[MOAB-dev] pull-request Tqdcfr.cpp

Lukasz Kaczmarczyk Lukasz.Kaczmarczyk at glasgow.ac.uk
Thu Sep 26 15:42:05 CDT 2013


Thanks,

There is problem with tag naming convention which you like to have there. There is a little bit lack of consistency, problem partially created by me. At some point it need revision, I will adapt to what you chose is the best.

I was able extract almost all information form cubit, there are still some problems with material parameters, however we use block attributes instead, so now is not an issue for us.

There is a problem with cubit as moab. We noticed that moab is not identifying the entities in a sideset (and node sets) if that sideset is set using triangles rather than surfaces on Cubit. I have to check that but it was reported by one of my users.

For example this will work,
reset
brick x 1 y 1 z 1
volume all size auto factor 8
volume all scheme Tetmesh
mesh volume all
Sideset 1 surface 4
Sideset 2 surface 6

But this is not working,
reset
brick x 1 y 1 z 1
volume all size auto factor 8
volume all scheme Tetmesh
mesh volume all
Sideset 1 tri 67 74 86 80 70 68 87 88 79 84 76 77 75 82 83 73 85 78 71 81 72 69
Sideset 2 tri 120 125 115 127 114 111 113 128 122 124 132 126 119 129 130 116 117 112 121 123 118 131

In addition:

Mesh refinement is working ok in our code, code pass tests with random edge spliting, however implementation in some parts is far from being elegant. If you would like copy that part of code, pleas be free to do that. We use entity handles as uids, it should work automatically for parallel mesh refinement, if instead of entity handles you will use gids. 

We have added mesh-smoothing based on barrier quality measure as well. It works with mesh refinement, but is not tested properly. 

Kind regards,
Lukasz

On 26 Sep 2013, at 21:19, Iulian Grindeanu <iulian at mcs.anl.gov> wrote:

> Hi Lukasz,
> Thank you for submitting it.
> Tim said he will look over, sorry about the silence. He is on travel right now.
> 
> I compiled on my machine, and it passed all tests in moab; 
> 
> So I will approve it, but Tim will look over it.
> 
> Thanks,
> Iulian
> 
> Hello,
> 
> I've made pull request in bitbucket with small changes in src/io/Tqdcfr.cpp. I don;t know if someone that monitoring.
> 
> We have been working on boundary conditions form cubit, definition of almost all of them you can found in,
> Line 1568
> https://bitbucket.org/likask/mofem-joseph/src/cf172dd94bcfd8b972c0ab43a669b8256e9035aa/mofem_v0.0.1/common.hpp?at=master
> 
> Each meshset with cubit NodeSet, SideSet for which boundary conditions are specified (in cubit) has tag bc_data, which could be interpreted as in above file. You could consider if you like to add that to moab.  
> 
> For example, dirihlet bc in cubit is represented by structure,
> struct __attribute__ ((packed)) _data_{
>     char name[12]; // 12 characters for "Displacement"
>     char pre1; // Always zero
>     char pre2; // pre-processing flags for modification of displacement bcs. They should not affect analysis, i.e. safe to ignore; 1: smallest combine, 2: average, 3: largest combine, 4: overwrite or no combination defined (default)
>     char flag1; // Flag for X-Translation (0: N/A, 1: specified)
>     char flag2; // Flag for Y-Translation (0: N/A, 1: specified)
>     char flag3; // Flag for Z-Translation (0: N/A, 1: specified)
>     char flag4; // Flag for X-Rotation (0: N/A, 1: specified)
>     char flag5; // Flag for Y-Rotation (0: N/A, 1: specified)
>     char flag6; // Flag for Z-Rotation (0: N/A, 1: specified)
>     double value1; // Value for X-Translation
>     double value2; // Value for Y-Translation
>     double value3; // Value for Z-Translation
>     double value4; // Value for X-Rotation
>     double value5; // Value for Y-Rotation
>     double value6; // Value for Z-Rotation
>     };
>     
> 
> The recent changes are about BlockSet attributes, we use them for specifying material and model parameter.
> 
> All of that works quite well, we use that for mesh smoothing and solid mechanics.
> 
> Kind regards,
> Lukasz



More information about the moab-dev mailing list