CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Phoenics (http://www.cfd-online.com/Forums/phoenics/)
-   -   Help on EXPERT (http://www.cfd-online.com/Forums/phoenics/51226-help-expert.html)

Tim July 17, 2001 17:13

Help on EXPERT
 
Hi,

I am new to CFD and PHOENICS and I am trying to simulate natural convection around a series of cylinders. I have been having trouble with convergence so I have tried using EXPERT. I have gotten expert to work with EXPERT=T, NOWIPE=T, and by LOAD(030) into the Q1 file (note that in POLIS it says LOAD(30), which does not work...). However, for some of my simulations that I did get convergence for, using EXPERT resulted in an increase in time and number of iterations before convergence occurred. Is this something that occurrs from time to time? Secondly, I have no information on expsol, expdtf, inres, etc... Where can I find information on this? I perused through the polis but found nothing on these. If I turn expsol to true, what does it do? What does it adjust?

Thank you for any insight on these questions, Tim

Mike Malin July 19, 2001 05:30

Re: Help on EXPERT
 
EXPERT has fell into disuse at CHAM as a dynamic relaxation device, largely because it has been superceded by SARAH (Self-Adjusting Relaxation AlgoritHm).

SARAH is documented on-line at the following web address:

http://www.cham.co.uk/phoenics/d_pol...nc_s.htm#sarah

SARAH can be set in the VR Editor or in the Q1 input file by means of the PIL Command SARAH=0.001, for example. Typical values of SARAH range from 0.00001 to 0.1, depending on the application. Like EXPERT it does not always lead to improvement.

Tim August 9, 2001 16:31

Re: Help on EXPERT
 
Thanks Mike, really appreciate the input. Is there any way of determining what a suitable value for SARAH would be or is it simply trial and error? Also, will the initial values used for the false time step relaxation factors affect convergence is using SARAH?

Thanks, Tim

Mike Malin August 13, 2001 10:02

Re: Help on EXPERT
 
The optimim value of SARAH depends on the application, but typically I start with 1.e-3 and if PHOENICS diverges or oscillates, I then use the graphical interrupt facility to lower SARAH to 1.e-4 or even 1.e-5 if need be. If PHOENICS doesn't recover, I can then start again with the lower SARAH value.

I don't believe that the initial DTFALS values should have an effect, but I have not experimented with them, so I cannot be sure without trying.


Tim August 15, 2001 10:50

Re: Help on EXPERT
 
Thanks again Mike!

I have tried using SARAH starting with 1e-3 and I found that it was too large and changed it in flight. Found out that a SARAH value of 1e-5 is what is needed.

Thanks again, Tim

PattiMichelle August 21, 2001 14:15

Re: Help on EXPERT
 
Isn't that funny? I thought that EXPERT was newer than SARAH. I got the impression from the documentation that SARAH was much older. I guess they need a little more updating. Patti


All times are GMT -4. The time now is 11:27.