I know that there could be cancellation errors and other magic stuff… :) … I was joking
Btw, I checked your code carefully and I do not see errors in the implementation. I re-checked it carefully together with Luca and we do not see errors in the implementation. Now, I would like to isolate the function that returns the maximum error (obviously the error could be split in several functions...). As you can see in what I showed last time there is an error in the JxW and this has to be isolated and fixed. Once we got a simpler code to use as failing test we can try to fix the problem or open an issue. Thank you for your test and work, Mauro P.S I will read your new code as soon as possible. On Tuesday, May 24, 2016 at 8:11:12 AM UTC+2, Praveen C wrote: > > Dear Mauro > > I am getting rather large errors for the integral of div(v) and the other > one. Here is the output > > Running with mappingq > Setting RotatedSphericalManifold > Setting SphericalManifold > Setting RotatedSphericalManifold > Setting SphericalManifold > Setting SphericalManifold > Setting SphericalManifold > Grid has been saved into grid.vtk > Surface mesh has 96 cells. > Surface mesh has 48000 degrees of freedom. > Div integral = 8.68643e-22 > n*v integral = 2.6536e-21 > Surface area = 12.5664 > Relative error in surface area = 3.67531e-15 > > Running with mappingmanifold > Setting RotatedSphericalManifold > Setting SphericalManifold > Setting RotatedSphericalManifold > Setting SphericalManifold > Setting SphericalManifold > Setting SphericalManifold > Grid has been saved into grid.vtk > Surface mesh has 96 cells. > Surface mesh has 48000 degrees of freedom. > Div integral = 4.13656e-09 > n*v integral = 3.24529e-09 > Surface area = 12.6084 > Relative error in surface area = 0.00334306 > > Could you run the attached code and see if you get the same result ? > > Note that these integrals are performed on each cell. The value you see > above is the maximum taken over all cells. If you add the integrals from > all cells, that may be zero due to cancellation on the sphere, but the cell > integrals could still be wrong. Note that integral on each cell must be > zero which is what I am checking in my code. > > Thanks > praveen > > On 24-May-2016, at 3:33 AM, Mauro Bardelloni <[email protected] > <javascript:>> wrote: > > Dear Praveen, > > I noticed that the divergence theorem is safe. :) > > Mapping Manifold: > \int_{\partial K} v \cdot n = -8.51746e-21 > \int_{K} \div v = -3.74547e-21 > Mapping Q: > \int_{\partial K} v \cdot n = -5.70692e-22 > \int_{K} \div v = 3.09366e-21 > > BTW, as you pointed out there is something wrong. I tried to calculate the > surface of a unit sphere: > Mapping Manifold: > Area = 12.7092 > Error = 0.142811 > Mapping Q: > Area = 12.5664 > Error = 9.78773e-12 > > Probably there is an issue in the JxW. I printed it for every cell and I > found that > for the Mapping manifold is: > -> 0.529549 > -> 0.529549 > ... > -> 0.529549 > -> 0.529549 > while for the MappingQ is: > -> 0.523599 > -> 0.523599 > ... > -> 0.523599 > -> 0.523599 > > > > > -- 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.
