Here is a though one (at least for me) I been trying to model hydrogen dispersion within a neutraly stratified atmosphere and I been having some trouble for more than amonth. First I tryed steady-state; this of course had convergance trouble. So I'm now trying to set-up a transient model. I'm having start-up problems. Here are the specs of my model definition. Any feedback would be inmensly apreciated.

•Domain size; L,W,H 750,350,250 meters

•Turbulent Flow: k-epsilon , 0.01-0.0002 (from ref)

•Inlet: U=7m/s and Temperature profile T/Z -2.45 K

•Density: Ideal(T)

•Hydrogen being model as a scalar from a pipe in center of domain with W=1m/s.

•Bouyancy; ON (having doubts on datum definition need help)

•Algorithm; PISO and Implicit Time-Disc

•Initialization from steady-state (no H2) run

•Time step 0.1s (having trouble defining this)

To whom ever responds... THANKS A LOT!

haoa

I forgot to add this:

Boundary conditions;

•Inlet at east

•Outlet (split=1) at west

•No-Slip wall at bottom

•Symetry sides and top

USING STAR-CD

Thanks!

A couple of things you might want to look at:

I'd start with a time step that is 0.1 of the minimum cell transit time, the time it takes a packet of fluid to cross a cell.

First find min over all cells of (dx/u, dy/v, dz/w), then set your dt as 0.1 of that. The code should eventually take over and adjust (grow) the time step as needed...

You might check your references for the velocity distribution across an atmospheric boundary layer. It can be several meters thick. This would change your upstream boundary condition.

Good luck!

As Jim mentioned, you'll need to specify the atmospheric boundary layer velocity profile. A quick google has provided:

http://personalpages.umist.ac.uk/sta...bl/regions.pdf

Look at the section containing equation 16.

You'll need similar profile descriptions for other turbulence quantities.

More quick and lazy googling finds:

http://www.dnv.com/binaries/Applicat..._tcm4-8313.pdf

I'm sure there's lots more out there.

What I've done in the past is to start the model simple, ensure that the defined ABL velocity profile does not degrade unduly as it passes over an 'empty' flat grounded domain. To support the sheer size of the ABL (sometimes up to 300m high) you should investigate the application of the correct roughness height to the ground domain plane. Once you're happy with the ABL then introduce the contaminant source into your model.

If you've got a point source of hydrogen then you'll need bucket loads of grid and a nice high order scheme to ensure that the intial spreading is not swamped by good old numerical diffusion.

Good luck, it's tough old problem but one that you should find lots of reference to in the literature. Hey, maybe the guys at CD support will be also be able to give appropriate help and advice!

Jim and Harry; Thanks for your feedback.

I used a uniform approach velocity with the idea that the BL would be fully developed by the time it got to my hydrogen source. (this is not of the same size of the actuial (ABL). But changing it to a log vel profile is a good idea.

•Do you think that the inlet vel-pprofile is truly critical?

Also my mesh is tetrahedral and is refined considerably (from 15m far from source to 4m near the source) keepping the size changes under 20%.

•The high order scheme that you suggest; do you think QUICK for all variables should work, or only for my scalar (H2?)

Thanks A LOT.

Hello,

I fully agree with Harry regarding the neccessity of having good start profiles.

Way back at my time at the university, I was dealing with atmospheric flows. I was calculating the wake of wind turbines wrt. atmospheric stratification. I remember that it was very important to have profiles of velocity, k and eps corresponding to atmosperic bl type profiles as boundary conditions, otherwise the effect of creating the bl profile was mixed with the effect of the wind turbine wake. We created the profiles by doing a 2-d simulation of a very (I remember about 2km long or so) long flat plate with appropriate roughness and stability conditions set.

Btw, the profiles are only valid in the lower part of the atmosphere (depending on stability, maybe about 200m above ground), above we assumed a constant velocity with very low turbulence level. A good starting point for information on atmospheric flow and tubulence is the book "Atmospheric turbulence" by Dutton, Panofsky and maybe "A first course in turbulence" by Tennekes, Lumley.

Hth,

-Thomas

Thanks for all your comments... I have the articles and the text books... I will revisit them.

Hope to payback the favor soon!

thnx!

Hector A. Olvera

I need some help:

I'm running the H2 dispersion in a stratified atmosphere; the scalar equation (H2) is not converging at 80 corrector steps and it has negative values (extremly small ones x*EE-175). Does any one have any suggestions why this could be hapening?

thnx!

What numerical scheme are you using? Is it bounded?

I'm using PISO I have and Inlet and Outlet boundaries in east and west. no-slip on ground and symetry all the rest...

i the residuals are 0.99 approx at 30 corrector steps