CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   Time in CFD (http://www.cfd-online.com/Forums/main/3520-time-cfd.html)

Newbie May 25, 2001 21:48

Time in CFD
 
Dear all.

I am solving a multiphase problem with an ALE code.

I need to simulate several hours (more than 10)of the process. In my machine it takes weeks to arrive to the final time.

Are there some techniques to make it faster?

Any help is very appreciated. Thanks

Arturo

John C. Chien May 25, 2001 23:39

Re: Time in CFD
 
(1). It is not uncommon to hear that a cfd run is taking a couple of weeks to converge. (2). If you are not lucky, a few months without a converged solution is not unusual. (3). Just make sure that you have the UPS battery power source installed. In the last few years, power out used to be common even at large companies.

Jim Park May 26, 2001 07:41

Re: Time in CFD
 
Which ALE technique are use using (code name? Or did you write your own?). Is it implicit or explicit? Are you interested in the details of the transient or just the final answer?

From your message, I would guess that you need the details of the transient. So your time step needs to be at least small enough to capture those. If the time step you're using is smaller than that, you MIGHT be able to shorten the calculation time with a larger time step, depending on the code, the computer, your knowledge of those, and a lot of luck.

John's surely right. A good power supply can save the whole calculation if your wall socket goes dead!

Optimist May 26, 2001 12:36

Re: Time in CFD
 
-Can you use multi-grid?? -Can you parallelize the code? -Is the code written in a language that performs well? -Has the code been profiled to see where the problem is??? -can you get access to a better computer (ie. are you with a university...)

These are probably the first things one might look at. I'm not too concerned about power-outs I just save my results every couple of days. I guess if you don't have the storage space then that might be a problem.

Everything is problem dependent, i guess except the faster computer. :)

Newbie May 26, 2001 21:13

Re: Time in CFD
 
Thanks Jhon and Jim I'm using an ALE explicit code and I'm interested in the transient details. I think that the transient process are in a time scale greater than the one that the code itself choices. But I don't know if I can do more on that except ask the developers for a change in the time step choice.

Thanks a lot for your answers, it made me ask questions I didn't even thought.

Newbie.

John June 3, 2001 21:27

Re: Time in CFD
 
Why not use more computing power?

Take a look at tsunamictechnologies.com

John


All times are GMT -4. The time now is 23:52.