I think even this was solve-able if one really wanted to.
Assuming the mesh is overall dense enough for the particular manufacturing case one could intersect the mesh with a surface and locally re-tesselate the mesh to yield a precise enough curve. Similar strategies are used inside programs like Zbrush to provide higher local resolution for finer details.
I think even this was solve-able if one really wanted to.
If a grid (mesh) dense, then production it seems has to be normal - I so think. ZBrush works with a huge set of grounds:
ZBrush is used for creating high-resolution models (able to reach 40+ million for use in movies, games, and animations
This is boring topic/thread because most of us already know the strengths and weaknesses of meshes, NURBs, and T-Splines. I suggest that we start new, related thread about optimal user interface of the new SubD kernel that will be included in Rhino 7. In my opinion, the SubD kernel should not compete with Mudbox, ZBrush, 3D-Coat, Claytools, or Sculptris. Rather, it should be optimized for designing outer surfaces of aircraft and boats. When you design these surfaces, you want lots of freedom, but you also want some constraints so that all engines fit inside the surfaces. It is natural to constrain the surfaces with curves. The ideal user interface would have 3 types of constraints:
- The surface must pass through entire constraining curve and it must satisfy chosen continuity at the curve (G0=kink, G1, or G2).
- The surface must be inside constraining curve. Slider pulls the surface away from the curve.
- The surface must be outside constraining curve. Slider pulls the surface away from the curve.
Simply ignore threads you find boring :o)
What I see is a great piece of information: Dozens of experts explain a person on bidirectional Google translate in what way Nurbs and Meshes differ. It might well be that this passionate consolidated effort didn’t quite reach the thread opener – but this info should be quite helpful for others.
But hell froze over, and I agree with your supposition regarding the opportunity to craft Rhino SubD around accurate design intent!
I just want to decide on NURBS. Why SubD isn’t accepted in production?
SubD gives a smooth result (maybe it’s difficult to control), but the grid can have any density (as needed). Well, still unnecessary polygons.
Because most products can’t be accurately designed, developed and then engineered without modeling in Rhino, Alias, SolidWorks or Creo. Again, see the Shimano groupset crank from a while back. Do you know the design and engineering tolerances and production constraints on those things? Polygons? No chance!
So the main problem in accuracy?
The main issues are 1. control within the given design and engineering constraints and, in all cases I worked on, 2. downstream data handling for final prototyping and toolmaking.
If you are a student, you learn the basics in the relevant industrial design courses or, later, in the design studio of your employer, and then take it from there.
Ok, Rhino has one function, and SubD has another. However, the Sub is used more massively.
There is some very true statements in all threads. I can confirm @Lagom point of view, although the others are right as well. In my understanding the problem is that in the whole chain of product development you sum up all tolerances. I asked my more experienced workmates why we model in a thousand of a millimeter tolerance setting, when production has 0.02 mm in tolerance. Its simple, if the data model already has 0.02 mm the maximum error can be 0.04mm (+ others)
When working with highly reflective materials (like car-exteriors), this error can be visible and often is. At least in class A where a shape is determined by the flow of light and quality standards are high this makes a clear visual difference. If you model based on highlights, you notice a change when pulling a cv for 0.01 mm on car-part-based scale.
In addition to that (I’ll have to say it again and again…) a lot of engineering related analysis and features are based on a fact that you have a full mathematical description of your shape. Sure you can create a lot of this for polygon modelling as well, but is it feasible?
I simply don’t know if you can do the same with polys only, I rather guess no. But even if it can reach the same quality, you cannot replace a process developed over decades from hundreds of engineers, just because some start up tech tells you its possible with their new software wonder. Innovation is important but there is also a lot of lying involved. But you know, what’s not feasible for some industry, may feasible for others. It depends on the problem and on your preference.
In addition if you can do better as others, then its valid to use your process. If you create it better with poly’s than even a narrow minded person will accept it. However you need to achieve at least 101 % of the same quality as with traditional approaches. If you can’t do, its not better. Its not better if you to it in half the time with 90 % of the quality. At least this is the mentality I’m used to.
I’ve seen the use of mesh in 3D printers. What is the quality of output? Or is it also like production?
@Lagom, what do you think?
For example, ZBrush:
every printer has information on its resolution. Usually if you own good industrial printer you can achieve very good quality. Sometimes even better than molding. However its not fast enough (yet). So for serial production its not suitable. In addition to that we feed in surface data for the same reasons as already mentioned.
However when looking at this skull, this might be a better job for zbrush. I’m talking about classical industrial design, and this skull does not fit in that. There are always situations to break the rule, such as reliefs or very organic forms
If you give the printer NURBS and SubD - where it will be better? I meant it …
This doesn’t matter, since you convert it to something discrete anyway (this was already said a couple of times ). In practice this is up to the software, but in theory it doesn’t matter, Same if you print it 2d. In the end the process of printing is stepwise. Production meshes and modelling meshes often is something different. I still believe you can achieve better visual quality with Nurbs in “classical” product design, but as a conclusion you can’t really say what’s better or not. Its has pros and cons, and it depends on the problem. More accurate is Nurbs, because it has more “information” for discretisation.
You can’t. Practically all 3D printer software only accept mesh formats, mainly .stl. The meshes are sliced to form polygon regions which are then filled with material by the printer layer by layer.
Again, lacrosse sticks, Swiss watches, aircraft seats or bike group sets aren’t done with polygons, for reasons explained a million times before above. Is that so hard to understand? And, no, you don’t 3D print a hospital bed or a car bumper or an American football helmet. Are you deliberately obtuse?
We do (partially) print car parts for Volkswagen concept design and early prototype models. Our printers have a resolution under 0.02 mm and cost around a million of euro. The final design after an expert did some polishing sometimes looks better than some final series molding part. we however model with beziers and feed in surface data, which then is converted into a very dense mesh for printing.
please don’t offend me