[archNUOVO]

Hello all, I am looking for a CFD program which is transparent in its action and can simulate urban airflow accurately (e.g. supports LES or advanced flow model). I would also be exploring natural ventilation possibilities of buildings.


With the word 'transparent', I meant the processes that are going on in the background are expressed in numbers and progress bars. For instance, when the meshing process starts it indicates, shows what percentage is completed at any stage of meshing or solving process, indicates how long will it take to complete meshing or solving process, approximate time left for the rest of the calculations (solver) etc.



The program would be best in automatically utilizing maximum RAM or employs processors in its best use. Manually tuning the RAM and Processors, are also acceptable, instead of automatic operations. However, no such options in the program seem frustrating for me.


Besides, I would be happy with a program which has a great GUI and favours academic research purpose.
Any help from the experts would be highly appreciated.




Regards,
Tanvir M

[FMDenaro]

I am not sure that your requirements can be found in a unique code....

1) the grid generation is a software independent from the CFD code.
2) LES formulation can be found in Fluent as well as OpenFOAM, Saturne_code and FDS. However, consider that often academic LES codes are own-made for specific problems.
3) Are you sure to have the required computational power for LES of large scale urban flow?

[lcarasik]

I think you're asking for something that doesn't outright exist and is by definition too open of a question.

[archNUOVO]

@Filippo Maria Denaro Thank you for the reply. Appreciated.
Could you please suggest me an open source grid generation software?
I wish to explore OpenFOAM’s LES code and I would have access to high computing power soon with 128 GB RAM.
Suggestions regarding academic LES codes would be highly appreciated.


@Lane Carasik
I just wanted to learn if that kind of program is available or at least parts of what a user might like to have, for large scale simulations.

[lcarasik]

Originally Posted by archNUOVO

@Filippo Maria Denaro Thank you for the reply. Appreciated.
Could you please suggest me an open source grid generation software?
I wish to explore OpenFOAM’s LES code and I would have access to high computing power soon with 128 GB RAM.
Suggestions regarding academic LES codes would be highly appreciated.


@Lane Carasik
I just wanted to learn if that kind of program is available or at least parts of what a user might like to have, for large scale simulations.

I think you're going to end up disappointed depending on what type of detail and how big of a domain you're attempting to model.

[sbaffini]

All the processes involved in your task are highly non linear, for which even estimates can be largely wrong. Imagine suddenly launching a CFD run while your machine was already full, because of a previous run. Both your runs will end up swapping, or maybe just one (whcih one?), or maybe one per arbitrary time windows. All the "time to completion" estimates a code could provide, at a certain cost, will end up completely wrong.

If I remember well, Fluent has a trivial estimation of the time to completion for the iterations... but that's not an actual feature, it's nothing different from what you could do by averaging the time/iteration of the previous iterations and multiplying that for the number of remaining iterations. Is that an accurate estimate? Maybe. Maybe not. Depends.

Meshing, by itself, can also have a lot of trouble in providing a reliable estimate of the required time to completion. Not only because of the fact that most meshing processes are not such a one shot thing. But also for those meshing processes which are one shot, it is near to impossible to a have, at a given point in the process, a clear estimate of the remaining amount of work. That's not how these things usually work.

If by automatic RAM and process allocation you refer to a feature by which the code would actually use the most suitable amount of both to solve your problem...well, this is also unusual. The code typically knows nothing of your problem until you submit it. But, to submit it, you need to first pick up a number of processes, to which some RAM is usually bonded. So you end up in Catch-22 (https://en.wikipedia.org/wiki/Catch-22_(logic)).

Even trying to do this, would require tasting the specific system at hand to an extent which is not typically allowed or convenient: launch a job, wait for it, analyze the input to get a measure of the required resources, analyze the system you're running on. Then, the same process can't usually launch another job and you have to exit to get in line for the next actual job run. Highly inefficient. The key here is that most large codes are developed to work without modification on the largest possible set of machines and the largest possible set of use cases.

Finally, a great GUI and a great research attitude are conflicting requirements (not in theory, but experience says they are in practice).

Despite all, however, it's plenty of codes out there:

https://www.cfd-online.com/Wiki/Codes

and you might certainly find something close to your needs. Those already cited by others above surely are the current most preminent open source ones.

Edit: obviously, here i'm talking about engineering estimates, numbers whose errors can be given a clear bound. If you're just talking about some marketing feature which could or not work according to an unknown set of conditions... well, that's always possible.