search

firemodels / fds

Fire Dynamics Simulator
https://pages.nist.gov/fds-smv/
Other
664 stars 624 forks source link

Implementation of Moinuddin and Li heat transfer model #945

Closed gforney closed 9 years ago

gforney commented 9 years ago 
Add eddy-viscosity-based convective heat transfer coefficient model of
Moinuddin and Li.  For constant coefficient Smagorinsky model a basic
question is how to cheaply apply the wall function damping factor to the
turbulent viscosity in the first off-wall cell.  The reason is that the
friction velocity is needed and this is computed in the WW model at the
staggered velocity location.  For the moment I have added a crude damping
function based on the idea that Cs for a channel flow should be roughly
0.05 and we use 0.2 by default.  The damping factor is therefore
(0.05/.2)^2 = 0.0625.  Let's see how far this gets us before we do too much
work to add the wall function.  I'd rather spend the time getting this
model work with the dynamic Smag model for use in FDS6, which may not
require a damping function.  If it does, we plan to follow the work of
Stoll and Porte-Agel as we work toward rough surface heat transfer models
for atmospheric flows.

Original issue reported on code.google.com by randy.mcdermott on 2009-12-17 22:28:08

gforney commented 9 years ago 
Changed my mind about the crude damping factor.  This would likely not work because
you still have a high eddy viscosity near the wall and this affects the strain rate.
 The right way to do this is just to add the damping function.

Original issue reported on code.google.com by randy.mcdermott on 2009-12-17 22:47:12

gforney commented 9 years ago 
Jason,

I think the right thing to do here is to store the friction velocity from the WW calc
in a wall array.  Would this not also be help for the soot deposition model?

If I have u* then I can compute the damping function in the wall loop in
COMPUTE_VISCOSITY and this should do it.

R

Original issue reported on code.google.com by randy.mcdermott on 2009-12-17 22:51:14

gforney commented 9 years ago 
FYI, the next release of FDS is going to be a minor one (5.5), so if you need to 
make a slight change in BL formulation, you can do it now.

Original issue reported on code.google.com by mcgratta on 2009-12-17 22:51:37

gforney commented 9 years ago 
Randy,

Storing it makes sense.  Soot makes use of it and no point in computing it twice.

Original issue reported on code.google.com by drjfloyd on 2009-12-18 04:00:31

gforney commented 9 years ago 
Jason,

Storing U_TAU in a wall array is easier said than done, unfortunately, because it is
really computed at edges, which have orientations and we would then need to
interpolate to the face center.  For the moment I have added the constant damping
factor back but now it is done in COMPUTE_VISCOSITY so that the velocity field will
also feel the effect.  Khalid sent me their implementation.  There are a few
deficiencies in their approach.

1. They do not account for the damping factor in the momentum equation, which of
course is important for convection.

2. They compute their 'yplus' from VELCON and MU (the turbulent viscosity), when
strictly speaking this should be computed from TAU_W from WW, which accounts for the
true velocity gradient evaluated at the wall (du/dy)_w and the molecular viscosity.

3. They do not account for the composition and temperature dependence of CP, as you
have done.

So, I think the implementation we have coded up currently should be better.  But we
will wait and see how things shake out once they are able to rerun their tests with
the latest code.

R

Original issue reported on code.google.com by randy.mcdermott on 2009-12-18 15:06:11

gforney commented 9 years ago 
Hi Jason and Randy,

I tried to run my case (presented in the draft paper) using fds executable from: 
http://code.google.com/p/fds-smv/downloads/list
(SVN 5414). I set H_EDDY=T on MISC as per your suggestion.

However, I got some error message which is shown in the attached file. I will 
appreciate it if you kindly comment on this.

Regards,

Khalid

Original issue reported on code.google.com by khalid.moinuddin@vu.edu.au on 2010-01-08 04:39:00

gforney commented 9 years ago 
Khalid,

Please attach your FDS input file.

Thanks.
Randy

Original issue reported on code.google.com by randy.mcdermott on 2010-01-08 13:36:56

gforney commented 9 years ago 
Sorry Randy, I missed your message. Please find the attached file.

Regards,

Khalid

Original issue reported on code.google.com by khalid.moinuddin@vu.edu.au on 2010-01-20 05:35:36

gforney commented 9 years ago 
The code is choking on the line (in the HEAT_TRANSFER_COEFFICIENT function)

YY_GET = YYP(IIG,JJG,KKG,:)

because we pass IIG=-1, JJG=-1, KKG=-1 on Line 862 of wall.f90 (SF%BACKING=VOID).

Kevin said he would take a look.

Original issue reported on code.google.com by randy.mcdermott on 2010-01-20 14:42:03

gforney commented 9 years ago 
I just added 

IF (H_EDDY .AND. IIG>=0) THEN

There are special cases where the back surface of a wall is set to a certain 
temperature, in which we want to compute a heat transfer coefficient to account for

convective loss off the back side.

Original issue reported on code.google.com by mcgratta on 2010-01-20 14:59:57

gforney commented 9 years ago 
Okay.  Now seems to work fine for me, though I am not quite convinced this is the
solution we want.  Are we happy using the old HTC on the back side for that special
case?  Because that is what this change reverts to.

Khalid, can you please check the latest code and verify that this is working for you?

Thanks!

Original issue reported on code.google.com by randy.mcdermott on 2010-01-20 16:49:29

gforney commented 9 years ago 
The fact that we have no info about the back side other than a surface temperature 
limits what we can do. The HTC just reverts back to its simplest form.

Original issue reported on code.google.com by mcgratta on 2010-01-20 16:53:17

gforney commented 9 years ago 
Hi Randy,

Last time I downloaded win32 executable installer from FDS downloads 
http://code.google.com/p/fds-smv/downloads/list

Is there any win32 executable available with the latest code?

Kind regards,

Khalid

Original issue reported on code.google.com by khalid.moinuddin@vu.edu.au on 2010-01-20 22:01:24

gforney commented 9 years ago 
http://fds-smv.googlecode.com/files/fds5_SVN5462_Win32_test.exe

Khalid,

I just posted this as deprecated.  Let me know if you have any problems.

Cheers,
Randy

Original issue reported on code.google.com by randy.mcdermott on 2010-01-20 22:12:09

gforney commented 9 years ago 
Thanks, Rnady.

The case is running. I will contact you once the run is complete.

Regards,

Khalid

Original issue reported on code.google.com by khalid.moinuddin@vu.edu.au on 2010-01-22 02:44:15

gforney commented 9 years ago 
Jason,

I just made a few major changes.  When you get time could you have a look and see if
you approve?

1. I rearranged HEAT_TRANSFER_COEFFICIENT and made calls to WW to calc friction
velocity.  So, now you have U_TAU(IW), the friction velocity, as a wall array. 
Hopefully this will be useful for soot deposition.

2. I added Van Driest damping (easy now that we have U_TAU) to the viscosity near the
wall.  Now, H_EDDY should actually be working correctly.

Khalid, please run your test case with the latest code.

Thanks.
R

Original issue reported on code.google.com by randy.mcdermott on 2010-02-17 23:56:03

gforney commented 9 years ago 
H_EDDY removed from source.  No verification.

Original issue reported on code.google.com by randy.mcdermott on 2011-08-26 19:52:25