So I want to write a UMAT in which the free energy function will have a fourth and possibly fifth invariant which has a fibre orientation tensor in it. The fibre orientation tensor for every gauss point is available from Moldflow and after some operations is available in a text file.
My free energy function could be anything like
U=C10(I1'-3)+1/D1(J-1)^2+C3(I4-1)^2+C4(I5)
where I1=trace(B')
I4=trace(B'A)
I5=trace((B'^2)(A)
where B'= Deviatoric left cauchy green tensor and A = Fibre orientation tensor which I obtain from Moldflow.
The main problem is how to input this tensor in the UMAT so that it is read on every gauss point when computations are made. Notice that A is not constant infact varies over several gauss points.
Dear A.Wasay Khan,
First of all you should keep in mind that the actual input of a UMAT is standard and it is defined by Abaqus. So the simple answer is that you cannot change the input. However, there are two ways to "go around" that. The first one, which I strongly suggest, is to use the STATEV array, meaning the array that contains all the user defined state variables to store the values of the tensors you are discussing, update them during each increment and use them as you wish for your calculations. The other way is to use external files that you are going to read and write to, but such a practise is computationally inefficient and much more complicated without offering any advantage.
So the general guidelines to perform your calculations is like follows:
1. you check if the current step and increment are the first ones, in that case you want to initialise the variables, for example I4 = 0.0_8, I5 = 0.0_8 etc.
2. then the rest of the calculations are performed.IMPORTANT: keep in mind that before exiting UMAT you have to update all the values of the STATEV array. When you update these values, STATEV from the last converged increment will be the input/initial value for the next increment.
Hope this general information helps you.
Regards,
Anargyros
Dear Anargyros A. Karakalas,
Many thanks for the answer. Maybe my understanding is very limited,but as far as I understood you mean I should be using SDVINI right ? because in my case inputting all the fibre orientations is a really tedious task. The total number of fibre orientations are 25873(each at gauss point) in number with each having six components.
Can you kindly elaborate more on the 2nd approach ? Maybe its really useful in my case.
Thankyou very much.
Kind regards,
Wasay
Dear A.Wasay Khan
You are welcome.
Okay, I see. Well you could use SDVINI, but I was thinking at first that you could define as STATEVs the orientations you would like. However, this was before knowing that you have that kind of many fibre orientations. So, assuming that the orientations are given as a six-component array and that they must be defined only at the initialisation of the analysis (for the rest they can be retrieved), I would do the following:
1. prepare a .txt file where each line would have all the required information for the assignment of the orientation, for example the coordinates of the point or some kind of identification and the six components that define it.
2. inside the UMAT subroutine add a read command that will open and read the .txt file and that will be called only during the first time increment (to reduce subsequent computational cost).
3. when UMAT is called, is called for each integration point. Thus, you could have six components reserved in STATEV array that will store the respective six components of fibre orientation for each integration point. Once read and stored they can just be recalled at any time increment without having to re-open and re-read the .txt file. This happens through just reading the respective STATEV component.
Check it out and let me know if you have any more questions.
Best regards,
Anargyros
Dear Anargyros A. Karakalas ,
Many thanks for the answer. Helps alot .I would try it out.This was what I was actually looking for.
Kind regards,
Wasay
- Badges
- Science topic