Why SolidWorks Needs a System

The history of history-based modeling is not that long, really. We can talk about all sorts of variations, but it boils down to two things: Pro/Engineer, and then SolidWorks and everything that came after, including Solid Edge and Inventor.

History-based modeling needs to start taking a back seat to newer technology. We continue to cling to it for some reason, probably just the lingering popularity of SolidWorks. In order to really improve, SolidWorks itself would have to gut or at least add another internal system, which would mean practically starting over. There are some people who believe SW is just about perfect as it is. I guess fanboys will always be fanboys. The product going forward will not change significantly, because any serious change will mean disrupting their own customers, which I think they finally figured out wasn’t a good thing. They’ll keep recoloring the icons, changing the renderer, and adding duplicate features until people stop buying it. Loyalty is a valuable thing until it starts inhibiting progress, and I think we’re there.

I think DS thought they were going to force the issue a few years ago with the V6 stink. I didn’t hang around to see how that one ended, but it looks like V6 doesn’t have much if any foothold in SolidWorks territory 5 years later.

Personally, I was hoping Onshape would do something more creative to help abolish history, but alas. They did several cool things, but pretty much bowed to the history-based deities and punted on that front. This is why, even though they have done some really innovative things, they still get accused of making an on-line copy of SolidWorks.

On the other end of the change spectrum from Onshape is Siemens PLM and their Synchronous Technology. In my 4 years with Siemens, I think the only thing I really accomplished was to find a way to communicate about Synchronous with history-based users. I wrote a free downloadable book on the topic. To boil it down to a couple of run-on sentences, Synchronous Technology essentially takes the intelligence out of the model (so in history terms, you’re working with a “dumb” model), and allows the user to define what changes and when it changes, by allowing selections of faces to determine the “feature” (whatever you have selected is the “feature”, and yes, the “feature” can include faces from multiple parts). Whatever you want to change just does what you tell it to – move 1 inch, move up to this face, rotate 45 degrees, etc, and it only does it when you tell it to. There is no such thing as “rebuild”. There are also no “persistent” relations (unless you force them to be such). And your selected faces can be the same or change as you need them to. Yeah, this oversimplifies it to some extent, but it should help your average CAD user who has been stuck in history-based modeling for 30 years understand it.

Anyway, Synchronous is a better system for prismatic models, and if you want to extend the idea to more complex models, you have to move to NX. Plus, you get to keep your history-based methods for the cases when you really need them (stuff you really want to drive with a sketch, like sketch text, or possibly fillets, or other detail type of features). Synchronous really helps eliminate parent/child relations, and just make changes that in some cases you would never even consider doing in history-based modelers because it would just be easier to remodel the part (such as rotating a feature out of plane, or only changing part of a feature, etc).

So Solid Edge already has a “system” built in. This is why Dick Gebhard’s Resilient Modeling has such a headstart in Solid Edge. You can still apply RMS on top of Solid Edge, but if you use Synchronous, you have inherently less need for it.

Now we have to talk about Pro/E. Remember Pro/E had a reputation of being difficult to use, and fairly rigid when it came to procedures. Back in the early 90’s, people who taught Pro/E taught some kind of method, or structured procedure. Something like a skeleton model, master model, or references built around planes. It was structured, it was tough to use, people failed to follow the procedures sometimes and made a mess of things. But it worked if you followed the procedure. Models were less likely to lose parent/child relations if you built everything around planes and sketches (like Stoltfus’ SSP method).

Ok, that’s all pre-amble. Now we can talk about SolidWorks. SolidWorks appeared with the advent of Windows NT, when Windows was entering the world of technical computing. You no longer needed a $20k unix workstation to do CAD. Along with the dumbing down of the IT OS/hardware requirements, SolidWorks was going to dumb down the requirements for 3D CAD usage. If Pro/E required a top 10% user to be successful with it, SolidWorks required a top 80% user. SolidWorks cleared the way, and got rid of all those pesky rules. You could do things in any order (sketch, feature, or feature, sketch). You could do object/action or action/object.  You could sketch right on model faces rather than creating planes. In fact, you could sketch on faces of other parts in assemblies. So the rule book was out the window, everything was easy, and the software, hardware, and OS was cheap and easy.

This is great until the models start getting more complex. This is the point where all of those people who stopped reading this too-long blog already need to listen extra hard. It turns out that just modeling without any system leads to the chaos that it sounds like it would. There was a reason Pro/E was hard. There was a reason Pro/E users looked derisively down on SW converts. If you have a history-based model with intelligence built in, that intelligence needs to be structured, or it’s just chaos. SolidWorks freed the market by loosening the restraints, damn the chaos. Ease of use as a mantra really gets people’s attention, but discipline is less sexy. Regardless if you do it or not, history-based modeling really needs some structure – rules and procedures. This is why SolidWorks users generally are superstitious about links, and why they do silly things like deleting all links between files at the end of projects.

But the thing is that one set of rules isn’t going to cut it for every type of design. Don’t forget that as much as we as SolidWorks users think we have in common, what we do and how we do it are very different. This is probably one of the places where Pro/E really could have done better. It’s also the blind spot that I think SSP and RMS have. Everybody is used to working with their own problems, and they will solve their own problems as if they are universal to all CAD users. You can’t dump the rules, and you can’t establish just one set of rules. Each type of design has to develop it’s own structure. Machine design has a lot of prismatic shapes, reused data, motion, custom or off-the-shelf parts. Plastic and cast parts are much more complex as individual parts, but there are fewer of them in an assembly. Structural design uses a different method yet. Some assemblies are going to use both types of data so you will have to adopt a hybrid system.

In my view, SSP works great for machine design. Any type of mechanism that can be mostly explained in one or 2 2D views should work pretty well with SSP. SSP is really a sketch master. You might call it 2.5D in the CAM sense.

Master model as I use it is a 3D, usually surface master. For me, this works best for plastic assemblies, or parts where an overall shape is composed of several individual parts, like a car body.

In my view, RMS is trying to be everything – it includes sketch and model masters – and on top of that it has a lot of rules. I’m not sure it built on any particular design process or type of design (eg plastic assembly vs machine design vs electronics enclosure and structural assembly), so it might have the best chance of being adoptable as a general system that can do anything if you eliminate the unnecessary parts.

Horizontal modeling I think is trying to be a very general concept that you will have to apply via customization to any specific situation in which you find yourself.

Does all of this sound at all familiar? Hasn’t the CAD world been through this cycle of freedom => chaos => system before? Doesn’t PDM sound an awful lot like this cycle? PDM is generally configurable for different systems, and is essentially a tool for enforcing best practice rules. I think the PDM solution is a good model for solving the problem of needing an enforceable variable modeling system. It would have to be CAD package specific, and configurable within the CAD package. It would have to enforce structure and best practice rules.

CAD users will not give up their freedom easily, but the promise of editable data is quite a prize.