Fred Hopengarten wrote:
> From:K1VR:
>
> Kurt -- So what is FEA software? How much does it cost? Do they have a
> web page? -- Fred
Hi Fred,
Some guys are good at asking the easy questions that require difficult
answers. Congratulations!
Since you asked via the reflector, I gotta answer in kind.
The idea of any TT subscriber getting and using FEA software, if not in the
too hard box, is definitely in the very difficult box.
Readers who do not want to make a career move by trying to obtain and learn
how to get useful info from software that is clearly intended for users with
lots of nut & bolts between their ears, should immediately move on to the
next post!
What is FEA software?
FEA stands for Finite Element Analysis. In essence, it is the mechanical
equivalent of NEC or MiniNEC software, depending on which software you get.
The "Finite Element" name describes the basic nature of how the user models,
and the software analyzes a structure. The "Finite Elements" are very
clearly defined segments of the total structure.
Each segment has it own set of geometric and material properties. These
properties are created by the user and stored in a database to be applied by
the user to each segment in a model. The property definition and database
and interface functions vary widely between the different programs. The
important thing to recognize is that user is usually required to determine
all of the material and geometric properties for all sections that will be
used in a model. In my software, to describe one material, like aluminum, I
have to enter data into 4 fields, plus one for a description. To describe a
section geometry, I need to enter 7 calculated values plus the description.
The user takes his choiced of "finite elements" and connects them together,
describes where the entire structure will be constrained (another word for
held in place) then applies loads to them in the X, Y, &Z axes.
Then runs the analysis and reviews the results to see if they make any
sense, makes manual calculations to verify that the analysis makes sense
(I.E. he didn't tell it to do something stupid, or something it couldn't
do).
Then, reviews the results, runs a few more models to see if it is behaving
correctly, then sifts through a pile of data to find the stuff he is really
interested in.
FYI, the people who usually run this software are dedicated experts in its
use. They are FEA professionals, which I am not. I was lucky enough to have
the assistance of 2 pro's that we use at work to help me get over the hump
in understanding how to use and interpret the results. It only took me a
year to get comfortable using it. On top of that, I'm using a very simple,
stupid, piece of software. Really bare bones, and with real limitations that
need to be fully understood. I've since run enough stuff and had the runs
verified by the pro's (costs me lots of beer when they come to town) to feel
that I am getting good information when I run inside the limitations of the
software.
There are two basic types of FEA code. One is linear and the other is
non-linear. Linear analysis is only accurate when the displacements or
deflections in the structure are small. The non-linear codes can account for
the significant geometry changes that occur when the segments of the model
move alot and change angles etc. I use a linear code because it is the
simplest and costs the least. That means that I have to be very careful when
interpreting results to figure out when I'm in the grey zone. Then I have to
pull some cheater moves to pull the analysis back into the accurate zone,
where I know it is thinking like I am. Then post process the runs to derive
the really important data and make the corrrections for the cheater moves to
get back to the few reliable values I need.
The entire genre of FEA codes have experienced a tremendous boom in
popularity over that past 8-10 years. It has moved from a questionable
curiosity to a valuable, reliable, technical tool, when used by capable
fingers and minds. Kind of follows the development of the NEC codes! There
are as many, if not more, people having the same discussions about the
accuracy and limitations of the FEA codes as there are about the NEC codes!
They just happen in different worlds that don't often collide!
How much does it cost?
The first version I tried to use was given to me by a friend. It was
shareware he found on a bulletin board. After I determined that I could
really do something with it, I upgraded to an enhanced version for $100. My
main Pro consultant runs software that costs around $18,000. It does
non-linear, has parametric modeling capabilities, has optimizing algorithms,
and stuff that I don't know about. When I have him run a verification model
of something new I'm trying to do, it takes me about 1-4 hours to set up the
run, 10-180 seconds to run the analysis, and another 1/2-1 hour to get the
info out of the results. He does the whole thing, including a fax in my
direction in about 10-15 minutes.
After I get sorted out on the first runs of a new project, It only takes me
1/2 -1 1/2 hours to run each iteration.
I'm sure you could find software that costs anywhere between the $100 and
the $18,000.
Do they have a web page?
The big expensive guys do have sites. Look for COSMOS, ANSYS, NASTRAN, but
you are wasting your time, cuz you can't afford it!
For the program I use, I don't know, haven't looked for it.
Here's the info to contact them:
Grape Software Inc
259 Columbia Drive
Winnipeg, MB, R3Y 1G7
Canada
204-489-9327
This is old info so I hope it is still good. They are really good capable
guys. I think they are associated with a university and use their software
as a learning tool for their engineering students.
Warning!
This stuff is not for "amateurs." The FEA world has its own unique
nomenclature. You will learn a second language before you get a useable
result!
I thought I started trying to use it with a fairly solid mechanical
background, and ended up getting a year (part time) of additional education
before finally getting to my original objective of designing my towers. Good
thing I wasn't in a hurry! Really happy to have learned something new along
the way, that's always fun!
73, Kurt
--
YagiStress - The Ultimate Software for Yagi Mechanical Design
Visit http://www.freeyellow.com/members3/yagistress
--
FAQ on WWW: http://www.contesting.com/towertalkfaq.html
Submissions: towertalk@contesting.com
Administrative requests: towertalk-REQUEST@contesting.com
Problems: owner-towertalk@contesting.com
Search: http://www.contesting.com/km9p/search.htm
|