FDS restart again

[chen...@gmail.com]

I edited the source code, in every time step I stop FDS and restart
FDS again. Without any other changes and without any input change
either, comparing with FDS results without stopping from start to end.
The results are very different. I compared the temperature curve
output by device ( I used the input file from the testing examples).
Can anybody tell me what I am missing here?
How can I do to fix the problem?

[Glenn Forney]

You asked what you are missing. Do a google search on the "butterfly
effect" and you should see the problem. As you will read from your
search, non-linear model results in general and FDS results in
particular are sensitive to changes in input. In order for the restart
option to work perfectly, the restart file MUST contain the ENTIRE state
of the FDS simulation and represent it EXACTLY. As you can imagine this
is VERY difficult. The day shift can answer this more precisely but it
is my understanding that the restart option was set up for "emergencies"
- say when your computer crashes and you don't want to start an FDS run
from scratch - so with a restart file you can continue the simulation
from when you last saved a restart file.

fyi: The butterfly effect gets its name from a science fiction
story/idea that someone goes far back in time and accidentally steps on
a butterfly. When they return to the present, the present is completely
changed - an illustration how small changes can propagate over time
resulting in big changes later.

Since you say you can modifying and build your own FDS code, I would
figure out how to read in data and modify FDS's state while it is
running without doing a restart. Even this would be dangerous
(numerically) - that is you need to know what you are doing!

By the way, as the author of smokeview, I'll give you another
perspective - I have no idea how all these restarts would effect
smokeview's ability to read in and visualize data properly!

--
Glenn Forney
National Institute of Standards and Technology
100 Bureau Drive, Stop 8663
Gaithersburg MD 20899-8663

Telephone: (301) 975 2313
FAX: (301) 975 4052

Pre-decisional and sensitive information. Not for attribution, distribution, or reproduction.

[dr_jfloyd]

hmmm... In the context of chaotic systems (those very sensitive to
initial inputs), I always thought the butterfly effect related to the
concept of a butterfly flapping its wings in Asia causing a tornado in
Iowa. Though that particular time machine story (Ray Bradbury:at the
end of the story the main character returning from hunting about to
die naturally dinosaurs steps out of the time machine to discover he
now live in a fascist state and finding a butterfly crushed on his
boot.) has been around for a long time. I wonder which came first.

In terms of the restart, there is some variable that we have missed
writing to the restart file. As Glenn noted, the main intent of the
restart was to allow a use to recover from a power failure. It my be
time to reexamine that file writing process. Are results utterly
different (as in different average behavior) or just timewise
different? A useful test would be to do what you are doing but by
slowly adding complexity to the input. Start with INERT surfaces and
just blow ambient air through a vent, then hot air, then a fire, then
add conducting surfaces, etc. If you can figure out what set of
physics causes the code to be different, it will help greatly in
locating what we are missing.

[Glenn Forney]

I was remembering the sci-fi short story you referred to which is
referenced here - http://en.wikipedia.org/wiki/A_Sound_of_Thunder

[dr_jfloyd]

Yep that's it. Wiki is awesome: http://en.wikipedia.org/wiki/Butterfly_effect
suggests the Bradbury came first (1952) but the term was made popular
by Lorenz in his chaos/weather work in th 60's.

On Nov 16, 12:07 pm, Glenn Forney <glenn.for...@nist.gov> wrote:
> I was remembering the sci-fi short story you referred to which is

> referenced here -http://en.wikipedia.org/wiki/A_Sound_of_Thunder

[chen...@gmail.com]

Yes, that is a good suggestions!
Some of variables, the restart file are missing, that is what I guess.
I was going to run the test to find it out. Hope I can make it.
Thanks everybody for the efforts!