XANSYS Message: 38958 [Go back to message list]
[bookmark on del.icio.us]
No rating yet
Subject: Re: Superelement query
Author: Sheldon Imaoka
Date: 2002-06-19 14:21:00Hi Raghavan,
Here are a few comments to your posting which may hopefully be of some
use:
> I have learnt now , perhaps the hard way , that ANSYS merges nodes
in the
> "use" phase or higher level "generation" phase solely on the basis
of node
> numbers irrespective of their global coordinates. This I consider a
little
> user-unfriendly.
Yes, you are essentially correct. When you read in a superelement
with the SE command, the nodes are defined, too. You can verify this
by defining MATRIX50 in a blank/new database, then read in the
superelement with SE - with NPLOT, you'll see that nodes are defined,
along with the superelement, even though you didn't explicitly define
those nodes before reading in the superelement. (Technically
speaking, the nodes are actually not 'merged' during the USE pass for
this reason.)
The superelement, like other 'regular' elements, are defined by nodes
for connection points. That is why the node numbering is important,
just like any other 'regular' ANSYS element.
While I understand what you mean by this not being user-friendly,
there are two points I'd like to bring up:
1) Node numbering is necessary for any element (superelement or
otherwise), so there is no easy way to circumvent this issue, I believe.
2) Top-down substructuring allows the user to forget about keeping
track of node numbers of different superelements, so I'd recommend
this approach in your case.
> The user has to take care to ensure that the master nodes that are
meant to
> be structurally merged from different superelements are all numbered
> identically.
Not entirely true - you can define node offsets with NUMSTR when
defining the superelements. Then, a user can use CP or CPINTF to
connect coincident nodes from different superelements.
> OTOH if the criterion for merging is global coordinates the user
will not be
> required to put efforts to meet the above. In fact merging on the
basis of
> global coordinates makes a lot of physical sense and promotes
realization of
> all the benefits that are associated with the use of this technique.
Yes, it may be easier for nodes defining connections, but in the event
that master nodes are not connected to any other element (such as
those which loads are applied on), then ANSYS must create its own node
numbering. As you can see, this is actually a more cumbersome
approach since the user would not have control of node numbering. (As
noted above, use top-down substructuring instead for this situation.)
> This I reckon to be the bottoms-up substructuring approach. I have no
> knowledge of what you refer to as top-down substructuring approach.
Top-down substructuring means that a user creates the entire model
first, then selects portions of the model to create superelements. In
this manner, node numbers are unique, and the user doesn't have to
worry about it at all. This technique does not lend itself to
superelement reuse (such as with SETRAN), but it makes preprocessing
*much* easier, IMHO. Please refer to the Advanced Analysis Techniques
Guide, Ch. 7.4. "Top-down Substructuring" for details on this
approach. I personally prefer top-down substructuring to bottom-up
substructuring because of this benefit of ease-of-use.
Best regards,
Sheldon
ANSYS, Inc.
http://www.ansys.com/ (Official Site)
http://ansys.net/ansys/ (Cheap imitation)
"This email is not being provided in my capacity as an
ANSYS employee. I am solely responsible for the content."