This is a read only copy of the old FEniCS QA forum. Please visit the new QA forum to ask questions

Multistep time integrator?

0 votes

I'm about to start writing multistep time integrator based on FEniCS, probably some variation of BDF. And I'm wondering if anyone else is/was working on such subject?

Gryphon project did tremendous job for one-step Runge-Kutta methods, but I can't find anything about multistep. Maybe I'm looking wrong, or maybe it's just weird idea to use BDF?

My problem is very stiff, non-linear, modified diffusion equation modelling bioheat transfer.

Any comments will be very appreciated :)

asked Jan 30, 2014 by sawickib FEniCS Novice (190 points)

1 Answer

0 votes
 
Best answer

Hey, I'm implementing various methods for reaction-diffusion-systems, in fact, today/tonight I want to implement a BDF2 and BDF3 method.
Well, for stiff problems BDF-methods are quite good (as you know, considering you want to implement them ;-) ), although I can't compare them to implicit Runge-Kutta methods, so I think it's at least worth a try. You can always compare them.

answered Jan 30, 2014 by bobby53 FEniCS User (1,260 points)
selected Jan 31, 2014 by sawickib

Thanks for your comment.
So I will try, and we'll see who will be the winner in my case.

commented Jan 31, 2014 by sawickib FEniCS Novice (190 points)
...