Droplets do not stick on walls

[GoogleCodeExporter]

Application Version: 5.4.0 Serial
SVN Revision Number: 4198
Compile Date: 16 June 2009
Operating System: Linux

The attached case fills up a tube with ultra-fine water mist droplets. 
Looking at the MPUA boundary file, I can see droplets only on two walls 
out of four. After 5 seconds, the W-VELOCITY and MPUV slices get skewed; 
also the DROPLET FLUX Z slice shows droplets exiting only from one corner 
of the tube.

I also commented out a DEVC because the code complained that 'W-VELOCITY' 
is not an appropriate quantity for DEVC (this did not happen with SVN4031).

Original issue reported on code.google.com by jukka.va...@vtt.fi on 17 Jun 2009 at 1:30

Attachments:

• Mist_generator.fds

[GoogleCodeExporter]

I'll take a look at it.

Original comment by mcgra...@gmail.com on 17 Jun 2009 at 2:31

• Changed state: Accepted

[GoogleCodeExporter]

Because your droplets are so small, they act much like massless particles. Near 
the 
wall, the normal component of velocity is close to zero. Because of the way the 
extrapolation was done, droplets that were near the positive x and y walls 
extrapolated a velocity that was small but enough to push them into the wall. 
Droplets on the other side extrapolated to zero normal velocity. I made the 
extrapolation more precise, and now no droplets hit the walls. Mathematically 
this 
is correct, but physically not. FDS does not account for surface tension or 
other 
forces that would make the droplets stick. 

Show the case to Simo and ask what he thinks. Also note that in 5.4.0 we're 
doing 
the velocity BC more correctly, so small droplets and massless particles get 
close 
to the wall and slow down moreso than before when we used a "half slip" 
velocity BC.

Also note that 'W-VELOCITY' is now rejected if you use XYZ with no STATISTICS. 
Simo 
can explain.

Original comment by mcgra...@gmail.com on 17 Jun 2009 at 9:09

• Changed state: Fixed

[GoogleCodeExporter]

Whatever you did seemed to fix the flow field skewness, thank you very much. 
Just a 
few minor comments to add.

The MPUV slice shows high local values close to walls. I'm not sure that makes 
sense 
(and perhaps it is a Smokeview issue). However if it does, and FDS can support 
large 
water mist concentrations close to walls without contact between drops and 
wall, 
then FDS may have a little cooling problem. Is there a threshold value for 
droplet 
normal velocity that pushes the drops to walls for most practical cases ?

Original comment by jukka.va...@vtt.fi on 18 Jun 2009 at 9:14

[GoogleCodeExporter]

In FDS 5.4.0 (which is the version you are using even though we have not 
officially 
released it) makes use of the Werner-Wengle model for the tangential velocity 
BC. 
The effect of this is to have zero normal velocity (as before), but also zero 
tangential velocity at the wall. The droplets are essentially being "trapped" 
near 
the wall with no (or very little) air velocity to move them into the wall or 
along 
the wall. No doubt this has some element of truth, but we don't have the 
physics in 
the model to make these near wall droplets stick. Ordinary sprinkler droplets 
still 
stick to the wall as before because they have enough momentum of their own and 
do 
not "go with the flow" as we say here in the US. I think that once the droplets 
get 
to sizes down to 10 microns, we might have to consider additional mass transfer 
effects near walls, similar to what we are exploring with smoke. You can see 
the 
effect all of this has had on MASSLESS particles by running the "device_test" 
case 
in the Controls folder of Verification.

Original comment by mcgra...@gmail.com on 18 Jun 2009 at 12:20

[GoogleCodeExporter]

In dealing with the subgrid droplet physics near the wall, my understanding is 
that
the "friction velocity" u* = sqrt(tau_w/rho) is typically used as the velocity 
scale.
 Just FYI, this quantity is readily available from the calculations that occur in the
routine for the WW model.  We would just need to output u* if it is of any use.

Original comment by randy.mc...@gmail.com on 18 Jun 2009 at 12:33

[GoogleCodeExporter]

Would u* be in the normal direction? If so, yes, it would be fairly easy to 
implement. 

Original comment by mcgra...@gmail.com on 18 Jun 2009 at 12:38

[GoogleCodeExporter]

u* is just a scale (like the speed of sound). It could be used in the normal
direction but hasn't someone already developed a model for how the droplets 
should
behave near the wall?  My thinking was simply that u* was probably a component 
of
whatever model is standard for near wall, multiphase LES.

Original comment by randy.mc...@gmail.com on 18 Jun 2009 at 12:46

[GoogleCodeExporter]

OK -- Jukka, do you know of simple models of droplet attachment near solid 
boundaries? It would only take a very small normal component of velocity to 
make 
these droplets stick.

Original comment by mcgra...@gmail.com on 18 Jun 2009 at 12:50

[GoogleCodeExporter]

Can't say I do at the moment. Seems though that there is plenty of literature 
available on the subject (see e.g. Marchioli et al, Int.J.Multiphase Flow vol 
29 p. 
1017 and references therein). What soot deposition mechanism(s) are you 
exploring, 
can you give a reference ?

Original comment by jukka.va...@vtt.fi on 22 Jun 2009 at 7:49

[GoogleCodeExporter]

Jason Floyd has been doing the soot deposition work.

Original comment by mcgra...@gmail.com on 22 Jun 2009 at 11:48

[GoogleCodeExporter]

I was just curious whether the soot deposition stuff could be directly applied 
to 
fine droplets.

Meanwhile, it would perhaps be best to close this issue as I don't have 
anything 
concrete right now with which to demonstrate a potential cooling/wetting 
problem. If 
something comes up, then we can open a new issue.

Original comment by jukka.va...@vtt.fi on 23 Jun 2009 at 10:21

[GoogleCodeExporter]

We will probably introduce into FDS 6 the notion of mass transfer to walls, 
somewhat 
analogous to heat transfer in terms of its empiricism and complexity.

Original comment by mcgra...@gmail.com on 23 Jun 2009 at 11:57

[GoogleCodeExporter]

Regarding Comments 1 and 2, it looks like the latest source again accepts DEVC 
with 
XYZ and no STATISTICS for 'W-VELOCITY'.

Original comment by jukka.va...@vtt.fi on 13 Aug 2009 at 1:21

[GoogleCodeExporter]

Yes, your original DEVC line was proper. FDS now reports the time history of 
the W 
component of velocity at the point XYZ.

Original comment by mcgra...@gmail.com on 13 Aug 2009 at 1:35

[GoogleCodeExporter]

Kevin, due to your recent appeal on the discussion group, I'm repeating my 
earlier 
request to close this issue. If I did not say it explicitly, I verified the fix 
you 
made and observed that no droplets were sticking to any walls. Whether this is 
something that is desired or not, I cannot say. The issue was related to the 
cup 
burner and ultra-fine mist. There is no experimental evidence to indicate that 
droplet sticking to the cup or chimney walls would make any difference with 
respect 
to the extinguishing concentration or the droplet aerodynamics close to the 
flame. 

Original comment by jukka.va...@vtt.fi on 18 Nov 2009 at 8:58

[GoogleCodeExporter]

Agreed. Thanks for letting me know. K

Original comment by mcgra...@gmail.com on 18 Nov 2009 at 9:07

• Changed state: Verified