Monday, March 2, 2009

Ode to Abaqus

The past 18 hours have seen me work for perhaps seven or eight hours trying to select the boundary conditions for a finite element simulation in Abaqus because the server is running slow. When I finally selected everything last night i went to bed. This morning started with a three hour input file editfest trying to get everything to converge. Now on the sixth simulation of the day I anxiously wait for it to finish so that I can see the pretty pictures of mass diffusion concentration changes.

Simulation six of today: failure. The minimum time increment wasn't small enough.

The best part of these high tech computer programs are when the problem is a simple error message like: TIME INCREMENT REQUIRED IS LESS THAN THE MINIMUM SPECIFIED. Finite element has to be one of the coolest tools available to engineers today. However, understanding coding and boundary conditions and convergence is no simple task, especially as the problems become more complex.

Simulation seven of today: failure. Same reason.

In Analytical Methods class the other day we learned about finite element theory and it was interesting how stability criteria fix the time step based on the size of the element. It's fantastic how computers automatically change the time increments or concentration changes based on convergence and material properties because if I had to do one of these 40,000 node problems by hand I would never finish yet the computer let's me know I have a problem in only a few minutes.

Simulation eight of today: failure. Same reason. This time I'm changing the maximum allowable concentration change per element instead of the time increment.

1% inspiration, 99% perspiration - Edison. I have the physical results here in my hand that I am supposed to simulate so I can tell you what we should see but actually making it simulate that is a whole other sport. 

FE tip of the day: In Abaqus Check the .dat file first for errors by scrolling to the bottom (my record is 13 fatal ones) they should be either specified throughout the file or will be listed on the .msg file. Second, check the .msg file for errors by first scrolling down to the bottom of the file and then if it is not listed at the bottom by carefully scrolling through the file. Since these files are often huge look for errors at changes in text. I mean there may be dozens of pages of warnings followed by a change to time increment changes and in between there could be an error about elements unusable for the time increment changes or maximum concentration changes. 

Simulation nine of today: failure. Too many attempts at normalized concentration change. I changed the concentration change too much.

No comments:

Post a Comment