About Hybrid Modeling with ACIS

Issue #902

 

by Ralph Grabowski with Brian Rohde Exact modeling (or precise modeling) kernels have been used for over 30 years. "Much of the world we live in has been designed with these precise or b-rep [boundary representation] modelers," said Brian Rohde, senior product manager at Spatial. They do not, however, handle organic entities (like body parts and plants), point clouds, voxels, additive manufacturing, and geological entities (like underground ore bodies) well. Organic shapes are well-represented through polyhedral, polygonal, and faced models -- approximating surfaces with polygons, usually triangles but could also be quadrilaterals or other convex polygon. Because it is an approximation, it is fast, and does well with complexity (by reducing it) and very large models. Because polyhedral modeling uses approximations, it suffers from accuracy, and so makes simple shapes unnecessarily complex, a like sphere.(See figure 1.)  Because it uses flat polygons, it takes many of them to approximate curved surfaces. "You can get very close with polyhedrons, but it is not perfect," said Mr Rohde. "All [exact modeling] needs to do to store and describe [a sphere] is to know its radius and its position." Figure 1 Polyhedral modeling (at right) cannot model spheres exactly as does b-rep (at left) For Spatial, hybrid modeling takes the best of both worlds. For example, a replacement rib cage is best designed with exact modeling and then 3D printed, whereas the natural irregularity of human rib bones are best modeled with polyhedral modeling. (See figure 2.) Another example is designing services (precise) inside tunnels (polygonal). Spatial's hybrid modeling combines precise and polyhedral modeling in a single body, maintaining the original representation when possible. Spatial sees hybrid being really useful for the fourth industrial revolution that involves custom manufacturing, fast delivery, minimal inventory, and higher margins. Examples include integrating internet-of-things into physical things. It's also used in hybrid manufacturing that mixes additive (like 3D printing) and subtractive manufacturing (like milling and polishing machines). The limitation today is that Spatial cannot freely go back and forth between the two formats. Currently some simple polyhedral bodies can be converted to b-rep entities (a.k.a. reverse engineering). In the meantime, both formats can coexist in a single model, and you can perform operations, like Boolean, using the polyhedral objects on the b-rep model, like cutting away or adding on.   Figure 2 Combining organic polyhedral (bone) with exact b-rep modeling (implant) For the future, Spatial is looking to adapt hybrid modeling to architecture, a discipline that parent Dassault Systemes keeps saying it wants to go into. One example is integrating scan data of buildings into a 3D model, or using it for augmented reality to find where the pipe is in the wall. [The presenter in the Webinar spoke about using robotics in the future to scan, manufacture, and insert replacement teeth many times faster than today. I had a tooth implant recently and so speak from experience. Two aspects were not addressed by the seminar. Major: it takes months to make the gum build up additional tissue in which to place the socket; the body will not be hurried. Minor: the new artificial tooth needs to be color-matched to surrounding teeth.]   Q&A Q: Will hybrid modeling become part of the ACIS kernel? A: Polygonal modeling has been in ACIS since last year. True hybrid modeling will be released in a little more than two months. Q: What is the ratio between a sphere and its polygonal representation in terms of the space it takes up in the disk? A: In precise modeling, the primary definition of a sphere is its radius, along with its position in space. This takes up very little space on disk [a few bytes]. Whereas in the polygonal case, it will depend on the level of accuracy. An octagon is a very crude representation but you haven't taken all that many facets to represent the sphere. But you can also take that sphere and represent it with thousands or millions of triangles [which takes up much more memory, such as kilobytes and megabytes].  Q: Are we going to be seeing new tools in our CAD package to handle reverse engineering using this new paradigm of hybrid modeling? A: It takes much more to do reverse engineering than just the conversion of polygons into b-rep. I think that over time this is going to become more common in CAD packages. Reverse engineering is not all that different from bespoke [custom] manufacturing, in that you have a scan or some other representation of an object or body part from the real world, and then you represent that in a 3D modeling environment -- measure it, perform analysis, customize it, make use of existing parts, and then manufacture it.  Q: What performance advantage do polyhedral models have? A: The performance improvements depend on the shape of the entities being modeled. A sphere will be more complex, but a bone will be more accurate and contain less data than b-splines and solid models. But operations like offsetting, fillets, and Booleans slow things down; the same operations with precise modeling is more efficient. Hybrid modeling lets you use whichever approach is best for the data. Q: Do you think parametric and b-rep modeling will disappear in the future, and that polygonal and hybrid modeling will become de facto? A: I think that is possible, and I hope it will happen, but I don't think it is possible because of the sphere example. As storage becomes larger and processing power more powerful, then the drawbacks to polyhedral representations might become less significant.  Q: How will tolerances be managed in hybrid modeling? A: That's a perfect case for hybrid, because your design is in the precise form. But the measurements might be from a laser cloud scan or high-precision metrology machine, and then overlay them to see if they are within spec. Q: How is the stitching of two models done in hybrid modeling? A: Take a precise-modeled sphere, place it on a 3D scan of a bone, and then do a Boolean operation to join them. At the surface where the two meet, there is a set of triangles on the bone's surface. The triangles have to be modified to line up with the sphere. Our Boolean operation adapts the triangles at the point of intersection between the sphere and the bone. The sphere becomes a polyhedral model. http://www.spatial.com And One More Thing... ASCON Group's Renga Architecture is getting frequent updates since its release last year, and now v2.3 imports 3D models from the company's KOMPAS-3D MCAD software.  No longer displaying models only in shaded gray (although it did look classy), Renga now does wireframe and color rendered display modes. Other new functions include tables, multiple views per floor, column and beam styles, and rotated beams (see figure 3). The company keeps dropping hints about Renga Structure, "our upcoming building structures design system."  Figure 3 Renga Architecture rotating a beam Price is e1,599, with 50% off for competitive upgrades from other CAD systems. Download the demo version from http://rengacad.com/en. Even More News We're on Twitter at @upfrontezine with late-breaking CAD news and wry commentarythroughout the day, such as.... upFront.eZine (@upFronteZine) May 18: Google Home sounds as creepy as pundits promised Google would eventually become. Video showing how Google Home replaces a need for parents. upFront.eZine (@upFronteZine) May 18: New video calling app from Google lets you spy on callers before you agree to accept their call. upFront.eZine (@upFronteZine) May 18: Depressing to see Android users no longer need to be alive to respond to my texts, etc. They just need to ensure Google chat bot is running. Letters to the Editor Re: In the Year '96, '96 Great retrospective. What a bizarre, twisted business CAD has been since the advent of the PC. So many wrong turns for so many companies, and the future doesn't look much better as everyone chases cloud-based CAD/BIM while neglecting the day-to-day needs of typical users.      - Larry Leake (via WorldCAD Access)     Leake Associates - - - I enjoyed your recap of 1996, but I was surprised to see no mention of Computervision or CADDS 5 amongst all the others. In '96 CV was still the "elephant in the room" with a large share of the CAD market. Of course, just a year later in 1997, there was the surprise acquisition of Computervision by PTC.     - Steve Huffman     Draftware Inc. The editor replies: I condensed what I reported on that year in upFront.eZine. I had no relationship with Computervision or PTC, but would have liked to have reported on them. - - - Congrats on 900 issues! Also, thanks for visit back to 1996; I'd not thought of the names of some of these products and companies for many years!  Since your eZine was the inspiration for my now-retired GIS Monitor newsletter, I feel obligated to offer one correction. Autodesk did not sell MapGuide; it turned it open-source in 2004 and later submitted it to what is now the Open Source Geospatial Foundation, OSGeo.     - Adena Schutzberg, principal     ABS Consulting Group Inc. - - - That was a lot of digging, Congratulations, son.     - Herbert Grabowski - - - Thanks for the history. Based on a forum post, the C4a patch for AutoCAD Release 13 was released on or about August 2, 1996. If I recall correctly this was the final patch issued for R13 and it was from that point on a pretty stable version, despite the historical lore of how bad R13 was in general.     - R.K. McSwain The editor replies: The worst part was then-ceo Carol Bartz insisting it was a marketing problem, not a technical one. - - - Very good history Ralph. I remember those days well. I recall quite a bit of those stories since a lot of them were right up my alley. For instance, about five years ago, I finally threw out the binder that contained the entire AutoCAD Release 13 bug list. It was over 400 entries. I hang onto stuff far too long.     - Ron Powell - - - Like a time machine -- loved it!  And the Zager and Evans reference, which resonated immediately without having to look for your footnote.     - Lynn Manning The editor replies: I grew up in Northern Canada, where there wasn't much available in the way of music, so I remember hearing it at the local swimming pool -- oddly enough. Mr Manning responds: Didn't think it would ever be warm enough in Northern Canada to swim :) . That was a pretty strange song, though. But obviously memorable! - - - I have lived through/survived CAD since 1984. I was intrigued by all the buying/selling/reselling/buying back/selling again of all the various companies and/or specific products that you outlined. Sometimes you can't tell the players even with a program!     - Bill Fane - - - I was an expert witness for Bentley Systems in the Baystate lawsuit. Baystate had a very unique theory in the case: basically, they said, "Since the Cadkey PRT documentation was copyrighted, you could not write software based on the information contained in the document." I had a very hard time not breaking out laughing when they rolled out that line of argument. I asked their lawyer if he was really asking me that.     - Scott Taylor     Taylor Made Software The editor replies: I'm thinking CADkey wanted any means to stop Bentley, and that was the best they could think of. Thank You, Readers! Thank you to readers who donate towards the operation of upFront.eZine.  Solid Modeling Solutions, USA Ross Goulter, Australia Robin Capper, New Zealand Should you wish to support upFront.eZine through PayPal, then the suggested amounts are like these: $25 for individuals > paypal.me/upfrontezine/25 $125 for small companies > paypal.me/upfrontezine/125 $500 for for large companies > paypal.me/upfrontezine/500 If PayPal.me does not work in your country, go to www.paypal.com and make payment to the account of grabowski@telus.net. Or you can mail a cheque (US$ or CDN$ only, please) to upFront.eZine Publishing, Ltd., 34486 Donlyn Avenue, Abbotsford BC, V2S 4W7, Canada.   Notable Quotable "My own personal litmus test for AI breakthrough is whether a computer can distinguish between sarcasm and irony. So far, none can."     - Andrew Orlowski, The Register     http://www.theregister.co.uk/2016/05/19/you_wanted_innovation_we_gave_you_clippy/

*7224