
Quote:
|
All times are GMT -4. The time now is 01:58. |