Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Species #857

Closed
gforney opened this issue Jun 24, 2015 · 18 comments
Closed

Species #857

gforney opened this issue Jun 24, 2015 · 18 comments

Comments

@gforney
Copy link
Contributor

gforney commented Jun 24, 2015

Please complete the following lines...

FDS 5.4.1
Win XP

Describe details of the issue below:

The TG.fds case mix air and a tracer gas (same MW). Air supply is 0.0432
kg/s, tracer gas supply is 0.000432 kg/s. This should give a gas mass
fraction of 0.000999 kg/kg. With the default transport scheme I get a
fairly accurate result.  With the transport scheme suggest in issue 878 I
got a mixture with gas mass fraction of 0.00072 kg/kg.  Any suggestion how
to improve the accuracy. 

Could a future version of FDS have the possibility to ramp MASS_FRACTION
with respect to time ?

(tracer gas technique is a common method to analyse the air flow pattern
within a building or a room, but a steady state flow field has to be
established before tracer gas is injected in the air supply)

Attached DEVC.xls shows the mass fraction results 

Finn

Original issue reported on code.google.com by Finn.Drangsholt on 2009-09-30 10:58:25


- _Attachment: [TG.fds](https://storage.googleapis.com/google-code-attachments/fds-smv/issue-883/comment-0/TG.fds)_ - _Attachment: [DEVC.xls](https://storage.googleapis.com/google-code-attachments/fds-smv/issue-883/comment-0/DEVC.xls)_
@gforney
Copy link
Contributor Author

gforney commented Jun 24, 2015

We'll take a look at the case. As for MASS_FRACTION, you can ramp it using RAMP_MF.

By the way, I suggest you not use MASS_FRACTION to introduce gas species. Use 
MASS_FLUX, which is controlled by the same RAMP_MF parameter. It is more accurate 
because it accounts for both the convective and diffusive transport of the gas.

Original issue reported on code.google.com by mcgratta on 2009-09-30 12:35:38

@gforney
Copy link
Contributor Author

gforney commented Jun 24, 2015

Finn,

As Kevin mentioned, all seems ok if you use

&SURF ID='SUPPLY', MASS_FLUX(0)=0.24, TMP_FRONT=20.0, COLOR='SKY BLUE' /

However, this is a good case to help us verify the MASS_FRACTION bc with FDS6.  I'll
keep you posted.

Thanks for bringing this to our attention.

Randy

Original issue reported on code.google.com by randy.mcdermott on 2009-09-30 13:28:22

  • Labels added: Milestone-FDS_5.5

@gforney
Copy link
Contributor Author

gforney commented Jun 24, 2015

Finn,

There were definitely problems with the way I was handling the
SPECIFIED_MASS_FRACTION case in FDS6, so I thank you for posting this case.  These
are now fixed, I believe, with SVN 4846.

However, your case still requires that you use a specified mass flux for your air
inlet (see attached input file).  To achieve your desired mean mass fraction of 0.001
kg TG/kg total, you cannot lose any of the tracer gas to the air inlet.  As Kevin
said, if you specify a mass fraction of 0 TG at the air inlet (which is a boundary,
not a cell center) you will have a diffusive flux OUT of the domain.  A nice
diagnostic tool to see this is to use MASS_FILE=T on DUMP, which is set in the
attached input file.  I find this to be the safest of all the diagnostics.  It gives
the total mass in the domain for all species at each time step.

Please let me know if this is working for you with the latest code.

Thanks.
Randy

Original issue reported on code.google.com by randy.mcdermott on 2009-09-30 17:21:30


- _Attachment: [TG.fds](https://storage.googleapis.com/google-code-attachments/fds-smv/issue-883/comment-3/TG.fds)_

@gforney
Copy link
Contributor Author

gforney commented Jun 24, 2015

I guess I have to wait for the next official FDS revision since I don’t have a 
compiler. Anyway, thank you for clearing up this issue

Finn

Original issue reported on code.google.com by Finn.Drangsholt on 2009-09-30 17:45:36

@gforney
Copy link
Contributor Author

gforney commented Jun 24, 2015

Finn,

I don't think so.  As I said, regardless of the fixes I made to this latest SVN, you
need to set mass flux of the air vent.  This you can do with 5.4.1 and I think it
will work fine.

Randy

Original issue reported on code.google.com by randy.mcdermott on 2009-09-30 17:51:53

@gforney
Copy link
Contributor Author

gforney commented Jun 24, 2015

I tried the case you sent me. I get 2500 oC i the top of the domain and no flow true

the domain.  With FDS6=.FALSE.  it’s ok

Finn

Original issue reported on code.google.com by Finn.Drangsholt on 2009-09-30 19:00:14

@gforney
Copy link
Contributor Author

gforney commented Jun 24, 2015

Sorry.  I should have tried it with 5.4.1 before suggesting it would work.

Kevin, I think we need to release 5.4.2 soon.  This case points out some serious
issues with FDS6 options.

R

Original issue reported on code.google.com by randy.mcdermott on 2009-09-30 20:09:07

@gforney
Copy link
Contributor Author

gforney commented Jun 24, 2015

Finn,

5.4.2 has been released.  Please let me know if this helps your issue.  Thanks.

Randy

Original issue reported on code.google.com by randy.mcdermott on 2009-10-20 16:25:16

@gforney
Copy link
Contributor Author

gforney commented Jun 24, 2015

Enclose follows a case (AE-test.fds and DEVC-AE.xls) where I use VOLUME MEAN to 
calculate the mass fraction in the exhaust with FDS6=TRUE and FDS6= FALSE on the 
MICH line. Steady state result should be 999 ppm. With FDS6 = FALSE the result is 
1000 ppm. With FDS6=TRUE the steady state result is 940 ppm (average of values 
between 2000 and 3000 sec.)  With FDS6=TRUE  the calculation speed is 3 time 
FDS6=FALSE.

Finn

Original issue reported on code.google.com by Finn.Drangsholt on 2009-10-23 19:03:21


- _Attachment: [DEVC-AE.xls](https://storage.googleapis.com/google-code-attachments/fds-smv/issue-883/comment-9/DEVC-AE.xls)_

@gforney
Copy link
Contributor Author

gforney commented Jun 24, 2015

Thanks Finn.  I'll take a look.

Original issue reported on code.google.com by randy.mcdermott on 2009-10-23 19:23:08

@gforney
Copy link
Contributor Author

gforney commented Jun 24, 2015

Finn,

I have been looking at your case for the last couple of days trying to see if I could
find some fat to trim from FDS6.  First, let's look at the specifics of your problem.
I have modified your input file slightly (attached).  I found an ID conflict (which
might not matter, but I changed the SPEC ID to 'TRACER_GAS' and SURF ID to
'GAS_SUPPLY' just to be safe).  Next, to save time I set RADIATION=F on MISC. And I
added MASS_FILE=T on DUMP (this is the mass output I like to use).  Also, if you
really want this to be a tracer the MW should be exactly the same as air or else you
will have non-negligible buoyancy effects.

Okay, with all that said, your problem is certainly valid.  First, the concentration
issue: the default in FDS6 is to clip any mass fractions that are out of bounds.  In
theory the flux limiter prevents this, but in practice in 3D it is possible to have
a
boundedness violation with the FL scheme and so we need a safety net.  The fastest
thing to do is clip the mass fractions, which is usually ok, but for a super long run
time of a ppm level tracer you are seeing the mass vanish.  The default in FDS5 is
a
more costly boundedness correction.  I have added this as an option to the new scalar
transport scheme.  The flag CLIP_MASS_FRACTION=T (FDS6 default) can be turned off (if
you set all options individually).  Note that CLIP_MASS_FRACTION=F is the FDS5
default.  As far as I can tell, both the FDS5 and FDS6 have the correct boundary
condition (see attached mass plots for the first 20 sec).

Next is the timing issue.  There are several places where I tried to make FDS6 run
faster, but in the end the dynamic model is just more expensive, both in the
calculation of the turbulent viscosity and in the fact that the time step is much
smaller for CFL_VELOCITY_NORM>0.  To somewhat alleviate this problem, I have made
several changes to the FDS6 options. The new defaults are:

1) CFL_VELOCITY_NORM=2 (L2 norm of velocity vector)
2) CHECK_VN=T, but based on MU_DNS instead of MU_TURB
3) VN_MAX back to 1.0

We had cranked down on this Von Neumann criterion because of some stability problems
we were having with DNS of cup burners.  However, these new options seem to work with
that case (see Issue 872) -- Jukka, please check this.

I spent most of yesterday trying an idea to develop a "sparse dynamic model" where
we
compute the coefficient on a coarser grid and interpolate to the finer grid.  But it
is really amazing how important it is to use all the information on the finest grid
to get the energy transfer correct.  If I used the sparse model the turbulent energy
spectrum looked horrible for wavenumbers larger than the Nyquist cutoff for the
sparse grid.  Long story short -- that idea does not look like it will work.

So, for the cases I was running (to 20 sec) I was still seeing factors of 2.?
differences in the timing.  But what I think is the more important question is which
simulation you want to believe.  I have attached images of the two cases at 20 sec.

My bet is that reality looks a lot more like the FDS6 image than the FDS5 image. 
With OpenMP coming on line, factors of 2 may be more tolerable as we press forward
with developing better models.  That doesn't mean I am happy about it or complacent,
I will speed things up as much as possible.  But, what I need from our users is
objective tests (with data) that distinguish one model from the other.  How do we get
the answer right for the lowest cost?  So, hopefully the clipping issue is solved
with CLIP_MASS_FRACTION=F.  But this only says that FDS6 will get the SS mass
fraction correct.  We already knew what that was.  So, there is a different reason
to
do CFD: number of air exchanges/hr or something.  How do we know we have that number
right?

I am going to mark this as fixed for now.  But please verify upon next release.  I
am
going to see if I can post a "test" release like Glenn does for Smokeview.  I'll let
you know when that is up.

Best,
Randy

Original issue reported on code.google.com by randy.mcdermott on 2009-10-27 21:05:22


- _Attachment: [AE-TEST.fds](https://storage.googleapis.com/google-code-attachments/fds-smv/issue-883/comment-11/AE-TEST.fds)_ - _Attachment: [fds5_tracer.pdf](https://storage.googleapis.com/google-code-attachments/fds-smv/issue-883/comment-11/fds5_tracer.pdf)_ - _Attachment: [fds6_tracer.pdf](https://storage.googleapis.com/google-code-attachments/fds-smv/issue-883/comment-11/fds6_tracer.pdf)_ - _Attachment: AE_FDS5.png
![AE_FDS5.png](https://storage.googleapis.com/google-code-attachments/fds-smv/issue-883/comment-11/AE_FDS5.png)_ - _Attachment: AE_FDS6.png
![AE_FDS6.png](https://storage.googleapis.com/google-code-attachments/fds-smv/issue-883/comment-11/AE_FDS6.png)_

@gforney
Copy link
Contributor Author

gforney commented Jun 24, 2015

Finn,

What platform do you use?  I have posted a test version of FDS 6 for 32 bit Windows
on the download site.  This is simply SVN 5003 with FDS6=T set by default.  However,
it is now possible to overwrite the FDS6 options.  So, for example, you could set

&MISC FDS6=T (redundant), CLIP_MASS_FRACTION=F/

This implements all other FDS6 options but uses the fully conservative boundedness
check for mass fraction.

Hope this helps.
Randy

Original issue reported on code.google.com by randy.mcdermott on 2009-10-28 16:57:32

@gforney
Copy link
Contributor Author

gforney commented Jun 24, 2015

Note that the installer is deprecated but you can still get to it.

http://fds-smv.googlecode.com/files/fds6_beta_5003_win32.exe

Original issue reported on code.google.com by randy.mcdermott on 2009-10-28 17:26:36

@gforney
Copy link
Contributor Author

gforney commented Jun 24, 2015

Have a look at the attached files
Finn

Original issue reported on code.google.com by Finn.Drangsholt on 2009-10-30 16:27:23


- _Attachment: [vent.fds](https://storage.googleapis.com/google-code-attachments/fds-smv/issue-883/comment-14/vent.fds)_ - _Attachment: [TLK-2K.avi](https://storage.googleapis.com/google-code-attachments/fds-smv/issue-883/comment-14/TLK-2K.avi)_

@gforney
Copy link
Contributor Author

gforney commented Jun 24, 2015

(No text was entered with this change)

Original issue reported on code.google.com by Finn.Drangsholt on 2009-10-30 16:33:25


- _Attachment: [Results.pdf](https://storage.googleapis.com/google-code-attachments/fds-smv/issue-883/comment-15/Results.pdf)_

@gforney
Copy link
Contributor Author

gforney commented Jun 24, 2015

Finn,

Thanks for sending this.  I would argue the results look better with FDS6.  You might
want to try a finer mesh.  I put together a multimesh case (attached) which should
run in a reasonable amount of time.

Please let me know if/when you feel comfortable closing this issue.  Is the steady
state tracer concentration satisfactory?

R

Original issue reported on code.google.com by randy.mcdermott on 2009-10-30 17:29:04


- _Attachment: [vent_multimesh.fds](https://storage.googleapis.com/google-code-attachments/fds-smv/issue-883/comment-16/vent_multimesh.fds)_

@gforney
Copy link
Contributor Author

gforney commented Jun 24, 2015

The steady state concentration changed from 0.00940 kg/kg to 0.00960 kg/kg with 
latest FDS6 - and yes I think you can close this issue

Finn  

Original issue reported on code.google.com by Finn.Drangsholt on 2009-10-31 11:12:53

@gforney
Copy link
Contributor Author

gforney commented Jun 24, 2015

Thanks.

Original issue reported on code.google.com by randy.mcdermott on 2009-11-02 12:59:55

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

1 participant