I have a problem regarding an ABAQUS simulation with UMAT. 

At exactly every 1033rd (this number depends on the mesh size, for a mesh with approximately double the mesh size and half the elements, this is every 2073rd increment) increment, ABAQUS receives a total time, time(2) (or array(3) from posfil), of ~ 1e-320, which suggests underflow. The UMAT right before this URDFIL gets the correct total time, time(2), as it should be. All the other values are also correct at the end of the UMAT.

The problem now is, that a total time of basically 0 should not exist after the initial step. Unfortunately after such a corrupted increment ABAQUS deletes the fil file (I don't know why this happens) and makes a new one, so all the results are corrupted and unreadable beyond that point. My question is: What can I do to fix the error and what can cause the error of this time change from the correct value to 0 between UMAT and URDFIL?

What I already found out: It's not an error of not enough RAM and it's not an error of a specific input file, as this also happens with other input files (with the same fortran code of course).

I use UMAT, URDFIL and SDVINI for my simulation. The programs I use are ABAQUS 2025, Intel OneAPI 2025.3.0 Fortran Compiler and Visual Studio 2022.

Any help is appreciated, thank you in advance! 

Also important: The .odb file is never corrupted and the .msg file always shows the correct values.