Surface Contact - wrong displacement

[Anonymous]

Hi, please see figure below depicting solid FEA model for linear statics. Surface to surface test model was incorporated. On the right you might see the displacements. Why bottom part does not contribute? Please assisit with setting surface contact properly (under setup) also please advise what is the practical advantage/dosadvatnage/consequences of choosing between surface contact and using compressive gap elements without gap?

[AstroJohnPE]

Hi Mickey,

My suspicion is that the part on top is not statically stable without the contact elements, and therefore you have received a garbage answer. But a little more information about the results would be helpful to determine what the real issue is, such as:

1. What are the displacement values? (If reasonable, they may be correct but the model is missing something -- like gravity. If large, the model may not be statically stable and you have a garbage result.)
2. What is the approximate size of the parts? (This will help us understand that the displacement of A units is small or large compared to the size of the part.)
3. What is the displacement of the "fixed" part? (Right-click on the part in the browser and choose Isolate). A blue color only tells us that the displacement is between 0 and 1/10 (approximately) the maximum displacement. This seems possible depending on the details of the analysis.
4. What are the loads and constraints applied to the model? Is gravity applied which would resist the overturning moment of the force we see on the right side of the upper part?
5. What are the forces in the contact elements? ("Results Contours > Other Results > Element Forces > Axial Force"). This will show how much of the surface is in contact. If no points are in contact, you have a garbage answer. (You can also use the legend thresholds to hide elements with 0 contact force: "Results Contours > Legend Properties > Range Settings > Less-Than > -0.001").

Here are some differences between using surface contact and creating gap elements:

• Surface contact is automatic. The user needs to create the gap elements (draw lines).
• Surface contact requires a matched mesh and 0 gap between the surfaces. Gap elements requires a matched mesh and some gap between the surfaces because you cannot draw a line that is 0 units long. (I suppose gap elements could be used if the mesh is almost matched, as long as the gap elements are almost perpendicular to the surfaces.)
• Surface contact adjusts the stiffness of each contact element based on the area surrounding each node. Gap elements use the same stiffness for all gap elements. So surface contact gives a more accurate result when the area around the nodes varies over the surface. (I think the area around the node always varies. Think of two rectangular surfaces in contact. Nodes along the edge have half the area of nodes in the middle, and nodes at the corner have 1/4 the area of nodes in the middle.)
• Surface contact has some variations (interferrence, friction) that gap elements do not have.
• Gap elements have some variations (work in tension, nonzero positive gaps) that surface contact does not have.

[Anonymous]

Hi John, thanks for contribution.

That is true it is test file only. But even when upper part is constrained, the results are the same - bottom part acts wrongly, as it would be rigid.

Nevertheless I would like to ask you how to perform, proper set up linear static analysis with few contact surfaces in the system? Please tell me how you usually perform it for solid elements.

[AstroJohnPE · Accepted solution]

I would do it as in the attached.

[Anonymous]

Thanks John.

However please note, your resultant reaction value is depended of the amount/stiffness of the additionally added spring supports. What about this consequence? What if one would like to readout proper reaction, which in this case will not be correct.

[AstroJohnPE]

True, the correct reaction of 0.27 lbf in my model was reduced by approximately 0.00000002 lbf by the weak springs. (In other words, the support reaction in the weak springs was very small.) This is insignificant compared to the accuracy of a typical analysis. (Known solutions can get within a few percent with carefully made models. I suspect that most models are within 5% to 20% of the exact answer.)

In summary, the weak springs can be placed on the model where they have an insignificant effect on the results (place them at points of least displacement, and use a low stiffness such that the amount of support they provide is negligible) but provide the beneficial stability occasionally or often required for the solution.