i mean paper and implementation in MFront may not exactly the same for some or several input parameters.
unfortunately, you did not post for just simple deflection result, no model sketch problem, no simple input parameter. Maybe someone else can give an insight about the problem.
p.s for specific to TFEL/MFront materials i suggested to their official forum, previously at SourceForge is lost so i can trace ‘k’ and ‘beta’ related parameters, please ask them to recover.
Yesterday I finished some testing on Mazars which I post. MFront_2.0_v2.8.rar.
Overall response agrees with MAZARS but not completely as I expected.
Seems like the available model for Calculix is the Original Mazars (1980) and not the “Revisited” from HAMON François (2013).
I say that because the Damage initiation surface in the bi-compression area has reveal as completely flat.
That corresponds to the original Mazars behavior where K was not yet introduced. I have change K value (0.7,0.5,0.3) without any difference in the Damage initiation. ¿?¿?
Maybe it is not correctly read from user input. ?¿?¿.
i’m not sure, only try to read from Mazars codes of MFront, it seems revisited ones. Some parameter has been fixed internally, probably related to shear in damage variable and lead to flat like. Posting at official MFront forums may help and clarify.
I have managed to get the first signs of cracking in a MAZARS rectangular concrete beam (D=0.3). They appear equi-spaced in a pattern similar to those observed experimentally. It seems that the most damaged spots are not the ones closest to the center too. What I do not understand is why they do not extend. It is a very shallow demage. To go to higher damage values I would have to go to dynamic explicit but I have a problem.
The only compression support doesn’t work in dynamic explicit. ¿is it me or ccx limitation?.
Without it, I don’t think supported beams can be modeled with this type of material. At the minimum tensile stress in the support, the demage reduces the stiffness there and the convergence fails. It’s the area sustaining the model.
as i mentioned before, damage material is much sensitive. I have small experiences by several tests: all factors will significantly affect the result i.e Mazars material parameter, loading scheme, support boundary conditions, element type, mesh size, solver incrementataion, etc.
previously, the author of MFront suggest me to do compares and use another concrete damage material. I checked again Mazars and FichantLaborderie materials, both have similar in CalculiX solver behavior and results FichantLaborderie seems more stable, it can reach similar result for case of boundary conditions type of displacement and force loads. Material parameters are the most important things, convergence is hard to achieved but response by stress/strain history given reasonable result. Damage index or scale still shown below 1.0 (around 0.7 in my case, probably due to mesh size), however equivalent plastic strain shown it has been reached the limits and meaning separation occurs.
edited
switch from linear element (C3D8) to quadratic type (C3D20) achieved convergence and highest damage scale
here i do compare table 1 you cited, some is missing, another i’m not sure.
