Unexpected air flow for a jet-fan blowing in the negative X direction

[GoogleCodeExporter]

Application Version: 5.2.5 Serial
SVN Revision Number: 2828
Compile Date: 10 Dec 2008
Operating System: Windows XP Pro

Dear All,

I'm working on a simulation involving several jet-fans. Each jet-fan is 
represented by a simplified casing plus an obstacle (with POROUS variable 
set to .TRUE.) at the "discharge" end, which is supposed to give air the 
required velocity. 
Some of my jet fans are required to blow air in the direction opposite to 
the positive X direction.
The jet of air created by the jet-fans blowing air in the negative X-
direction is strongly "tilted" (i.e. it is not symmetrical about the 
longitudinal fan axis as one would expect). 
I am attaching a much simplified input file which displays the problem, 
together with a screen shot ilustrating the extend of the "asymmetry" in 
the flow.
The problem does not occur if the airflow is positive X-direction. Also, 
removing the casing solves the problem of the asymmetry.
Improving grid resolution to 10 cm does not seem to help much.

Did anyone experience similar problems? I'd most appreciate suggestions on 
how to overcome this malfunction. 

Kind Regards,

Piotr

Original issue reported on code.google.com by psma...@gmail.com on 29 Dec 2008 at 2:09

Attachments:

• jet_fan.fds
• jet_fan_0738.jpg

[GoogleCodeExporter]

I'll take a look.

Original comment by mcgra...@gmail.com on 29 Dec 2008 at 2:17

• Changed state: Accepted

[GoogleCodeExporter]

A quick fix is to do the following:

&OBST XB= 9.0, 9.8, 6.0, 6.0, 2.40, 2.85, COLOR='BLUE', SAWTOOTH=.FALSE. / Fan 
casing 
&OBST XB= 9.0, 9.8, 6.4, 6.4, 2.40, 2.85, COLOR='BLUE', SAWTOOTH=.FALSE. / Fan 
casing 
&OBST XB= 9.0, 9.8, 6.0, 6.4, 2.40, 2.40, COLOR='BLUE', SAWTOOTH=.FALSE. / Fan 
casing 
&OBST XB= 9.0, 9.8, 6.0, 6.4, 2.80, 2.85, COLOR='BLUE', SAWTOOTH=.FALSE. / Fan 
casing  
&OBST XB= 9.4, 9.4, 6.0, 6.4, 2.40, 2.85, COLOR='RED', SURF_ID='BLOW1' / 

The SAWTOOTH=.FALSE. tells FDS not to impose a velocity "slip" condition on the 
fan 
casing. The trouble with any obstruction that is zero cells thick is that the 
velocity BC is applied on one side or the other, which explains why the air was 
being blown sideways. Usually, the effect is not that dramatic, but in your 
case it 
was. Second, I moved the RED obstruction into the center of the fan casing. 
That 
helps, too.

Long term, we need to ensure consistent velocity BCs for thin obstructions. I 
will 
keep the case open to work on the long term solution, but hopefully, you can do 
what 
you need to do now. 

Randy -- will the alternative differencing of the momentum equation eliminate 
the 
bias that is apparent in the input file above? The current differencing scheme 
uses 
the vorticity vector and stress tensor at cell edges with no regard to 
orientation.

Original comment by mcgra...@gmail.com on 29 Dec 2008 at 5:50

• Changed state: Started
• Added labels: OpSys-All, Performance

[GoogleCodeExporter]

I doubt that using the conservative form of the momentum equations will help.  I
think part of the problem is that the obstructions are snapped to cells and 
that in
the staggered scheme (conservative or not) the momentum control volume is 
staggered
relative the obstruction.  So, you end up with a partially blocked cell either 
way. 
I think that Ruddy's material point formulation will ultimately be the most
consistent fix for these kinds of problems.

Original comment by randy.mc...@gmail.com on 29 Dec 2008 at 6:52

[GoogleCodeExporter]

I think that you are thinking about porous boundaries. The issue in question 
here is 
when you have a thin obstruction and you try to apply a tangential boundary 
condition to it. 

Original comment by mcgra...@gmail.com on 29 Dec 2008 at 7:10

[GoogleCodeExporter]

Thanks Kevin - it does resolve the problem.
As regards the location of the "red" OBST - the reason why I've located it at 
the 
end of the casing was to add tangential velocity components which would account 
for 
the deflector plates installed on the jest-fans and the fact that some of the 
jet 
fans are not exactly aligned with the axis of the coordinate system. 
One thing I'm not sure I understand is why the problem that you have explained 
only 
manifests itself when the direction of the flow is in the "negative" direction? 
I 
should say that I've got a similar effect for the negative Y direction as well!

Regards,

Piotr

Original comment by psma...@gmail.com on 31 Dec 2008 at 1:15

[GoogleCodeExporter]

The reason for the direction dependence is that the boundary conditions are 
being 
applied in the order -x, +x, -y, +y, -z, +z. For a thin sheet, the boundary 
condition is applied at the cell edge, in which case the same BC is seen on 
both 
sides. This is not right -- there should be two BCs associated with each edge. 
When 
the sheet is not zero cells thick, this is not an issue -- the BCs are unique.

Original comment by mcgra...@gmail.com on 31 Dec 2008 at 1:15

[GoogleCodeExporter]

There is still a fix needed to the tangential velocity BCs on flat plates. I am 
putting this issue on hold to remind myself to fix it.

Original comment by mcgra...@gmail.com on 10 Feb 2009 at 6:19

• Changed state: OnHold

[GoogleCodeExporter]

This bug had to do with boundary layers over thin-sheeted obstructions. Should 
work 
properly in FDS 5.4.0, which should be out in a few weeks.

Original comment by mcgra...@gmail.com on 6 Jun 2009 at 6:44

• Changed state: Fixed

[GoogleCodeExporter]

Original comment by mcgra...@gmail.com on 23 Nov 2009 at 10:26

• Changed state: Closed