[overture] Re: channel grid
- From: Bill Henshaw <henshaw@xxxxxxxx>
- To: overture@xxxxxxxxxxxxx
- Date: Tue, 28 Jul 2009 20:44:41 -0700
Hi Kyle,
The smallest scale theory (see e.g. papers on the Overture web site) indicates
that the number of grid points you need is proportional to the square root
of the kinematic viscosity divided by the local size of the vorticity (or
gradient of the velocity) :
h = C sqrt{ nu/| vorticity| }
Here C is some order 1 constant. Thus the grid spacing you need to resolve
the flow depends on the local solution. You can thus do a sample
computation to get the approximate size of the vorticity and then choose h.
You can also use boundary layer theory to estimate the size of the
vorticity if you have a boundary
layer flow. I think that in 3D it may be that for incompressible flow,
h >= C Re**{-3/4}
It is always a good idea to repeat your computation with half the
mesh spacing to see how the answer varies.
...Bill
Kyle Schmidt wrote:
Hello everyone,
I was wondering if anybody has done any grid sensitivity on a channel
for any Laminar Re number; specifically Re=100? I have a simple channel
with an inlet of a given velocity and outlet, and my vertical lines are
stretched to give me a more accurate reading of the velocity profile
near the walls. If anyone has, how many vertical and horizontal lines
did you use for an accurate result and how did you go about finding the
correct line numbers?
Kyle
If a command file is necessary let me know and I will provide one.
Other related posts:
