I used an one element 3D cube to make some tests with Abaqus user defined materials. I made an observation with the Abaqus fully integrated C3D8 element with an simple elastic material model. I do really understand this observation.
The model is shown in the Figure cube.jpg. The nodes in the x-z base and node 3 have a minimal
constraint to avoid rigig body motions. Some combined tension and shear is applied with the upper nodes 1, 3 with a prescribed displacement in x-direction, same for both nodes. Poisson's ratio is chosen with 0 to avoid lateral effects in z-direction.
The observation is, regardless of this Abaqus computes lateral displacements for nodes, i.e. nodes 7, 8, 1, 2, see Fig. U.jpg . There are also strains and stresses in z-directions as eigenstress without corresponding nodal forces.
That contradicts my mechanical intuition telling me there should be no lateral effects with Poisson = 0. This intuition is also confirmed with another programm taking identical model
data.
So what might be wrong? My mechanical intuition, or did I oversee something with the model.
Or is there some pitfall with the C3D8 element - seems to be od for me.
Appended is the inp-file.
The problem setup is not clear from the description. Could you please share all the applied constrains and forces on each node?
The lateral effects are caused by inhomogenuity of the stress-strain state.
If to add BC U1(2)=U1(4)=0.005, U1(5)=U1(7)=0, the state becomes homogeneous, and with Poisson = 0 all lateral displacements are equal to zero.
Answering Victor: This seems to be pure tension. But what I want is combined shear with tension. This is actually not homogeneous and leads to a nonhomogenous state within a fully integrated C3D8 element. Lateral z-stresses look like a locking effect but the trilinear trial functions (Abaqus theory manual - solid isoparametric quadrilaterals - first order brick) should be flexible enough to avoid this, or am I wrong? My independent implementation with such a C3D8 does not have such a lateral effect.
Answering Chennakesava: Node 1 prescribed x, node 2 zero y, node 3 prescribed x - same as 1 and zero z, node 4 zero y, node 6 zero x and y, node 8 zero x and y. All other dofs free. No forces. Added a slightly modified .inp corresponding to this.
Ulrich Häussler-Combe There is a hourglass mode x*z for U3 which can not be eliminated owing to lack of sufficient number of constraints (only U3(3)=0).
I see, it is not always obvious how to constrain rigid body motions with 3D strucutures. The only explicit constraint for z(3) is prescribed with node 3 which looks like allowing for rotations around the vertical y(2) axis. But this should be avoided with the x(1) constraint of node 6 to make the model stable. Applying a test with a z(3) force in node 7 confirms this.
Regarding hourglassing: the model is fully integrated with 8 integration points. Does this permit for hourglassing? May hourglassing be connected with eigen-stresses?
In the model in question, the constraints are set correctly to avoid rigid body motions and they are not enough to avoid so called "hourglass" mode. However, I doubt that we can call this a hourglassing, because, as far as I know, full-integrated elements are not susceptible to it.
- Badges
- Science method