Hi Kevin,

> I've created the surface from the curves and as expected
> it has warped, buckled and twisted into all sort of unsual
> shapes.
> However the nose part of the aircraft is looking pretty good so far.

Yup, as we previously discussed if you try to make cross sections go through big changes, like pull an elongated piece out from a single surface construction, it will tend to be much harder to control and generate things like that.

So I would recommend not doing that since it causes problems.

Notice how the front part is formed nicely where the cross-sections are not going through any major changes in shape? Keep that system going towards the back, focusing just on making a smooth tube shape and do not attempt to build the protruding tail fin out as part of the same single surface construction.

Instead, get a smooth body piece built first, then build that tail fin as a totally separate construction, then they can be blended or filleted together.

- Michael

>you can also use "add point" on either side of the corner point, then delete the corner point in the middle.

Thanks burrman. That is the answer to the problem !

>If you don't have transparency you can do that by separating them slightly in depth so that they
>are not stacked right on top of one another, some details here:
My top and bottom blueprints are not stacked right on top of another, however I can't model stuff in MoI without using transparency as I can't see my blueprints while modelling.

>If they are transparent then there currently isn't any way to have them turn off automatically, but it should be possible to set up a keyboard shortcut that would toggle a >particular image on or off with a single keystroke, let me know if that would be helpful.
That would be very helpful, what script do I have to enter to get this to work ?

>It is on my todo list to handle reversed view images work more easily, I think that probably the best way
>would be to remember if an image was placed in a reversed view or not, and if there is an image placed
>in both the regular view and the reversed view I will skip drawing the one that is opposite from your current view.
>That should make it work with transparency as well.
That sounds great, look forward to it.

> Problem 2. Cant join the surfaces created together which I need
> to do to create a good mesh when exporting out to my
> polymodelling software.

>Typically that means that they do not have edges that are close enough.
Yes, seems to be a problem with software as that edge was used to create the new surface so why isn't it a perfect match suitable for joining ? Possibly it springs apart due to tensions in the nurbs surfaces.

>Yup, as we previously discussed if you try to make cross sections go through big changes,
>like pull an elongated piece out from a single surface construction,
>it will tend to be much harder to control and generate things like that.
To be honest I don't believe I am asking the software to go through big or radical changes in cross sectional shape but a subtle change of curvature. However it is still buckling and twisting so I'll give this model a go with just using a tube for the fuselage and add the fin, tail hook fairing underneath and cockpit as seperate surfaces and see if that works.

Thanks everyone for your input, I'll have another go at this model as suggested by Michael.

Hi Kevin, re: background image toggle:

> That would be very helpful, what script do I have
> to enter to get this to work ?

Use this script - this script targets a specific image name, change the name that is at the front of the script inside the ' ' quotes to target a specific image.

script:var imgname = 'F9F-5 Side.jpg'; var images = moi.view.getBackgroundImages(); for ( var i = 0; i < images.length; ++i ) { var img = images.item(i); if ( img.fileName.indexOf( imgname ) != -1 ) img.hidden = !img.hidden; }

If you set that up under Options / Shortcut keys, that will allow you to toggle a particular image on or off with one keystroke.

- Michael

Hi Kevin,

> Yes, seems to be a problem with software as that edge was
> used to create the new surface so why isn't it a perfect match
> suitable for joining ? Possibly it springs apart due to tensions
> in the nurbs surfaces.

It looks like that particular edge was fine, however the one on the bottom that is supposed to touch it does not touch it. Here I deleted the surfaces and zoomed in:






When you do a Network the surface that is generated will be kind of an average of those curves, if you have a bottom piece that does not touch the side one, then the averaging action will pull the created surface away from the side.

To fix this, make sure that your initial curve network has ends that match up exactly and then the generated surfaces will also match exactly and should join together.


If your curve network has gaps in it, then the surfaces created from them will also have gaps.


Building the main fuselage as one long simple tube instead of a patchwork of pieces would probably avoid this as well though.

- Michael

Hi Michael

Thanks for the script, that will be very useful to me and easy to change the names for each project.

Ok cheers for the explaination about making sure my curves touch exactly to avoid joining problems later on. I wasn't aware that creates problems.

Ok
I've redone the fuselage as a simple tubular revolve using the side profile of the aircraft. I've then add the tailhook fairing to the underside of the plane and the two surfaces join Ok. So far so good. I used a two rail sweep to create the fairing underneath the aircraft and then the scaling rail.

Everytime I use the scaling rail I get a little tear in my nurbs surface just near the end. Is this a display error or likely to cause meshing problems later on ? (see attached image).
edit: Just did a check the tear does create meshing problems. I'm pretty certain the scaling rail has a bug as my curves all look good.

Sometimes it may help to see other surfacing models. Here's a link to a good 3dm of a nice car creation. You can download and evaluate it.

http://moi3d.com/forum/index.php?webtag=MOI&msg=1408.1

Thanks Burrman

I've now downloaded Mercedes and I'll have a good study of it, there may well be things in it that will help me on my project.

Hi Kevin,

> To be honest I don't believe I am asking the software
> to go through big or radical changes in cross sectional
> shape but a subtle change of curvature. However it is
> still buckling and twisting

Yes I've noticed that you expect for this to work, but from the standpoint of the mechanisms involved that would be considered a big change.

Maybe "radical change" is not the right word for me to use to describe the problem - really the underlying problem is when you look at a shape and use your human intelligence to match different features of it together, the surfacing mechanisms just do not have the same matching intelligence built into them so they are giving you results that you do not expect.

Take these shapes for example:



Then consider these different regions on the curves:




You're expecting for the green region on the left to match to the green region on the right profile, and the blue region on the left to match to the blue region on the right.

You'd like for the blue part to kind of stay nailed down and only the green part to sort of stretch up.

The surfacing methods just do not have this sort of subtle pattern matching between changing parts and automatically knowing which pieces to nail down and which to stretch. Instead they will just average the shapes together more evenly which will tend to produce results like you're seeing.

You can't rely on the surfacing tools to do this kind of "sub region of single curve" pattern matching between changing things for you, you have to break things into separate constructions so that you more directly control the matching. That's why the fin needs to be done as a separate piece.

Does that help to explain it a bit more?


Really your project here is not something that I would call a slam dunk to do with NURBS instead of polygons. With NURBS modeling you definitely have to break things down into components and it can be difficult to identify the components in something that has a totally smooth skin at the end.

Wanting to have an output of a totally smooth blended skin is starting to get quite similar to construction of organic shapes like a face for example, it is moving towards an area where polygon/sub-d type techniques can be easier to work with.

Similar to a car, it is kind of a somewhat gray area that I would not term is automatically easier to do in NURBS. To do it with NURBS will involve using much more of the "low level" and advanced NURBS toolset of creating all custom formed surfaces, trimming, blending, working a lot at a surface level instead of a solid level. The things that go really really much more quickly with NURBS than polygons is when you're leveraging all the solid modeling type tools which is not really the case here.

At any rate, in this case and in the case of doing a car, the strategic approach is very important, some strategies like trying to build things incorporating more than one "feature" all at once is not likely to work very well, it is too hard to control that.

I can understand that it would be cool to just draw a few cross-sections through the whole plane including several features (such as fuselage and fin both) and then just do a loft or sweep and have a perfectly formed result spit out automatically... Unfortunately things do not work that way, the sort of "artifical intelligence" pattern matching is just not in there to make that happen. You've got to break things down into components, and recognizing the components is a big part of the strategy that is needed to make this kind of model work right.

It is definitely a difficult model to work with, it is not something that I expect is going to be a really easy and quick thing to do, although once you get a workable strategy planned out it can get a lot easier...

For things that fall more directly into the solid modeling path like more mechanical parts, stuff that may have edges in it instead of a totally unbroken smooth skin, those are the kinds of things that are much easier to do with a NURBS toolset, this does not really fall into that area though.

- Michael

Hi Kevin,

> Everytime I use the scaling rail I get a little tear in my
> nurbs surface just near the end. Is this a display error or
> likely to cause meshing problems later on ? (see attached image).

That looks like the kind of dropoff that can happen if the scaling rail did not traverse the entire length of the sweep, it basically can abruptly stop influencing the scale at that spot.

Here's an example with a one-rail sweep. Here is a one rail sweep with the scaling rail off to the side:




When the scaling rail is applied, the profiles will scale out away from their normal plane until they hit that scaling rail:




If the profile is not able to find any intersection with the scaling rail, it will just stay its normal size, which may cause an abrupt change in shaping.

Since the scaling rail is applied in this manner, it is not a bad idea to extend it a ways past the start and end of the sweep so that you know the entire zone of the sweep is totally covered with it.

If that is not what the problem was, I will probably need to look at the .3dm model file to see what is happening.

- Michael

Ok thanks, I'll double check that there is not some miniscule distance between my scaling rail and curves which is causing the problem. I've had it happen quite a bit while modelling so I need to spend more time checking those osnaps snapped to where I wanted them to.

Just an observation.
Much of this seems to have been covered in a forum thread back in January.

Has anyone been making reference to that thread?

Brian
(One of the images I posted on that thread)

Michael

Even after breaking my model down into sub components I'm still getting warping, twisting and deformation of my surfaces. To break my model down into even smaller components (to keep nurbs happy) makes it extremely difficult for the user to get all his curves right so that the shape of this aircraft is captured especially once the surfaces are trimmed there is no mechanism to adjust or tweak them. Also the more sub pieces that are created the more shading errors I get when it is converted to a mesh in 3DS Max. By shading errors I mean the joints between the surfaces are visible depending on the light direction.

Thankyou for taking the time to write such a detailed analysis of nurbs v poly and whether this project is right for nurbs. You've confirmed what I suspected, it is not.
I was really hoping that it was so that I could save time and model more accurately with nurbs. As you can see, I am already a skilled polymodeller http://www.websitetoolbox.com/mb/felixdk?forum=107238 and was hoping nurbs would be less time consuming.

At this point I head back to polymodelling unless some ground breaking piece of software eventuates in the future that makes it easier and quicker. Some might say dream on !

I should however still be able to use my investment in MoI to do some of the detailing of the aircraft. For example on the F9F it has a speed break underneath with about 100 holes in it which is almost impossible to do with polys so nurbs will do that easily. MoI will also come in handy for doing cockpit, landing gear details where it is far quicker to boolean stuff.

Thanks for the post Brian. I have seen that model and thread before. It is a very simplified model of mk22 spitfire and only vaguely resembles the real thing. It is relatively easy for me to create a similar model.

There is however a huge difference between making it look like the real thing and vaguely resemble it. Making it look like the real thing is doing stuff that nurbs doesn't like to do unfortunetly.

Hope there is something in this thread for other users on their own projects.

.

kevjon

Sorry, but I think you are missing something about Nurbs.

.You might like to have a look at this.

http://gallery.rhino3d.com/Default.asp?language=&g=7

(Michael was involved with the creation of very recent Rhino app versions you may find. MoI is his "Encore")

Brian

The underlying thing in this thread has been the need to use NURBS properly to get the result.
Trying to model polymodel style with NURBS just wont work. There has to be A change in technique and approach. If a polymodeler wants to use NURBS they have to LEARN!

It's not a NURBS problem. The REAL airplane would be modeled in NURBS by a skilled NURBS modeler and would take a sufficient amount of time. You cant just start "throwing curves around" and wonder why it doesnt work, then make statements about the problem with NURBS. It's not fair to the others who view this forum or are experienced enough to actually model an airplane with a NURBS modeler.

With some time invested one could become skilled enough to model an "AIRPLANE", but lets be realistic.

In a sense,

If you want technical data files with an ability for a machine to MAKE something accurately, you use Nerbs/Cad stuff.

If you want to make an "Image" of something, for some, poly modelling may be better/easier.
It's all in the mind set.
For me, coming from the likes of Hexagon poly modelling to MoI was like the world had become a modelling joy!
But I come (Very originally) from a background of engineering/ fitting and turning and mechanical drawing

Horses for courses.

Brian

PS I can/ have taken my MoI files into the likes of ViaCAD to produce viable dimensioned drawings "blueprints" which model makers could actually use.

Hi Kevin,

> Also the more sub pieces that are created the more shading
> errors I get when it is converted to a mesh in 3DS Max. By
> shading errors I mean the joints between the surfaces are
> visible depending on the light direction.

Yeah, this is related to getting creases between pieces that are built separately from one another, that's described in some more detail here:
http://moi3d.com/forum/index.php?webtag=MOI&msg=1398.18

Typically 2 surfaces that are just constructed adjacent to one another will not be guaranteed to have a completely smooth shape to each other. In the future I do expect to add some more tools to edit surfaces to achieve smoothness to existing ones, but that kind of "smoothness in all directions" is really the great function that subdivision surface modeling provides much more easily and automatically. I mean that is like the primary function of subd.

Currently in MoI to make pieces smooth to one another you would generally use a fillet or blend to build a small smooth connector piece, but that does not work so easily when you want more of a totally smooth skin like thing which is more of what you have in this situation - like more where the different components are very blurred and not really very distinct as different pieces.

That kind of heavy "blurring" between everything tends to be a indicator to lean towards subd for that task.


> and was hoping nurbs would be less time consuming.

It definitely can be if you use it with the models that are most suited for it, that will be stuff that has a more mechanical design to it rather than a kind of "smooth skin" design to it.

I think you've got the right idea - things like instrumentation, cockpit interior, landing gear, parts that have a lot of holes drilled in them, those are the projects where it will become a slam dunk to use NURBS instead of polys. Those are the kinds of projects where the focus on booleans makes things come together more quickly and the "smooth blended skin" main function of subd does not really help...


> At this point I head back to polymodelling unless some
> ground breaking piece of software eventuates in the future
> that makes it easier and quicker. Some might say dream on !

Eventually you will see more of a combination of SubD tools being combined with NURBS surfaces together - that will be where you will have making a subd cage be just another option in creating a base NURBS surface which can then be trimmed or whatever same as the result of a sweep. You can see this in action currently with the TSplines plugin for Rhino. That will be coming into play more frequently in the future but it will take a while before that will get into MoI, MoI is still primarily focused on the type of "boolean oriented" models right now and probably will be still for a fair while yet.


In addition to using MoI for boolean-oriented pieces, one other workflow that may appeal to you is to use some of the surfacing tools in MoI to create a kind of rough base form quickly, then take that into your polygon package and use some of the newish style retopologizing tools to rework it into a sub-d friendly form and then continue to work on it as polygons. That can give you kind of a jump start because you can get a kind of basic rough outline of your shape pretty quickly with the NURBS tools by drawing something like 5 or 6 2D curves.

That helps to leverage one of the strengths of NURBS which is that you can kind of whip out some concept forms quickly by drawing just a few curves.

That kind of combo could be worth looking at I think.


Anyway, sorry that the NURBS toolset did not deliver on what you expected. I think it will be a lot more in line with your expectations for the more mechanical type pieces.

- Michael

Hi Brian

>You might like to have a look at this
Yep I've seen those before. None of them are accurate compared to the real thing and I can see where the user has simplified things in those models to suit what nurbs can and can't do.

>Michael was involved with the creation of very recent Rhino app versions you may find. MoI is his "Encore
Yep, I'm aware of that. I apologise if I have offended Michael or MoI users. Michael is the one who has suggested to me a couple of times that I would be better modelling this aircraft with polys than nurbs. I persisted in the hope that I could get there with Nurbs and perhaps speed up my workflow.

Hi Burrman

I totally agree with you that in order to become a skilled nurbs modeller you have to put the time and effort into it.

>The REAL airplane would be modeled in NURBS
This aircraft is from the late 40's early 50's. No nurbs then, just slide rules and very clever engineers.
Modern aircraft are indeed modelled with nurbs, but not with Rhino or MoI. The toolset of those modelling packages is much more comprehensive and have a huge price tag to go with it. A price tag that is way, way, way beyond the hobbyist, let alone the learning curve needed to learn such complicated software.

>You cant just start "throwing curves around" and wonder why it doesnt work.
I did my best but unfortunately I just wasn't getting the results I wanted which is good clean smooth mesh ready for uvmapping. Just couldn't get those surfaces to go where I wanted and couldn't tweak them to shape either. All the files and blueprints are there for someone to show me how it should be done. I'm still keen to see if it can be done and learn from it.

PS I've edited my previous post of this thread as I have no wish to upset anybody. Nurbs modelling of aircraft is however not for me but that is not to say it is not the perfect solution for someone else.

Hi Kevin,

> Even after breaking my model down into sub components
> I'm still getting warping, twisting and deformation of my
> surfaces.

Just another note looking more closely at your image there - I don't think this will really change any of the key points, but that is not really still the correct "componentization" strategy for what would be required to do this well in NURBS.

You've still got a very polygon-oriented mindset there where you want each component to end up having an "edge to edge" match with an existing piece.

That's not how you want to approach the components with NURBS, instead each component is going to be more of an extended piece that punches through the others, then you trim and fillet or blend them together to make your final connections.


So you've got this where you have stuff laid out constructed edge-to-edge:



NURBS componentization would instead be more like this (just very quick rough sketches for the pieces here):

Fuselage piece:



Canopy piece:



Tail fin piece:



All components:




Notice how each component is a fully independent individual extended piece and is not hanging directly off the edges of another piece.

Then those get booleaned or trimmed, which then creates a sharp edge curve between the components, and then you use either fillet to smooth the sharp edge, or you trim back a bit of a gap and use Blend.

That's more of the kind of componentization that I was trying to get at, sorry I did not make it clearer. Notice how the pieces are constructed separately and there is not a huge effort spent on building them to the exact same final edge - instead those edges are created when you intersect the pieces.

You seem to be avoiding intersection in your approach - the big difference with NURBS is that you want to use intersections into the approach.

Anyway, just some more information... But you seem to really want to control the edges directly - to build the surface directly by controlling all "skin" more directly instead of using intersections, that is certainly more suited to subd instead.

- Michael