On Sat, Jun 29, 2013 at 12:13 PM, subramanya gautam
<[email protected]> wrote:
> Hi Derek..
> I had it in my code as an mod.. I will send the code in a while
>
> Sent from my Windows Phone
> ________________________________
> From: Derek Gaston
> Sent: 6/29/2013 11:56 AM
> To: Subramanya Gautam Sadasiva
> Cc: libmesh-users; libmesh-devel
> Subject: Re: [Libmesh-users] assign subdomain ids in abaqus IO
>
> On Fri, Jun 28, 2013 at 11:55 PM, Subramanya Gautam Sadasiva
> <[email protected]> wrote:
>>
>> 1. ABAQUS elsets are not unique.. elements can belong to multiple elsets.
>> However as far as I can see,
>> libmesh elements can belong to only one subdomain.. being able to be part
>> of multiple subdomains would be very useful though..
>
>
> This is not an ABAQUS thing... most of our mesh formats (both input and
> output) require this (elements only being in one subdomain)... in particular
> ExodusII. This assumption is made in a LOT of places through the code and
> wouldn't be easy to change.
>
> If anything - I would say that what you need is maybe something closer to
> BoundaryInfo for element sets. Leave subdomain_id alone... and use a richer
> data structure to allow you have arbitrary groupings of elements. This is
> actually something we already do in our code... but we are reading our own
> proprietary formats for those element groupings and keeping track of that
> stuff ourselves.
>
>>
>> 2. MeshBase only allows a unique subdomain name - subdomain id mapping..
>> and giving subdomains ids depending on both elset name and element type
>> makes this very hard to use the subdomain_name() function. I do not
>> understand the original motivation for the implementation and I think a
>> simpler scheme should not pose too many problems.
>
>
> Why do you want your elements set names to have anything to do with element
>type? This really shouldn't be necessary with libMesh.
The current version of AbaqusIO in libmesh assigns subdomain ids based
element types.
>
>>
>> I have implemented code to enable the reader to handle generated element
>> and node sets (start, end , step) as well as surfaces that are defined
>> through element sets.
>
>
> Did you try to attach the code - or is the code somewhere we can take a look
> at it?
I have attached the code and also add an example that shows the
working of the modified Abaqus IO
The main differences are as follows.
1. Generated element and node sets.
Abaqus CAE can occasionally create node sets using a generate
command .. this is done by providing the start end and step of the
element numbers.
2. Internal element sets. ABAQUS CAE creates internal element sets to
construct surfaces or sidesets. These element sets should not be
involved in assigning sidesets.
3. Surfaces can be created using element sets. this is now supported
along with the one that is currently implemented in AbaqusIO
4. The abaqus element set name is set as the name of the subdomain.
5. The abaqus surface name is set as the name for the sideset.
6. No changes have been made to the reading of the nodesets. The
convert_nodeset_to_sideset thing is not making too much sense to me,
and I don't really need it so I am letting it be for now.
>
> Derek
------------------------------------------------------------------------------
This SF.net email is sponsored by Windows:
Build for Windows Store.
http://p.sf.net/sfu/windows-dev2dev
_______________________________________________
Libmesh-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users
