How to Build Impossible Models

I once sat at a table and listened to someone mock engineers, explaining why we have such a lack of creativity. I blew my top, because I had heard it before from this particular guy. He obviously had some very narrow definition in his head of what he thought creativity was, and was so unbelievably arrogant as to not see it in anyone else. I’ve rarely been so professionally indignant. I got up, said my peace to a table of about 15 people, and stormed out of the restaurant. Walked back to the hotel.

In his defense, I have talked to at least one engineer since then who felt that they didn’t have an ounce of creativity in their body (but that was a municipal waste water engineer, not a consumer product engineer).

The fact is that engineers are very creative, or more to the point, very imaginative. I participate in a lot of community music events, with other amateur musicians. Two professions dominate those groups. Guess what they are: The largest group is school music teachers, which makes sense. The second largest group is engineers of various kinds.

Engineering is the act of taking scientific concepts and making things actually happen in the material world with those concepts. So we are primarily problem solvers, and to solve problems, we must be able to imagine solutions. Part of problem solving is design – providing a means to fill a need. Design means a lot of things beyond just doodling with a colored pencil. You start from a problem or a need, and you develop a device or process or other means to fill that need. Just because my friend couldn’t see it doesn’t mean it isn’t there. I’m sure most of you have seen a lot of examples of creative engineering or design done by engineers.

Anyway, that’s a long preamble for a situation I found myself in recently. Maybe I should start by clarifying what I mean by “Impossible Models”. I’m really talking about stuff that mid-range CAD is just not equipped to handle, like most of the models shown in this article. They’re not literally impossible, else I couldn’t show anything, but they are maybe better labeled… improbable. Let’s go with improbable.

I like to take on modeling projects that stretch my capabilities sometimes. Sometimes the projects are for a book, or a blog post or some other purpose, but there’s nothing that keeps you sharp with the software better than taking on a project that some people might think is impossible.

I want to walk you through my stream-of-consciousness type of thinking as I worked on one particular model over the course of a few sessions. I made 3 different runs at it using drastically different approaches each time. In the end, I probably had 20+ hours into the part, but only 6 into the actual finished part. The other 14 weren’t totally wasted, I did learn a lot of things that you can (and can’t) get away with on this kind of model.

And by the way, this sort of “failure” is the source of any sort of success that I’ve ever had with this kind of work. You rarely get things right the first time when you’re doing something impossible or improbable.

You have to approach this stuff with a lot of patience. If you think you’re going to put a picture of something on the SW Forum and have one of the generous geniuses there turn your lightbulb on for free – that’s not going to happen. This kind of work doesn’t work that way. You have to know what tools are available and how they work and how they don’t work before you can do the impossible.

My task was to make a decorative piece something like a wreath, but covered with leaves. I’ve done interpretive stuff like this before, like the candleholder model, or some of the scrolly viney leafy things I’ve done for furniture design and even jewelry. This time, it was a tapered U shaped drape covered by leaves.

When I start a project like this, I start by thinking of different ways I might accomplish the end goal. And then I wind up thinking of things I want to avoid. So in this model, I feel like I need to make it look organically variable rather than with computer-like precision. It probably needs to be patterned in some way without looking like a repeated pattern. There are a lot of things you can do with patterning, and a lot of things you can do with making perfect surfaces look organically imperfect. Sketching with splines can introduce some variability that avoids the perfection of lines and arcs. So even though I’ve narrowed it down a little bit, there are still a lot of choices to be made.

One of the things you need to remember as you head into a project like this is that sometimes, stuff just doesn’t work, and you can’t fix it. You’ll know what I mean when you run into a problem, and you just can’t make a particular loft work regardless of what you try. For this reason, one of the most important things to remember about making impossible improbable models is that you need 3 or 4 tricks in your back pocket for every complex stunt you want to try to pull off. You can almost count on it as a rule that when you’re pushing the limits, something somewhere is not going to work, so you need to be prepared. If you don’t consciously have alternative methods, you at least need to be armed with a lot of patience, flexibility, and imagination.

First Attempt

So. The first attempt at this part was going to be risky, but if it works, it will save me a lot of time. I start out by making a bit of a “storyboard” with words to plan the technique.

The leaves have to follow the U shape. So the patterned leaves are concave on the top (inside of the U) and convex on the bottom (outside of the U). This means you can’t just use a dumb pattern, which would make them the same on the inside and outside of the U. So in this case I made a reference surface, and extruded up to the reference surface, and patterned the feature so you get the patterned intelligence.

First, I’ll make a leaf as a solid. Pattern the feature around, add a Freeform feature to it, tilt the back end of the leaf down using Move Face (to make the tips of one group cover the ends of the next group), then pattern the bodies, rotate the patterned bodies, and then pattern again lengthwise.

That process gives you a straight tube of leaves. From there, use the Flex feature to taper the tube, then use Flex again to bend it into a U.

Great plan. Gutsy. Lots of features you don’t see around the house every day. Unfortunately, it creates so much bad geometry, it doesn’t pass Verification on Rebuild, drops a lot of faces from the display, and it can’t be translated. Bummer. Hours and hours, and I just can’t fix it. Specifically, I think the problem with this one was the second Flex feature. I removed fillet features and tried to line things up such that they “should” work, but it just didn’t work. Time to move on.

Second Attempt

The second attempt was even more risky, but I should know quickly if it was going to work or not. This past summer I wrote for the Mastering SolidWorks book about the new SolidWorks feature called the Variable Pattern. That feature would be perfect for this situation. You have to do patterns with intelligence because the U is curved differently around the circumference.

The idea of the Variable Pattern is beautiful. You can use a table to drive certain dimensions parametrically, and almost like configurations, except all of the pattern instances exist at the same time. So, you should be able to set up a pattern to go around the circumference, and then go axially along the U.

The only problem with the Variable Pattern is that SolidWorks can’t rotate things without flipping them. So I could get half of the leaves to work. I thought I could try to just use two Variable Patterns, and get the other half of the leaves on the second pattern. But it didn’t work that way. Variable Pattern would have been a brilliant way to solve this modeling problem, because all of the changes could (theoretically) be driven, including the circular patterns, driving the pattern around the U, making the leaves smaller … it all is supposed to work in theory, but it just doesn’t.

When stuff doesn’t work, it doesn’t do any good to get your panties in a bunch. You can stamp your foot and scream on the forum all you want, but that just annoys other people. SolidWorks has always had the problems of flipping stuff. Flipping sketches, mates, planes, everything. I think this is a Parasolid problem because Solid Edge also has this sort of difficulty.

Again, time to move on.

Third Attempt

By this time, I’m starting to think about giving up patterning altogether. Just model each leaf individually. There are only ~120 leaves on the part, so there would at least be a point where I could say it was done, and I know I can do it that way. It would just take the better part of 3 days to do all of that work. Maybe I can scale that back some without being so drastic. Use what works.

So here’s how I finally did it. I started with a reference surface of half of the U. Then I used an Extrude, with the Offset From Surface end condition, and then used Move Face to move it some amount like 0.020″. Then I patterned the first extrude (without patterning the Move Face) to get additional leaves at a lower level, then I patterned that again. I really should have used the Freeform feature to add some organic variation. Let’s talk about the Freeform feature, because it’s not something you hear about that much.

Freeform

Freeform is the closest thing SolidWorks has to tug-and-pull modeling that you find with subdivision modeling. Here’s how it works:

You start with a face of a solid or surface. Then you place control curves on it (pink and green in the image), and then control points on the curves. You can control the tangency direction at the end of the curves, and the position of the points. The points drive the curve like a spline, and the curve drives the surface like a loft. So you tug and pull the arrows and the points. This is actually very good for making flat surfaces into wavy surfaces, but it’s tough to get fancy with it.

Lining Up the Leaves

Getting the leaves lined up along the curving U shape took a couple of attempts. At first, I couldn’t get the leaves to pattern around the circumference in a uniform way. The last pattern instance would always be at some wonky angle. So I used one of the sketches I had used in the Variable Pattern attempt to line up the leaves in that model.

Each leaf sketch for the initial extrude was lined up with the centerline of the U or horn shape. This kept the patterns from getting too squirrely. Each leaf was somewhat shorter and narrower than the previous ring of leaves.

Patterns

From there, I just repeated those steps for a total of 9 levels of leaves. Each folder holds the same features, repeated with the offset distance being a little less each time to achieve the taper.

The only other feature of interest here was the wraps. The wrapped pieces were just a bunch of surfaces and the Intersect feature. Intersect is still kind of new. I drew S shaped splines, and extruded them, then offset the original horn-shaped loft, and use Intersect to find a solid region bounded by all those surfaces.

Design Intent

Notice I haven’t even mentioned Design Intent. That’s because hang design intent. Edits at this level either work, or they don’t. If they don’t, just delete them. You’re not modeling a sales demo piece.

I’ve known some people who make a “practice” model where they do all of the real work of the design, and then make the “real” model where all of the design intent stuff is cleaned up, things are named properly, etc. That’s great if you’re going to revisit the model and change it often, but it might also just be a great OCD way of wasting time. Do you really need to do that? Can you point to an actual benefit?

In The End

In the end, everything worked out ok. The Tools/Check worked, Verification on Rebuild didn’t show any problems. No missing faces, no bad translations. I’m working on an old (8 year old) computer, so the rebuild times are higher than they would be on another computer. The last method took the longest to model, but it takes the least time to rebuild, and I spent far less time checking or fixing or testing on the last one than the first one.

I did leave out some details, but this should give you an idea of what it takes to make a model that some people might consider impossible in CAD. I still believe that this sort of stuff is beyond what mid-range CAD should be expected to do, but it’s still fun trying to make it work anyway.

I think this might also be a good time to reference the Rings Of Fire graphic that I made some years ago. There are some features that you are pretty safe with, like Extrude, Revolve, and reference geometry. These are placed at the center of the rings. The further out (closer to red) you get, the more risky the features become. The Flex feature is orange. Using the Flex feature twice on the same geometry should be in a new circle colored black.

Remember when you sit down to a project like this, you have to use your imagination. Imagine different methods to create the geometry you need to create. Try to imagine what difficulties you might encounter and how you are going to get around them.

And finally, no whining. Really. This stuff is hard, but not impossible. If you can’t do it, hire someone who can. The software doesn’t always do what you think it should do. You have to try to be creative and workaround issues like that. Should the software always work? Yeah, maybe. But it doesn’t, and crying doesn’t change that. You might wind up bald like me, but the first time you succeed against a beast like this, you’ll find it very fulfilling.