generatively designed crank spider

after multiple messages back and forth with the ebay seller to make sure the crank was exactly what i wanted i got this.

every cloud has an aluminum lining so I decided to use my mates big brain (and his cnc machine) to mill a new one.
Fusion 360 has just come out with generative design which takes force input and then uses an algorithim to find the optimum amount of material to perform the task. This is obviously massively simplified but it's worth a google.
Enjoy the following image dump.

6061

early iteration

I was a little bit worried because the later iterations seemed to be held together with aluminum toothpicks and faith. But this one seemed viable

All toolpaths

rearside complete

and it fits! Pretty pleased as I modelled the spline with a pair or calipers and a lot of assumptions

And the front. We added 2mm material overall to be on the safe side as i needed it asap. Lime bikes were putting me in debt.

in all honesty a bit of fettling was required as i had modelled the angle of the crank from the face incorrectly, nobody's perfect.

cut the tabs and filed.


Tada. Unfortunately it looks a lot better from the back since much more material is removed as it flares towards the spline profile. All you need is good ideas, better machines, and sick mates. Get to it!

Excuse the shite editing, I'm lazy! will put future skill points into patience and photography.

Bike is a 1993 kona hahanna, for more details see: https://jmwl.blogspot.com/2021/03/my-bike.html
p.s the blog is back baby

Middleburn make spiders that fit some chainrings. I love your can-do approach, but what was it about this chainring that made you go to all this effort?

Middleburn make a 110bcd road double but my bikes 1x10 and the spider is £70 (with shipping) which is mental. The aluminium stock cost me £40 overnight shipping and I don’t pay for machine time so ostensibly i did it because it’s cheaper. Beige answer is I like to make stuff and in this case the spider was a perfect excuse to do that. Also I only had to model the spline profile and chainring holes and the computer filled in the gaps. My mate has to figure out all the tool paths and deal with bullshit nuance of his machine which is most of the effort but also good practice, so everyone’s a winner. @Skülly
Edit: something else I forgot to mention is that cnc machining the first one of something is a much longer process than any subsequent parts. Once you know that your code is good you can run multiples at the same time and churn them out.

So cool. How much does it weigh?

Into this. Big ups for following it through to it's logical conclusion.

I will let you know on Sunday!

Appreciate it

Nice work - this is extremely cool. I've been thinking about something like this, since I want to put 165 mm cranks on my road bike with 11-speed Athena and the Middleburns are one of the more attractive external BB options available. I'd like to continue to use the stock chain rings but they don't make a 135 BCD spider so some custom work would be required.

I'm curious as to why the generated model turned out to be so irregular. What were the input forces specified?

It's super interesting how the algorithm has calculated where to remove the most material... I'm guessing there's more material left on the other side of your spider to where the drive side crank arm sits over, to take into account the added lateral load when pushing down on the NDS crank arm? Does that make any sense at all?

I've dabbled with algorithmically generated architecture stuff, but it's really nice seeing something physical being made/used!

Loving the use of generative design with cnc - this is great fun.
Excellent work - I look forward to hearing how you get on using it in anger.

Sometimes the ‘toothpick’ bits arise due to loading case anomalies so a bit of intuition is essential or you’ll be in trouble the first time you hit a tree root or load case that wasn’t part of the optimisation!

It was 1009nmm(millimetres not a typo*) The chainring holes were also under load and a bit of lateral load, although we thought the crank and bb would take most of that. All total it feels like a pretty huge number.

https://www.researchgate.net/publication/22663484_Forces_applied_to_a_bicycle_during_normal_cycling

I believe I got 318nm out of here somehow ps I have no idea what I’m doing. I’m also puzzled why it’s so irregular since the model itself is perfectly symmetrical. although there are moments of similarity it’s not the rule. The final spider is only the 11th iteration for material loss concerns previously alluded to and it didn’t seem to be trending towards similarity. Perhaps it may be a more meta symptom of machine learning in that the algorithm hasn’t yet got perfect logic or more likely the machine knows something we don’t or perhaps this is way above my head. It seems maybe @Rik_Van_Looy might know a bit more than me.

If mine works I’d be open to making some more if my mates cool with it. (maybe out of titanium #fullbodycume) although warning they will be trick as fuck but they’re likely to be pretty expensive as well.

• attached pic is of a spider modelled to hold 109nmm because we’d misread it the first time
  
  

Edit: just realised this spider is very symmetrical. We never tried this but I’m not sure if you run the same model twice that you’ll get the same results which is interesting to think about

1 Attachment

great post
great blog
need more of this shit on the internet

There is! although as you can see the al gore rhythm is pretty keen to fuck it off. There is some lateral load applied to the model but I had more concern over this than my mate so we’ll see if it explodes. I hope I’m wrong

1 Attachment

Great, another thread that reminds me how thick I am.

Excellent work, one day I'll figure out how to ask someone how to do this for me.

It’s survived the biggest 5cm bunny hop you’ve ever seen so here’s hoping to many more. We did actually leave a pretty conservative 2mm overall material (offset on each side) which is why the spider doesn’t look exactly like the model.

144bcd spider for SiSL2 arms?

need more of this shit on the internet

Yeah completely agree. This reminds me of reading about Joris Laarman’s bone chair, designed around how bone grows under load

Great stuff! Love the attitude of “Why order this for 70 quids if I can spend 40 on material and a theoretical 200 on labour and machine time” :D
See also: how I ended up buying a 3D printer because my frame was missing internal cable guides.

Unreal! This is super cool!

Do you have the means to force test it? I've done some generative design in the past and it's always super satisfying to see if you can get good correlation between your model and what you actually make.

Confused, and a little aroused.

Yeah the correlation between digital input, especially ‘theoretical’ inputs and real outputs can be almost spookily accurate. Speaking of digital accuracy it’s especially cool when the cnc has an analogue indicator on it and you zero the spindle by ‘hand’ because you can set it to move a micron at a time and watch the precision on the indicator. For reference spider web silk is around 3-8 microns. But to answer your question I’ve only got my legs, I’ll keep you updated

Awesome! Thanks for sharing

Awesome! Don’t understand any of the tech behind the design process but the end result is a thing of beauty.

Does your cnc machine log processing time - it would be interesting to know how long it takes to cut that out of the block of stock?

Isn't it now time you changed your name to Spiderman?