Basic FEM

[Windvang]

A new plug in for Rhino is available, able to do some basic FEM, it works with Rhino solids, no complicated meshing or exporting. I have very little FEM experience, so I can not commend on the results.

A free beta version is available for download. http://www.scan-and-solve.com/

[ACuttle]

Thanks for the info, I'll have to give it a look over.

[MikeJohns]

Thanks Windvang.
Have you tried validating the results at all, or compared them with one of the major FEA packages? It looks very limited for the moment. Ironically a shell based approach for Rhino would be more useful for a principally surfaces CAD package.

I would have liked to see the mesh/grid able to be refined in critical areas/volumes, otherwise it's not practical for large models.

[Ad Hoc]

Thanks Windvang,
This is interesting. The problem i have with freeware like this is when the makers come up with statements, like below, to promote their software to get anyone using it:

"..No meshing or conversions are required and no prior knowledge of analysis or finite elements is needed.."

FEA is not for those have no structural background. If you do not understand the basics or even are able to do some quick hand calculations to "get an idea" of the results, then you are playing with fire. Classic 'GIGO'.

It is bad enough when poeple use hydrodynamic software creating all sorts of shapes and "conculsions", with endless pretty colours without prior knowledge of hydrodynamics and indeed understanding any or none of the limitations of the software being used. Using any technical "number crunching" software requires a background in the subject matter and the limitations.

The software must ALWAYS be the slave, not the master. It is just a tool, not a magic wand!

Mike, good point.
I tried looking to see if there is any link with NAFEMS, but couldn't find any.

[daiquiri]

Originally Posted by Ad Hoc

"..No meshing or conversions are required and no prior knowledge of analysis or finite elements is needed.."

I agree with Ad Hoc that the above hype is absolutely inappropriate, or even dangerous. Performing a FEM with a glossy software without sufficient prior knowledge about material science, material mechanics and FEM can lead to design errors with very bad consequences.

Mike Johns is also right when he points out that a FEM software which uses solid elements for meshing thin-walled structures is inefficient at best or unreliable or completely unusable at worst.

Be careful.

[Windvang]

@ Ad Hoc

I discovered that the mother company is: http://www.intact-solutions.com/technology.php
the principles of the software are published here http://www.intact-solutions.com/Scan&Solve.pdf.

I agree just having software like this doesn't make you an good engineer, just like having marine design software doesn't make you a naval architect. I don't think you can blame the software maker for delivering affordable and easy to use software and that it is the responsibility of the end user to make good use of it.

[terhohalme]

Fortunately, I have so much structural training that I dared to try it. Fun...

[DCockey]

Originally Posted by daiquiri

I agree with Ad Hoc that the above hype is absolutely inappropriate, or even dangerous. Performing a FEM with a glossy software without sufficient prior knowledge about material science, material mechanics and FEM can lead to design errors with very bad consequences.

Mike Johns is also right when he points out that a FEM software which uses solid elements for meshing thin-walled structures is inefficient at best or unreliable or completely unusable at worst.

Be careful.

I agree. Unfortunately too many of the users of even sophisticated FEM and CFD software are expert at putting the input together, running the software, and making pictures of the output without understanding the basic assumptions and limitations inherent in the modeling and analysis. Physical and numerical aspects of the results are confused, and for many analysts numerical is more "real" than the real world.

[Ad Hoc]

Originally Posted by Windvang

@ Ad Hoc
I discovered that the mother company is: http://www.intact-solutions.com/technology.php
the principles of the software are published here http://www.intact-solutions.com/Scan&Solve.pdf.

I had a quick look at their report. Still worrying.

"..By definition, meshes must adapt to the smallest geometric detail in geometric model, leading to excessively large meshes, and making accurate meshing impractical for any geometric model with small features or geometric errors..."

Meshes do not need to be the smallest geometric detail, but generally do, because that is one of the pruposes of using the FEM. It does not lead to large meshes....only if performed by someone not well versed in FEA and/or structural analysis.

This also is based upon the assumption that the users is letting the FE software decide the mesh, ie automatic. This too is prone to errors.

The "mesh" is performed on the outside of the model, so it still uses a mesh, but the 3D mesh is produced automatically (and 'projected' onto the model) and one has no idea of the density.

Their conclusions are telling:

"..Despite the perceived computational overhead, performance of Scan&Solve™ is already approaching that of commercial FEA systems, and is likely to exceed them on models containing small features, errors, and noisy boundaries..."

In otherwords, no, it has yet to be validated by NAFEMs.

One example it gave was of a 2D plate with a hole, for a SCF. Yet it was plotting results with Von Mises....er...why?? Why plot Von Mises of a 2D model??....Ergo, what is the user actually looking at and how much understanding is there, if the user is unaware of what Von Mises is, and what it is used for. Also the 2D does not state how the SCF is computed other than the mesh size reduces. The mesh size is critical in establishing SCFs and then "hot spot" stresses.

So what does "...We then solve for displacements and stresses inside the thin solid and compare results to the analytic solution..."...."inside" the solid, how, how does it know what the internal geometry is like when the 'mesh' is on the outside surface only?

Too many unanswered questions.

[DCockey]

I also looked at the white paper at http://www.intact-solutions.com/Scan&Solve.pdf but probably from a different viewpoint. My tentative observations based on a very brief study of the white paper follow.

There are many, many other methods for solving systems of partial differential equations in addition to the very well developed conventional finite element methods. The Scan-and-Solve method appears to use higher order, non-local basis functions based on B-splines and interpolates the boundary conditions where the boundary doesn't coincide with the mesh points. The resulting systems of equations for the unknown coefficients / nodal values are not as sparse as with classical finite element methods and would generally increase solution times and difficulty. To offset this they appear to use an evenly spaced rectangular grid which is facilitated by the interpolation of boundary conditions. The even nature of the grid provides a better structured set of equations.

In many respects this methodology appears to be more similar to some higher-order finite difference / finite volume methods than to conventional finite element methods. It also has some similarities to modal methods. It probably would be best if it is not refered to as a finite element method.

"how does it know what the internal geometry is like when the 'mesh' is on the outside surface only?"

The internal geometry is the space inside the boundaries, and the current version of the software appears to assume linear, isotropic elastic behavior which is modeled by the basis functions. It's fundamentally different than using discrete elements as in FEM methods.

For the current implementation which appears to use only a uniform mesh density the ability to resolve fine geometrical detail will be limited by the mesh size. It would seem that holes would need to be at least several mesh spacings in radius to have an effect on the results. An interesting question is how thick do plates and shells need to be compared to the grid? Numerical experimentation would be interesting.

Another limitation of the current implementation is boundary conditions are limited to applied stresses or zero displacement. Moving beyond these boundary conditions is very likely to significantly complicate the "pre-processor".

A potentially very significant limitation for marine use of the current formulation implementation is the constraint of isotropic materials which precludes accurate analysis of almost all composite and wood structures. This limitation will require some very significant extensions of the formulation as well as a different pre-processing.

Obviously this software is not mature enough yet to be relied on for critical engineering decisions. As with any analysis method verification and correleation with both theoretical and experimental results will be needed. Also the display of the results is rather rudimentary. Better access to the details of the results as well as the mesh characteristics will be needed.

However the overall approach is very interesting and potentially very useful, particularly for preliminary design cycles when the fundamental geometry is being considered. The facilitation of a uniform mesh for computational efficiency by using interpolated boundary conditions is potentially a significant advance. I don't expect it to replace FEM analysis in the foreseeable future

[MikeJohns]

Originally Posted by DCockey

I agree. Unfortunately too many of the users of even sophisticated FEM and CFD software are expert at putting the input together, running the software, and making pictures of the output without understanding the basic assumptions and limitations inherent in the modeling and analysis. Physical and numerical aspects of the results are confused, and for many analysts numerical is more "real" than the real world.

Absolutely and it's hard work understanding what's behind all this.

Reality is removed several steps from FEA results and it's essential that the jockey has an insight into the results. They should know what to expect and more importantly why they should expect them.

[whoever]

i think solidworks is the best.

[MikeJohns]

Originally Posted by whoever

i think solidworks is the best.

You mean cosmosworks professional I presume, not the cosmos lite bundled with solidworks? Even then it's meshing is not the best, it takes a lot of time to set up a decent mesh with proper node sharing between components of the model. . I have much better meshing tools in other packages I use. Automatic mesh refinement and a powerful mesh generator saves a lot of time. Cosmosworks lacks offset nodes too. Its not as transparent as I'd like and its very 'graphics' oriented.

I do like the ease that you can get shells created in rhino directly into cosmosworks without having to go through IGES, it opens native rhino files.

[Ad Hoc]

Originally Posted by MikeJohns

You mean cosmosworks professional I presume, not the cosmos lite bundled with solidworks? Even then it's meshing is not the best, it takes a lot of time to set up a decent mesh with proper node sharing between components of the model. . I have much better meshing tools in other packages I use. Automatic mesh refinement and a powerful mesh generator saves a lot of time. Cosmosworks lacks offset nodes too. Its not as transparent as I'd like and its very 'graphics' oriented.

I do like the ease that you can get shells created in rhino directly into cosmosworks without having to go through IGES, it opens native rhino files.

I've never used the cosmosworks professional, nor the 'lite'. But i use the orginal Cosmos/M. Have done since 1992. It has lost its "magic" since being part of the Solidworks umbrella. Thus i still use the 'older' but better original ~v2.9 of cosmos and have not botherd to update it to the solidworks. The v.2.9 does what i want and always has, excellent FEA, RIP.

[alidesigner]

The problem with FEA that's bundled into CAD programs is that they are marketed by people who are not FEA specialists who claim that if you can build a model then you can analyse it with the wizards. I have seen a naval architect analyse a lifting lug and believe the results even though the loads were wrong.

If you dont know the theory behind FEA and the elements you are using, and if a part failure could kill someone (ie lifting lug), then dont use it. It's not something you can learn in a few hours.