CAD Tool For Design Tiny Aircraft

[Le Chaud Lapin]

Hi All,

I have never really used a CAD program to design anything. I toyed
with AutoCAD back in 1987 but nothing more.

I'd like to design a small model aircraft, about one meter in length.
Even though it's small, it's still complex. There are many mechanical
pieces.

The most important feature I need, by far, is interdependencies of
paramters. [There is probably a fancy name for this]. In other words,
if I change an artifact of the aircraft from one material to the
other, I would like the change to manifest in every aspect of the
aircraft that depends on the material. I guess this is standard
feature. I would like to be able to program interelationships also,
preferrably in C++, but a scripting language will do.

The other important feature is that I need the tool to be "3D-aware"
from the outset. I'm hearing others in rec.aviation.piloting that
AutoCAD is not entirely 3D-aware. I don't know what that means, and I
am definitely not interested in finding out by trial and error.

I post to CCS because the presentation of SolidWorks on its website
gives me the feeling that they understand these issues and attacked
them head on, but any CAD package would do.

Finally, I prefer cheap over expensive.

-Le Chaud Lapin-

 

[Martin X. Moleski, SJ]

I have never really used a CAD program to design anything. I toyed
with AutoCAD back in 1987 but nothing more.

I'd like to design a small model aircraft, about one meter in length.
Even though it's small, it's still complex. There are many mechanical
pieces.

RCCad? It's 3-D.

http://www.rccad.com/

The most important feature I need, by far, is interdependencies of
paramters. [There is probably a fancy name for this]. In other words,
if I change an artifact of the aircraft from one material to the
other, I would like the change to manifest in every aspect of the
aircraft that depends on the material. I guess this is standard
feature. I would like to be able to program interelationships also,
preferrably in C++, but a scripting language will do.

I don't know whether it supports scripts.

The other important feature is that I need the tool to be "3D-aware"
from the outset. I'm hearing others in rec.aviation.piloting that
AutoCAD is not entirely 3D-aware. I don't know what that means, and I
am definitely not interested in finding out by trial and error.

A design program from a different standpoint:

http://www.davincitechnologies.com/AirplanePDQ.htm

It is CAD. I don't know whether it does 3-D.

Marty

 

[Bo]

Hi All,

I have never really used a CAD program to design anything.
I am definitely not interested in finding out by trial and error.

I post to CCS because the presentation of SolidWorks on its
website gives me the feeling that they understand these issues
and attacked them head on, but any CAD package would do.

Finally, I prefer cheap over expensive.

-Le Chaud Lapin-

My recommendation is to get an overview introduction through a
school's CAD training class. CAD is only one small part of any design
& engineering project.

Cheap: As is commonly said, you get what you pay for. SolidWorks at
$4000 US may be considered expensive. A more important consideration
in the end, is when you have to supply 3D Solids files to people you
collaborate with and manufacturing companies for CNC work. You will
need to send them files in the format they need to do their work.

Learning to model in a 3D CAD program will NOT give you the elements
of mechanical engineering, design, & aircraft engineering specifics.
I would expect you to spend even more time learning engineering
issues, than the CAD side of the project.

Bo

 

[TOP]

The word you are looking for is parametric. SolidWorks fits this bill.
There are several ways to make components in SW work together. At the
most basic level geometry in one part can be tied to that in another.
To this can be added equations that relate different dimensions. To
this can be added design tables which are nothing more than
spreadsheets built into a part or assembly of parts that drives part
dimensions. And to this can be added control from an external program
like Excel (the most common) to Access to a custom written API
program.

As to cheap, well the question there is whether this is a hobby
interest or a business interest. If it is a hobby, no doubt the cost
of SW at 3,995 plus yearly maintenance might be a bit high, but for a
business it isn't much at all. In addition SW requires a fairly high
end PC to do the kind of thing you are talking about.

As with anything as complex as airplane design (you didn't say it had
to fly, but I am guessing it will) to do the things mentioned in the
first paragraph will require some training, some practice and probably
more questions on this forum.

TOP

 

[Jerry Steiger]

I'd like to design a small model aircraft, about one meter in length.
Even though it's small, it's still complex. There are many mechanical
pieces.

Finally, I prefer cheap over expensive.

If your model plane doesn't require very organic shapes, then Alibre might
be a good choice and is cheaper than SolidWorks. If you need more organic
shapes and smooth transitions, SolidWorks would be a better choice, but it
can take a lot of work to get it right and the models will be less robust.
(You'll change a parameter and some feature far away may break.) If you need
really nice shapes, and don't have the time or patience to mess around, you
might need to go to the expensive guys, like CATIA and UGS. Jumping into
them from a non-CAD background would be really scary.

Jerry Steiger

 

[jon_banquer]

Personally, I prefer a pure geometry based modeler.

Users should have the option to use "a pure geometry based approach"
or a parametric approach in one package. Unfortunately at this point
they don't have this option. There is no reason KeyCreator shouldn't
add parametrics. Using parmetrics is often faster when creating parts
from scratch. Why Kubotek refuses to do this for KeyCreator is beyond
me.

The just announced KeyCreator V7 looks very disappointing because not
enough progress appears to have been made on direct dimension editing.

From what I can tell the only progress in this area is that faces can

now be angled when making direct dimensioning editing changes.

How robust do you feel "direct dimension" editing in KeyCreator is
now?

Where do you feel improvements need to be made?

Simple dimensional changes affecting many components may take longer, but
it's far easier to make large sweeping changes if necessary, or switch
to an alternate design approach.

Probably true but depends a lot on the skill of the user.

BTW, it appears to me that SpaceClaim is far ahead in regards to "pure
geometry" changes compared to KeyCreator. Too bad SpaceClaim insists
on a licensing scheme that will never work in machining job shops.
Same deal think3 tried. There is also no demo of SpaceClaim to try. At
least KeyCreator has a downloadable demo.

Has Bob Bean been removed yet? He's really holding KeyCreator back.

 

[Le Chaud Lapin]

A fully 3D program is, IMHO, a must for any kind of design, anything
else is silly. A parametric modeler, however (like SWX and many
others) can be very cumbersome to use... and I've used a lot of them
over the years. Yes, if the design constraints are set up correctly
from the start, minor changes can be ridiculously easy... but if not,
or if you don't have a clear idea of where you're going from the
start, you can find yourself boxed into a corner and have to start
from scratch.

Personally, I prefer a pure geometry based modeler. Simple
dimensional changes affecting many components may take longer, but
it's far easier to make large sweeping changes if necessary, or switch
to an alternate design approach. Most of my work nowadays is large
machine design (though my degree is in aero engineering), for which I
use KeyCreator (formerly Cadkey). Same price range as SWX, though,
which I don't define as "cheap".

I read all the responses and looked around the 'Net, and it seems that
SolidWorks, if not what I'm looking for, is create by people who had
the mindset I was looking for.

But now I am confused. I thought parametric modeling was good.

I program computers from time to time, and being able to change the
structure of a component and have everything that depends upon it
change accordingly is simply invaluable, so I cannot see why this
would be bad. That's precisely the behavior I want.

For example, in my miniature aircraft I envision, there is only one
fuel tank, and it's cylindrical, but its radius and length are a
function of several other parameters.

I am guessing that, like in programming, there is an art to
structuring the interdependencies so as to minimize likelihood of
running into dead-end that you mention.

Finally, I was really surprised to learn that parametric modeling was
not fundamental in all CAD programs. I cannot imagine what it would
be like to try to optimize a design without it. What do people do
without parametric modeling? Tweak every single component manually
during optimization phase?

[I am going to give Alibre a look also.]

-Le Chaud Lapin-

 

[jon_banquer]

A fully 3D program is, IMHO, a must for any kind of design, anything
else is silly. A parametric modeler, however (like SWX and many
others) can be very cumbersome to use... and I've used a lot of them
over the years. Yes, if the design constraints are set up correctly
from the start, minor changes can be ridiculously easy... but if not,
or if you don't have a clear idea of where you're going from the
start, you can find yourself boxed into a corner and have to start
from scratch.

Personally, I prefer a pure geometry based modeler. Simple
dimensional changes affecting many components may take longer, but
it's far easier to make large sweeping changes if necessary, or switch
to an alternate design approach. Most of my work nowadays is large
machine design (though my degree is in aero engineering), for which I
use KeyCreator (formerly Cadkey). Same price range as SWX, though,
which I don't define as "cheap".

I read all the responses and looked around the 'Net, and it seems that
SolidWorks, if not what I'm looking for, is create by people who had
the mindset I was looking for.

But now I am confused. I thought parametric modeling was good.

I program computers from time to time, and being able to change the
structure of a component and have everything that depends upon it
change accordingly is simply invaluable, so I cannot see why this
would be bad. That's precisely the behavior I want.

For example, in my miniature aircraft I envision, there is only one
fuel tank, and it's cylindrical, but its radius and length are a
function of several other parameters.

I am guessing that, like in programming, there is an art to
structuring the interdependencies so as to minimize likelihood of
running into dead-end that you mention.

Finally, I was really surprised to learn that parametric modeling was
not fundamental in all CAD programs. I cannot imagine what it would
be like to try to optimize a design without it. What do people do
without parametric modeling? Tweak every single component manually
during optimization phase?

[I am going to give Alibre a look also.]

-Le Chaud Lapin-- Hide quoted text -

- Show quoted text -

Suggest you find your way ASAP to www.kubotekusa.com and view their
video on direct dimension editing to see what can be done without
parametrics.

 

[TOP]

Creating a 3D parametric model can be likened to programming. Some
people make spaghetti code and some make nice tight robust models.
This is called capturing design intent.

It can also be likened to a database capturing spatial information.

You can iterate in a non-parametric world too. It takes good revision
control.

TOP

 

[Le Chaud Lapin]

Suggest you find your way ASAP towww.kubotekusa.comand view their
video on direct dimension editing to see what can be done without
parametrics.

Ok, I just watched the video, and I barely understand anything, as I
am an ignoramus when it comes to CAD. However, it seems that the "dumb
geometry", as the presenter calls it, allows "dumb dimension-based
editing", but after you are done fiddling with "witness lines, etc.",
you have your model, and nothing else.

Parametric modeling, OTOH, as I understand it, allows the programmer
to define constraints, and let those constraints rest in a sack that
is carried around with the model. If that is the case, I *absolutely
love* this feature! The power of this approach should be apparent, I
think, no?

Now I think I see what TOP meant in his response to your post, about
spaghetti code. I think the preference for the models depends on the
approach to designing systems. Some people think in terms of
relatives. Some think in terms of absolutes. I think in terms of
absolutes. I'd rather walk around in woods for 2 or 3 days working out
the kinks of a system in my head before I commit to anything, even if
I think I already have 40% of the answer. Only when I am sure that
the remaining irregularities are so minor that they will not impede
the march toward finalization of the design will I commit. Then I
employ the tool bear down upon my preconception of the system to see
that it is correct and to optimize it. I guess this is why I prefer
parametric. It seems like it is the right tool for the tightening
process during optimization.

Incidentally, that is the whole reason I've decided to fiddle with CAD
to make minitature plane, to see how much cost reduction can be
achieved by rethinking the system as a whole and not simply trying to
get better prices on conventional components.

-Le Chaud Lapin-

 

[Dale Dunn]

But now I am confused. I thought parametric modeling was good.

Parametric modeling IS good. But you must use the right tool for the
right job. Some people's jobs are not well suited for parametric
modeling, or they never learned to use it well, so they think it's bad in
general. As TOP said, a well constructed parametric model captures design
intent and allows easy generation of new or altered designs. People who
structure models poorly or need to deal with radical changes to design
intent can be better served by non-parametric modelers. For optimizing,
parametrics are not just good, they should be a requirement. And if you
need to make radical changes in a hurry with no regard to capturing
design intent, that can be done fairly easily in a parametric modeler by
making models that would otherwise be thought of as horribly structured.

But you'll be wanting well structured models again when it is time to
make fine changes. This may require that the design be completely
remodeled, but that can be mitigated by structuring your models in a way
that can support radical alterations. Look into techiniques like skeletal
modeling or so-called horizontal modeling.

Be warned, there is a steep learning curve, even for tools with a UI as
streamlined as SW. A tool that is easy to use, may also be easy to use
incorrectly. This goes double for complex shapes such as an airplane will
have. If you do choose SW, we can recommend some good resources and
examples for study.

I have built and flown some radio controlled model airplanes, and I think
SW could handle modeling something of that complexity fairly easily, if
the models are structured well. If not, it will be a painful experience.
Having said that, even with a well-structured model, there will be
frustration with rebuild errors and the like. Hopefully, R/C airplanes
are similar enough in scope to your project that my opinions will be
useful in your decision.

 

[jon_banquer]

Ok, I just watched the video, and I barely understand anything, as I

am an ignoramus when it comes to CAD. However, it seems that the "dumb
geometry", as the presenter calls it, allows "dumb dimension-based
editing", but after you are done fiddling with "witness lines, etc.",
you have your model, and nothing else.

It's very important to understand that parametric data does not get
exchanged
between different cad systems. What you get is a "dumb solid" when you
open
your model done in Solidworks in another system like SolidEdge. All
the design
intent / parametrics you established in SolidWorks will be gone.

Now I think I see what TOP meant in his response to your post, about
spaghetti code.

His example is one sided and doesn't give you the downside of
parametric
modeling.

http://management.cadalyst.com/cadman/article/articleDetail.jsp?id=325125&pageID=1&sk=&date=

"KeyCreator is a nonparametric application, but that isn't necessarily
bad. It gives users the freedom to do all kinds of things to a model
that they'd never think of doing in a history-based system."

I use SolidWorks everyday. I don't use KeyCreator. I'm not foolish
enough to think that a parametric / history based approach to modeling
is the only approach or always the right approach and unlike most
others in the SolidWorks newsgroup I'm not a product loyalist. Dana
Hague had some very valid points in his post to you.

 

[Le Chaud Lapin]

I have built and flown some radio controlled model airplanes, and I think
SW could handle modeling something of that complexity fairly easily, if
the models are structured well. If not, it will be a painful experience.
Having said that, even with a well-structured model, there will be
frustration with rebuild errors and the like. Hopefully, R/C airplanes
are similar enough in scope to your project that my opinions will be
useful in your decision.

Ooooh...rebuild. I like rebuild. I guess it's the same concept of
compilation in programming. I was wondering in my OP if there were an
equivalent with parametric models. After all, with all the
interdependencies..something has to be synthesized from the expression
of interdependencies, just like in programming.

This is my first design of anything using more than a ruler and scraps
of paper, let alone an airplane. I know it will take years, so I'm
going to just take my time and learn what I need to know as I go.

-Le Chaud Lapin-

 

[Le Chaud Lapin]

Creating a 3D parametric model can be likened to programming. Some
people make spaghetti code and some make nice tight robust models.
This is called capturing design intent.

I think I get it now. Hit me while walking my dog.

Example:

1. One dimension of fuel tank depends on required fuel capacity.
2. Fuel capacity depends on mass of certain parts.
3. Mass of parts depend on geometry and density of material of those
parts and load requirements, etc..
4. Load requirements depend on configuration of other structures.

And it would seem that there is a right way and a wrong way, and
again, finding the right way is more art than science. "Reaching" too
deep into model to extract parameters to be used elsewhere might be a
bad idea. Deliberate indirection and hierarchy would be important.
There would also be opportunity for circular references. Also, there
should be some kind of "on/off, part is there, part is not there"
programmability.

You can iterate in a non-parametric world too. It takes good revision
control.

I was wondering about this. It seems like parametric is a superset of
non-parametric in some ways.

-Le Chaud Lapin-

 

[Dale Dunn]

Ooooh...rebuild. I like rebuild. I guess it's the same concept of
compilation in programming. I was wondering in my OP if there were an
equivalent with parametric models. After all, with all the
interdependencies..something has to be synthesized from the expression
of interdependencies, just like in programming.

The concept of rebuild isn't quite like compilation... well maybe. The
analogy breaks down rapidly, but I think you have the concept. Just in
case the analogy is leading to a misunderstanding: parameters are the
rules by which the CAD program creates and manipulates the individual
elements that make up the models. After a change to one of the
parameters, the rebuild process has to be repeated so that the model is
correct to the parameters. Features (as in "feature based parametric
modeler") are very closely analogous to subroutines, if the rebuild is an
execution of code to build a model. The parameters would then be
constants within those subroutines. For example: a hole feature generates
a hole in the model based on diameter and depth parameters, when the
model is rebuilt.

If you want to do automated optimization, you are going to need to
program some macros for SW, maybe add-ins. There is an excel spreadsheet
from about 8 years ago that uses Excel's optimazation tool, but it's
limited to a single variable, IIRC. I saw in you OP that you were
interested in programming some relationships, but these should be the
first things that require programming on you part.

Also, I see in your conversation with TOP that you might be interested in
some parameters being driven by masses and volumes. You will need to do
some coding for that too, at least in SW.

Other CAD programs may not need macros for this particular task. Some of
the add-on analysis tools (FEA, CFD) have such features, but that will
get very expensive very quickly.

Anyhow, parametric CAD seems to be the tool you need, and you will grasp
the mechaniics of using it very quickly. A good coder should easily,
almost intuitively grasp what it is to organize the models well. I think
you will be ok, but there will be bumps along the road (such as managing
strange dependency behavior). SW will serve for your task, but something
like Alibre may be sufficient and much cheaper. Demo versions of most CAD
packages can be had, and standard advice is not to pay any cash until you
have taken several for a test drive.

 

[TOP]

This isn't quite aircraft, but a while back I wrote a VB macro that,
in conjunction with a very crude model of a concrete mixer, was
capable of predicting the CG of aforesaid truck with a fraction of an
inch of what the customer was measuring. For such studies you make a
rude and crude model that will update quickly but that captures the
intent of the study you are doing.

I will many times not even use a 3D model if I can do it in Excel.
Nobody likes this because you can't see the pretty pictures, but you
can run through lots of scenarios very quickly that way. And with the
solver in Excel being what it is there is little that can't be done to
get close to the right answer.

TOP

PS Also search the NG for configurator. There are some very good ones
out there for SW.

 

[TOP]

Your walk in the woods method is something that I run into every day.
Unless you have a very special mind, spatial relationships are very
difficult to imagine and solve mentally. You can get the topology in
your head, but when it comes to parts bumping into each other in 3D,
most heads can't get around it. 3D CAD brings you down to reality in a
way that even 2D CAD cannot do because in many ways 2D CAD is still a
mental excercise (Thank you Gaspard Monges). Frequently you will
encounter people with ideas that don't stand the test of 3D. This
isn't just an associate with a quick scribble on an envelope, but even
many 2D drawings are simply cartoons. What 3D CAD is, is a way to
simulate reality realistically (well up to a point). There is a
continuum:

1. 2D CAD (catches and idea, still much is required in the head)
2. 3D CAD (captures the 3D constraints, will it fit, etc.)
3. Kinematics software (will it move the way I intend, what are the
rigid body forces)
4. FEA/CFD (How will it deform, How will air flow over it?)

The first is probably the quickest route to getting a specific idea on
paper. The next one is more flexible and more time consuming. The
third requires the work of the second plus additional work and the
last also requires the second and perhaps output from the third to
give good answers.

Since SW starts with 2D sketches for the most part it captures much of
1 and pretty much all of 2.

TOP

 

[Le Chaud Lapin]

Anyhow, parametric CAD seems to be the tool you need, and you will grasp
the mechaniics of using it very quickly. A good coder should easily,
almost intuitively grasp what it is to organize the models well. I think
you will be ok, but there will be bumps along the road (such as managing
strange dependency behavior). SW will serve for your task, but something
like Alibre may be sufficient and much cheaper. Demo versions of most CAD
packages can be had, and standard advice is not to pay any cash until you
have taken several for a test drive.

I think I will try Alibre. I was sure SW was what I wanted, but
Alibre looks good from website too.

I've been to the SW website about 5 times today hoping that, by magic,
the price would drop to $100 for full version. Strangely, this has not
happened, so I need get an eval version soon.

-Le Chaud Lapin-

 

[Jerry Steiger]

1. One dimension of fuel tank depends on required fuel capacity.
2. Fuel capacity depends on mass of certain parts.
3. Mass of parts depend on geometry and density of material of those
parts and load requirements, etc..
4. Load requirements depend on configuration of other structures.

And it would seem that there is a right way and a wrong way, and
again, finding the right way is more art than science. "Reaching" too
deep into model to extract parameters to be used elsewhere might be a
bad idea. Deliberate indirection and hierarchy would be important.
There would also be opportunity for circular references.

This is going to be very tricky. There HAVE to be circular references in
your optimization. When you change the weight of the fuel tank, you have to
reevaluate the size and weight of all of your other components to account
for the new load. But now you have changed the weight of the rest of the
components, so the fuel tank needs to change again. If you are lucky, the
solution converges and you end up with a design that works. If you start
from the wrong spot, it might never converge.

The good news is that you seem to have the type of mind set that would allow
you to work through this type of problem. The bad news is that it is an
extremely complex problem that requires a lot of deep knowledge in many
areas of design.

Jerry Steiger

 

[Martin X. Moleski, SJ]

This is going to be very tricky. There HAVE to be circular references in
your optimization. When you change the weight of the fuel tank, you have to
reevaluate the size and weight of all of your other components to account
for the new load. But now you have changed the weight of the rest of the
components, so the fuel tank needs to change again. If you are lucky, the
solution converges and you end up with a design that works. If you start
from the wrong spot, it might never converge.

The good news is that you seem to have the type of mind set that would allow
you to work through this type of problem. The bad news is that it is an
extremely complex problem that requires a lot of deep knowledge in many
areas of design.

Speaking only from practical experience with RC models (~14 years total),
the envelope for small aircraft is extremely forgiving for ordinary
flight regimes. Most RC aircraft can double or triple their fuel load
without noticeably affecting flight performance.

Look at the college competitions for evidence. A few years back,
a weight-lifting competition was won at around 19 pounds of payload
for an aircraft powered by a plain-bearing 0.40 ci engine. Engines
like that are usually used to fly ~5 pound trainers with a wingspan
between 40" to 48" or so.

Drat. The American Institute of Aeronautics and Astronautics
website is down right now ... Here's a report on the 2006
competition:

http://mae.eng.uci.edu/aiaa/DBF2006.pdf

If you're trying to fly an RC aircraft across the Atlantic with
a gross weight of 5 kg (11 pounds), then you DO need to consider
a multitude of tradeoffs such as you describe:

<http://www.progressiveengineer.com/profiles/maynardHill.htm>

An engineer friend of mine likes to say, "One observation is worth
ten thousand expert opinions." Regardless of what the design
software predicts, the product needs to be tested in flight to
see whether the theories work. (I'm assuming that the initial
post in this thread was about a flying model.)

Something funny happens as you go down in scale. It has something
to do with Reynolds Numbers and the volumetrics of small aircraft
(volume decreases far more quickly than area). This means that
the power-to-weight ratio favors the model aircraft and that,
as a general rule, the ratio of the strength of materials to
G-forces increases. A 1/4 scale model (using 1/4 of the linear
dimensions) has 1/16 the area of the prototype and just
1/64th of the volume.

One way to get in the ballpark when designing a new model is
to select dimensions from aircraft that are already known as
good flying designs. The airfoils that work well for full scale
do not work well on small aircraft (as a general rule--Clark Y
airfoils probably scale OK; fighters and bombers from WW II on
generally do not scale well).

So a lot depends on the kind of aircraft the other poster
wants to build, whether it is supposed to fly, and what kind
of performance is to be optimized.

Marty