Convergence of Anisotropic Hyperelastic Abaqus-Subroutine

Hello everybody,

i try to re-implement the anisotropic hyperelastic material of Holzapfel-Gasser-Ogden form already included in Abaqus (cf. Analysis User's Manual 22.5.3).

After successfully debugging the subroutine (Uanisohyper_Inv), i had some serious convergence problems, anyhow. My one-element test modelsdidn't even deform a bit.

The material uses the Heaviside Step-function to allow fiber-reinforcement only in cases of tension. That causes inconsistencies in the strain enery potential and all of its derivatives. To avoid convergence problems, i use a monotonicly increasing step-function composed of hyperbolic tangent (graph attached). 

Now my test-elements deform a bit more, but only until one tenth of the cubes side lengt or so.

I double checked all derivatives by hand and in Mathematica, debugged successfully with Visual Studio and everything seems to work correctly.

Anyone here, with an idea, what could still go wrong?

How would the inconsistencies be solved inside the built-in Abaqus material?

Is there a way to link the object file created within Abaqus to a self-written driver, to check the output of the Subroutine?

I already have the driver, it works great to check the routine without having to start Abaqus. All the variables passed in and out have the right values, but i would like to also check that for the object file created by Abaqus.

Thanks in advance for every answer!