I finally broke down and got myself a 3d printer, and immediately my internet feeds are full of crap that people have done. Funny how that happens! I keep seeing this multi-threaded screw, in the form of a cone, where the centre piece moves freely up and down. I had to have a go!
The outer shape can be any vaguely cylindrical shape, so I went with the Chess Queen, which also fits nicely with the two-colour scheme.
For extrusion 3d printing, there is a limit to the extent of overhangs, so the curve used for the outer profile has to keep this in mind.
The core is just a seven-fold star, then extruded and twisted through 360 degrees. The main problem I had was getting the core offset, to serve as the part to be subtracted from the outer shell, leaving a little slop for ease of turning. Mesh Offset failed miserably, leaving a non-watertight set of faces, but the new ShrinkWrap component includes an Offset option, which worked nicely. The gods smiled on my final Mesh Intersection/Difference booleans, which usually fail for me, although they take 40 seconds on my Mac.
I’ve seen this kind of thing popping up online the past couple weeks, a chess piece is a nice idea for a base shape. I’ve wanted to give it a go, so thank you for sharing your file!
I tried your definition, compute time on my pc was 1min 43sec.
Using a different workflow I got it down to 4.5sec.
Filleting the seven-fold star shape instead of rebuilding kept the curve simpler
Orienting the seven-fold star on to perp frames with a little rotation applied to each can then be used in a loft. This results in a simpler shape to work with than the twist component makes
Offsetting the seven-fold star curve by 0.3mm will create the gap between the parts, the offset shape can be oriented on each perp frame and lofted also
I booleaned using the breps instead of meshes. It was the mesh booleans which really slowed things down for me.
I have found revolved surfaces can sometimes behave better when they can be capped. In this case I removed the bottom line on the chess piece curve and capped it after. Booleans fail if this is not done.
Vastly superior! Thanks for playing with this and for all the improvements. I stopped when it was “good enough” for my purposes, and this is clearly better. I’m still learning this stuff.
For the actual printing, I used standard settings for PLA, except that I went for three layers of plastic for the outer surfaces, and a slightly denser infill (20%), to add a little strength, which is especially needed by the outer shell part, which extends unsupported to the crown. I also used a closer layer spacing for the top of the crown, where the surface is curved, but close to horizontal, in order to minimize the ‘steps’ effect. Once I got the outer profile right, minimizing overhangs, the prints came out nicely, and the fit is good. The core falls out under its own weight.
Printing three outer shells at once takes me about seven hours, and three cores is five hours.
I like it so much I also tried printing a couple. I used matte pla, .3mm layers and 3 perimeters for the first try. 1 piece took 4.5 hrs and I printed them together as seen in the photo. I think a full set would be fun to do!
Yes, in this case I am sure you are right. Sigh! But I have never been able to use the Rhino interface, and I lack the artistic skills to get things right the first time. Grasshopper appeals to my software engineering background, and I like the ability to hone the design gradually and to play with alternatives. In this particular case, I also had to make a couple of test prints to get the twist pitch and thread offset right. To each their own!
It is not something I have tried to support commercially. You are welcome to 3D print one for yourself. It can be done with a standard extrusion printer, since both halves start from a base and there are only modest overhangs. I used PLA.
This is one of the better passthrough chess pieces I’ve seen around so far, nice work both you and Bob!
Is it OK to ask for the .stl here? I currently cannot open grasshopper files.
I wonder if it would be possible to make an entire set like that as to store it with less parts, having the black and white pieces be stored easily inside each other
Scaling my stl file will also reduce the clearance between the two parts in proportion. You seem to be fine, but would need to back to Grasshopper if the fit was too tight.
I had to test and tweak a little, but it was easy enough to get it to work. I used an A1 Mini though, which admittedly can print very accurately (0.1mm clearance tests print correctly).
Initially I printed both parts at 0.16mm layer height, but around the wavy areas near the top that caused a rough surface, especially in the silk filament. That made the parts slide less smoothly.
Printing just those few areas at 0.2 and the rest at 0.16 helped a ton. At that point both parts would slide together with ease (with elephant’s foot compensation dialled up a little), but they would not fall out like in the vid.
To get it to fall out, I just had to tweak the X-Y contour compensation with 0.1mm for the inner part, at which point you get what you see in the vid.
Wow Michael - I am impressed. I found this thread by accident, and your method for constructing the inner and outer pieces just amazed me. I did something like this some time ago with a much simpler shape, and my GH file was far more complex (ugly) than yours: Twisted cone
I"m printing your queen now and will update this post in a few hours with the results. But I have a couple of questions about your GH script:
You said your use of rotated PFrames for making the Loft works better than a making a straight Loft first and then twisting it. Do you have any idea why this is? I’ve been twisting things for quite a while now and so far have not had any problems, so I wonder if you have found out what the differences are.
The issue of generating the proper space between the inner and outer shapes is critical of course. I did it by ScaleNU (X & Y only), but you regenerated the final Loft by using the Offset component. It seems to me ScaleNU would be easier and quicker - did you try this?
I think I can manage tweaking the Queen curve enough to make the other chess pieces - except for the Knight. I do prefer the standard design for all the pieces. I’m not above doing a copy/paste so I’m going to look for an STL file I can decapitate and re-use the head part.
Thank you for the reply. I’ll do my best to answer your questions.
Twisting introduces complexity. Similarly, if you project a planar circle on to a complex 3d surface, the number of control points needed to represent the circle can increase dramatically. See here, both models 1 and 2 were created using my script by adjusting the angle of the pFrames. The third model is the twisted version of the first model. Notice how complex the surfaces have become shown by the number of isocurves on the model. Exporting these models as STEP files you can see there is much more information needed to represent the third model.
ScaleNU and Offset are two quite different things and for good reason. Below, the original curve is red, the offset is green and the scaledNU is white. The scaled version crosses over the original curve and the models would clash if printed this way and would not fit. There may be exceptions to this but it would depend on the shape of the curve and exactly how much it is being scaled up or down.
Thanks for that extra explanation - it’s clear that for the general case of geometry that has both positive and negative curvatures offset is the only method that works. The geometry I used for the twisted cone was simpler than this, so for that scaling did work.
I had a little trouble at first getting the 2 parts together and apart but with a little effort I got them to go together and apart ok. I also changed your twisty part to have a wider base so that the twists show through the entire base. I’ve also got the pawn piece (no twists} and will start workout on the bishop next.
