Do CAD format Boat Plans still require Lofting?

A line on a screen is not the same as a line on a mould loft, and the mechanism that the eye and brain use to determine the fairness and accuracy is also different.

 

By the way, I also wanted to mention the example of how commercial media, both written and televised, constantly applies the term 'artificial intelligence' in ways that imply it's realization is about to happen, or has happened in limited cases. Many people believe this, since it's repeated often, and routinely, is totally wrong.

The best kind of computerized systems, be it design software, or medical diagnostic applications, are interactive, requiring repeated input from users with expertise, in interactive modes that trend asymptotically towards a workable approximation/solution, which is always confirmed/determined by those with knowledge and experience.

 

Gonzo,
you really nailed it this time.

Software differences.
To my knowledge, I am pretty sure I may be a little behind here, but there are two (major)kinds of software.

One kind based on NURBS(non uniform rational b-spline curves - yes, this is pure name dropping here - I love the sound of it, just like Gonzo does the smell of his pencil), which allow you to create 3D free forms(Rhino, Maxsurf, DelftShip, Autoship, etc). Here you can build your hull shell and interior structure. But...if after you are done, you decide to, say make the hull narrower by 3", you have to then regenerate the interior structure, to fit the modified shell. This is no big deal if you design a shell with 10 frames and 10 longitudinals(gross simplification), but if you have designed a boat consisting of 200 major parts of various shapes, this is a hassle.
The other kind of software is based on Relational Geometry(Multisurf, Solidworks, Catia, etc.). In this case, when you adjust the hull shell to make it 3" narrower, all the interior structure you have previously designed will adjust itself automatically to fit the modified shell. Of course, you also get free floating hydrostatics and weight schedule with that, so you immediately see how you affected you boat.
Beat that the "Sweet Batten" tribe ;-)
There are newer ones like SpaceClaim(non marine) which are better in some aspects, but I am not familiar enough to judge.
Extracting the CNC data( not need for offset table - but Multisurf will give you one if you insist) is similar in all cases, by unfolding the shell(in case of compound curvature you get stress maps to know how to form the shell to achieve compound curvature from a flat sheet of metal - expensive, and saving it as a 2D DXF which can then be sent for CNC cutting.
Alternatively, the hole hull can be 3D CNC cut from a block of foam by a multi axis router, with superhuman accuracy.

 

Ad Hoc,
I am not sure what you mean by: "A line on a screen is not the same as a line on a mould loft". Can you elaborate?

 

That may need a bit of clarification David. Final release surfacing is indeed done on a computer, but most styling sensitive design work is still done in clay (at least at the big 3 company I work for and most that I’m aware of), scanned as a Polygonal surface, and then rebuilt as a nurbs surface. What I found over the years doing both digital and manual surface development is that designs executed in a pure digital form tend to be a bit stiff and lack grace. There is no question that CAD is a powerful tool for surface development, but even looking at a design on a 20’ screen, you are still looking at a 2D representation of a 3D object, it’s just not the same.

 

A cautionary tale here, Some years ago i took the lines off a 1905 SE Saunders launch i was rebuilding, a customer wanted a replica to put a steam engine in.
I lifted the lines with great care & as much accuracy as could be had. The body plan was drawn full size on good quality ply. Now the easiest & most sensible route would have been to loft the boat full size from this & correct as necessary.
Instead the customer insisted on inputting the offsets data into AutoCad so he could do calculations construction details etc.
Now he was very proficient in Autocad but it did not appear to have the fairing capability or tools that a good yacht or ship software design package has.
The wireframe view was obviously suspect but i could not get through to him that it looked unfair.
In the end two other people tried to fair the lines using the same programme & eventually it was pronounced fair.
By this time exasperated i had walked away & washed my hands off it.
The molds were cut out & erected & then it was found that the "perfectly fair" hull was far from it. In the end to get the customer out of the fertiliser i came back & faired the erected moulds by eye. In some places i had to plant an inch on & in other places take an inch off.
Nobody would admit or explain why it was so awful. I presume (& stand to be corrected if im wrong) that Autocad when faced with a line of points which were not fair in a boat or ship sense simply drove a fair line through them all bumps & hollows included. The moral is as always garbage in garbage out!
Incidentally the hull came out ok in the end.

This was what sparked my interest in delftship, onwards with the learning curve!

 

Jarek has confused two different aspects of software. One is the mathematics used to represent geometry; NURBS, etc. The other is to what extent and how different geometric entities are linked.

NURBS is a method for mathematically representing arbitrary shaped surfaces and curves, and is the most common method used in design software for arbitrary shapes. Many, but not all, nominally different methods of representing arbitrary curves and surfaces including Bezier curves and surfaces and Coons patches are directly equivalent to NURBS or a sub-set of NURBS. Rhino uses NURBs. MultiSurf also uses a form of NURBS for arbitrary curves and surfaces. I'm pretty certain that CATIA also uses NURBS though they may also have different terminology. DelftShip does not use NURBS but rather a subdivision method. I'm not familiar with MaxSurf or AutoShip.

Software which enables setting relationships between entities so that a change in one entity causes a corresponding change in other entities, Jarek's example of frames moving when the surface is changed for instance, is frequently called "parametric" software. It can be implemented for NURBS as well as other methods of representing geometry. MultiSurf's "relational geometry" is one implementation of a parametric scheme. As Jarek mentioned this can be an advantage provided the changes which occur in the other entities are as desired. For example when the beam is narrowed should the house and cockpit become narrower by the same amount, should the house and cockpit stay the same width and the side decks become narrower, or should both the house and side decks become narrower? Which occurs in parametric software depends on what relations were established, explicitly or implicitly, as the design was developed. There may be added work to establish the relationships during the design process.

 

Some interesting points have developed in this thread.We all have experiences that encourage us to lean one way or the other and one thing I would suggest is that we all need to refresh our awareness of what is currently achievable from time to time in order not to lock ourselves into an attitude that may not reflect the state of the art.My favourite analogy is the true story of a man who was in charge of typewriter maintenance for a very large local company around twenty years ago,he believed he had a job for the rest of his working life.The advent of word processors and later computers made his highly developed skills redundant.
Ad Hoc's point about checking welded frame assembllies against a scrieve board doesn't seem like a strong argument in favour of traditional lofting as the section shapes can easilybe derived from the CAD model and cut directly into the scrieve board from it.I will happily profess a lack of experience of building in metal and if there is more to the matter I will gladly learn a bit from those with more knowledge.
Post #41 contains the essence of the process these days and where Gonzo gives examples of the time it takes him to extract frame details,he doesn't give details of the time taken to paint the lofting surface,draw the grid and then fair the lines.It gets worse if the design in question has been drawn by a traditionalist with a pencil to the outside of planking.Almost nobody takes the time to correctly deduct the plank thickness,which might not matter in some areas of the hull,but does matter with a flared bow or full stern and it takes a while.Giving a man with a CNC machine a CD with dxf files of the parts required is probably a faster way to go.

 

Keith66,
I do not blame you for walking away. That was the right thing to do with a customer like that.
1. As you rightly point out, AutoCAD is not the right software for this kind of a job. Was not and is not.
2. Reverse engineering is a skill, it will not happen automatically by throwing the body plan into any software, even the dedicated marine packages. You just have to know how to approach each case.
3. Your measurements were not spot on(I assume you are human), for which you cannot be blamed.
4. The hull itself was not perfect - I assume it was built by humans, at the beginning of the 20th century.
AutoCAD has a number of ways of dealing with points, including driving a curve through it, but this is not the whole story. Actually, this is the best way to produce an unfair curve, using more than three points, as was the case here.

 

That's why I said "final surface development".

I worked for a big 3 auto company on vehicles starting with the very beginning of the programs, and was generally in the studio(s) multiple times per week. The company I worked for used a combination of clay and math, iterating between them. A surface would be worked in clay for a while, then scanned. A math model of the surface would be built from the scan, checked against criteria, and modified as needed and desired. The revised surface would be milled into clay for evaluation and further revisions. This iterative process typically would go through many cycles, first in scale then in full size before the surface was released for final refinement on the computer.

I've always been somewhat mystified as to why clay models are not widely used for boat design. Clay is relatively easy to manipulate, and it almost as easy to add as to subtract material.

claydog, any experience with looking at designs in full size 3D "virtual reality"?

 

Seems like were discussing symantics again... good computer modelling software contains tools which help the designer acheive fair lines because of the visual difficulty of working on a small screen. I wont elaborate as all software is different, suffice to say they have means to exagerate and display curve inflections, developable surface rulings, a multitude of different shading modes and more - just like a shipwright needs experience to use his tools to get a fair line, so does a designer in his software.
Im lucky enough to have drawn my own set of plans in 3D, exported the files to CNC router, and now building. Ive had no problems with the fairness of the lines, but this is also owing to my build method using automatically fairing premade panels...

But why so hung up on the fair lines thing? theres so much more in the original question that started this thread... A CAD drawing with CNC cut files can save the builder many hours of work simply cutting frames, parts, developable plates, mould stations etc... a paper plan gives you no such option. If you like working with paper, its simple to get your plans CAD plans printed, if you only have paper plans, you cant go the other way if you wanted to. So all in all, the digital plans give more freedom...

 

DCockey,
thank you for the clarification, I am not a mathematician. I think though, in case of Multisurf, the situation is slightly different as the software has to translate Multisurf native objects to NURBS for export. The program can also create NURBS entities but they have different characteristics than the native Multisurf ones. I never had enough courage to ask Dr. Letcher for an explanation.
On the second point, yes the various relationships have to be user defined, and it does get complicated, the payback is being able to "clone" one complete 3D model and create a new one that is, say, two feet longer, one foot wider, with a slightly different cabin sole slope, somewhat longer cabin, a smidge straighter sheer, etc, etc, in a day or two.

 

Jarek: I agree totally with you that the drudgery of designing can be avoided in many ways by using computers. However, I have seen computer generated designs where the interior was larger than the hull. That is where the lack of experience is a problem. I think that the ideal system is to loft by hand for cosmetic reasons and do the rest with computers.
Ad Hoc correctly points out that there has to be some way of double checking parts as they are manufactured. That is a routine part of any manufacturing process. A scrieve board is a fast way of doing that with frames and other flat parts. It is possible to use laser measuring devices to check pre-determined points in the structure too. That has been used by auto makers for decades to check during manufacturing and to verify alignment after repairs.

 

gonzo,
as with everything, garbage in garbage out. Computers do not think. We are far from the Startrek level. Believe it or not I have had otherwise bright people ask me: "can you tell your computer to change the boat from inboard to outboard..."or some such nonsense?
Quite frankly, I do not understand Ad Hoc's point about parts checking( I work in aluminum as well, although in smaller boats where the shrinkage problems are less of an issue); yes, aluminum shrinks and warps - you need to try to minimize it - pulse MIG strongbacks,etc.
Yes, you need to check.
Yes, the parts might still be a bit off, I believe, for larger vessels they are cut with a surplus margin that can be trimmed during fitting, but this does not constitute the case for lofting instead of CNC cutting. When you loft a large part, necessarily out of sections(aluminum is not cheap) and then weld it, it will shrink and distort as well.

 

It is basically down to what is termed Brain Computer Interface and HCI (Human Computer Interaction). In appreciating a curve for its aesthetics visually such as a car, a woman or simple lines on a piece of paper compared to the same function of appreciation/review, but on a computer, different parts of the brain are used. That’s even before considering the resolution of a screen compared to film/paper etc.

I’m not advocating traditional lofting nor CNC. I’m merely pointing out that once you have your “parts”, whether they be made and cut from a traditional mould loft or a purely by computer, once fabricated, they still need to be checked. A computer won’t give you that either, since ‘parts’ inside a computer environment are just a 3D shapes, it is not a fabricated shape; with all the changes that occur during fabrication. As previously noted, it is basic QA/QC.

So a completely computer generated 3D shape is never trusted in the Auto industry?

Several reasons:
Most boats require plate development, i.e. not complex curves. Only composite boats can be made with complex curves.

Secondly what is the purpose of the clay model, is it to acquire a more faired curved or a more aesthetically pleasing curve? If a more faired curved, this is what the mould loft, in the traditional sense, did and can still do at a fraction of the cost and time. If it is aesthetics, stylists/designers have their own means for such and is not naval architecture as such since it is used for visual styling of the above water shape, not below water, which is hydrodynamically driven.

Finally the time/money required to even make a largish size clay model and the ‘tools' to map this shape back into a computer is beyond the budget/time scales of most companies. This ‘function’ would therefore be required pre-contract, since immediately post contract you have production shouting..where are the parts! But once you start, and the client changes his mind….back to sq1. A car which is a mass produced product is not a boat. A boat is ostensibly a one off and reflects the clients wants/needs and often changes during build.

See above comment.