The kind of things I’m working on lately are all surfaced on the outside, and all shelled, ribbed and bossed on the inside. Recent projects have included sports equipment, and an industrial handle. I can’t show the entire model for either project, but here are some images of the interesting areas. Both parts had the same approximate workflow, and the same problems. Getting the outside shape about 90% of the way there is relatively easy after years of experience of pushing this software. On parts this simple, the problems usually come from just one area, and usually a comparatively simple area at that.
On the sports equipment model, I was trying to use features to model over mesh data created in Zbrush. Just for reference, Zbrush lets you do insane stuff like this. It’s not product modeling, and it’s not CAD, it’s just insane 3D mesh. Mesh modelers don’t have the limitations of needing to break the shape up into artificial “features”. So trying to match the shape with features that was originally created with mesh was a little different. It’s like going to Southern Indiana and trying to fit in with their accent without sounding like you’re from Western Kentucky. In one area, I had to cut a section out and model it using another method to get the same shape using features that the original mesh data had. This is the “hack and whack” school.
The end product on the left looks ok, but when you see how I got there on the right, it’s a bit ugly. The edges represent what I had to do to get the shape correct. There were conflicts between using fillets and splines or boundary surface to smooth out an area, and sometimes the boundary surface could not be coerced to give the proper shape through 100% of the feature, which accounts for the elliptical patches on the right.
After the outside was done, the part had to be split in two and shelled in sections. I had two areas that could be shelled automatically (with the shell feature), but the middle section had to be shelled manually, and then the three sections joined. Shelling a part like this can take an 8 hour project and turn it into a 3 day curse fest. I don’t feel good handing over a model like this to the customer, but it is the price you pay for not using software that is intended for this kind of work. Using SolidWorks for this stuff is agonizing. Even finding the area of the model causing the shell to fail can add an hour to a project, especially when rebuild times for complex features is accounted for.
In the industrial handle project, I was reverse modeling from a combination of plastic parts and fully dimensioned autocad drawings. It always amazes me how clean things are in 2D drawings, especially when what is depicted is completely impossible in 3D. Anyway, I was thankful for the drawings because they give me the opportunity to make a 3D model much closer to the actual part, but there is nothing like a 2D drawing for imposing an artificial feature set on a model. In the end, I wound up fighting the software to make a couple of shapes, and especially fought false changes, where every now and then 90% of the model would just erupt in a sea of red marks in the tree.
Once the shape was created, the feature history was doing me no good, in fact, it was causing problems by failing randomly. This is why the “Feature Freeze” in SolidWorks 2012 is so important. It essentially begins to compensate for some of the implementation problems of SolidWorks tree management. But this customer specified SW2011. Bummer.
The problems I ran into with both of these projects are directly related to feature-based modeling scheme. Shapes are not always easily broken into areas appropriate for specific features. Very often in product design work, you need direct control over the shape. Fiddling with sketches and feature settings is indirect control at best.
To the left is a picture taken from Ed Eaton’s Curvy Stuff 201 powerpoint presentation. Just for reference, it was presented at SWWorld 2004, so the model was made in SolidWorks 2004 at the latest. This is the best illustration of the difficulty of using a feature-based model to make a general complex shape that I have seen. Ed took a physical bottle and marked it up with a marker to determine the flow of various areas of shape. This is important with NURBS modelers because all of the faces are essentially a mesh of curves that want to be perpendicular to one another. Ed is trying to represent the 2 directional mesh of the NURBS data using the marker lines. What he’s really doing here is establishing which areas would be made using different features.
To the upper right is Ed’s modeled bottle. The edges represent each feature. I made a version of the same bottle, working through Ed’s exercise, and came up with a somewhat different feature set, shown to the right. You can see I didn’t round off the sharp edge going around the bottle. Looking at it now, after 7 years, there are some things I would do differently, but it serves as an example of how different feature breaks on a model can subtly (or not so subtly) change the character of a product’s shape.
The importance of “going with the flow” is that if you have an edge that goes through a face at an angle to the NURBS mesh, that edge is going to look smeared, or like it has actual woven fabric going over it at an angle.
Below is an exaggerated surface showing this effect. It is a boundary surface with 6 splines. The splines have a peak that moves from left to right in each successive sketch, going front to back. The pink and blue lines are the U-V 2 directional mesh, and notice that the peaks in each spline want to follow the blue lines, not the artificial edge we are trying to produce by having a moving row of peaks. This sends little creases down the surface where you don’t want them. Just so it’s clear, this is highly undesirable.
There are two ways to get rid of this effect. One is to line up the U-V direction with the line of peaks, so effectively turning your planes and sketches by 45 deg. The other way would be to use a connector in the boundary surface to connect the dots between the peaks. This rearranges the U-V mesh. It could cause other problems too, like scrunching up the U-V mesh in the upper right corner, and stretching it out in the upper left corner. The connector is the row of pink dots that go across the surface in the image below along the peaks in the splines.
You never want your modeler to drive the kinds of shapes you can make, but when you’re doing NURBS modeling, and we all are, you just can’t avoid it. Your choice of features can have a very negative effect on the shape outcome. And sometimes the available techniques limit the kinds of control you can have over shape. Mastering this kind of control over shape takes a lot of practice, and more than anything, a lot of failure.
To me, this is why real product design requires direct U-V push and pull (direct edit) capabilities in addition to feature-based tools. You often just can’t get there with straight feature-based tools. What kind of tools am I talking about ? Tsplines is one you can get to play with your SolidWorks data. I keep mentioning other tools like SolidThinking and Siemens NX for this kind of work as well.
If you are interested in the model used to make the last two images, you can download it here (280 kb – SW2011 format). It uses configurations to show both examples.