[aspect-devel] Limiting dofs with AMR

[Wolfgang Bangerth]

On 05/26/2014 09:53 AM, Timo Heister wrote:
>> It seems like the "Coarsening fraction" and "Refinement fraction"
>> parameters don't have the effect I would expect. I want to stabilize
>> the number of potential degrees of freedom by setting these parameters
>> equal to one another. Intuitively, I would expect the number of cells
>> to stay the same after a refinement in such a case.

Just as a sidenote to what Timo said: if you marked a fraction x of cells for 
refinement and a fraction y for coarsening, and if you marked cells in such a 
way that deal.II can indeed honor your choices, then you would end up with
   (1-x-y)*N + 2^d*x*N + 1/2^d*y*N
cells (every refined cell is replaced by its 2^d children, and every coarsened 
cell is replaced by 1/2^d of its parent). In other words, to keep the number 
of cells constant, you have to choose
   x= (1-1/2^d)/(2^d-1)*y
which is much smaller than y.

But, as Timo pointed out, the fractions you specify in the input file are 
fractions of the error, not fractions of the cells.

Best
  W.


-- 
------------------------------------------------------------------------
Wolfgang Bangerth               email:            bangerth at math.tamu.edu
                                 www: http://www.math.tamu.edu/~bangerth/