FW: A question about FDS

[Reneke, Paul A. (Fed)]

Kai,

 

We are still looking at CFAST but the issue is that despite this seeming to be a fairly simple problem it is actually, in terms of the physics, a fairly complex problem. Zone models make a lot of simplifications. One is basically that the geometry isn’t that important. In your problem the geometry in the middle room is everything. One wall, between 1 and 2, is the hottest part. The wall between 2 and 3 is cooler and the air in between the two walls probably is not reasonably approximated by a single volume at a uniform temperature.

 

Kevin McGrattan graciously ginned up a very simple case that can be run in FDS. FDS may seem like overkill but I think it is probably what is needed for this problem. One of the issues that needs to be kept in mind is that we don’t have any test data to compare our answers to. However, FDS does explicitly do all the physics that is involved and so you can have a lot more faith in the results.

 

I hope this helps you answer your question. Sorry it took so long but thanks for asking.

 

Paul

 

 

From: Kevin McGrattan <mcgr...@gmail.com>
Sent: Friday, June 10, 2022 3:53 PM
To: Reneke, Paul A. (Fed) <paul....@nist.gov>
Subject: Re: A question about FDS

 

Three NIST offices 

 

On Fri, Jun 10, 2022 at 3:10 PM Reneke, Paul A. (Fed) <paul....@nist.gov> wrote:

Thanks.

 

From: Kevin McGrattan <mcgr...@gmail.com>
Sent: Friday, June 10, 2022 3:10 PM
To: Reneke, Paul A. (Fed) <paul....@nist.gov>
Subject: Re: A question about FDS

 

Yes. I will put together a silly little case just to be sure.

[stevelikai]

Many thanks Paul, I did not realized the issue is so tough either. I understand that FDS may do the job with heat transfer as I have been worked with Simo on the FDS solid solver for several years. The case I posted is just a test case while we are actually dealing with a scenario with 200 compartments. We have developed an interface to overcome the 100 limit of CFAST and in this case we probably need to consider an alternative solution. 

Cheers,

Kai

paul.reneke 在 2022年6月11日 星期六上午4:48:56 [UTC+8] 的信中寫道：

[dr_jfloyd]

I looked at this case.  I wondered if the underlying issue was the very different layer heights in the two compartments since CFAST only has an upper and a lower layer surface. To reduce time some I just ran for 600 s. I made compartment 2 and 3 shaft compartments but the temperature still dropped. If the compartment 2 and 3 are both a single layer, compartment 2 has a higher temperature than compartment 3, and both compartments are equal in size with the same area of wall connecting the two compartments, then having heat being transferred from the lower temperature compartment to the higher temperature compartment shouldn't happen. There is a fundamental error somewhere as the 1D heat transfer should only be transferring heat from 2 to 3. The next thought was perhaps an error with how wall heat transfer is being apportioned when there is a connection. I then ran the case (still with shaft) changing the wall fraction between 2 and 3.  As I reduced the wall fraction the temperature drop got worse. If I increased the wall fraction the temperature drop got better. 

[Reneke, Paul A. (Fed)]

No. The issue is that because walls are the whole room you have to come up with a method of averaging the amount of heat that goes into the back of the wall and that just isn’t straightforward. If we did 10 wall heat transfer, i.e., ceiling, floor, 4 upper wall segments and 4 lower wall segments and required that walls can only be connected to two compartments we could probably do this reasonably well but that would be a fairly significant recoding effort and without data to test against I would be a little concerned with the answer.

 

Paul

--
You received this message because you are subscribed to the Google Groups "CFAST" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cfast+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cfast/bcdab005-efd9-4acc-a4fd-418c39065350n%40googlegroups.com.

[dr_jfloyd]

I get it now, thanks. I was thinking too much of the FSSIM model I wrote for the Navy while at JH which does track individual surfaces between compartments. The approach taken in FSSIM (which would be a lot of work to implement in CFAST) was surfaces were defined as a wall type, area, connecting compartments, and an orientation (ceiling, floor, fore/aft, port/starboard). The radiation model used view factors based on the compartment dimensions (e.g. the current length, width, height in CFASTs inputs). The source term in the view factor equation for a surface was defined as the area fraction weighted sum of the e sigma T^4 for each surface belonging to that orientation. Not perfect, but generally by the time surface radiation heat transfer became significant the compartments were smoke filled and the gas temperature drove the radiation.