Hi Michael,

Have been playing around with loft for ship hull creation as a way to build a complete hull with as few joins as possible. The current hull for my model of HMAS Sydney has lots of networked surfaces. This has mostly worked with quite a lot of tweaking but when rendered and on close inspection still has some visible seams between surfaces.

I started experimenting with a loft which stops at about the point where the rudder is, a difficult shape common to most ship hulls. From there I networked. Unfortunately, the networked sections are more visibly "not right" when examined next to the smooth lofted section. Next I tried an almost full hull length loft, stopping just short of the stern. This is almost exactly what I want but with a couple of problems.

The first attachment shows a loft on loose setting. In profile the shape is almost perfect with no strange divits or wobbles. Very nice. Unfortunately, in the second attachment you can see the same loft in the other plane doesn't match at the edges. The third shows similar issues at the stern. A question: Would it be possible to have a setting for loose lofting which forces the loft to maintain alignment with the loft's outer edges? Perhaps it mights be possible to select edges which loose loft must follow, sort of a "loose loft with exact profile"? That would be ideal for this.

Next I experimented with full length loft on "normal" setting. Fourth attachment shows a much better edge alignment with adjoining surfaces (though probably still not actually "joined"?). Unfortunately "normal" loft doesn't follow the desired profile as well (attachments 5 & 6).

My preference is to use loose loft. To do so requires rejigging of the other curves and resurfacing.

Sorry for the long post. I'm interested in everyone's thoughts on this, particularly if there is an option I've missed or a change to my workflow which would help.

Hi Mark, thanks for showing the results of your experiments!

But it's pretty hard for me to follow along and also be sure I understand what is happening just from a text description and a screenshot.

It really helps me to understand what you are talking about if I can also examine the 3DM model file, so I can do things like zoom in, rotate around to look at things from different angles, repeat the loft with different experimental options, stuff like that.

Is it possible for you to attach a 3DM model file for each of these situations you are showing screenshots for? Ideally if you can have it zoomed in on to the lofted surface so that I can also delete it and reloft it to repeat what you did, something along those lines would help out a lot.


> Fourth attachment shows a much better edge alignment
> with adjoining surfaces (though probably still not actually "joined"?).

Normally a regular Loft should be joinable to its source curve, so this could be a bug.


re: Loose loft with fixed ends only - I understand what you mean by that but unfortunately I can't immediately think of any way to make that happen - the way Loose loft works is that it just takes the control points of the curves as being the control points of the generated surface. It's not really very feasible to do that only for some curves and not others because the surface structure is more of one unified thing, like all the control points in the U direction of a surface need to have a similar structure.

I have some ideas in the future on some possible options for Network, so it would also be good if I could see a 3DM file with your Networked version in it so I could select the curves and repeat the Network construction over here as well.

- Michael

Hi Michael

>Is it possible for you to attach a 3DM model file for each of these situations you are showing screenshots for?

Done! I've attached unlofted curves, lofted loose & lofted normal. Let me know if you need anything specific.

>Ideally if you can have it zoomed in on to the lofted surface so that I can also delete it and reloft it to repeat what you did, something along those lines would help out a lot.

Not sure what you mean here. Do you want me to do some specific screenshots of what I am attempting?

> Fourth attachment shows a much better edge alignment
> with adjoining surfaces (though probably still not actually "joined"?).
>>Normally a regular Loft should be joinable to its source curve, so this could be a bug.

No need for concern here. I added the "?" because I couldn't initially get the upper surfaces to rejoin with my new loft. I separated the already joined surfaces then rejoined everything and it worked fine.

>I have some ideas in the future on some possible options for Network, so it would also be good if I could see a 3DM file with your Networked version in it so I could select the curves and repeat the Network construction over here as well.

Sounds interesting Michael. My networked version has a ton of additonal work done on it (plate extrusions and booleaned portholes etc). It might be difficult to get anything useful from it. Also, hull half alone is 16mb! I might be able to find the hull just before the extrusions were done. Let me know if you need it.

Hi Mark,
WHile your waiting for his answer I just wanted to throw out that i interpreted this as:

If you zoom to the part you are refering to in the model, then save the file, it will open zoomed to the area you are refering to.

I wouldnt have to hunt and twist to know what part you want me to look at. :O

Hi Mark, thanks for posting those, that should help to clarify things. I'll try to check them out later tonight.


> Not sure what you mean here. Do you want me to do some
> specific screenshots of what I am attempting?

Well I just meant that hopefully I would be able to know which surface was the result of the loft so I would know how to recreate it.

If you attached something like a big model with 100 parts in it, it could possibly be difficult for me to know which piece of it you are talking about.

But if it is kind of streamlined down with extraneous parts removed, or having the particular loft framed nicely in the 3D view that can help me to know which piece is the one you are talking about.


> I might be able to find the hull just before the extrusions
> were done. Let me know if you need it.

Well, not if it is not convenient. It just may be difficult for me to understand the Network problems you were mentioning without being able to see the particular surface and repeat the Network command on it over here.


My guess is that I won't be able to actually do anything about this stuff but with some examples it can help me to at least get some gears turning to maybe be able to improve it in the future.

- Michael

Hi Mark,

Just had a quick look at this, if I may, the approach I would take would be to use a combination of Network and Blend ie. Network every second panel and use Construct>Blend between them to get tangency.

That's my first initial thought without getting into it deeper.

Hope this Helps

Thanks guys.

I've attached a screen capture of the section I am attempting to loft. I've now tried it full hull length right round the stern using both loose and normal. It is really close using either option. It would be a really smooth hull! Also, a one piece surface would make trimming and then extruding the "plates" a doddle. The networked version created extrusions with internal surfaces that I had to hunt down and delete.

>My guess is that I won't be able to actually do anything about this
>stuff but with some examples it can help me to at least get some
>gears turning to maybe be able to improve it in the future.

And that's one of the great things about this software. Even if there is no way to achieve this I am impressed that you give consideration to ideas from your user base. Doesn't happen much in the software world.

Hi Danny,

>Just had a quick look at this, if I may, the approach I would
>take would be to use a combination of Network and Blend
>ie. Network every second panel and use Construct>Blend
>between them to get tangency.

Hadn't even considered that as an option. I just tried it but couldn't get the blend to work (blend is one of those commands I struggle to understand). In the attached, should the selected curves blend or am I thinking about this all wrong? The second attachment shows what I got. I don't think I'm using blend correctly at all!

Hi Mark,

> I just tried it but couldn't get the blend to work...

To blend between two surfaces you have to pick the edges of the surfaces not the curve you generated them with, to do this, hide all your curves so you don't get confused between picking a curve and a surface edge, then pick the surface, mouse over the edge and pick again to select the edge, do the same with the surface you want to blend to and select Construct>Blend and a surface should be generated.
Play around with the options to see what affect they have.

If you need a mini tute let us know.

Cheers
~Danny~

Hi Danny,

I was on the right track. I didn't hide my curves and was selecting those instead of the surface edges. Must remember that for the future.

This method seems to have a different sort of limitation. The blend creates a gap as shown in attachment 1. This can be adjusted out somewhat but creates a curvature then at the bottom per attachment 2. Have I missed a setting?

My problem with the networked version is, I think, that the curves are traced from paper plans of varying quality which have been scanned, jpeged and coffee stained to within an inch of their lives. The accuracy of the resulting nurbs curves is questionable and any problems between the networked surfaces shows up at the joined edges. I have rebuilt my curves in the vertical which helps a lot. I did not try it in the horizontal. Perhaps I should have. If I rebuild all curves, will a hull as curvy smooth as the lofted version be the result? The problem in the horizontal that I can see is that if I rebuild the segments of the curve I get better but still wavy curve segments. If I join the segments and rebuild as an individual curve, the curve shape changes slightly, enough for the horizontal and vertical curves to no longer intersect. This would then require me to snap curve points.

Hi Mark I think I get it - if I understand correctly, you are doing a loose loft using these profiles:



But then the part that you want to nail down more is actually this side of it here:



Is that correct?

But basically Loft does not give you such specific control over the side edges like that, sweep or Network are the ones that give side edge control. But more pressure to constrain a shape tends to make for more compressed undulations in it as well which I'm sure you've run into a lot.


But do these pieces here have to be separate surfaces:



What about making a larger single Loft that includes those pieces all within it, that would help to make them touch properly. You can have a sharp corner in a loft profile, that's ok.

That would also involve making some of the rear sections be extended up somewhat extra, and then if need there to be a space in that area you would cut it out with a trim here:




- Michael

Hi Mark, attached is a 3DM file of the kind of thing I'm talking about - here you would do a Loft (Loft Style: Normal seems to match your top side curve good) with these taller profile pieces to make that upper part along with the lower part all in one Loft rather than as separate pieces.

It's ok to have a kink in the profiles, but it is also good in that case to make sure that each profile is made up of the same number of segments, so I split a couple of the back upper parts to make a segment arrangement that kind of continues those strips throughout the loft.

When you do a loft in one go like this with kinks in the profiles, the generated pieces will all be sure to touch each other.

So that would solve your problem of having gaps between pieces anyway...

- Michael

Hi Mark,

Had a better look at this Hull, and I'd like to correct my self in saying that this is more suited to lofting after the curves have been optimised like you said and some finishing off with some blends.
It was a nice exercise, some learning for all I think.

Cheers

That's the technique Michael! Works great. Why didn't I think of doing it that way?

This also works for that rudder area. I've extended the curves down and trimmed. With care I might be able to extend it right to the stern without need for any networking at all, for a one piece hull.

I've included a quick render and the result looks near perfect.

The key points I think are to extend and trim and to keep the same number of curve segments.

Thanks Michael and thanks Danny and Burr for your input.

That's great Mark, I'm glad that will work for you!

It tends to be a bit tricky to get that "build extended then trim" modeling instinct down, it just is a bit counter intuitive to think about building a part of something that you don't actually want at the end.

But it can greatly help to simplify the initial construction with a more regular topology so it tends to be a good habit to develop with NURBS modeling in particular.

Anytime you find yourself kind of sticking together smaller individual fragments and wishing they were all smooth to one another or behaving similar to being a single surface instead of separate pieces, that's a good warning sign that a "build extended then trim" approach could be worth considering.

- Michael

Michael has reminded me several times to try not to create the exact shapes you want all in one operation.

It's better to create a very close shape with unwanted areas and simply trim them away in a second operation.

Being that ship hulls and spaceship hulls are similar to a degree, I've learned a lot from this thread also! ;-)


-Will

From the Michael file
Something I don't understand : seems number of points are not accord to the sides?

Hi Pilou,

> Something I don't understand : seems number of
> points are not accord to the sides?

Are you just looking at the control points of the curves?

Some of those curves just happen to be drawn with a different number of control points in them.

There is no restriction for surfacing commands like Loft, Sweep, or Network that each profile curve needs to have the exact same number of control points in it. It is allowed to use curves that have completely different numbers of points in them.

MoI goes through some steps to automatically synchronize curves like that up to one another when constructing the surface.

- Michael

Hi Michael,

Unfortunately I've hit another snag.

The hull looks great until I try to do the very stern. At the stern the bottom points need to be on the hull's centre line. In the first attachment I have left the curves rounded (no corner point). The hull lofts the way I want it to. In the second attachment I have added a corner point and snapped it to the centre line. Attachment 3 shows the problems that result from doing that.

I am at a loss. I can't see any way to get the stern capped without having the rest of the hull come to the centre at some stage. I can't do that without a corner point snapped to the centre. I have tried this many different ways and have rather lost track of things now. What to do?

Hi Mark, can you post a 3DM file with your new curve setup in it that you are having problems with?

That would make it easier for me to check it out and possibly recommend a strategy.


So I know I just got done telling you lofting things in a larger sheet (and then trimming them), but if you have a piece that behaves more like a protrusion from a base shape, then that may not fit into that same category. It will probably be something like model the protrusion as a different pieces and either fillet or blend it in to the other piece.


The "build in one large piece" type strategy is for things that are more like one smoother sheet. It can change shape to a certain degree over it, but trying to pull out a thin protrusion is too much shape mutation, such things will likely need to be a separate piece for the protrusion.

- Michael