Blast from the Past: Is Parametrics History

[editor]: It’s interesting to go back and view points along the evolution of my conversion from history-based mogul to history-free supporter. When you read this, it’s important to remember that it was written in 2011.  At that time I was heavily involved in organic surface development using a tool that was in my opinion completely unsuited to that type of design. Still, companies contracted me to do the work. I think this type of tool is still needed in product development today, as evidenced by the spirited courting of Tsplines by multiple players.

The more overlap we get between products like your average NURBS modeler and tools like modo, Tsplines, 3dsMax, and so on, the better it will be for designers and engineers who need to produce organic shaped products for market. This is a valid market segment that is underserved by the mid-range market. No, it isn’t as big as the machine design market, but it’s bigger than the mold design market, which several CAD vendors have courted for years.

=========================================================

New CAD tools lately have been SpaceClaim, Moment of Inspiration, and even CosmicBlobs Pro, along with tools like T-splines, Free Dimension, modo, maybe some others. What do all of these have in common? Well, nothing, really, but in general, they center around making changes easier and concepting. Oh, and one other thing. None of these are parametric history based modelers (MOI has some history based concepts, I think). Most of these are for shapely parts, but SpaceClaim and MOI are also adept at prismatic parts.

So what is the industry telling us? I don’t know that either. You’d have to find “the industry” and ask him. Or her. or whatever. But, here is what I’m hearing, regardless if its only the little voices in my head that are saying it: Concepting and ease of change. Here’s what I’m not hearing: design intent.

You see, parametric history based modeling is great for making parts with intelligence. But what kinds of parts can you design with intelligence?  But what do you do when some of your parts look like this:

And the changes you want to make to them are more like silly putty than anything else.

This is a medical part, an interlaryngeal airway mask, which seals off against the back of your throat during anesthesiology procedures. It is blow molded from soft silicone. Developing geometry to fit the shapes of internal organs can be a challenge. This part is hollow, the outer loop functions as a balloon, but also allows gas to travel through the big hole at the end. It is shown half cut away and transparent here to help you see the internal features.

I’ve been working on a set of products like this for the past couple of years, and the changes that need to be made sometimes require either drastic rebuilds because they don’t fit the original “design intent”, or they might require an ugly hack just to get the geometry without worrying about the parametrics.

I’m going to guess that I’m not the only person that has this kind of problem. With shapes like this you really don’t want to have to deal with the overhead of parametrics. SolidWorks is so sensitive sometimes that just moving a couple of spline points causes a whole set of what seem to be unnecessary repercussions down the tree, things like a trim reordering the resulting bodies, resulting in a different selection to keep than you intended, or losing track of the edges of a planar surface selection, the list is really endless, these are just the problems I had today with this model. Of course this is just reflecting problems with the implementation of the technology, not the technology itself, but when it comes down to it, is there another tool which handles the implementation better? Would I stop having problems with parametrics if I were using say Pro/E? Probably not.

Still, I don’t think direct geometry editors can handle this type of work either. SpaceClaim certainly could not modify this model in the ways I need to model it, nor could SolidWorks make the changes if it were an imported model. You would need to work with Rhino, but there is no guarantee that it would be easier than working with native data in SolidWorks. Its an ugly problem. Parts that combine complex shapes with internal geometry and the need to morph organically are problematic no matter where you work on them.

So what do you do? What tools do you use to create and more importantly change geometry like this? Mesh and subdivision modelers are sophisticated, especially when it comes to morphing shapes, but not sophisticated enough in the ways needed for creating manufacturable geometry.

Have you ever seen the movie The Incredibles? 3D animated cg stuff. Its really a movie about what you can do with mesh modelers. The characters stretch and deform and flex in ways you can’t even imagine doing in an engineering CAD tool, but we still sometimes need to be able to do that kind of work with our parts. You can’t do that kind of stuff with parametric history based modelers. Not efficiently anyway. So what do we do? Use a mesh modeler like Maya or 3DS Max? No, they can’t hold the accuracy that NURBS modelers hold. So for the time being we are stuck. Changing shapes organically is too complex for parametrics, and too precise for mesh modelers.

I don’t think this is just a case of somehow making users smarter. Parametrics can be challenging even on far more easily defined prismatic parts. Certainly there are “best practice” rules you can follow to some extent to help make changes more likely to update properly, but the truth is that the SolidWorks execution of the technological concepts of parametrics for complex shapes is imperfect (it doesn’t always act predictably).

How is this going to work out? Will parametrics ever make the technological leap needed to make NURBS modeling practical from an organic manufacturing model accuracy and flexibility standpoint? Will mesh modelers progress in accuracy? This no mans land of organic CAD models is a lonely place to live, really half way between two very different worlds, neither of which really satisfies all of its needs.

4 Replies to “Blast from the Past: Is Parametrics History”

  1. I think that parametric surfaces would be best controlled by a coarse version of Tsplines. All of the nurbs surfaces should be untrimmed surfaces. These are then obligated to join at edges with the desired level of continuity. The edges should be as low order approximation as possible and have curvature and continuity constraints. SW can get confused about trims and will trash the entire model deleting many features, and sketches. The SW way of consuming sketches and curves must be flushed, untrimmed surfaces are always shared by two or more surfaces. The surfaces themselves should be of reduced order, conic being my favorite. The scary part is that these edges and guide curves will often be a smooth approximation of the projection of two mathematically defined sketches.

  2. Have a gander, should you get the chance, at the tools for this type of work (in a parametrically or history-based manner that allows complex surface creation and edit) in both NX (The work they’ve done with X-form and sync tech is incredible) or Delcam’s PowerShape with it’s surface edit/manipulation, morph and history tools.

    Tools have been out there for a while – just they rarely see the light of day compared to the more mainstream tools.

    Also “Cosmic Blobs”. Good lord, not heard that in a fair ol’ while.

    Al

  3. “Will parametrics ever make the technological leap needed to make NURBS modeling practical from an organic manufacturing model accuracy and flexibility standpoint? Will mesh modelers progress in accuracy?”

    Matt, I think with the big push of 3D printers that all Parametric CAD software needs to look at this seriously. They immediately need to be able to import NURBS or STL models and have the capability to add or subtract material to these bodies. But ability to edit compound surfaces and curves, etc. may take a bit of time to happen.

    Most CAD products can do FEA which is a form of MESH modeling… maybe they can start there and use this existing capability. If they can wrap a Parametric Model with a Mesh, how hard would it be to do the opposite?

    Bob

Leave a Reply

%d bloggers like this: