feat: Add elastic fragile tension constitutive law for concrete

[Pier53]

Hello everyone,

Please be lenient, as this is my first real public pull request. I hope I haven't made a mistake! :-)

I've started to implement ElasticFragileTension.

• Add a basic class ElasticFragileTension(ConstitutiveLaw)
• .get_ultimate_strain() (chapter bellow)
• Add marin() method
• Add marin_tangent() method
• Add tests
• Other ?

.get_ultimate_strain()

I had a trouble with .get_ultimate_strain() with the Parallel class because compression law returned epsillon_tension_max = 100 instead of 0.00.

compression = ParabolaRectangle(fc=45)      # ou Sargin, Popovics...
tension = ElasticFragileTension(fct=2.7, Ec=36000)
compression_tension = Parallel(constitutive_laws=[compression, tension])

compression_tension.get_ultimate_strain()

return :
[
-0.0035,
100
]

So the ultimate strains of the compression_tension made with Parallel class was wrong.

__	compression	tension	compression_tension
classe	ParabolaRectangle	ElasticFragileTension	Parallel
negative	-3.500 ‰	0.000 ‰	-3.500 ‰
positive	100000.000 ‰	0.075 ‰	100000.000 ‰

Maybe you've already talk about it in past ?
I've thinked of different solutions.

1. Replacing by None --> Wrong because we are going to insert None instead of float, and it's not clear to understand what it's mean to have a None Ultimate Strain...
2. Replacing by 0.00 --> Wrong idea (break constitutive laws when they are alone)
3. Replacing 100.0 by np.inf (or -np.inf)
4. Replacing by NaN --> It's like np.inf

So I've made a try with np.inf !

#--- Compression law ---
def get_ultimate_strain(self, **kwargs) -> t.Tuple[float, float]:
    """Return the ultimate strain (negative and positive).
    """
    del kwargs
    return (self._eps_cu, math.inf)


#--- Parallel law ---
def get_ultimate_strain(self, yielding: bool = False) -> t.Tuple[float, float]:
    """Return the ultimate strain (negative and positive)."""
    neg = [law.get_ultimate_strain(yielding=yielding)[0] for law in self._wrapped_laws]
    pos = [law.get_ultimate_strain(yielding=yielding)[1] for law in self._wrapped_laws]

    finite_neg = [v for v in neg if np.isfinite(v)]
    finite_pos = [v for v in pos if np.isfinite(v)]

    eps_n = min(finite_neg) if finite_neg else -np.inf
    eps_p = max(finite_pos) if finite_pos else np.inf
    return (eps_n, eps_p)

With this, i got good results :

__	compression	tension	compression_tension
classe	ParabolaRectangle	ElasticFragileTension	Parallel
negative	-3.500 ‰	-inf ‰	-3.500 ‰
positive	inf ‰	0.075 ‰	0.075 ‰

But this was introducing some trouble in BeamSectionCalculator so i introduce a "bracket" with a large strain of 1.0.

My test file : 2026_09_18_bending_strength_concrete_only.py

Marin

I add a minimum for eps to avoid to get an np.inf value in the discretize_law

eps_min, eps_max = self.get_ultimate_strain()
eps_min = max(eps_min, -1)
eps_max = min(eps_max, 1)

-> Do i need to add a specific marin method for ElasticFragileTension ?

Tests

All test passed.

Roadmap

• Waiting your review and applying it
• Add tests
• Next ?

I reallyenjoy doing this, but I'm not used to work on project on Github, so I need your advises ! 👍

[mortenengen]

Thanks for this contribution, @Pier53! I am sorry that you had to run into that issue with the strain limits. Indeed, I remember @talledodiego and I discussed what to do with the limits, and we decided to keep it like this awaiting a tensile law 😃 Thanks for elaborating on this in the message above.

Anyway, I will have a closer look after the weekend, and @talledodiego will have a look at your solution for the strain limit issue.

If possible, please go ahead implementing the __marin__ and the __marin_tangent__ methods for completeness.

[mortenengen]

Thanks for this nice contribution, @Pier53! I think the use of np.inf is a clever solution. I only have one comment which I would appreciate if you could fix in order to trigger the pipelines. Final approval awaits the outcome of the pipelines, i.e. whether they pass or not, and the review from @talledodiego.

[Pier53]

I think I made a mistake with the get_ultimate_strain method. Because my result with and without rebar are inconsistent/wrong. Please wait a bit before the review because I'm going to fix it soon.
But I'm not totally available this week. Thanks

I'm also going to fix this noqa bug.