Composite Steel/Wood Space frame analysis

[CGN]

Andrew has been of great help here in the forum and has always approach any responses in a very professional manner.

[darbikrash]

Well I sure didn't mean to start a controversy, and the last thing I expected was to be accused of name calling, certainly that was not the intent.

I have strong opinions about the use of codes that are not tier 1, mainly centered around the lack of verification models, small installed base of users, and the basic premise that it takes just as much effort to learn a high end code as it does to learn to learn any other, if the user limits themselves to using the comparable capability.

The upside is that if and when the user needs more capability, he can extend his training into the new areas using the same interface. Most analysts will "outgrow" a code very quickly, so why not start with the one that you can stay with for your entire career?

The simulation software industry is changing, they are developing standard pre and post processors, and adding very specialized functionality to these platforms such as CFD, multi-physics, dynamics, etc. An example of this would be PATRAN, which is now ported to support everything from simple 2D beams, composite laminates, non-linear (MARC) fluid/structure interaction (DYTRAN) dynamics, and the list goes on.

I realize most people have no need whatsoever for many of these tools, but you might someday, and re-learning new point solution tools every time you have a new problem is just silly in my view. Training is the most expensive part of simulation, why do it twice?

So I advocate shopping carefully for a structural analysis tool, pick one and stick with it.

Things I look for in a code center around long term development, support, and most importantly, a rich list of verification models that have direct traceability to a governing body like NAFEMS (www.nafems.org). These guys independently verify whether a code is returning accurate solutions, and give confidence to the user that the solution algorithims are based on sound practice.

How many of the codes suggested on this thread comply with NAFEMS verification? How many users know enough to ask?

At any rate, I did look over some of the suggested software in this thread prior to my post, including that recommended by Mr. Mason, and Multi Frame looks like a fine product, and I have no commentary on the suitability of this code outside of the remarks I have already made. I called no one a name, and do not wish to offend anyone. My purpose in responding to this thread was to provide some insight, based on my experience, on some of the issues involved with choosing a structural analysis code. Clearly, my input was summarily dismissed without a single comment on the content, and I responded to Mr. Mason's two sentence reply by pointing out that he had in fact trivialized the problem and made a remark to the effect of "using a sledge hammer to drive in a thumbtack". If this isn't flippant, I don't know what is. I meant no insult by this, it's just the way it came off to me.

If the philopsohy on this board is to try and find the cheapest code to get by, to search out free FEA downloads, and chortle about what a good deal we got, well, perhaps it's not worth it. However, if people are legimately interested in helping each other solve problems, I would be feel privleged to contribute in whatever small way I can.

John

[CGN]

John you contribution are more than welcome, is not about controversy or name calling on respect of previous comments is just that Andrew has always been very professional on his comments and always has been open on it's position of being part of Maxsurf, as you can see there is no many of us that can fluently understand FEA and it's reach as an analysis tool, and i very much welcome your comments and i hope you stay active in the forum and share some of your knowledge, NAFEMS??.....everyday there is something new to learn, thanks

[darbikrash]

Here is a better link:

http://www.nafems.org/

[nero]

"Oftentimes, very simple models, bearing little resemblance to the final CAD model, can provide more insight in a shorter period of time than can a more complex and highly detailed solution. Less qualified users performing more complex analyses may cause a regression, rather than an improvement, in effectiveness." from the nafems.org link above

This is what a FEA program like Structurix and some of the other "free" programs do. No meshing complications.

[darbikrash]

I agree, and this point was made in my earlier post (#11) above.

Let me try to clarify just because a program is NAFEMS approved, and is considered "Tier 1", doesn't mean you have to use a mesh to solve beam element problems. All of the major FEA programs support beam elements for simple space frame problems, and no mesh is needed.

A good analyst always uses the least (computationally) expensive approach to solve his problem, often this means not using a mesh, simply connecting 2D beam elements together and quickly solving the problem. In such a case, it would be relatively rare to rely on the importation of CAD geometry.

The comments regarding meshing no doubt spring from the diasters that routinely occur when CAD operators are given access to FEA tools without appropriate training. If you're not trained as an analyst first, it may seem easier to simpy "dump" the 3D CAD model into the solver, mesh it, and hope for the best. The software industry goes to great pains to further this myth, with products specifically marketed to these groups.

These meshes in the wrong hands usually do give bad answers, but they sure provide some great pictures!

I do not advocate using a tool that is inappropriately complex for the task at hand, and to be sure there is a progression of the analysis that would start with hand calcs, advance to perhaps a 2D beam element loadcase (no mesh) and when the global sizing exercise is complete, move on to some 3D work in areas identified as "hot". From there, fatigue, notch sensitivies, dynamics, etc can be studied as needed.

As stated earlier, how far you take an analysis down this path is up to you, and the "culture" within your work environment. Many perfectly good designs go into production without even a hand calc, some stop at 2D beam elements, and some go all the way through to dynamics. Depends on many things, among those is aversion to failure, and to some degree the sophistication (and available skill sets) of the design group.

My point, now thoroughly beaten to death, is that to me it makes more sense to use a software tool that is extensible into higher level functionality even if you don't think you need it right now. It is just as easy to use PATRAN for example to create a non meshed beam element problem (probably easier) as it is any of the so called specialty applications, free or otherwise.

So why waste your time? Again, the selection criteria should not focus on the cost of the software, (within reason!) this is not significant in the grand scheme of things. The cost for training is an order of magnitude higher than the software acquistion costs. And then the central point, is the software validated from a independent third party like NAFEMS?

If not, believe me even if it's free, you can't afford it.

[nero]

"My point, now thoroughly beaten to death, is that to me it makes more sense to use a software tool that is extensible into higher level functionality even if you don't think you need it right now. It is just as easy to use PATRAN for example to create a non meshed beam element problem (probably easier) as it is any of the so called specialty applications, free or otherwise." -darbikrash

Back in the late 80's and early 90's, I use to evangilize this point to graphic arts professionals. Start with the best software, learn it and the pay back is many fold in time and money saved. (also applies to an OS like Mac OS)

The clearly high level, well developed software (Photoshop, Illustrator, Xpress) are now the industry standard. For some insane reason corporations are running Microsoft.

So yes you are correct. However, sometimes tools are based on a single project with time constraints not long enough for the training needed for full-blown FEA packages. And for private users laying out $20K for a limited time license is unthinkable ... even if it would run on Mac OS.