Hi. I am solving a time dependent heat conduction equation using deal.II. I used the theta scheme for time discretization (say theta = 0.5 for Crank-Nicholson scheme) and the KellyErrorEstimator as my refinement indicator. The domain is [-1,1]x[-1,1] with initial value of 0 at all space except at [-0.1,0.1]x[-0.1,0.1], where the initial value is 1. As for boundary values, I use 0 for all boundaries at all time.
I used a starting time step of reasonably small (dt ~ 0.002) and with each cycle I decrease the time step by half. However, I realized that as the cycle increases, the errors in the solution increase, notably there are some jumps amd irregularities in my new grid. Impossible values (like negative temperatures) starts to appear. The density of the grid is not right at all, except at cycle = 0 (where I use only global refinement) I wonder if the errors have something to do with my time discretization or the error estimator itself? All the errors appear when I started using adaptive local refinement. I am a little confused about where the things went wrong. I would appreciate if you could enlighten me. Thank you! -- Regards, Soon Hoe _______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
