Continue to Site

Eng-Tips is the largest engineering community on the Internet

Intelligent Work Forums for Engineering Professionals

  • Congratulations IDS on being selected by the Eng-Tips community for having the most helpful posts in the forums last week. Way to Go!

Extremely slow solve time for linear elastic compression with pressure load

Status
Not open for further replies.

ereneel

Mechanical
Aug 16, 2018
3
I am running simulations on the pores of cortical bone. This involves complex geometries with extremely high element counts. However, the problem being solved is relatively simple using a single step with one iteration. I apply a pressure load with a total force of 1 N to the top surface. I assume a young's modulus of 17500 MPa and 0.3. For boundary conditions, I restrict the bottom surface in the z-direction and then select three nodes to restrict in all 3 directions to prevent sliding. If you want an idea of what my models look like:
boundary_conditions_swmpg2.png


Most of my models are about 5 to 7 million elements (quadratic tetrahedrons) that run just fine and finish in about 3 hrs. I set the job to run with 16 processors and 20 GPUGPU acceleration. I also used 90% of the memory on a computer that has 512 GB of RAM. I have a new model that is taking over 24 hours to run. The geometry is much larger than my previous models, but I manipulated my mesh settings to keep the mesh at 7 million elements. I usually get a warning message about 2500 distorted elements, but so far it has not been an issue. This larger model only gives a warning about 1500 distorted elements, so I don't think that is causing the increase in run time. This is the slow running model:
mesh_ucn4fn.png


Does anyone know what might be causing this new model to run so slowly? I was thinking of trying to keep the element size more uniform and go with a 10 million element mesh or testing a linear tetrahedron mesh to see if that will solve faster. One other difference I forgot to mention is that I only exported data for about 2 million elements in my original models and now export for about 5 million. I also discounted this as the cause because the monitor isn't even showing "Complete" which is when it starts writing the output.
 
Replies continue below

Recommended for you

Have you used a linear static pertubation step?

Also try the iterative solver.
 
Thanks for your reply. Would those help with this problem if it is completely linear? I'm pretty sure this is a linear elastic problem and there is no time stepping.
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor