Learning Forum

[Lovisa Svanborg]

Hi!

 

I have got this message and wonder what i can do about it? The model is quite simple so don't know why it isn't enoght memory.

 

**There is not enough memory for the Distributed Sparse Matrix Solver to 
 proceed using the out-of-core memory mode.  The total memory required  
 by all processes = 37827 MB.  The total physical memory that is        
 available on the system = 29418 MB.''

 

 

Regards

 

[mrife]

Hi @Lovisa Svanborg  The complexity/simplicity of the model is not necessarily an issue.  Mesh a unit cube fine enough and eventually it will need more RAM than is available.  

There are a lot of paths you can take to address this but I am assuming some you don't want to, or can't, take.  For example, how about installing more RAM on the system?  Or solve with multiple compute nodes.  

It would help to know what type of analysis this is and some details of the model.  Is this an assembly of solid parts (not beam and/or shells)?  If so try changing the Mesh default to use linear elements.  This will decrease the node count, and so reduce the memory requirement.  Then once solved you can post-process and determine where a finer mesh is needed to provide an accurate capturing of the strain/stress field (assuming some kind of structural analysis here).  Then go from there.

Mike

[Lovisa Svanborg]

Thank you. No I can't install more RAM. The geomentry is just a plate (circular). The mesh element order is linear and tetrahedros.

Lovisa

[mrife]

Hi @Lovisa Svanborg

Ok, then coarsen the mesh.  And why mesh with tets if this is a circular plate?  Are there small features?  Some other reason?

Mike

[Lovisa Svanborg]

I get a message saying that a recommendation could be to have tets. No other reason. No small features. I will course the mesh.

Lovisa

[Lovisa Svanborg]

Here is the message:

 

[mrife]

Hi Lovisa Svanborg

Just how thin is the plate compared to the diameter?  Perhaps a shell FEM would be appropriate.  Mike

[Lovisa Svanborg]

It is very thin. 2 mm compared to d=138 mm. How can I do that? Thanks

[mrife]

Hi, it depends on the CAD program you are using.  Usually there is some feature to extract a mid-plane surface from a thin body.  Or create a "surface body" from the face(s) of a solid body.  This chapter from the AIS course on analyzing thin bodies may be of help:

/courses/index.php/courses/geometry-representation-using-ansys-mechanical/lessons/analyzing-thin-structures-efficiently-lesson-2/

Mike

[Lovisa Svanborg]

Thanks! I Will take à look at it.

[Lovisa Svanborg]

I don't get the shell meshing to work. Can you recommend another mesh metod? (I had tethredrones)

Lovisa

[mrife]

Hi Lovisa Svanborg

What did not work?  

[Lovisa Svanborg]

I could't share the midsurface. It says ''There were no opbjects to share withinthe specified tolerances.''

[Lovisa Svanborg]

 

[mrife]

Ok maybe this is a dumb question, but I can't tell from the picture, but are those two circular areas?  Instead of a circular area and a ring?

[Lovisa Svanborg]

Not a dumb question. I have made a circular and then split the body to inner circular. This is because I want to fix the outer part in the model and have a pressure in the middle. Maybe I have done wrong. And Ii had problem first to split the body in a inner circular. Maybe the split is the problem when selecting midsurface?

[mrife]

Hi Lovisa

Possibly - if you select each surface and it changes to a different color does the selection make sense?  I.E. is the outer ring really a ring?

[Lovisa Svanborg]

I don't know If I have done the split right. How can I split a inner circle? Make a circle and select it to split. It just got a inner circle on one side and I want to split the hole body.