Mesh. 1

[jenhe]

Hi.
For about a half year ago I took a E-learning course to learn about mechanica, but when i took the course there wasent not much information about MESH, always use default mesh when i make a tutorial. So my quistion is, if I want the correct "stress von miss" result which mesh do i have to use then?? and where is the default MESH (settings)?.

If i go to the menu line: AUTOGEM/CONTROL, and in this box i can change the size of the elemet, if I change the element size i also change the result of my stress von miss?? So if i want to see the correct "stress von miss" result of my stell construction, which kain of MESH do i have to create??

Hope you understand.
Thanks for any help.

 

[nlinesk8s]

I have not tried this yet, but you can go to

http://learningexchange.ptc.com

and view a tutorial on Autogem meshing. You will need a userid, but it requires only a few minutes to register.

 

[jenhe]

Hi.
Thanks for your information, but when you create a static simulation, which mesh do you use then, default or do you create your oven mesh.? I think a have a problem understand mesh, any help will be great. Thanks.

 

[1Khaleel]

Jenhe,

I haven't used Mechanica for a few years now, but the principles of FEA remain the same.
There is no one right mesh/element size that results in correct results.
Often the finer the mesh the better the results.
However, you do not want to over do it, because a very fine mesh requires more memory/time.

I suggest you run a multi-pass analysis (MPA) with the default mesh then look for the Von Misses measure.
Does the measure converge? If it does, then you don't need to refine the mesh.
If it doesn't, then you need to refine the mesh near the area(s) of concern.

In previous versions (Wildfire series), you could refine the mesh by seeding (adding datum points) on the surface of concern.
Changing the Autogem settings will work too, but the changes will apply to all your models.

I hope this helps.

 

[integrisbrad]

I will reiterate 1Khaleel's words. There is no one correct mesh or element size. Starting in WF4, there are more mesh size controls and they continued to add capabilities in this regard, which has been a very welcome addition.

In my experience, meshing uniformity (size and shape: no long, pointy elements and no large elements connected to small elements) in the area of interest seems to be the biggest concern for p-elements in Mechanica. Maybe this is a general p-element statement, but I've only used Mechanica's p-elements. You can generally let the p-elements do their job at refinement if you have the uniformity mentioned previously. I have never really relied on any of the default measures for results accuracy. There are too many opportunities for errors in both boundary condition and geometric singularities. It's my understanding that those MPA measures are supposed to account for those situations, but I don't trust them to do so. In general, I've learned from experience in regards to the mesh that I need to create, and I will refine if my interpretation of the results warrant refinement. If running MPA, I create measures in the area(s) of interest on which to converge instead of using the defaults.

 

[brep]

There are many, many pages that could be written about this (!) but just a few notes from my experience using Creo Simulate (I started with RASNA Applied Mechanica back in 1995 BTW)

- in most cases the default mesh and SPA convergence will give as good (sometimes better) accuracy and do it MUCH faster. I recently did some tests for a big company comparing manually controlled meshes of various densities against default autogem. The defaults were as accurate, and were an order of magnitude faster. Moral of the story: you do not need to create the same uniform and dense mesh as with a 1st or 2nd order h-code like nastran. Let the polynomials do the hard work!

- the simpler the model, the more influence the boundary conditions will have, *especially* if they are "wrong"! Try simulating a flat square plate fixed at one end and under tension at the other. If you fully clamp the end, what are you actually simulating in real life? (hint: infinite stiffness, where does this exist????) Do some reading about Saint-Venant's principle and see why just making the elements smaller and smaller does not yield a "better" result. Moral of the story: there are many untruths/myths out there in the FEA world, you'll need to think for yourself to truly understand what you are simulating!

- the right mesh (in any FEA, regardless of h- or p- method) is the one that gives you results that answer your particular question. You are simulating a car crash: Are you interested in the stresses at the root of the bolt that holds the rear license plate on a car? No? Then why mesh the threads in this bolt? What else can you exclude and not influence the results that you need? This is the biggest issue for simulation in my opinion. Moral of the story: "All models are wrong, but some models are useful"

http://en.wikiquote.org/wiki/George_E._P._Box