Nominal stresses output

[venky_krishnan]

Hi,

The Abaqus manual doesn't explictly say so, but I believe it is not
possible for users to output the nominal stresses in a large
deformation elasticity problem. Is this because Abaqus standard uses
the updated lagrangian formulation and hence keeps track of only the
relative deformation gradient and not the total deformation gradient ?

If not, is it possible to output all the element jacobians and
deformation gradients at all integration points, so we can compute the
nominal stresses ourselves.

Thanks,
V

[BenZ]

I also think you cannot output this stress.

No, Abaqus provides the full deformation gradient.

Despite Abaqus uses FLA or FLT to write the equilibrium, you can use
any constitutive law (Cauchy stress related to strain rate tensor,
Piola-Kirchoff 2 related to Green-Lagrange strain.....), so this is
not the reason.
The reason is that when dealing with large deformation, the Cauchy
stress is commonly used to understand stresses on the current
configuration.

Using the UMAT/VUMAT you can output the jacobian, you can compute
the nominal stress.

BenZ.

Post by venky_krishnan
Hi,
The Abaqus manual doesn't explictly say so, but I believe it is not
possible for users to output the nominal stresses in a large
deformation elasticity problem. Is this because Abaqus standard uses
the updated lagrangian formulation and hence keeps track of only the
relative deformation gradient and not the total deformation

gradient ?

Post by venky_krishnan
If not, is it possible to output all the element jacobians and
deformation gradients at all integration points, so we can compute the
nominal stresses ourselves.
Thanks,
V

[venky_krishnan]

Hi,

This is regarding the use of reduced order integration elements in
Abaqus STANDARD for a quasi-static problem.

Should one be concerned about "hourglassing" if the global stiffness
matrix is non-singular and the solution converges.

For a fully incompressible material, I get the same numerical solution
using either fully integrated hybrid elements or reduced order hybrid
elements. So I guess reduced order integration is always better, as
long as the solution converges. Am I right in thinking so?

Thanks,
Venkat

[Fernando]

Hi Venkat,

Post by venky_krishnan
Should one be concerned about "hourglassing" if the global stiffness
matrix is non-singular and the solution converges.
For a fully incompressible material, I get the same numerical solution
using either fully integrated hybrid elements or reduced order hybrid
elements. So I guess reduced order integration is always better, as
long as the solution converges. Am I right in thinking so?

More or less. Reduced integration avoids many of the problems leading
to hourglassing and provide a faster stiffness matrix assembly (you
have 1 integration point instead of 8, for C3D8). The price is also
reduced precision in the output, so if you have high gradients in your
solution, or special materials (think UMAT), you need to be careful.
As always, there is not a solution that's always valid (that's why you
have a range of elements to choose from!)

Additionally, solution in /Standard may converge even in the event of
mild hourglassing (the solver is quite robust, you see). It is always
a very good idea to check for hourglassing visually, by plotting the
deformed mesh (for small deformations of the mesh, you may need to
scale the displacements to see the hourglassing).

Note that hourglassing is more likely to happen in very regular,
grid-like meshes, and that very often you just need to perturb these
meshes to get rid of it. So you know what to expect for incompressible
parts with simple, regular geometry...

Hope this helps,
Fernando