implementing two-crystal shape memory using UMAT

[NazBaba]

Hi all,

I am implementing the constitutive behaviour of two-crystal CuAlNi shape memory alloy during uniaxial loading using UMAT. My aim is to pass single crystal to polycrystal model. So, I am trying two-crystal problem firstly. I have no trouble with implementing single crystal and I get results of volume fraction and stress. However, in two-crystal problem, the solution doesn't converge. I use only elastic region and there is no plastic region in my problem. I transform only elastic matrices and transformation strains local to global in two-crystal program. (I calculate the consistent Jacobian by using the finite difference method.)





Thank you.
Nazım Babacan

[matsgd]

The description is a bit vague... But one thing you could do is to do the two crystals with identical orientations. That way you could at least see if the problem is due to many crystals (and errors in this implementation), or if it has to do with your presumed averaging of orientations...

Mats

[NazBaba]

Thank you for the reply. I try to mention my problem more detailed. Our model is 3-D. We use two-crystal. One crystal's local coordinate is same with the global coordinate. And the other crystal is rotated 10 degree around z axis. I implement transformation only in my fortran code (umat subroutine). Actually, I don't know the usage of *orinetation card in the input file exactly. I think it is not essential. In my subroutine, I use rotation tensor and I rotate our elastic stifness tensors and transformation strain tensor but I don't get results. After phase transformation starts, the solution does not converge. I tried single crystal which is 10 degree rotated crystal and I got results.

Nazım Babacan

[NazBaba]

In addition, I use C3D8 element in Abaqus with 8 integration points and I seperate elements with such as "if (noel.eq.1) ; rotation tensor= [1 0 0 ; 0 1 0 ; 0 0 1 ] for untransformed crystal. "

[matsgd]

If it does not converge, the simple answer would be that your Jacobian is incorrect, and it just happens, by luck, to work in some cases... How do you calculate your Jacobian?

Mats

[NazBaba]

Thank you very much Mats, I solved the problem. My Jacobian was not true. I thought about your suggestions again in another thread and I understood my mistake. Now, I think I get true Jacobian calculation. Thank you again.