Direct modelling mania!

02 December 2010

‘Direct Modelling’ used to be just another relatively unpopular 3D technology. Now it seems vendors are falling over themselves to offer some direct geometry manipulation. Martyn Day gets his facets manipulated

I recently attended a seminar on SpaceClaim, a direct modelling design system launched in 2007.

SpaceClaim was backed by industry star investors and headed up by Mike Payne, the man responsible for the development of Pro/Engineer and SolidWorks. And while the creation of a new CAD system is rare and always big news, this one was backed by a team with great form, so the industry took note.

History-based CAD has dominated the mechanical CAD scene since 1985, when Pro/Engineer was released by PTC. Nearly every other mainstream offering since has been a variation on a theme,
should that be SolidWorks, Inventor or Solid Edge.

All the hype was about the benefits of parametric modelling, the potential for automation and good old moving from 2D to 3D and very few addressed the intrinsic problems of historybased modelling, or the huge complexities in creating and perhaps more importantly, editing fully parametric assemblies.

History-based modelling is a function of the computers of the time. Slow hardware meant it made sense to break down the construction of a part into individual step or features.

An edit to these steps would require a rebuild of the part to update each subsequent steps to arrive at the final form.

While powerful, it did give rise to issues with complex relationships. Edit a feature too drastically and things could quickly get out of hand.

There have always been alternatives such as CoCreate Modelling and IronCAD which offered ‘direct’ modelling.

These didn’t have the overhead of history-based systems and allowed model geometry to be directly manipulated. They also excelled at working with imported data. The problems came when working with non-prismatic shapes, such as surfaces.

Direct Modelling tools du jour were hopeless while history-based systems had the advantage that, by breaking down the construction of the geometry into individual steps, you could create and work with complex models. There’s clearly a huge mismatch in these two approaches.

History-based systems got out of the gate first, achieved market prevalence very quickly and so it has remained.

The likes of Pro/E, SolidWorks, Solid Edge and Inventor flourished and the direct editing tools never eally gained much traction outside of a few core industries - until recently that is.

Mike Payne

So SpaceClaim isn’t a new idea, it’s a new take on an old idea.

While it doesn’t resolve the inability for direct and history-based models to live together, it does offer great ease of use (aimed at non-CAD users and occasional users), an inherent ability to edit legacy data, and has the potential to become a competitor to the existing players.

My take is that the investors were hoping to sell out early and make a tidy sum, as had happened to Revit, their previous venture, but that didn’t happen. Instead the industry has gone ‘Direct Modelling’ mad and now there is a lot of hype to contend with.

PTC, which was first to be invited to see SpaceClaim, ended up buying CoCreate for $250 million. Siemens developed Synchronous Technology for Solid Edge and NX and Autodesk is developing Fusion.

Despite saying that nobody was asking for direct modelling in 2009, even SolidWorks has some capability in its Instant3D tools.

Within the space of three years, direct modelling has come from the backwaters of CAD to a key sales pitch function for all the players. I pointed this out to Mike Payne at the event, to which he replied acidly “Imitation is the sincerest form of flattery”.


The clear fact that’s being brought out of all this frenzy is that, despite years of development, the geometry problem is still not solved.

History-based modelling can’t do everything and new tools are required. The most recent developments have been trying to marry the history-based and the direct modelling abilities to provide some sort of workable system, as opposed to the design intelligence taking a one way trip to oblivion.

Autodesk, Siemens and PTC are all developing intelligent tools which allow the two methods to function together.

The history tree is the recipe of how a model is made and from it derives the geometry.

A direct modeller plays with the geometry without recourse to any parameters thus breaking the link. The new solutions attempt to rationalise the changes made to the geometry and topology and update the features as defined in the history tree.

I feel technologies are merging here to produce a new breed of even more intelligent design systems

However it’s still possible to make such edits that break the geometry.

In all the demonstrations we have seen, while impressive, the demo conditions appear uncomplicated. The direct editing of a complex fillet with various radii would cause most of these systems to keel over.

So while history-based parametric modellers are learning new tricks, developers are also making dumb geometry, less dumb.

At this year’s Isicad/ COFES Russia in Moscow, Dmitry Ushakov, LEDAS director of
product management, gave a demonstration of it’s Variational Direct Modelling (VDM) constraints system which when included into a direct modelling solution looks at the design intent and can create constraints and boundary elements in a featureless model.

Similar feature recognition technology is also being developed by companies like Geomagic that take dumb scanned data and automatically build feature-based solid models for intelligent editing.

I feel technologies are merging here to produce a new breed of even more intelligent design systems.


So, while Mike Payne has been at the core of history-based parametric CAD development, his latest venture, SpaceClaim has help push a signifi cant rethink in the industry as to how fl exible these design systems can and should be.

Comments on this article:

Thanks, Martyn, You make an excellent point about only seeing facile geometry in most direct modeling demos. We did that for our first couple releases too, and then we figured out how to finish the job. The reality is that in many cases those simple tweaks don't cut it, and you need higher-powered tools. Frankly, based on how we see SpaceClaim used today, those demos do a poor job showing the real power of direct modeling. Slice and dice, hack and stack, deleting and recreating irrelevant parts of the model -- the type of stuff that would terrify a seasoned history-based user -- that's where the real power is. But perhaps the most exciting part of the story is that direct modeling changes the role of 3D in engineering companies. Rather than being a tool used by highly-trained CAD specialists, everybody who wants to visualize ideas and answer questions in 3D can do so. We're seeing manufactures dramatically reduce time-to-market and product development costs by moving from typing-pool workflows to those where everybody can work in 3D the same way that everybody can use email today. We now have customers who have deployed SpaceClaim to thousands of engineers to foster innovation and improve their competitive position. It's exciting to watch, and I'm looking forward to the next few years. -Blake SpaceClaim marketing guy @bcourter on twitter

Posted by Blake Courter on Thursday 02 2010 at 04:14 PM

I'm fortunate (I think?) to interact with ~8 CAD systems on a regular basis. I have come to appreciate and not appreciate the different implementations of the technology. I see the pluses and minuses of both parametric and direct modeling. With regards to direct modeling, I think vendors should make a strong effort to educate designers and engineers of the benefits of direct modeling. Not a criticism, just an observation. Many do a great job showing specific functionality and expect that engineers "get it". I've talked to numerous users over the years and when I ask about direct modeling, the reaction is usually.. "it's really cool, but...." What follows is a variety of questions and misunderstandings about how direct modeling fits into their day. Think about it, we all have alot of tools on our machines, the last thing I want is yet another tool, unless it solves a big problem and it works very well with my existing tools. Being a "simulation" guy, I "got" direct modeling almost instantly. All of the vendors are now showing how direct modeling can be used in the simulation space. Personally, I think it could be shown way more as I don't think everyone really appreciates all of the problems that direct modeling really solves in the simulation space. It's classic product marketing, tell me the problem you are going to solve for me, show me how I can solve it using your tool and then tell me what you showed me. I think too many questions are left on the table for prospects out there to come to their own conclusion about how direct modeling can solve problems on a daily basis and how others have incorporated it into their busy day. Not criticism, just observation.

Posted by derrek cooper on Friday 03 2010 at 07:43 PM

Good article. I've used solid modeling since the dawn of time, including AutoCAD's Advanced Modeling Extension (AME) (boolean-based parametric), AutoSolids (a more advanced "AME"), SolidWorks and Inventor (feature-based parametric), and now back to AutoCAD (boolean-based semi-parametric, semi-direct editing). I've also looked at moving up to Solid Edge (feature-based, parametric or direct), and SpaceClaiim (feature-based, direct). What I've found is they all fit well in certain places, and none fit well everywhere. Boolean-based is fast and easy, but gets cumbersome above medium complexity. Feature-based parametrics are very powerful...once you get the hang of it. Direct editors are great for imported geometry and playing with concepts, and are helpful once your historical data becomes unmanageable. If I could only pick one, it would be Inventor/SolidsWorks (with AutoSolids a close second). Besides the power, you edit the same things (profiles, dimensions, etc.) you used to build the model. With direct editors you have to build it one way, throw away that knowledge, then edit it a different way. For designing from scratch, that's sheer nonsense. Your mention of Spaceclaim's goal of being bought out was interesting. I felt the same way since Day 1. Everything I could find to read was about the company, no details about the product. That's changed lately, probably due to what you mentioned about the competition. Even so, when I tried to use the product a few months ago it looked great on the surface, but using it was "another matter". I'm sure they can get there if they'll focus on the product instead of a buyout. Hope so.

Posted by Calvin Stigler on Friday 03 2010 at 09:46 PM

Martyn: Direct marketing mania - you got that right! Thank you for commenting on this important development in the CAD industry. At the beginning of the year I titled a blog: 2010 - The Year of 3D Direct Modeling. And that is exactly what has taken place. For years history-based 3D has stifled the creativity of designers. History-based 3D CAD is great if you already have your design fully developed in order to lock down the design and communicate it to your partners. But if you really want to create something new and try many design iterations, complete with analysis - nothing beats direct modeling. In 2005 Kubotek USA (developers of KeyCreator 3D CAD) - developed Facelogic technlogy, the ability to recognize and edit complex geometry. In 2007 we developed Direct Dimension Editing, using dimensions to drive changes to geometry without without the need for history and regardless of the origination of the geometry. This technology breakthrough has spurred on new companies (Spaceclaim) and a renewed interest in Direct Modeling. We see this technology continuing to evolve and in time may eventually replace history-based technologies. With this you are seeing most of the history-based CAD companies jump on the bandwagon. We welcome this competition, we know that the end user ultimately benefits as does the creative process within the thousands of our customers companies. Thanks for reading, Scott Sweeney @ Kubotek USA - Masters of Geometry and developers of KeyCreator Direct CAD, Kubotek Validation Tool and Spectrum CAD Viewer

Posted by Scott on Monday 06 2010 at 06:11 PM

I do agree with you, that technologies of direct modeling and parametric modeling "are merging to produce a new breed of even more intelligent design systems”. What I’m not agree with is that you are using “parametric modeling” and “history-based modeling” as synonyms. Traditionally they appeared as a single approach – parametric feature-based approach of PTC. But, actually, what is the remarkable in the approach of PTC is “parametric”, and what is to be thrown away is the “feature-based” (or in another words “history-based”). The task is not to marry “history-based modeling” and “direct modeling” (this the wrong way), but to marry “parametric modeling” and “direct modeling”. Parametric modeling might be neither “feature-based” nor “history-based”. As well as parametrics don’t need to live only in 2D sketches (as it happens now in traditional parametric feature-based approach). Were we have a really powerful parametric solver (that can resolve in real time corresponding systems of equations containing hundreds of thousands of equations) we could implement parametric approach in 3D space, and to work with solid parts as a whole. In this case both direct modeling methods and parametric modeling methods could find their natural place in one common workplace. The good news that we – in Cloud Invent has such a super solver (we called it Cheetah). Ad we do believe that this will be the Real CAD Revolution – the real Great Unification of direct modeling and parametric modeling.

Posted by Nick Sidorenko on Friday 01 2011 at 04:31 PM

Very Interesting article. The struggle as I see it, is about embedding intelligence into CAD models. Design is in my opinion is a form of geometric animation. Look at the difference between intelligently structured models and geometric models in animation. Animating a pure geometric model is like animating a corpse and animating an intelligent models with skeletal structure and constrains is so much easier. So the trick is to build that non biological skeletal structure and constrains with the CAD model. This can be done with “genetic modelling” used for generative design. Here are some examples of it : Genetic models can now be built only on history based parametric CAD. It is difficult to embed design intent in direct modelling because direct modelling deals only with the end design - which a particular (very limited) geometric expression of design intent. For a long time CAD was about moving stuff from paper to screen. Then it was about going from 2D to 3D. Now it is about how to do it best in 3D. The discussion about direct modelling does not make sense. Because, it is purely about geometric massage. It makes it easier now allowing kids to massage. CAD should be about building variational intelligence into models - that are aligned with intent.

Posted by Sivam Krish on Friday 29 2011 at 06:04 PM

Thank you for this short review, MatthewThe only thing I'd like to add to it is that the Constraint Manager (on the right, but it can also be aceathtd to the left) differs from traditional feature-tree. We manage constraints, not features. Constraints are all independent (no order), features usually form some dependency (tree-like order) this is an important difference that allows constraint-based approach to solve quite sophisticated parametric problems that cannot be expressed in traditional feature-based modeling systems.

Posted by Ester on Monday 25 2012 at 08:28 PM

Leave a comment

Enter the word you see below: