Influence of domain size.

[woodfire]

Hello everyone,


I am studying the influence of domain size on
FDS output like temperature, velocity etc. at different locations in a
compartment with centrally placed fire source in different time
instants of simulation. The size of compartment and wall thickness as
well as grid size remain fixed. What I am trying to investigate is, to
what extent should I extend my computational domain size beyond the
wall of the compartment in the area where I have an opening. It is
logical to have extended domain in areas having opening, so as to
capture the influence of fire.

So, I gradually vary the extension of my domain
size, keeping everything else same. So far, I have observed greater
variations in temperatures and velocities inside the compartment for
different computational domain sizes, especially in the areas nearer
the fire source. I was expecting convergence as I take larger domains
but that is not the case. Has anyone done such studies before? FDS
does not specify how much should the domain be extended beyond the
external walls of a compartment. Does anyone have idea regarding such
provision? How does the size of computational domain (not compartment)
influence the temperatures and velocities inside the compartment? By
the way, I am extracting data from slice files for the comparisons,
taking a time average of 2 seconds and 20 seconds.

Thank you in advance.


Best regards,
woodfire

[Hostikka Simo]

There is a paper on this very topic in the 9th IAFSS symposium (next
week):

Computation Domain on Simulation of Small Compartment Fires
Yaping He, Chris Jamieson, Alan Jeary and Jian Wang, U Western
Sydney, Australia

Simo

[Kevin]

I have never done a study like the one you describe, but just as an
FYI, I recently modeled 55 compartment fire tests performed by
Steckler, Quintiere and Rinkinen at NIST (then NBS) in 1979. This
dataset has been used by many for model validation. I posted the
results in the FDS Validation Guide, and the datasets and input files
are in the Repository. I suggest that if you conduct a study like the
one you describe that you use one or two of the Steckler cases so that
you can study the sensitivity issue as well as compare to real
measurements.

[woodfire]

Thank you Simo and Kevin.

Kevin, how can I access the input files and datasets that you
mentioned? and where can I find more information on these experiments
like the boundary conditions, materials used, ignition source used
etc.?

I find it interesting what you suggested because it will be easier if
I know an experimental value for a parameter, temperature inside the
compartment for example. I find it difficult to decide what domain
size is reasonable for a simulation because I observed greater
variations in temperatures inside the compartment and velocity at door
and window, with varying computational domain sizes and I think it is
quite difficult to attain convergence especially in the areas near and
just above the fire source.

Best regards,
Woodfire.

[Kevin]

I suggest that you follow the instructions in this wiki:

http://code.google.com/p/fds-smv/wiki/AccessingSubversionRepository

It will enable you to download onto your computer everything in the
FDS repository. If you do not understand everything, find someone to
help you. You may not be familiar with some of the terminology. You
can also get the information directly via

http://code.google.com/p/fds-smv/source/browse/trunk/FDS/trunk/#trunk/Validation

However, I suggest you "check out" the entire repository (the first
suggestion) because then you can refresh the information if we make
changes. The actual test report is available from fire.nist.gov -->
BFRL Pubs On-Line. Search under Steckler.

> > > woodfire- Hide quoted text -
>
> - Show quoted text -

[woodfire]

Hello Kevin,

Thank you very much for your support. I have looked into
the test report and input files. At first glance, I couldn't
understand few things, which I will try to mention specifically:

1. In the test report, the experimentally measured temperature and
velocity data are given in table B. For each test case, 4 or 5
different columns of data for temp. and velocity are given. I couldn't
figure out where these values are measured in the test compartment.

2. In the validation guide, it is said that only maximum temperature
inside the compartment is considered. Is it the maximum temperature at
each height level inside the compartment? How can I filter out these
data from FDS slice file output?

3. Do you think imposing OPEN-Vent B.C. at the external wall, for
example at ZMAX or ZMIN will make any difference to the measured
temperature or velocity inside the compartment? If I want to expose
the external walls of my compartment to passive ambient conditions,
then this is the right thing to do, I think. Please share your views
on this.

Thank you once again. Once I finish running the
simulations, I will update you with the results.


Best regards,
Woodfire

On Sep 17, 4:08 pm, Kevin <mcgra...@gmail.com> wrote:
> I suggest that you follow the instructions in this wiki:
>
> http://code.google.com/p/fds-smv/wiki/AccessingSubversionRepository
>
> It will enable you to download onto your computer everything in the
> FDS repository. If you do not understand everything, find someone to
> help you. You may not be familiar with some of the terminology. You
> can also get the information directly via
>

> http://code.google.com/p/fds-smv/source/browse/trunk/FDS/trunk/#trunk...

[Kevin]

> 1. In the test report, the experimentally measured temperature and
> velocity data are given in table B. For each test case, 4 or 5
> different columns of data for temp. and velocity are given. I couldn't
> figure out where these values are measured in the test compartment.

I scanned the report and converted all the data into electronic files,
all of which are in the repository. Be aware that it is possible that
a few numbers may not have scanned correctly, so check your results
carefully. There is a single vertical temperature profile for each
test measured inside the compartment. Then, there are vertical
profiles of both temperature and velocity in the doorway. The Delta X
and Delta Z on each page indicate the spacing of the instruments
laterally and vertically, respectively. There is also a diagram in the
report explaining where the instruments were located. So far, I have
only used the interior temperature profile and the centerline velocity
profile for comparisons with FDS.

>
> 2. In the validation guide, it is said that only maximum temperature
> inside the compartment is considered. Is it the maximum temperature at
> each height level inside the compartment? How can I filter out these
> data from FDS slice file output?

I use the program called fds2ascii (which comes as part of the
standard FDS installation) to convert "slice" files of temperature and
velocity into time-averaged text files. For example, this "slice" is
really a vertical profile of temperature inside the room (x=2.5,
y=1.1, 0<z<2.18).

&SLCF XB=2.50,2.50,1.10,1.10,0.00,2.18, QUANTITY='TEMPERATURE' /

>
> 3. Do you think imposing OPEN-Vent B.C. at the external wall, for
> example at ZMAX or ZMIN will make any difference to the measured
> temperature or velocity inside the compartment? If I want to expose
> the external walls of my compartment to passive ambient conditions,
> then this is the right thing to do, I think. Please share your views
> on this.

Ken Steckler told me that the floor extended outside the compartment,
which is why there is no OPEN boundary at ZMIN. However, I did OPEN
the boundary outside the compartment on all other sides, especially
the top (ZMAX). I think that the exhaust plume does effect the doorway
profile and to some extent the internal environment, but I will leave
it to you to tell me exactly how much. As for exposing external walls
to ambient conditions, you do not need to do this explicitly. It is
assumed by FDS that the exterior of the computational domain is at
ambient temperature. The compartment walls have a specified thickness,
and heat is assumed to be lost to an infinite, ambient void.

[woodfire]

Hello Kevin,

If I understood it correctly, the FDS output that you have
put in the repository is just at the locations mentioned in the input
file. But what about the time averaging? Did you averaged the data
over 1800 seconds? In the test report, the data given are average of
360 reading, each taken at 5 second interval, i.e. 1800 seconds. This
prompts me to raise one more question: is it logical or wise enough to
compare data averaged over such a long time for validation or
sensitivity analysis purposes? Normally, I am comparing data averaged
at short interval like 20 or 10 seconds and they show big variance
with different domain sizes. And that makes me think, shorter the time
interval for averaging, greater the variations in data recorded by FDS
and I think it is really difficult to attain convergence for short-
time-averaged data or say instantaneous data, e.g. 500 seconds or 1100
seconds.



Best regards,
Woodfire

[Kevin]

The objective of the experiment was to measure the steady-state (time-
averaged) velocity and temperature profiles. To do this, they heated
the compartment for enough time so that the wall temperatures stopped
changing, then they took their meaurements and averaged the results.
In FDS, we use a simple trick (TIME_SHRINK_FACTOR) to run 1800 s of
simulation in 180 s by reducing the specific heat of the compartment
walls by a factor of 10. Essentially, we just want to accelerate the
heating of the walls to steady state.

Short time averaging makes no sense at all. In these kinds of
experiments, it is not unusual to see the fire "move" about the pan
over tens of seconds, producing short term variations in temperature
and velocity. I am interested in comparing my long time-averaged
results with the experimentalists long time-averaged results.

> > and heat is assumed to be lost to an infinite, ambient void.- Hide quoted text -

[Boris Stock]

> In FDS, we use a simple trick (TIME_SHRINK_FACTOR) to run 1800 s of

This parameter isn't mentioned in the User's Guide. It belongs to MATL?

[dr_jfloyd]

&TIME TIME_SHRINK_FACTOR=x
where x is the amount to accelerate the simulation. SPECIFIC_HEAT is
divided by the factor as well as any time based &RAMP functions.

[Kevin]

This parameter has not been put into the User's Guide because we fear
that it will be misused. In the Steckler compartment simulations, we
were only interested in the steady state flow field, something that is
not typical in a fire simulation where time-dependence is important.
So we created this simple trick to allow the walls to heat up roughly
10 (in this case) times faster. The actual Steckler experiments lasted
about 30 min, which we accelerated to 3. We worry that people will
think that they can run their FDS simulations 10 times faster by doing
this. We could mention it, but we'd have to include many warnings (and
even then we will probably regret it!).

On Sep 18, 12:35 pm, Boris Stock <Boris.St...@gmx.de> wrote: