ASDConcrete3D AutoRegularization with user-defined Mander-based curves

[n.dgn]

Hello,

I am using ASDConcrete3D with user-defined backbone curves generated from a Mander stress-strain model.

My compression envelope (Ce/Cs) is obtained from a Mander curve, while the tension envelope (Te/Ts) is defined from the cracking strength and includes a residual stress plateau for numerical stability. Damage curves (Cd/Td) are also user-defined.

The stress-strain curves are not derived from experimental fracture-energy data. They are generated analytically from the Mander model and then extended with residual branches to avoid convergence issues after severe damage.

My question concerns the -autoRegularization option.

From the documentation and from Dr. Petracca’s explanations, I understand that when user-defined stress-strain curves are provided, the reference characteristic length should correspond to the specimen length from which the stress-strain data were obtained.

However, in my case there is no physical specimen. The curves are generated numerically from a Mander model.

Therefore:
1. How should the reference length be selected when using Mander-based user-defined curves?
2. Should the same reference length be used for all meshes?
3. Is it acceptable to use AutoRegularization in this situation, or is it more appropriate to disable it and accept some mesh sensitivity?
4. What is the recommended workflow for ASDConcrete3D when the backbone is Mander-based rather than fracture-energy-based?

I performed analyses with element sizes of 0.50 m, 0.25 m and 0.125 m and observed noticeable differences in peak strains, both with and without AutoRegularization. I am trying to understand what should be considered the correct approach.

Thank you.

[STKO Team]

Dear user, the topic of mesh sensitivity with strain-softening materials is quite difficult to describe in a short post (we're planning a new webinar dedicated to this topic)
I'll try to answer all your questions:

• 1. How should the reference length be selected when using Mander-based user-defined curves?

  If no specific reference to a test campaign is given, use the size of a typical concrete specimen, more or less 150 mm.

• 2. Should the same reference length be used for all meshes?

  Yes, the reference length DOES NOT depend on the mesh, it's a physical quantity. The material model then will regularize the input backbone curves using the reference length (given by the user) and the element length (automatically obtain by the OpenSees FEM model).

• Is it acceptable to use AutoRegularization in this situation, or is it more appropriate to disable it and accept some mesh sensitivity?

  You should always use regularization

• I performed analyses with element sizes of 0.50 m, 0.25 m and 0.125 m and observed noticeable differences in peak strains, both with and without AutoRegularization. I am trying to understand what should be considered the correct approach.

  Using regularization will not necessarily give the same result for all meshes (only in some simple cases) but will converge to a unique solution with mesh refinement, while without regularization, you will have a more brittle material as the mesh size decreases