Good afternoon

In UMAT, you typically have strain and stress as your main parameters in addition to "props" for, but in PPR or XN model, you define delta_n and delta_t based on your displacement vector. Meanwhile, it is better use "zero thickness" element which give you zero Jacobian if implemented in ordinary elements and that is a source of error.

The other reason is interface elements work better with Lubatto integration in contrast with traditional Gauss method and that is an issue that you can address only in UEL.

Of course with some considerations, you can use UMAT for these kind of constitutive models, but you should be aware of what you are doing, especially in the case of considering path dependency in these type of potential based models.

Good luck

Hi Reza,

Thanks for your comment.

Regarding the delta_n and delta_t, don't you think that we can also use cohesive element COH3D8 (or any Abaqus built-in cohesive element) and treat our defined TSL as a constitutive model in UMAT? So in that case, you will get info such as delat_n delta_t and also you will not have the zero-jacobian problem?

Regarding the integration, I think Abaqus built-in element formulations are already equipped with proper integration/quadrature scheme.

I mean, I don't really see a strong point that enforces us to implement cohesive element in UEL.

In fact, I just found CZM implemented in UMAT in thesis of Marco Musto, "A fractional rate-dependent cohesive-zone model".

Thanks

Hi,

I've got couple of questions about implementation of a constitutive law for cohesive elements using VUMAT. I thank if you answer them, my questions are as follows:

1) If we wan to implement the traction-separation law; is that correct to consider the following?

S33=tn; S13=ts; S23=tt

2) Please confirm whether it is correct to state that strain components are equal to displacements?

3) How to define the separation? I mean do I have to consider element deletion like for bulk elements?

4) If my assumptions in questions 1-3 are correct, I am wondering that there is no necessity to use cohesive elements, I am curious to know the importance of using cohesive elements in this case. 

Thanks for your help in advance. I am looking forward to hearing from you.

Regards,

Mehdi

Ditho Ardiansyah Pulungan

please let me know if you find some solution regarding your query.

Actually i am trying to implement a simple VUMAT traction separation code for built in cohesive element in abaqus by using the relation, separation=straininc*charlength. but while implementing i am getting the error as "zero or negative initial dilatational modulus caused by bad material data". i have changed my charlength=1 as it tends for constitutive thickness in traction separation approach. although the dilation error being removed, still i am not able to validate with with my GUI solution of abaqus.

Hello,

I am not sure if that answer your question, but perhaps you could get some insight in this paper:

Article Assessment of cohesive traction-separation relationships in ...

where they discuss about some inconsistencies that may arise from the built-in CZMs in Abaqus.

One way to deal with this issue, is to post-process the UEL results, even though it can be time consuming. One example on how to do this (quite simply) can be seen in the highlighted figure in:

Article Analysis of a castable refractory using the wedge splitting ...

that was done by one further analysis after convergence. Although I was using imposed displacements in other part of my geometry as boundary condition, I had as result the displacements for every node of the PPR cohesive elements. I then made a zip-like simulation, only with these displacements applied, to have the traction separation law for every time step, and every element. With such approach, I did not need to change the UEL to get any of the inside calculations, and was faster than asking for the needed variables.

I hope it helps.

Best regards,

Rafael