DNS of Re 3900 cylinder flow

cfdnewbie · May 30, 2013, 13:01

Hello fellow researchers,
I'm looking for DNS results of the standard RE 3900 cylinder test case. I have done a literature search on this topic, and the "largest" DNS I found is in here: http://www.cfm.brown.edu/people/gk/jfm_re10K.pdf (with a reference to a paper by Ma with a similar resolution). Do you happen to know of any more recent DNS of this flow, preferably with a higher resolution? I'm looking for the best reference possible for this case, and I am trying to judge whether it is worthwile doing a higher resolution DNS of this flow.
Thanks a lot, have a great weekend!

Far · May 30, 2013, 15:04

Mesh in journal paper is too coarse

cfdnewbie · May 30, 2013, 15:53

Do you have a recommendation for what the mesh should look like? Why do you think the mesh in the paper is too coarse?

Far · May 30, 2013, 16:05

1. It should be hexa/map mesh

2. It should follow DX, DY and DZ requirements of DNS mesh

cfdnewbie · May 30, 2013, 16:11

Quote:

Originally Posted by Far

1. It should be hexa/map mesh

hm, why would you say that? there are two in the paper. One is on hexas, the other on tetras. As long as the resolution is close to a DNS, it doesn't make a difference.

Quote:

2. It should follow DX, DY and DZ requirements of DNS mesh

It seems that the resolution they are using does that, doesn't it? They have a grid point at 0.32 based on the friction velocity...that sounds sufficient to me, or am I misunderstanding somehting?

The only thing that bothers me somewhat is that they use only about 14mio of DOF for this case...that seems not too much for today's standards, but of course that was in 2000....

Far · May 30, 2013, 16:13

There are two meshes. One is low resolution tetra and second is high resolution tetra

cfdnewbie · May 30, 2013, 16:19

The second tetra (9272 elems) has about 24mio grid points, the second hex mesh they are talking about has less than that (about 14mio). So the highest number of DOF is indeed 24 mio, which seems to me not too much for a DNS - and certainly easily surpassed with today's codes.

Far · May 30, 2013, 16:26

Quote:

Originally Posted by cfdnewbie

The second tetra (9272 elems) has about 24mio grid points, the second hex mesh they are talking about has less than that (about 14mio). So the highest number of DOF is indeed 24 mio, which seems to me not too much for a DNS - and certainly easily surpassed with today's codes.

Could you please explain what is meant by 9272 elements and how it is equivalent to 24 million grid points.

cfdnewbie · May 30, 2013, 16:30

They are using a 5th order Jacobian basis in each hex. Most likely, that's 6*7/2 DOFs per element, at least that's the standard approach on tets. So each 2D plane has about 9000 * 21 = 189000 DOF. Times the Fourier nodes (128), thats about 24 mio.

May 30, 2013, 13:01	DNS of Re 3900 cylinder flow	#1
cfdnewbie Senior Member cfdnewbie Join Date: Mar 2010 Posts: 557 Rep Power: 20	Hello fellow researchers, I'm looking for DNS results of the standard RE 3900 cylinder test case. I have done a literature search on this topic, and the "largest" DNS I found is in here: http://www.cfm.brown.edu/people/gk/jfm_re10K.pdf (with a reference to a paper by Ma with a similar resolution). Do you happen to know of any more recent DNS of this flow, preferably with a higher resolution? I'm looking for the best reference possible for this case, and I am trying to judge whether it is worthwile doing a higher resolution DNS of this flow. Thanks a lot, have a great weekend!

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
flow around a cylinder	pXYZ	Main CFD Forum	14	July 25, 2011 11:05
1 Was FSI for a flow over a cylinder	vmlxb6	CFX	17	May 16, 2011 03:29
reference calcultion, flow around a cylinder	butch85	Main CFD Forum	3	January 31, 2011 17:10
Flow past cylinder (Re=10^5)	lbeaudet	Main CFD Forum	9	June 2, 2009 04:59
Flow around a cylinder at subcritical regime.	yousef	FLUENT	0	March 18, 2006 10:18

May 30, 2013, 15:04		#2
Far Senior Member Sijal Join Date: Mar 2009 Location: Islamabad Posts: 4,558 Blog Entries: 6 Rep Power: 54	Mesh in journal paper is too coarse

May 30, 2013, 15:53		#3
cfdnewbie Senior Member cfdnewbie Join Date: Mar 2010 Posts: 557 Rep Power: 20	Do you have a recommendation for what the mesh should look like? Why do you think the mesh in the paper is too coarse?

May 30, 2013, 16:05		#4
Far Senior Member Sijal Join Date: Mar 2009 Location: Islamabad Posts: 4,558 Blog Entries: 6 Rep Power: 54	1. It should be hexa/map mesh 2. It should follow DX, DY and DZ requirements of DNS mesh

May 30, 2013, 16:13		#6
Far Senior Member Sijal Join Date: Mar 2009 Location: Islamabad Posts: 4,558 Blog Entries: 6 Rep Power: 54	There are two meshes. One is low resolution tetra and second is high resolution tetra

May 30, 2013, 16:19		#7
cfdnewbie Senior Member cfdnewbie Join Date: Mar 2010 Posts: 557 Rep Power: 20	The second tetra (9272 elems) has about 24mio grid points, the second hex mesh they are talking about has less than that (about 14mio). So the highest number of DOF is indeed 24 mio, which seems to me not too much for a DNS - and certainly easily surpassed with today's codes.

May 30, 2013, 16:30		#9
cfdnewbie Senior Member cfdnewbie Join Date: Mar 2010 Posts: 557 Rep Power: 20	They are using a 5th order Jacobian basis in each hex. Most likely, that's 67/2 DOFs per element, at least that's the standard approach on tets. So each 2D plane has about 9000 21 = 189000 DOF. Times the Fourier nodes (128), thats about 24 mio.