Grid size & Mesh/Multiple Meshes

[bicky]

I am learning how to use FDS and have to model a fire located within a
speciality shop of a shopping centre. The shopping centre dimensions
are180m x 77m x 6m with speciality shops opening onto central mall. I
am confused on how you determine the correct grid size can please
advise? (for this and future projects). Also how do you determine
when/if you use single mesh or multiple meshes?

Thanks

[Kevin]

Users Guide -- "Choosing Optimum Grid Dimensions"

Users Guide -- "Multiple Meshes and Parallel Processing"

Start by reading these sections. Then ask specific questions if
something is not clear.

[drjfloyd]

Some guidance on grid size is given in 6.3.5 of the User's Guide. It
is only guidance.

Single vs. multiple mesh: If I can afford to run my simulation
(runtime and memory) using one mesh, then I will do so. Otherwise I
will use multiple meshes.

If you are doing this for actual life safety work vs. a school
project, please carefully consider if you are qualified to use FDS for
such a purpose.

On Sep 2, 1:59 am, bicky <markqui...@hotmail.com> wrote:

[Lenny]

This one of the fundementals of CFD.

Follow Kevin and the Dr's advise. It is always sound. Read the manual,
do some google searches, there is a lot of information out there on
this.

These are some of the rules of thumb i use myself

Check your machine an operating system. Roughly 1GB of ram runs 1
million cells.
Run some of the test cases provided, and do some mesh size sensitivity
testing. See the differnce in key variables (say temperature)with mesh
size.
Start with a small room, with a course mesh. Refine the mesh. Add to
the model. (Much easier to debug).
Try to have the meshes be a cube, stretched meshes can influce the
output.

Single mesh vs multiple. If I could I would use a single mesh not
always practical with large models.

[Franck Didieux @ LNE]

In addition to what have already been said about grid sensitivity, you
have also to consider the duration of computation you can afford and
the trueness of results you require (ideally compared to experimental
data). Sometimes, there is just no compromise.
On such a large case, you might expect a very very long calculation if
you use a too much refined grid, but you have to be very very cautious
with the results if the grid you use is too coarse... Try, and check
vs experimental as much as you can.
If you just have two days to solve such a case, it might be wise to
decide not launching the calculation at all... (except on a very
powerfull workstation of course)

Single vs multiple meshes : if you are a new user, try single mesh
first on simple cases. Using multiple meshes requires experience, and
may lead to spurious results, even if used correctly. From experience,
there are some cases where using multiple meshes may lead to numerical
instability and unreal behaviours.

[/SD]

The size of the domain "only" leads to high computing times, so it is
better to start getting some experience studying small cases first.
Then check yourself the effect of changing grid size or using multiple
meshes on such cases.
A good way is to learn from someone already skilled in using FDS.

Remember that FDS will not tell you "you are making wrong choices to
model your case".
So please be extremely careful if your results will be used to design
the fire safety of a real building.
In this case, check your results by making comparison with other
calculations as much as possible (zone models, hand calculation, etc.)

Good luck