How to do a unit enforced displacement check

manuelz

Hi, I just bought the book called "Practical Finite Element Analysis for Mechanical Engineers" and I'm trying to learn from it alongside the Mecway free edition.

I want to implement one of the mathematical validity checks metioned in Ch 16.3: Unit enforced displacement check. This check should allow me to "determine nodes on the structure that are either loosely connected or have very small stiffnesses". To do so, I need to:

- remove all constraints
- apply a unit enforced displacement or rotation in an element close to the CG of the structure, while restraining the other five components of this node to not move.

I'm using the basic graphics tutorial example.

- I have removed the loads and constraints
- I have added a "displacement" for 1 unit in X in one node around the center of the structure

Some questions:
1) What type of analysis should I use? Needs to be nonlinear? Can it be linear?

2) How do I selectively constrain degrees of freedom in a node? Fixed supports for example seem to only allow me to constrain faces, and the nodes of those faces are fully constrained. The "what's wrong message" says that I should use beam elements instead.

3) I'm not sure what element are being used, but, how could I change them?

Those are the doubts that I have for now.

Thanks,
Manuel

Victor

Hello Manuel

1) It can be linear. Even if the displacement is large, linear is still OK just to see if its doing rigid body motion.

2) You can use a separate displacement constraint to constrain each translational DOF and a separate node rotation constraint for each rotational DOF. Fixed support is a shortcut to constraint both displacement and rotation, but a single node on a 3D solid or 2D plane element doesn't have rotational DOFs so it can't have rotation constrained in that simple way so fixed support doesn't work.

3) I don't understand what you mean by "element". If you mean beam/truss/continuum/shell, etc. then it's determined by the shape, analysis type, element properties (beam/truss) and material (spring). For example a quad4 element in a 3D analysis type is always a shell and in a 2D analysis type, it's always a plane continuum element.

manuelz

Thank you for your answers Victor, they are of great help to me. My third question was badly formulated; after reading a the manual and your example, it's much clearer now.

disla

Hi Manuel,

Mecway also has a very useful tool called "Open Cracks" under the view tree. It helps a lot to find nodes/elements that are disconnected.

JohnM

Manuel-

Congratulations on your interest in FEA, welcome to the career that has kept me fed for almost 40 years.

I have a strong opinion about learning FEA. I say "just do it". I get many interns and new graduates that have been subjected to an "FEA class" in college, only to find that no one needs to derive a shape function any more or understand the math of inverting a matrix. I think many professors would disagree, but I have convinced at least a few to modify their curriculums to be more practical.

At my company, new engineers are pointed to a lot of example problems. Try http://simcommons.com, and start with the "clip" problem at the bottom of the Structural/Thermal Analysis list. Most of our young engineers become quite proficient going through these 5-minute exercises. All of these models work in the demo version of Mecway.

It is not my intention to discourage you from understanding the more technical aspects of finite element analysis, and since your book has the word "practical" in the title that may be a good start. But if you have 10 hours to learn, I would spend 9 hours simply "doing problems", and you will find that serves you quite well!

Good luck,

JohnM

manuelz

disla, I guess what you suggest is a software-related way of finding disconnected nodes, whereas what the book suggests is a software-agnostic way of doing it. Thanks for the tip!

JohnM, I have seen the link that you suggest, and I think is a great resource! The clip example that you mention looks really interesting and makes me wonder if that kind of approach (of moving an object over another while having a contact defined) is possible to do in a static analysis? Indeed, the book so far looks very practical and software-agnostic, but I really like it because of the section where it explains how to validate the analysis, sort of guardrails in case you know how to do something, but don't really know all the details behind. Thanks for the suggestions!

JohnM

Manuel-

You often start with a simple static analysis of a single component and then extend as necessary. It's important to have a problem statement to define what you trying to learn. For example, using a Static Analysis you can simply load the clip contact surface with a pressure, and determine the insertion force and an estimate of the stresses. But if understanding contact pressure at the interface were important, this would need to be extended to Nonlinear.

I would not base your approach on complexity/analysis type, I would consider the goals you have for the analysis and the time to model/run/interpret. There is always this tradeoff between time and accuracy, and it's usually your own experience that tells you the right approach.

manuelz

Understood, first gotta define the problem, the goals and the timeframe, then see what can I use to get that on time. Thank you for your insights JohnM!

JohnM

By the way, a simple way to find "loosely connected" stuff is to run a modal analysis.
Don't delete constraints - simply run a modal with a good number of modes, and disconnected/loosely elements will show themselves as low or zero frequency response.

manuelz

Oh right, before reading about the unit enforced displacement check I read about that in the book. However, I wonder why do you say that constraints shouldn't be removed?

JohnM

A modal analysis will find 6 - 0Hz modes for each DOF in an unrestrained body. If you have N parts, this will be 6*N modes. Normally if you are checking for loosely connected parts in an analysis, there is no need to let go of (delete constraints) on bodies that are already properly held. What you are typically looking for with this approach is a 0 or near 0 Hz response of a component that is not properly restrained, after all is set up (including the constraints).

manuelz

Thanks for the details John.