Learning Forum

[Kiyean Kim]

Hello, team.

 

I am new to Ansys and APDL, so i am sorry if my question is trivial, and thank you in advance.

 

I am trying a multi-loadstep mechanical simulation, and I defined a contact condition.

In the 1st loadstep, I perforemed a simple deformation simulation and it works.

But the problem is in the 2nd loadstep, where I want the deformed contact to be fixed as it is.

(And i would like to roll it back to original contact condition in 3rd loadstep, after this problem is solved)

 

I tried using "contact step control" so that the contact condition is dead in the 2nd loadstep.

Meanwhile in the command snippet, I used the command below:

 

*GET,nmstp,active,,solu,ncmls

*IF,nmstp,eq,0,then

*ELSEIF,nmstp,eq,1,then

ALLSEL

DDELE,ALL,ALL

ALLSEL

D,moving_parts,UX,%_FIX%,,,,UY,UZ

D,pdlinnersurf,UX,%_FIX%,,,,UY,UZ

--------------------------

The named selections "moving_parts" and "pdlinnersurf" share the contact areas. (one contains the contact surf and the other contains target surf)

 

The solution results in an error, which the msg is:

 

 ERROR: Worker process(es) with rank(s) 12, 3, 9 have encountered a FATAL error.

 The information below was gathered from the file*.out output file(s).

 Please review the worker process output file(s) listed below for more

 details on this error.

 

 

 Worker Process Output File .\file12.out:

 

 *** FATAL ***                           CP =    1543.031   TIME= 15:17:22

 The displacement boundary condition on node 12345 is modified on        

 multiple domains.         

----------------------------------

 

When I looked into the nodes, they seem to be in the contact areas.

So I think the "dead" contact condition somehow still has some constraint in the region, which could not be overriden with displacement constraints.

I assume that this is because "contact step control" is based on "elements", rather than "nodes".

 

For more information, step 2 was solved appropriately when solved with a single-core processor. The error occurs when using parallel processing.

However, with geometry of slightly different parameter, it works.. It drives me crazy and I have no clue.

 

 

I would appreciate any suggestions or clues.

 

Thank you,

Kiyean

 

[Kiyean Kim]

.

[wrbulat]

Hello Kiyean,

 ==> The displacement boundary condition on node 12345 is modified on

 multiple domains.   

 

Yes, this seems to be a message resulting from domain decomposition used by the distributed solver.  You already discovered that switching to the shared memory parallel (SMP) solver (which doesn't decompose the model when it solves) works fine. It's usually a little slower and doesn't scale well with number of processors selected, but if it doesn't result in excessively long turn-around, I'd just continue using it.

 

I don't know the internal logic used to decide where/how to split the mesh for domain decomposition (~random, I  suspect) and I don't know of anyway to control that. I can see how you might hit the error for some models and not others.

 

I couldn't find record of the error you reported after a cursory search in our records... you may be the first user to report it to us.

 

Best,

Bill