Applications and Limitations of FDS
frans dommelen
I'm a mechanical engineer (so not a fire specialist) and at the moment
involved as FIDIC engineer in a project.
FDS is suggested to be used for the fire related designs (fire
protection, fire extinguishing and mechanical heat-smoke-extraction).
I have studied a lot of the information available on the FDS website
(at NIST), but the limitations of using FDS for these designs is not
clear enough to me.
Therefore I have the following questions to the (specialist) members
of this group:
1.Can FDS be used to judge which fire extinguishing system is best
applicable from fire extinguishing point of view (sprinkler or
watermist or gas extinguishing)?
2.Can FDS be used to decide quantity of heat- and smoke extraction in
relation to applied fire extinguishing system?
3.Can FDS be used to decide fire protection provisions of structures
(columns, beams and slabs) in relation to applied fire extinguishing
system?
My main interest for your answers is in proper application of FDS and
it's limitations (reliability and unreliability of the results of the
simulations).
Awaiting your replies and thanks in advance.
Frans van Dommelen
Barbro Maria Storm
First of all, as a non-native speaker of English I'm not sure what
FIDIC is an abbreviation for, and thus I don't know what your role in
this is.
That said, any simulation with FDS should be done by a trained fire
engineering specialist with experience in simulation. He/she will be
able to tell you the limitations of your specific project and scope.
In my opinion it's simply impossible to answer that question for every
single possible building with varying ventilation and
extinguishing-systems and coverage.
--
Barb.
Fire Engineer
S.Desanghere
I think FDS limitations and capabilities are not expressly mentioned
because it has no sense to do so (I mean in a general way). I suggest
having a close look at the V&V guides of FDS in order to see if there
are cases similar to what you are going to study.
In my opinion, FDS (when correctly used !) is pretty good at
simulating hot gas flows. So, it can provide very useful insight when
dealing with smoke management.
The calculation of thermal actions to structural members is correctly
handled by FDS, but you must know that it strongly depends on the heat
release rate of the fire YOU define.
Fire extinguishing is much more complicated than the two previous
topics, so it is still very difficult to rely only on numerical
simulations. To me, it can only give "qualitative" answers.
Remember that it is the job of the user to give the right parameters
to the model. A good model is efficient only in good hands. You have
to be sufficiently experienced in fire physics to be able to validate
any numerical result.
If the safety of people is involved, you cannot do without a fire
specialist, at least to validate your work.
Good luck
Sylvain
On 16 déc, 11:21, frans dommelen
Kevin
It is difficult to say what FDS can or cannot do. The reason is that
any mathematical model uses simplified physics that may or may not be
appropriate for a particular application. That is why we, the
developers, try to explain as best we can the physical assumptions in
the model, and we publish verfication and validation studies that
quantify the accuracy of the model for various applications. That
being said, I will give you some advice on your topics:
> 1. Can FDS be used to judge which fire extinguishing system is best
> watermist or gas extinguishing)?
these systems. In most cases, FDS is used to assess the RELATIVE
performance of the system. By this I mean the following. Suppose you
have a particular rack storage configuration that you are trying to
protect with a sprinkler system. Next, suppose you use FDS to simulate
a particular fire scenario. You would have to judge, based on your
experience or experimental data, whether or not FDS is providing an
acceptable level of accuracy. This would be your decision to make.
There is no formal guidance on what constitutes "acceptable." If you
decide the model is acceptable for your application, you then compare
calculations in which various input parameters have been changed. The
model can provide insight as to whether or not a change in sprinkler
spacing, flow rate, ceiling offset, etc., provides better or worse
protection. For complicated storage arrangements, I do not think it is
possible, given the simplifications in the model currently, that you
can predict outright whether or not a given design is going to provide
adequate protection for scenarios with complicated combustible loads.
However, the model can provide insights as to whether one protection
strategy is better than another. If you want to know if any given
design is going to "work," you have to test in a full-scale
experiment. The model is intended to reduce, but not eliminate, the
need for full-scale experiments.
> 2. Can FDS be used to decide quantity of heat- and smoke extraction in
and it would output the change in these quantities after the
application of an extinguishing system. This only means that FDS
predicts and outputs these quantities. The cautions in the answer to
Question 1 still apply.
> 3. Can FDS be used to decide fire protection provisions of structures
> system?
the temperature of exposed structural elements like columns and beams.
It does not predict loss of structural integrity, however. As stated
in the answer to Question 2, FDS can account for a change in the
thermal environment due to an extinguishing system. When I say "can",
I mean that FDS has a mathematical model that accounts for it. It does
not imply that FDS can do it to a particular level of accurately. As a
developer, I can tell you what I have added to the model, but I am not
the one who says that the model is acceptable or sufficiently
accurate. That is your decision.
szilagyics
The FDS SV Survey file was really interesting report from the works
with FDS.
Can you give a weblink from it?
Frans,
The FDS is a mathematical model not the real world. It's really
important what we use in the input data (like material properties and
geometry) and maybe more important how we assess the results. So I
thought it's a complex work.
Regards,
Csaba
Rein
know/report about the limitations of FDS [*]. This lack is even more
dramatic with any other fire model out there; FDS is by far the best
validated and the most transparent so far. However, I wish more work
could be made available on this. Note that the V&V extensive document
from the FDS developers is a great starting point. However, FDS is
applied currently in many more different scenarios that those
addressed in the V&V document. This other applications rarely see the
public light.
From my particular and limited experience, I can provide openly one
case where fire modelling with FDS worked well (#1) and one case where
fire modelling with FDS (&CFAST) did not work well (#2):
#1 Experimental Data and Numerical Modelling of 1.3 and 2.3 MW Fires
in a 20 m Cubic Atrium, Building and Environment 44, pp. 1827–1839,
2009. http://dx.doi.org/10.1016/j.buildenv.2008.12.010 (Open access
version: http://hdl.handle.net/1842/2761)
#2 Round-Robin Study of a priori Modelling Predictions of The
Dalmarnock Fire Test One, Fire Safety Journal 44, pp. 590-602, 2009.
http://dx.doi.org/10.1016/j.firesaf.2008.12.008 (Open access version:
http://hdl.handle.net/1842/2704)
Cheers
G.
[*] There is very few brave people in this world that would risk to
publish a report/paper showing that their results are wrong
Kevin
Guide
http://fds-smv.googlecode.com/svn/trunk/FDS/trunk/Manuals/All_PDF_Files/FDS_5_Validation_Guide.pdf
The intent of this Guide is to provide references to validation work
by others and a detailed description of work done by the developers.
However, with the exception of yourself and a few others, I have
received very little feedback from my request to have papers and/or
citations to FDS validation work:
http://groups.google.com/group/fds-smv/browse_thread/thread/8278fff87e8fbf5a#
Nevertheless, there are a considerable number of papers, reports, and
student theses cited in the Guide that we have dug out of the
literature. What more would you have us do?
On Dec 17, 8:13 am, Rein <rei...@gmail.com> wrote:
> This is a very interesting discussion. I agree that very few people
> know/report about the limitations of FDS [*]. This lack is even more
> dramatic with any other fire model out there; FDS is by far the best
> validated and the most transparent so far. However, I wish more work
> could be made available on this. Note that the V&V extensive document
> from the FDS developers is a great starting point. However, FDS is
> applied currently in many more different scenarios that those
> addressed in the V&V document. This other applications rarely see the
> public light.
>
> From my particular and limited experience, I can provide openly one
> case where fire modelling with FDS worked well (#1) and one case where
> fire modelling with FDS (&CFAST) did not work well (#2):
>
> #1 Experimental Data and Numerical Modelling of 1.3 and 2.3 MW Fires
> in a 20 m Cubic Atrium, Building and Environment 44, pp. 1827–1839,
> 2009.http://dx.doi.org/10.1016/j.buildenv.2008.12.010(Open access
> version:http://hdl.handle.net/1842/2761)
>
> #2 Round-Robin Study of a priori Modelling Predictions of The
> Dalmarnock Fire Test One, Fire Safety Journal 44, pp. 590-602, 2009.http://dx.doi.org/10.1016/j.firesaf.2008.12.008(Open access version:http://hdl.handle.net/1842/2704)
> > > accurate. That is your decision.- Hide quoted text -
>
> - Show quoted text -
clauten
I do not believe that it is accurate to say for the Dalmarnock round
robin that FDS "did not work well". More accurately, different FDS
users conducted fire growth simulations with different material
properties and got different results--as one would expect. The scatter
in the predicted HRR curves is an obvious user effect, not necessarily
attributed to any inherent deficiency of FDS. Since Cone Calorimeter
data were not available for each material in the study, participants
had to rely on literature data for material properties. Due to the
dearth of this data in the literature, the best one can hope for is to
"bracket" the HRR, i.e. provide lower and upper estimates of HRR based
on a sensitivity study--and that appears to be what happened in the
round robin FDS modeling. To shrink the "error bars", more accurate
material properties would have been required.
Chris
On Dec 17, 5:13 am, Rein <rei...@gmail.com> wrote:
> This is a very interesting discussion. I agree that very few people
> know/report about the limitations of FDS [*]. This lack is even more
> dramatic with any other fire model out there; FDS is by far the best
> validated and the most transparent so far. However, I wish more work
> could be made available on this. Note that the V&V extensive document
> from the FDS developers is a great starting point. However, FDS is
> applied currently in many more different scenarios that those
> addressed in the V&V document. This other applications rarely see the
> public light.
>
> From my particular and limited experience, I can provide openly one
> case where fire modelling with FDS worked well (#1) and one case where
> fire modelling with FDS (&CFAST) did not work well (#2):
>
> #1 Experimental Data and Numerical Modelling of 1.3 and 2.3 MW Fires
> in a 20 m Cubic Atrium, Building and Environment 44, pp. 1827–1839,
> 2009.http://dx.doi.org/10.1016/j.buildenv.2008.12.010(Open access
> version:http://hdl.handle.net/1842/2761)
>
> #2 Round-Robin Study of a priori Modelling Predictions of The
> Dalmarnock Fire Test One, Fire Safety Journal 44, pp. 590-602, 2009.http://dx.doi.org/10.1016/j.firesaf.2008.12.008(Open access version:http://hdl.handle.net/1842/2704)
Rein
fabulous job. Your work exceeded in quantity and quality my
expectations and predictions long time ago.FDS is the best tool
currently available for fire modelling. It is the state of the art.
Chris: I did not say FDS did not work well in the Dalmarnock round-
robin. I said "fire modelling with FDS (&CFAST) did not work well".
The differences between 'fire model' and 'fire modelling' are
addressed in the paper and are in my opinion essential concepts for
fire safety engineers. The conclusions of that paper should apply to
any fire model and not just to FDS and CFAST.
Cheers
G.
> > Dalmarnock Fire Test One, Fire Safety Journal 44, pp. 590-602, 2009.http://dx.doi.org/10.1016/j.firesaf.2008.12.008(Openaccess version:http://hdl.handle.net/1842/2704)
frans dommelen
Thanks a lot for your responses, and here I like to clarify some by
you raised questions.
For Babro:
FIDIC, the International Federation of Consulting Engineers (the
acronym stands for the French version of the name) represents globally
the consulting engineering industry. (see http://www.fidic.org/)
At the project I am working as independent FIDIC engineer between the
investor/employer and the contractor (design & build). Basically I as
part of a whole team have to review first the designs first and
afterwards the execution of the works. As the Fire Protection is not a
separate trade but integrated into the architectural, structural,
mechanical and electrical trades. So I will not work with the FDS
itself, but have to review/comment similaation(s) and give advises to
the client. The FDS is planned to be used to justify if exemptions to
the local actual regulations are first of all possible and second may
be permittable by the local authorities.
For Sylvain:
Thanks for clear staement of your opinion.
And as explained above I'll not do the FDS work at all, only need to
get at least an understanding of it's proper application and
limitations. Therefore I've gone through all the available info on the
NIST-FDS website. But to get some feedback from people working with it
posted my question.
For Kevin:
Thanks a lot for your extensive answer, making clear that full scale
experiments stay necessary (so their results can be used for other
situations mostly depending on how far to go with "remodelling" them).
Clarification for question 3: Regulations define equirements for fire
resistance of structures (in the project: concrete and steel). The
foreseen steel structures themslves can not fullfill these
requirements and need additional provisions. Practically that means
fire protective painting and/or covering, which for the exposed parts
have influence on the appearance of those structures. my question is
may any fire extinguishing system be considered giving (partly)
protection to the structures. If yes it could mean improving the
overall appearance of structures (including applied extinguishing). My
main concern is that the structures are a statical feature, generally
including statical fire protection, and the extinguishing system is a
dynamical feature (which implies possible failure). So how far does it
seem acceptable to replace statical protection by dynamical
protection.
Hope the above makes things more clear.
For Csaba:
Thanks for pointing out that FDS is a model (only) and not the real
world.
For Rein:
Thanks for stating clear that FDS is by far the best in it's field,
more information about V&V is highly necessary and at the same time
providing additional V&V info.
Thank you all for your attention and in advance for further replies.
Frans van Dommelen
Kevin
Thanks for clarifying your 3rd question, even though I don't have much
to say on the subject as this seems more like a regulatory, rather
than a modeling, issue.
K
On Dec 18, 10:16 am, frans dommelen
<tfc.fidic.frans.domme...@citromail.hu> wrote:
> Dear members of FDS group,
>
> Thanks a lot for your responses, and here I like to clarify some by
> you raised questions.
>
> For Babro:
> FIDIC, the International Federation of Consulting Engineers (the
> acronym stands for the French version of the name) represents globally
> the consulting engineering industry. (seehttp://www.fidic.org/)
dr_jfloyd
The answer to your third question is going to depend on the
regulator. Some projects I work on, one can take credit for fixed
suppression systems, for some projects one cannot take credit for
them, and other times it depends upon how redundant the fixed system
is (such as multiple supplies of water and multiple, independent
risers). That is a question whose answer may be project dependent.
szilagyics
Have you a new FDS-SV-Survey? I have found a one on my computer from
2005. This is a pdf file from different projects, maybe it can give
some information to somebody who is new in FDS where the FDS is in
useable.
Thank you for your attention,
Csaba
Kevin
> > > > Frans van Dommelen- Hide quoted text -
frans dommelen
Off course my 3rd question is related to regulations, but I think
that's valid for all design. I mean requirements and defining limits
for acceptance is somehow coming from regulations, also for all fire
related designs.
My main point with that question,
"3.Can FDS be used to decide fire protection provisions of structures
(columns, beams and slabs) in relation to applied fire extinguishing
system?", is:
3a: How accurate can FDS model/simulate/predict parameters important
for structures (mainly temperatures) under (by regulation) defined
fire circumstances.
3b: Same question if fire extinguishing (sprinkler/watermist/gas
extinguishing) is applied.
Thanks you all for your attention.
And wish you "Merry Christmas and Happy New Year"
Frans van Dommelen
welchs
> more accurate material properties would have been required.
I concur with Chris's comments here but I would go beyond the solid
phase to look at the gas-solid coupling. As we all know many of the
processes involved here are highly non-linear and coupled. Thus if we
can't accurately resolve the details of the temperature, soot and
species fields, and related optical properties, we ought not to expect
to be able to reliably reproduce the right heat fluxes without resort
to substantial empiricism. This is further compounded by the need to
be able to do this for real world fuels (i.e. mixtures of volatiles)
and at sufficiently small scales (or with pdf models sufficiently and
adequately resolved in some compositional space, e.g. mixture
fraction). All of this is hugely challenging, and all applies to fire
models in general, not just FDS. One small example, I recall
grappling with upward flame spread prediction over simple materials
using SOFIE - in one full-scale scenario accidentally turning off the
soot model completely eliminated flame spread, even though predicted
gas temperatures were higher (the exposure being the combined
influence of soot concentration via the interacting fields of optical
properties and temperatures). Due to the feedback processes inherent
in fire growth, the resulting behaviour is so drastically different
that effectively you have a no spread/rapid spread scenario, so it's
not just a matter of getting the time of fire development wrong by
some percentage. All this could apply to DFT (ref #2 above), even
though it was somewhat artificial in having a large fuel load near the
fire source, which once ignited goes to flashover - since the large
fire is only relevant once the small fire is established and spreading
on the second item and the uncertainties in that are huge, whether
ignition occurs "globally" over reasonably large areas or more locally
and spreads; i.e. it is not only due to material properties
uncertainties that I would not expect us to be able to reliably
"predict" that in a quantitative sense. Incidentally, though this may
be OT for this forum, I was one of the teams which was keen to join
the DFT Round Robin but forced to withdraw due to the challenge of
making the SOFIE flame spread predictions behave! It is still a
source of great regret that we were not able to provide a RANS
contribution to the exercise, though any such predictions would have
been subject to all the same constraints mentioned above.
Furthermore, having just re-read the overlapping priori/posteriori
debates on here* and the Edinburgh blog** I fully agree with all of
the insightful comments posted there, particularly those of Bart
imploring us to do sensitivity analyses if we are to make any sense of
results where small initial differences *can* produce wildly different
final outcomes; I do understand the distinction on fire model/fire
modelling but nevertheless I would not be comfortable with the
statement "fire modelling did not work well" if this is implied to
refer to deterministic models, as IMHO the task set for them is simply
too difficult!
Finally, in the same spirit, I think we have learnt many useful
lessons here in our attempts to do sensor-linked modelling for
FireGrid (Koo et al IAFSS9, FSJ paper in the pipeline). This is
telling us that even when we are continuously directing our model via
(pseudo) measurements from the fire the lead time for reliable
predictions is very debatable. In a well behaved fire we might have
confidence in a few minutes hence estimate of whether or not flashover
will occur, but in practice uncertaities arising e.g. from interaction
between burning items - one burns out, so spread to a second is on a
knife edge and everything beyond that is different, OR, the glass
drops out when a gust of wind arrives, OR, an aerosol explodes, OR, a
fireman opens the door, etc. - mean that in real-life scenarios our
"predictions", even though they extend to the likelihoods of a range
of different scenarios developing, have to be tempered by a healthy
dose of reality - they might all be completely wrong and can only ever
be indicative of the potential of a hazard arising, rather than an
indication of what will happen. All the more so then with pure
"blind" predictions...
* http://groups.google.com/group/fds-smv/browse_thread/thread/a9459d7427dd4c39/60cca8912232e640?q=#60cca8912232e640
** http://edinburghfireresearch.blogspot.com/2009/08/blind-vs-open-fire-modelling.html
welchs
profile definition (sorry, new here!): Stephen Welch, lecturer in fire
modelling at Uni of Edinburgh*. The SOFIE work I referred to was
published at the 4th ISFEH, 8-12/9/03, pp. 259-270
Kevin
very complicated. For that reason, we are trying to decouple the two.
On the solid side, there is a project underway to develop the
necessary guidance to take benchscale measurements of thermo-physical
material properties and apply them within existing pyrolysis models.
The end goal of this project is to demonstrate a robust methodology to
describe the decomposition of a solid exposed to a fixed heat flux --
like a virtual cone calorimeter with no flames. We want to eliminate
as much as possible the uncertainty of the gas phase calculation -- we
just want to know how well we can describe the solid phase.
Next, there are efforts underway to evaluate how well the models can
predict the gas phase environment, in particular the heat flux to
solid surfaces. This typically involves looking at experiments where
there is a specified HRR and no reacting surfaces. How well can the
model predict the heat flux?
Once we understand how well the models can handle both parts of the
problem separately, then we can combine them and look at cases in
which we predict, as opposed to specify, the HRR.