Tutorial: Ship hull modeling using NURBS

For discussion of Computer Modeling. Virtual Ship building. Computer Aided Design and Drafting, CAD/CAM, CGI, and the techniques behind them.

Moderator: ArizonaBB39

User avatar
Jean-Paul Binot
Posts: 160
Joined: Thu Jan 13, 2005 4:37 pm
Location: Upcountry Thailand

Tutorial: Ship hull modeling using NURBS

Post by Jean-Paul Binot »

Hi all,

The recent hull modeling tutorial gave me the incentive to share my own attempts, that have followed a somewhat different path. My attempts at virtual modeling (for ships and other things) have been concentrating on using Rhino3D and more recently MoI, that are both outstanding NURBS engines.

The only real problem area when modeling ships with NURBS is to generate fair and smooth hulls. All the rest is fairly straightforward n-gonal shapes (boxes and cylinders mostly) that can be made easily from basic NURBS techniques, but ship hulls are semi-organic and can be rather hard to get right. Of particular concern (at least to me) is continuity. One does not want kinks, sags or any other imperfection. Hulls need to be smooth and fair, and accurate too, according to plans and blueprints of the real thing.

Here is one way I found to generate near-perfect smooth hulls from simple curves. This is a generic, somewhat WW2 Italian looking, cruiser hull, but the principle can apply to many other shapes. The model is made of three ovals that are waterlines: the keel, about flotation level and top. The sheer can be made by trimming later. All the rest are half stations, including stem and stern, at reasonably regular intervals. It is important to keep the curves as simple and clean as possible. With this setup, one can generate a lovely hull in one go of the Network command. Et voil�!

Image

Then, one can trim the top above the sheer line, rebuild the keel around the rudder and propellers, as the hull here is not smooth and continuous but displays hard angles etc.

Image

You do not need to have an oval shape for the bottom closed curve. It can be square, which is more appropriate for a warship if the hull has a keel that is integral to it (many WW2 ships have that). In that case, the network method I recommend generates a troubled area at the stern because the flat portion under the hull and along the keel goes on across the stern, which is definitely undesirable.

Image

The remedy is simple enough: the Network command generates two edges that materialize the sharp edges of the keel, all the way across the stern. This makes it easy to remove that portion and then blend the edges, which produces a smooth stern. Alternatively, one can model a pointier stern if so is desired. There are refinements too in the other delicate area on the other end of the ship: the stem. One should not want to make it pointy. Even the sharpest prows are not razor-edged. They are blunt to a point if seen from up close. Other possibilities include a bulge etc., for which specific techniques can be used.
Image
Jean-Paul Binot
User avatar
Jean-Paul Binot
Posts: 160
Joined: Thu Jan 13, 2005 4:37 pm
Location: Upcountry Thailand

Re: Tutorial: Ship hull modeling using NURBS

Post by Jean-Paul Binot »

I am working on a ship hull, a T3 Fleet Oiler of the US Navy, using my usual method of basically Network command to generate in one go the entire half ship. This works quite well but there is an unwanted kink showing in the bow area. There are several ways to handle this, such as using a blend between the two halves, like in the stern, where a kink is wanted by the way. But surely, Network should work well with such simple flowing curves. The curves over there are smooth, fair and all that and there should be continuity from one side of the hull to the other. Such continuity happens correctly with the flat bottom, but not at the bow, as the curvature there is too abrupt for MoI to handle.
Image
Image
Image
Image
Image
Jean-Paul Binot
User avatar
Jean-Paul Binot
Posts: 160
Joined: Thu Jan 13, 2005 4:37 pm
Location: Upcountry Thailand

Re: Tutorial: Ship hull modeling using NURBS

Post by Jean-Paul Binot »

So I reworked the model towards a more extensive but still fairly 'natural' or intuitive network of curves, with much better results. The bottom is very flat, and the bulbous bow is smooth as desired. There is still a very annoying but barely detectable kink near the bow, probably due to the pointy shape of the lowest closed curve. That should be easily fixable.

Image
Image
Image
Image
Jean-Paul Binot
EJFoeth
Posts: 2907
Joined: Wed Jan 21, 2009 1:51 pm

Re: Tutorial: Ship hull modeling using NURBS

Post by EJFoeth »

Of particular concern (at least to me) is continuity. One does not want kinks, sags or any other imperfection. Hulls need to be smooth and fair, and accurate too, according to plans and blueprints of the real thing.
What you really want is continuity of curvature. When we look at the hull we always make a hull curvature plot. It is very difficult in any generic CAD program to get a good smooth hull as they do not have to right tools for hull fairing (plugins don't really help). You rarely need the rational part of the spline but you can model a perfectly round bilge keel. At work we developed our own CAD software where we can fair a hull based on loci of radius of curvature points and we have four CAD engineers specialized at this task! It is a lot of work though; they sometimes need over a day to fair a hull that the shipyard thought was pretty good :smallsmile: We're going to build a plugin in Rhino so that we can use all tools and the GUI plus specific hull fairing tricks. Oh, we have a vacancy to help with this plugin...

We use the hull files to make solids and prepare the form for milling and computational fluid dynamics; some of these programs run into errors when your splines do not match by 0.00000000000001 mm... :Mad_6:

But for modeling, your hull looks great! Some sanding is allowed for the hobby! :thumbs_up_1:

Image
User avatar
DrPR
Posts: 1689
Joined: Sun Mar 07, 2010 12:01 am
Location: Corvallis, Oregon, USA
Contact:

Re: Tutorial: Ship hull modeling using NURBS

Post by DrPR »

EJ,

I have worked from several Tables of Offsets that were generated in the early 1940s, before computers or even electronic calculators. They worked on drafting tables with pencil and paper and used slide rules for calculations. There is a great similarity in the plans and tables, so I assume there were some formulae or standard methods for generating the numbers if the Tables of Offsets.

Do you know how this was done?

Phil
A collision at sea will ruin your entire day. Aristotle
EJFoeth
Posts: 2907
Joined: Wed Jan 21, 2009 1:51 pm

Re: Tutorial: Ship hull modeling using NURBS

Post by EJFoeth »

I am still 'old enough' to have used the draughting table that was required for some drawing exercises (I studied Naval Architecture & Maritime Engineering). I think they were rolled out of the universities about 8 years ago much to the distress of the older employees (I couldn't care less about draughting tables now). The lines themselves are the three views plus, as Fritz pointed out, the (or more) diagonal(s). That helps smoothing out the surface greatly. I haven't made any full lines drawing since that class exercise, I only made a few small modifications that I gave to a proper draughtsman to fair and for number crunching.

I suppose the table of offsets are a simple measurement from the lines drawings, but when calculating the surface of a cross section you could use the planimeter:

Image

(Wikieven has an entry!). It's a mechanical integrator that gives you the surface area just by moving one end of the device over a closed curve. (math checks out, this is really a brilliant device). There are even planimeters that give you inertia of a surface, something you need when calculating the stability of a ship. Of course, there are mathematical tricks to calculate this from your offsets (Simpson's rule for example). When you have the table of offsets you can get surfaces and integrate them to calculate buoyancy and center of buoyancy and so forth. I suppose that in the earlier days the slide rule was used, another mechanical device. I've never learned to use it but it works using logarithms, turning multiplication into addition (also a great tool). Normally you'd have two calculation offices calculating the same value and if all went well, they produced the same result!

Now it's "easier" as we work with surface files and increasingly more with solids and you can get very accurate numbers. Changing a hull shape can be performed by deforming the surface, not requiring any change to the basis hull form. Some of the CAD programs we use have simple strength calculations included. So we draw the model setup for testing, add material properties, estimate forces and you can see the deformation and stress. (The great part is that it is a "simple" CAD program that does this, not a dedicated strength analysis tool)

Some programs use the same techniques as the mechanical integrators to obtain results, although more modern methods have surfaced. Well, I use the Gauss-Legendre quadrature and that's still 18th century but very easy and fast to implement numerically nowadays :big_grin: find it fascinating how much is already so old. I think people in the 19th century experienced much larger changes in daily live than people claim we do now because they have a new iPhone. Anyway, although those 3D files seem great, we really spend a lot of money on our geometry files. Individual settings between CAD programs can result in forms that fair well in one program and not at all in the other. When clients are not careful, the surfaces that look continuous actually have gaps between them that are invisible to the eye. We create calculation meshes for analysis that look like this

Image

The meshing programs we use are all developed elsewhere (these are quite complicated) and if your surface isn't any good, these programs don't run or start meshing the inside of a ship. You can run special CAD fixing tools to weed out small errors to a degree but that is really a stop-gap measure. The meshing programs that give the best results are more sensitive. By using our own tools we can reduce the processing time greatly; we don't really want to spend 4 weeks polishing a hull to create a good calculation mesh! So, using CAD has its benefits but certainly its drawbacks.
Last edited by EJFoeth on Mon Feb 25, 2013 2:09 am, edited 1 time in total.
User avatar
Fritz
Posts: 125
Joined: Fri Aug 04, 2006 12:03 pm
Location: Salem, MA, USA

Re: Tutorial: Ship hull modeling using NURBS

Post by Fritz »

First off.. Jean-Paul..Thank you. this is exactly what I was hoping for. A discussion on how to make a decent hull with different methods.

Second, Evert, I love your shot of the planimeter! THAT brings back some memories! Or maybe not.. as I sit here and look at the one we still have in the office. And here's the truly scary realization: We still use proprietary programs for hydrostatics and stability that are based on the Simpsons rule. When we compare the results to the volumetric results from the output from the Cad program, they are within 8 cubic inches... for at 36 Lt vessel. Has to make you wonder just how much Simpson, Newton and Cote knew back in 1750. Or for that mater, how much uor current algorithms owe to them.
Best Regards

Fritz K.
EJFoeth
Posts: 2907
Joined: Wed Jan 21, 2009 1:51 pm

Re: Tutorial: Ship hull modeling using NURBS

Post by EJFoeth »

They knew a lot (Newton invented calculus, more or less), but these integration rules are exceedingly simple ;) Modern algorithms often work the same using a few tricks in between. I think we have some planimeters lying around as well... Nowaways I think people track a line with the camera of their smartphones and use GPS to calculate the path length and enclosed area :big_grin:
tone
Posts: 1
Joined: Sat Oct 31, 2009 8:59 am

Re: Tutorial: Ship hull modeling using NURBS

Post by tone »

Hi, Jean-Paul. Sorry to "email you" in a BBS thread, but the email address I had for you has gone stale and someone contacted me asking about your Potemkin model.

Please drop me a line in PM (I just registered here, and it did not afford me use of PMs).

thanks!

tone
Post Reply

Return to “Virtual Ship Modeling”