Do you want a faster mesher or a faster fea solver? 2

[ZeroExperience]

If you had to pick one, would you pick a faster mesher (faster cad defeauturing, midsurface extraction, volume and surface mesher, etc) or would you pick a faster solver (Ansys, Abaqus, MSC Nastran, etc)?

 

[rb1957]

things are so damn fast these days (i'm solving models of >40,000 elements in <10 minutes with a pretty basic desktop) i'd pick the code that is easiest to use, pre- and post-processing.

 

[inline6]

i wish there was a CAD system that could defeature the geomerty and extract midplanes properly without redoing the parts but the 3D models i get are not made for FEA they are done for production purposes and are constructed as such and can not be defeatured, but we make the best of them. If only we had to analyse a plate with a hole like the vendor demos...lol
There is an awful lot of manual midplaning and editing with a fabrictaed assembly of hundred and hundreds of parts plus one needs to judge how these parts are connected to determine relatsic load transfer and stiffness. the FEA systems in the CAD programs are very primitive and despite vendors claiming they can help us they have not convinced us except for quick and dirty analysis of single casting part.

 

[rb1957]

that's pretty olde school ... de-featuring and using 2D elements on the mid-plane ... i'd've thought everyone used 3D elements to analyze fittings, on 2D elements for large scale structures (then why worry about mid-plane ? OML should be good enough).

but i agree with FEA in CAD programs ... it's best avoided, it's very limited, very "canned", quite open to making pretty pictures.

 

[inline6]

not sure what OML is but on some structures by not using midplanes can mean some areas have incorrect section properties e.g local stiffeners, box type structures There are some d^3 factors so even on plate structures half thickness can produce significant errors if outer or inner surface is used so it is best to be consistent and stick with midplanes in most cases. The midplanes are not the time consuming part it is just geomerty in general.

 

[rb1957]

OML = outer mold line = outside surface ... in my business with small thickness skins, thickness isn't as significant

 

[jagad5]

I'd pick faster mesher over faster solver. When I'm creating a model I want to be able to compare mesh options during the day. Then I can submit it to run the solver overnight. We often run models with up to a few million elements (with assembly load steps and contact between parts in the assembly.) These models we build often take as much as a couple days to solve so finding out on Monday morning that the mesh I used Friday afternoon had something wrong with it is a major hassle. Reducing the meshing time 25% would be of more practical significance than reducing the solve time by 25%.

Doug

 

[rb1957]

so i guess it depends on whether you spend more time solving or meshing !

 

[racookpe1978]

As an FEA newbie then, can I ask what is the ideal (3d ?) mesh for analysis?

I thought every mesh had to be "drawn" across the (outside) surface approximating all of the solid part, then developed through the cross section of each region of the part so all of the physical geometry is "filled" and every analysis cube touches its neighbor so the diff. equations "match" at every touching surface; and the mesh has to be cubic/near-cubic in shape everywhere - not a mix of triangles and cubes, with different number of cubes in different regions.

 

[GregLocock]

"with different number of cubes in different regions"

no, that is not a sign of a good mesh, in fact a durability engineer will often refine the mesh at hot spots while leaving the bulk of the mesh undisturbed.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[GregLocock]

Reason being St Venant and all that

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?