All, The version of the code I attached has line 70 already set to "1.0" (Domain 2) instead of "0.5" (Domain 1). Domain 1 is the version which will not work as it has the corner. My apologies.
On Wed, Aug 15, 2012 at 2:00 PM, Jason Sheldon <[email protected]>wrote: > Hi All, > > I have been developing a fluid-structure interaction (FSI) solver using > deali.II (somewhat similar to example step-46). I've run into an issue > with the application of some constraints when attempting a monolithic > coupling strategy. > > > First a little background information. FSI requires the coupling of > different physics models to predict their interaction. There are two main > ways to handle this, with a partitioned scheme or a monolithic scheme. > > In a partitioned scheme, the governing equations for each physics model > are solved iteratively until convergence is reached. In a monolithic > scheme, the governing equations for each physics model are solved > simultaneously. > > I am using a formulation with three distinct objects, the solid, the > fluid, and the mesh itself. > My governing equations are written in terms of the solid displacement and > velocity, the fluid velocity and pressure, and the mesh displacement. > > On the interface between the solid and fluid, there are two strongly > enforced constraints: > > - The solid velocity and fluid velocity are equal on the interface > (no-slip condition) > - The solid displacement and mesh displacement are equal on the > interface > > For a partitioned scheme, implementing these is a simple matter of calling > "set_inhomogeneity" for the constraint matrix, using the known value from > the last iteration. However, in a monolithic scheme, as everything is > solved simultaneously, these values cannot be simply set. Instead, the > dofs must be linked across the interface to have the same value using > "add_entry" for the constraint matrix. This is very similar to how the > periodic boundary conditions are handled in the step-45 example problem. > > I have been running into an issue that *these constraints do not work on > corners. * The errors I am getting say that entries are missing from the > matrix I am trying to distribute to. Although I believe I have set up the > constraints and flux sparsity pattern properly. I've attached a basic code > with similar structure to my FSI code which reproduces this problem. > > The domain being solved on in this example is a 2x2 grid where the lower > left hand square is defined as the solid region, and the rest is the > fluid/mesh region. In the attached image this is labelled "Domain 1" > > If you run the attached code it will crash when attempting to > distribute_local_to_global in the upper right cell. > > The domain in the image labelled "Domain 2" does not crash the code. To > see this, simply change the "0.5" in line 70 to "1.0" and rerun the case. > > If you place the attached code in your examples directory in deal.ii it > should compile and run. > > Any assistance with this issue would be greatly appreciated. > -- > -Jason Sheldon > -- -Jason Sheldon
_______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
