I am doing an FE simulation of a polycrystal ( uniaxial). I currently am using C3D8 elements to simulate the same. I would like to reduce the simulation time, Hence I am planning to use a C3D8R ( reduced integration) element.
Is this a good idea ?
I am running my Abaqus simulation along with a UMAT ( for crystal plasticity). There are many options along with C3D8R elements like the value of *hourglass stiffness, *section controls etc, Could you help me set this up ?- any suggestions or references would be appreciated.
I just need to obtain the equivalence of global stress- strain in both cases ( C3D8 and C3D8R) .
Thanks,
Prithivi
The whole purpose of developing "Reduced Integration Elements" was to increase computational efficiency without loosing the accuracy. But you should understand, "there is no free lunch in this world". Therefore, reduced integration elements come with their own cost. To simply answer your question, yes you can use it. But you've to pay attention to the caveat associated with it. To understand and to utilize it properly, you've to read theory books (ziknozwich is the god book), and abaqus analysis manual, carefully.
P.s. There's no pre-made recipe regarding how to handle hourglass effect, especially when it comes to umat
The whole purpose of developing "Reduced Integration Elements" was to increase computational efficiency without loosing the accuracy. But you should understand, "there is no free lunch in this world". Therefore, reduced integration elements come with their own cost. To simply answer your question, yes you can use it. But you've to pay attention to the caveat associated with it. To understand and to utilize it properly, you've to read theory books (ziknozwich is the god book), and abaqus analysis manual, carefully.
P.s. There's no pre-made recipe regarding how to handle hourglass effect, especially when it comes to umat
Hi Prithivirajan
yes, usually u can use C3D8R element instead of C3D8.
good luck.
Yes, I used the reduced hour integration in the past. It provides the same result.
Dear Veerappan,
When solving a mechanical problem using the finite element method there are:
1. Nodes where the displacement is calculated by minimizing the residual
2. Gauss points where the constitutive model is integrated.
The computational time depends on both the number of nodes as well as the number of Gauss points. A ratio can be calculated of the number of Gauss points per node to give optimal results as well as low computational costs. Low numbers of Gauss points result in a "checkerboard" like deformation map which might be misleading (although many a times this checkerboard cannot be seen because commercial softwares show an interpolated map rather than the exact Gauss point values, which is why it looks so smooth). Global stress strain values might converge even with a stiff mesh but local accuracy is not guaranteed.
Best regards,
Harris
- Badges
- Science topic