UMATI and UMATL - "STRANGE ISSUE" with history variables

[ikoutrom]

Hello,

I am creating a user-defined material model (user,2) which needs to store 10 history variables. I set n1=10 in UMATI2, then wrote the main code in UMATL2.
I presumed that n1 is the size of hn() and the size of h1() (i.e., for n1=10, hn and h1 are assigned 10 components each).

While running some simple analysis, I noticed that one of the history parameters spuriously changes during computation (meaning that, while it is stored with the proper value in the correct location of h1, when it is called in subsequent steps of the computation by using the appropriate component of hn its value has become zero).

More strangely, by changing the value of n1 to 20, the computation proceeds normally and I do not see any problem with the history variables.

I would be grateful if you could provide me with any ideas on what may be causing the issue. I am running a single-element analysis (plane stress conditions)
and the only user-defined component is the material model (I am not using user-defined elements which could be spuriously accessing and changing history terms)...

[FEAP_Admin]

The proper sequences it to set n1 in umatiXX.f.

Then in umatlXX.f, hn( ) will always contain the values from the previously converged global time step, t_n.
Your umatlXX.f should always compute what it thinks the correct value is for the current time step, t_n+1,
and store it in h1( ).  The values from h1( ) are copied over to hn( ) once the global problem has converged
and you have moved on to the next time step using the TIME command.

If you are not seeing this behavior please elaborate more and if possible post a simple test case.

[Prof. R.L. Taylor]

Can you check if running the problem as a plane strain one the problem goes away when you set n1 = 10.

For plane stress there are additional parameters used and we may have a bug related to user models.

[ikoutrom]

Professor Taylor,

Running the problem as plane strain when I set n1 = 10 does indeed complete the analysis WITHOUT any error.

Yannis

[Prof. R.L. Taylor]

FEAP asks for an extra history variable for plane stress to store the thickness strain that makes the stress zero.  Thus, if your problem is plane stress you need to increase the n1 value to store this.  We will make corrections to have future releases do this automatically.