- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi, I have written an Abaqus (standard) umat for an elastoplastic model with explicit integration and automatic error control. Results with this user material matches exactly with the results published in literature.
The issue is that it takes very long time to complete the job even with 100 elements. I have checked continuum Jacobian at the end of all step increments which is found to be correct and the quadratic convergence is also achieved within 3-4 iterations.
Could anyone explain this? Does the way how I code my umat in fortran affects the Abaqus job computational time?
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
It may help if you can show your code (or some of the CPU intensive part of the code).
Jim Dempsey
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Do you have a separate "main" program that can run the umat subroutine by itself? I've found that is a useful way to test and debug an Abaqus user subroutine by itself. I would expect the umat routine to be called for each element integration point, so you could simulate that from the main program by calling the umat routine within a loop, which might help investigate the slow performance.
Regards, Greg
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Can you use VTune for CPU profiling?
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Jim and Greg,
Thank you for your responses. Greg's advice worked for me and I could make my umat code computationally more efficient.
Regards,
Vikram
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page