PROBLEMS WITH UMAT, negative eigenvalues etc.

[ramona]

Hey,
I'm a student of mechanical engineering and I'm working on my diploma with Abaqus UMAT. I got two proplems while running a job with my umat. The umat is very long and complicated, the constitutiv equiation is about biological material.However it runs with another, university (selfmade) FE-programm!
First problem: There is sometime- but not everytime I run the same job- an error code "illegal memory reference" ( signal 11) and I absolutly don't know what that means...
Second problem is, that if the job runs, datacheck completes sucsessfully, but the analysis after that is abortet because of negative "eigenvalues" of the system matrix...
Is anybody used to this problems and may give an advice to me?
Thank you very much!!!!!
Ramona

[Jorgen]

What you have noticed is that it is never easy to write and debug a UMAT subroutine. I have never seen an "illegal memory reference" from within a UMAT subroutine. I assume that you can get that error, for example, if you create an array of numbers, and then write numbers to the array PAST THE DECLARED SIZE OF THE ARRAY. Although, off course, I am not sure if this is what caused your problem.

I have seen the negative eigenvalue problem that you described. It is actually a common warning/error in FE simulations that are not unconditionally stable. I recommend that you check two things: first, make sure that you predicted stress/strain response is stable. Secondly, make sure that you calculate a reasonable Jacobian. If you return an incorrect Jacobian from your UMAT, then you can "fool" ABAQUS to think the solution is unstable.

Best of luck,
Jorgen

[mds_polymer]

hi
I just want to notice that "negative eigenvalue" some times comes from bad situation of boundary conditions
I had the same problem and when I changed the boundary conditions this problem solved :D