>
> Dear Prof. Bangerth,
>

Thanks for your quick reply! 

>  You mean you are wondering why the "grad u square" term grows with the 
size of the problem but "u square" does not? 

Yes, that's what I am confused. Our u_exact have exact "grad u_exact 
square" and  "u_exact square". And our numerical solution "u" and its  
"grad u square" and  "u square" should be approximation of the exact one 
though with adaptive mesh. So I think  "grad u square" should not have so 
huge jump with different adaptive meshes, because they are all the 
approximation of   "grad u_exact square".

> I don't know either (and the one place I looked at in your code seems 
correct to me), but what happens if you do global refinement? Does it 
increase with a fixed ratio from one step to the next? 

I change the code in attaching, and the output is:
Cycle 0:
   Number of active cells:       20
   Number of degrees of freedom: 89
 u square 0: 0.0305621  grad u square 0: 0.256542
Cycle 1:
   Number of active cells:       80
   Number of degrees of freedom: 337
 u square 1: 0.0489011  grad u square 1: 0.341164
Cycle 2:
   Number of active cells:       320
   Number of degrees of freedom: 1313
 u square 2: 0.0540201  grad u square 2: 0.361384
Cycle 3:
   Number of active cells:       1280
   Number of degrees of freedom: 5185
 u square 3: 0.0549874  grad u square 3: 0.365557
Cycle 4:
   Number of active cells:       5120
   Number of degrees of freedom: 20609
 u square 4: 0.0553762  grad u square 4: 0.367069
Cycle 5:
   Number of active cells:       20480
   Number of degrees of freedom: 82177
 u square 5: 0.0555421  grad u square 5: 0.367712
Cycle 6:
   Number of active cells:       81920
   Number of degrees of freedom: 328193
 u square 6: 0.0556186  grad u square 6: 0.367986
Cycle 7:
   Number of active cells:       327680
   Number of degrees of freedom: 1311745
 u square 7: 0.0556576  grad u square 7: 0.368119


and the ratio 

[image: ask-8-1.JPG]
are:
u_square:            1.8410    2.4038    1.3149    1.2287    1.1168    
0.9720
grad_u_square:   2.0652    2.2766    1.4646    1.2336    1.2843    0.9891

And the results without adaptive mesh seem like approximate the exact 
results correctly. So I am confused why the results of "grad u square" with 
adaptive mesh are so strange? Maybe the reason is that we refinre mesh 
where the gradient of u is huge, so only "grad_u_square" can be very 
different with changing adaptive meshes? If so, how to deal with this 
problem? How to approximate   "grad u_ecaxt square" correctly with changing 
adaptive mesh? Maybe change the value smaller of scale of refine cells and 
coarsen cells, to make the change smaller?

Thank you very much!

Best,
Chucui
 

-- 
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 [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to