working with UMAT for non-linear elastic behavior in composites

[ataur]

Hi Jorgen n all,

I have to start to work with UMAT for my next step in my research.I have to work on user define material for cohesive zone modeling in composites using UMAT with traction-separation data.As all of us know that ABAQUS supports only for linear elastic behavior using trction separation data for cohesive zone modeling in composites which is very rare in practical approach of composites.Now, I have to write a code in UMAT to define non-linear elastic behavior of cohezive zone element in my model.Can anyone suggest me how to start working with UMAT or can u guys suggest me any examples available related to my work??

Thanks in advance.

[Jorgen]

Interesting project. Unfortunately I do not know of any relevant examples that you could study...

One thing, though, depending on what types of simulations you want to run you might want to consider writing a VUMAT instead of a UMAT. A VUMAT is easier to write and allows you to include element failure and progressive damage failure.

Best of luck,
Jorgen

[ataur]

Hi Jorgen n all,

Thanks alott for the speedy reply Jorgen.Well, I will be running a simple static analysis for my model.I have the UMAT material(ABAQUS) to learn how to write a code for the non-linear elastic behavior of the cohesive zone element in the composite.I have to write a code only for implementing material properties of the cohesive zone in the composite using the non-linear traction separation data.So, can I use UMAT for implementing non-linear elastic behavior of the cohesive zone element in the composite using the relevant traction-separation data available?


Thanks in advance.

Ataur.

[Jorgen]

Just so I understand, when you say "cohesive zone element" are you referring to a cohesive zone in a physical sense, or cohesive elements in a finite element sense?

If you know the Young's modulus as a function of strain then you can use that info to create a UMAT.

- Jorgen

[ataur]

Hi jorgen,

In my case I am implementing a cohesive element between the fiber and matrix assuming that it behaves as cohesive zone between fiber and matrix in the composite(thickness of 1micrometer-cohesive element).I have the properties of fiber and matrix but the material property of cohesive element is available in terms of non-linear traction-separation data which is determined from a simple pull-out test.Traction-separation data is nothing but when 'force(stress)' is applied on a point in the domain of CZE, the point starts moving from the original position and that displacement of the point from the static position is termed as 'separation'.I dont have any information relating to young's modulus versus strain for the cohesive element.

Thanks in advance.

Ataur

[Jorgen]

Recent versions of Abaqus has something they call "cohesive elements". From your description it sounds like these elements are perfect for what you are looking for. Have you looked into using this feature? It might not be necessary for you to write a UMAT or VUMAT.

- Jorgen

[ataur]

Hi Jorgen n all,

Thanks for ur reply Jorgen!!

I have a paper published in which the concept of cohesive element has been used.But, as I have told you that ABAQUS is limited only for linear traction separation data in which we can assign traction separation data as the slope of the linear curve in the terms of Knn,Kss,Ktt.Now, in my case I need to implement non-linear traction separation data as the mechanical behavior of the cohesive element which cannot be implemented directly in ABAQUS as you have mentioned. So, I thought I should go for UMAT in order to assign the non-linear mechanical behavior of the cohesive zone which is given as non-linear traction separation curve.In simple words, I have to implement a non-linear curve of traction-separation data as the mechanical behavior of the cohesive element in the composite in ABAQUS. I hope I have given you a brief scenario of my job!!!!

[ataur]

Hi Jorgen n all,

Thanks for ur reply Jorgen!!

I have a paper published in which the concept of cohesive element has been used.But, as I have told you that ABAQUS is limited only for linear traction separation data in which we can assign traction separation data as the slope of the linear curve in the terms of Knn,Kss,Ktt.Now, in my case I need to implement non-linear traction separation data as the mechanical behavior of the cohesive element which cannot be implemented directly in ABAQUS as you have mentioned. So, I thought I should go for UMAT in order to assign the non-linear mechanical behavior of the cohesive zone which is given as non-linear traction separation curve.In simple words, I have to implement a non-linear curve of traction-separation data as the mechanical behavior of the cohesive element in the composite in ABAQUS. I hope I have given you a brief scenario of my job!!!!

As you have told me if I have young's modulus as a function of starin, but I have a similar data ie., I have traction(stress) as a function of separation(strain) which is given as a non-linear curve.Can I use this curve as my input to implement mechanical behavior of cohesive zone in my composite?

[Tompa]

Hi Ataur
I have implemented material models for cohesive elements in ABAQUS, and it works very well. You can either do it with a continuum approach (stress-strain) or with a traction-displacement approach. Which way you choose depends on you material, if the cohesive zone in your material has a neglible thickness you can choose the traction-displacement approach otherwice you can choose the continuum approach.

As always when you make a UMAT it is important to get the Consistent Jacobian (ddsdde) right. I suggest you start up with the formulation and decide how you will do the stress(or traction)-update and how to calculate the Consistent Jacobian (also called, Algoritmic Tangential Stiffness, ATS etc.). Afterwards I suggest you make yourself a simplified FE-program where you can implemented the material model (either in a fortran-environment or MatLab, or Octave, or similar). This will probably save you a lot of time since it can be time-consuming to debug and test convergence rates etc. directly in ABAQUS. (I have seen scarying examples of projects were guys have spent years trying to get UMAT's running without any progress, when implementing them directly in ABAQUS ).
Once you have a material model running it usually takes pretty little effort to change is so that it works in ABAQUS.

The latest UMAT for cohesive elements I developed for non-linear geometries included both softening, an advanced non-associate flow rule, damage (degeneration of elastic stiffness) etc and it worked surpricingly well with good convergence rate. So I think ABAQUS has made an good job with their cohesive elements. (It's just a shame their material models are soo poor.)

Good luck with the development

[ataur]

Hi Tompa n all,

Thanks alott for ur reply.:)

I have also implemented material models for cohesive zone elements in ABAQUS assuming linear elastic traction-displacement approach as my cohesive zone material has a neglible thickness(1microm). But in real case the cohesive zone for my material is not linear elastic but it is non-linear elastic. As you know that, ABAQUS is limited for linear elastic cohesive zone and so I am looking for UMAT to implement non-linear elastic property in cohesive zone for my material using traction-separation response.It is really very tough to write an UMAT as I am new to UMAT. My question is that if it is sure that I should imlement non-linear elasticity for my material property in the cohesive zone using UMAT only or I have other options too???. I hope I have given you a clear picture for my case. And aslo I appreciate if you could let me know if your UMAT that you have implemented is for non-linear material property using continuum approach (stress-strain) or with a traction-displacement approach???:confused:

Thanks alott in advance!!!!:)

[ataur]

Hi Tompa n all,

Thanks alott for ur reply.:)

I have also implemented material models for cohesive zone elements in ABAQUS assuming linear elastic traction-displacement approach as my cohesive zone material has a neglible thickness(1microm). But in real case the cohesive zone for my material is not linear elastic but it is non-linear elastic. As you know that, ABAQUS is limited for linear elastic cohesive zone and so I am looking for UMAT to implement non-linear elastic property in cohesive zone for my material using traction-separation response.It is really very tough to write an UMAT as I am new to UMAT. My question is that if it is sure that I should imlement non-linear elasticity for my material property in the cohesive zone using UMAT only or I have other options too???. I hope I have given you a clear picture for my case. And aslo I appreciate if you could let me know if your UMAT that you have implemented is for non-linear material property using continuum approach (stress-strain) or with a traction-displacement approach???:confused:

Thanks alott in advance!!!!:)[/QUOTE]

[Tompa]

I have implemented non-linear UMATs for cohesive elements both with continuum approach and with traction-displacement approach, both ways works fine (altough I most say that I prefer the continuum approach since the traction-displacement approach can give you slightly unfamiliar measures when you use non-linear geometry, nlgeom=on).

As you mentioned ABAQUS only provide linear elasticity for the traction-displacement approach. And if this isn't good enought for your simulations I see no other way then to implement your own UMAT (I have no experience with UHYPER, but I wouldn't think it is an option here).

The models I have implemented are elasto-plastic, so a purely elastic model should (hopefully) be easier to implement.

If you can provide some more info regarding your material and applications, then maybe I can help you out. For example tension-tests, shear-tests, dilation of the material during shear (if any), some explanation of possible applications-nlgeom=on/off, you mentioned the thickness of the cohesive region but in what scale is the surronding material.

If you, for some reason, don't want to publish everything here you can email me through this site.

Take care

[Mike]

Hi All:

I am coding a VUMAT for 3D cohesive element COH3D8 with traction-separation lows in ABAQUS. I have the following questions and hope to get some quick answers from the experts here.

1. How to extract the traction(stress) and separation (displacement of strain if constitutive thickness 1 is used ) corresponding to Mode I, Mode II and Mode III? In UMAT/VUMAT for 3D elements, there are 6 stress/strain components, but COH3D8 requires only 3 traction/Displacements?

2. If the default thickness of 1 is used, the strains can be used as relative-displacements in the constitutive law, is this correct?

Thank you all and I wait anxiously for your reply,

Mike



Hi Ataur
I have implemented material models for cohesive elements in ABAQUS, and it works very well. You can either do it with a continuum approach (stress-strain) or with a traction-displacement approach. Which way you choose depends on you material, if the cohesive zone in your material has a neglible thickness you can choose the traction-displacement approach otherwice you can choose the continuum approach.

As always when you make a UMAT it is important to get the Consistent Jacobian (ddsdde) right. I suggest you start up with the formulation and decide how you will do the stress(or traction)-update and how to calculate the Consistent Jacobian (also called, Algoritmic Tangential Stiffness, ATS etc.). Afterwards I suggest you make yourself a simplified FE-program where you can implemented the material model (either in a fortran-environment or MatLab, or Octave, or similar). This will probably save you a lot of time since it can be time-consuming to debug and test convergence rates etc. directly in ABAQUS. (I have seen scarying examples of projects were guys have spent years trying to get UMAT's running without any progress, when implementing them directly in ABAQUS ).
Once you have a material model running it usually takes pretty little effort to change is so that it works in ABAQUS.

The latest UMAT for cohesive elements I developed for non-linear geometries included both softening, an advanced non-associate flow rule, damage (degeneration of elastic stiffness) etc and it worked surpricingly well with good convergence rate. So I think ABAQUS has made an good job with their cohesive elements. (It's just a shame their material models are soo poor.)

Good luck with the development