RE:
I will redesign the part so it'd take inserts well (we're doing thousands of those in SLS and FDM).
In case you missed it, I already have a model on Printables that is designed to use a threaded insert.
RE: Tension pulley broken
For comparison:
RE: Tension pulley broken
ChatGPT gives a good explanation of the benefits of increasing the thread length. So apparently there's little to no benefit in going longer than the 5.7mm insert.
Practical Takeaways
Increasing thread engagement improves load distribution and reduces stress on individual threads, but not in a directly proportional way.
For M3 threads in steel, 3 to 5 threads are usually sufficient to fully develop the bolt's tensile strength.
Beyond 5–6 threads, additional thickness has little benefit for tension loading — the bolt will likely fail in tension before threads strip.
Summary
Force per thread does not reduce linearly with nut thickness.
Tripling thickness will reduce the per-thread force, but not by a factor of 3.
Most of the load is carried by the first 3–4 threads, regardless of nut thickness.
Aim for at least 1× bolt diameter in thread engagement (so ~3 mm for M3) for good strength.
RE: Tension pulley broken
ChatGPT gives a good explanation of the benefits of increasing the thread length. So apparently there's little to no benefit in going longer than the 5.7mm insert.
But that reply does not take the lateral (tilting) forces into account, right? It only seems to discuss axial force. For the tilt component (due to the laterally displaced pulley), I would expect a linear reduction of forces in proportion to the lever length (i.e. thread length).
RE: Tension pulley broken
ChatGPT gives a good explanation of the benefits of increasing the thread length. So apparently there's little to no benefit in going longer than the 5.7mm insert.
But that reply does not take the lateral (tilting) forces into account, right? It only seems to discuss axial force. For the tilt component (due to the laterally displaced pulley), I would expect a linear reduction of forces in proportion to the lever length (i.e. thread length).
Good point. I've been following up with the specifics of this scenario. I don't want to just dump a load AI text here, but the final thoughts are encouraging. I'll now add the detail about the lateral forces to see how far it's analysis can go.
🧠 Final Thoughts
You're solving two problems at once: increased load safety and eliminating galling.
Just ensure the brass insert is:
Properly secured (press-fit, heat-set, or bonded)
Thread class matches the M3 bolt (ISO metric coarse)
This is a smart, effective improvement, especially for repeated assembly/disassembly or in environments where galling is a chronic issue.
RE:
It's been an interesting read! In my opinion, this is the sort of thing that AI is good at. Here's its summary after I described the offset loading. (click on the images to see all the text)
Finally I asked whether a longer insert would be beneficial in this scenario.
RE: Tension pulley broken
In case anyone wants to see everything that ChatGPT had to say on the topic, here's a link to the discussion: https://chatgpt.com/share/687f62d9-62ac-8003-bb3f-ce64f21cac34
I'll be the first to admit that ChatGPT is often caught 'bullshitting with confidence', but I didn't see anything that stood out as plain wrong.
RE:
It's been an interesting read! In my opinion, this is the sort of thing that AI is good at. Here's its summary after I described the offset loading. (click on the images to see all the text)
[...]
Finally I asked whether a longer insert would be beneficial in this scenario.
[...]
Seems pretty solid -- although I am missing a quantitative justification for the final assessment. Please take ChatGPT's tendency to "please the user" into account. You might want to try opening a new chat where you ask the very same questions, but tell it that you have designed for a 10 mm long insert. I'm wondering whether that will come out as the "ideal" solution then...
RE: Tension pulley broken
Seems pretty solid -- although I am missing a quantitative justification for the final assessment. Please take ChatGPT's tendency to "please the user" into account. You might want to try opening a new chat where you ask the very same questions, but tell it that you have designed for a 10 mm long insert. I'm wondering whether that will come out as the "ideal" solution then...
That's an excellent question, and you are right to be skeptical. (hehe - I'm picking up its style)
You'll see from the chat link that I tried to be agnostic, but since I did specifically ask about 5.7mm it may well be trying to please me I guess. But you can see my questions in that link - it could be an interesting experiment to see if you get the same responses, and maybe you could initially suggest the longer insert.
RE:
I had not seen your chat link before I wrote the previous post. Thanks for sharing it!
Sure enough, when I enter the same questions but describe that I plan to use a 10 mm insert, that's what ChatGPT encourages me to use. I don't get the exact same recommendation table it provided to you, but the one below. Go figure...
In general, I don't trust ChatGPT for anything quantitative if it requires math (i.e. the specific ready-made result cannot simply be looked up somewhere). Where I find it most useful is when a large body of specific factual knowledge needs to be combined with basic reasoning. Programming tasks which require knowledge of complex APIs or large libraries are a nice example.
RE: Tension pulley broken
Haha - that's very telling isn't it. What good is an obsequious AI?
Anyway, they both seem to agree that a brass insert is better than the stainless nut.
I'm inclined to ask it again, but this time suggest a shorter insert. I have some here that are 4mm long - let's see what it thinks of those.
RE: Tension pulley broken
It was enough to just start a new session in order to get completely new answers.
Here's the final summary when I lead the discussion with a 4mm insert.
RE:
I don't trust a chat GPT one bit. BUT I just had threaded connections course as part of my Mechanical Engineering degree.
I will print your design and report the results, thank you!
The thread loading in such connection looks like that:
I can write a bit more later, when I will have a bit of time to look it up in my notes.
The threaded insert is more susceptible to being sheared out than a square nut. CNC kitchen had tested it.
RE: Tension pulley broken
In all of the failed parts that you've had, have you tried a bit of grease or anti-seize on the bolt? There are lots of folks out there running their tensioners dry (as they came) and haven't had failures yet. I understand the desire to design something perfectly (I share this impulse), and I agree that the use of a thin square nut is poor design...but if there is something small that can make it work, why not start there?
-J
RE: Tension pulley broken
The threaded insert is more susceptible to being sheared out than a square nut. CNC kitchen had tested it.
I tried to take this into account in the design - there is a 1.5mm thick shoulder for the end of the threaded insert to sit against. This is why the insert has to go in from the pulley side. I think this should be pretty strong. There are reports of people using inserts already, and I suspect that they have simply pushed them into the original mount (complete with nut slot) from the bolt side, and I've not seen anyone reporting failures. So I think it's very unlikely to pull out.
RE:
edit: Seems I'm late to the party, but that's ok...
Strength isn't the problem; steel would only help with longer thread area but is still at risk for galling on a threaded part. For the tensioner, aluminum may well be the better option if you're doing both machined. Brass or bronze would do the trick really nicely for the pulley. The biggest value I see here is that you'd also be able to safely step up to a larger diameter bolt (e.g. part walls can get thinner without losing necessary strength).
Inserts may be the best overall solution though, all things considered. And a redesigned pulley was shared already too.
Galling is the word I was missing, thank you! In my opinion that's the mechanism I was trying describe.
That's main reason, why I suggested CNC milling those parts from steel, not from aluminum.
Another idea is using steel for tensioner and phosphor bronze for the pulley. This will create a good self lubricating effect.
I will redesign the part so it'd take inserts well (we're doing thousands of those in SLS and FDM).
I also have Siraya Tech Fast Mecha and even some POM filament lying around... Maybe I will give those a go?
If the parts will be milled out of metal, it would allow for some dovetail that will make camming the pulley out from the tensioner less likely.
I did tolerance test before printing the tensioner, and 0.05 was very tight and 0.0 doesn't fit, so when I saw "out of tolerance" dimensions on the pulley, I was convinced it's by design. I will give your suggestion a try.
RE: Tension pulley broken
Honestly, I'd make the hole the right size that it requires threads cut in - maximize the amount of shoulder. I don't like that you can see so much of the insert through.
The threaded insert is more susceptible to being sheared out than a square nut. CNC kitchen had tested it.
I tried to take this into account in the design - there is a 1.5mm thick shoulder for the end of the threaded insert to sit against. This is why the insert has to go in from the pulley side. I think this should be pretty strong. There are reports of people using inserts already, and I suspect that they have simply pushed them into the original mount (complete with nut slot) from the bolt side, and I've not seen anyone reporting failures. So I think it's very unlikely to pull out.
RE:
Honestly, I'd make the hole the right size that it requires threads cut in - maximize the amount of shoulder. I don't like that you can see so much of the insert through.
You've forced me into an experiment!
That's a 14lb doorstop hanging via the tension screw, suspended by a bit of coat hanger through the spindle holes.
In case you're wondering if that's strong enough, I asked ChatGPT to do some sums for me.
A 30cm belt tuned to 105Hz is likely carrying ~1N of tension (depending on the assumed linear mass density). Since the idler has a loop of belt around it, each side of the loop carries this same tension, making a total tensile force of 2N. That would be equivalent to hanging a weight of ~0.2 Kg from it.
My 14lb weight converts to 6.35Kg. That's a decent factor of safety.
If that much force was applied to the pulley, each side of the 30cm belt would vibrate at around 600Hz.
I think it's ok as is 🙂 (Assuming ChatGPT isn't bullshitting me again).
Seriously though - if the hole wasn't a clearance hole it would compromise the ability to use a screw to pull the insert in and align it properly while the plastic is soft, and it would also make life difficult if the insert was very slightly off-centre.
(Oh, and I've just realised that I only screwed the tension screw in about half way along the threaded insert. Obviously I didn't try turning the screw while it was under that load (I only have so many hands), but it's still turning freely now.)
RE: Tension pulley broken
I think ChatGPT is bullshitting me again!
I think it was way off with the linear mass density.
Checking......
RE: Tension pulley broken
Hmm - if I use the linear mass density that I measured myself (by weighing the two lengths of GT2 belt that I removed), then it turns out that my 14lb weight is almost exactly equal to the tension being carried at 105Hz. I even checked it by suspending my weight with a loop of belt (30cm/side) and measuring the frequency. Bang on 105Hz.
So no factor of safety at all. That'll teach me not to be so smug. I need a heavier weight to check with.