Hello All, Reviving an old thread as I wasn't able able to resolve my problem.
To quickly restate, I'm solving a reaction-diffusion problem using a framework much like Step-52. My equations are stiff and hence I'm using implicit methods for solving the resulting ODEs. Unfortunately, it seems that even implicit methods are not good enough for this stiff problem, the simulation doesn't converge for time steps as low as 0.001! I can definitely try reducing the time step further but, I was wondering if there is any other efficient way to deal with this issue within the dealii framework? I have tried MATLAB's ODE15s and it works good for this problem but, I couldn't find an equivalent C++ implementation. I know about GSL's (GNU Scientific Library) ODE suite and I think I will use that as it includes some advanced ODE solving algorithms. I was wondering if anyone knows about similar or better implementations? On a similar note, are there plans to implement more ODE solving methods in upcoming releases of dealii? Regards, Vaibhav On Friday, September 16, 2016 at 7:59:37 PM UTC+5:30, Bruno Turcksin wrote: > > I have created a pull request here: > https://github.com/dealii/dealii/pull/3123 > > Thanks again > > Bruno > > 2016-09-16 10:01 GMT-04:00 Bruno Turcksin <bruno.t...@gmail.com > <javascript:>>: > > 2016-09-16 9:22 GMT-04:00 Vaibhav Palkar <vaibhav...@iitgn.ac.in > <javascript:>>: > >> > >> The error being exactly zero is indeed intriguing. I'm still looking > out for > >> the reason why that is happening. I didn't understand what you meant by > a > >> "ramp in time". Could you please elaborate? > > Since you are using a low order method in time, I am just wondering if > > the solution is linear in time. This is one way to have the exact > > solution at each time step and it could explain why the error is zero > > (assuming that there is no spatial error) > > > >> Step-52 works fine, I'm getting exactly the same results as the online > >> documentation. I looked at time_stepping.templates.h, I think there is > a bug > >> around line 730, delta_t_guess should be assigned the value of > new_delta_t > >> (the coarsened time step) rather than delta_t (old time step). > > Good catch!! Patch incoming. > > > > Thanks > > > > Bruno > -- The deal.II project is located at http://www.dealii.org/ For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en --- You received this message because you are subscribed to the Google Groups "deal.II User Group" group. To unsubscribe from this group and stop receiving emails from it, send an email to dealii+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
