[Getdp] ... non-linear iteration convergence ...

[Kubicek Bernhard]

What is done in commercial CFD sometimes is that one splits the T^4 like that: T^3*T
The T^3 is taken constant for each matrix, while the T is included linear in the system matrix.
Maybe there is a way of doing that in getdp...

Also, a good initial guess can help a lot...

greetings, bernhard
-----------------------------------
Bernhard Kubicek
Energy Department
Electric Energy Systems

Bitte beachten Sie unsere neuen Kontaktdaten!
Please note our new contact details!

AIT Austrian Institute of Technology
Österreichisches Forschungs- und Prüfzentrum Arsenal Ges.m.b.H.
Giefinggasse 2 | 1210 Vienna | Austria
T +43(0) 50550-6258 | M +43(0) 664 6207748

bernhard.kubicek at ait.ac.at | http://www.ait.ac.at
________________________________________
From: getdp-bounces at ace20.montefiore.ulg.ac.be [getdp-bounces at ace20.montefiore.ulg.ac.be] On Behalf Of Matt Koch [mattkoch at scitex.us]
Sent: Tuesday, September 29, 2009 12:59 AM
To: getdp at geuz.org
Subject: [Getdp] ... non-linear iteration convergence ...

Hello All,

I have made some progress on my radiative heat exchange problem from a
long time ago. This is true radiative heat exchange with configuration
factors, not just a simple T^4 boundary condition. However, I am facing
convergence problems in my non-linear iteration and have to go to
relaxation factors as small as 0.01 and smaller in order to make things
move, and even then it often fails to converge. Does anyone have any
advice on such an issue? Are there other approaches to solving a
non-linear problem in GetDP, i.e. other than "IterativeLoop[ItnCnt,
ItnErr, ItnRlx]"? For example, I stumbled across a JacNL notation
somewhere, but have no idea how to set this up?Any thoughts would be
much appreciated.

Thanks,

Matt Koch, Ph.D., P.E.
Science & Technology Consultants (SciTeX)
www.scitex.us
mattkoch at scitex.us
(830)-249-9499 (o)
(830)-446-1839 (c)