Appropriate FOS for Timing Belts

Has anyone done any independent load testing on HTD timing belts and pulleys?

Referencing this document (page 43)…

The recommended ratings from Gates seem fairly conservative (which is fair for their target market). For example, a 44T, 15mm pulley (close to what’s on the AM14U, or the 2013 kitbot) would have a load rating of 17.7 Nm. Two CIMs without a gearbox put out 4.8 Nm, so a gearbox reduction of 3.6 would put the setup right at the brink of failure by the books (this is assuming a very worst case scenario of course… but bears consideration).

Is there an appropriate safety factor (which would be less than 1) to use with these numbers? Or are we just shooting in the dark?

Can we crowdsource a safety factor by gathering wheel diameter/pulley diameter/gearbox ratio/motor count on several datapoints? Fill out this survey, maybe we can!

I think it’s worth mentioning that those numbers (afaik) are meant for when belts might start skipping, not when belts would break.

While it isn’t quantitative, we’d start stripping 9 mm belts in our drivetrain by the end of a competition in 2019. We’d usually go a whole season with 15 mm belts. 2019 had 6" pneumatic wheels driven by 24t pulleys, with a 24:9 torque reducer (24t:9t) from a 3 Cim 2-speed Ball shifter (Vexpro), driven by 2 NEOs.

Short of sticking a bar through the spokes, you don’t usually stall a drive train. If you assume that at most 1/4 of a robot’s weight (~156#/4 ~39#) is on any given wheel, and 8 inch wheels with a coefficient of friction of one, the maximum torque being driven by a belt on such a drive train is about 13 foot-lb, or very close to that 17.7 Nm.

1 Like

Crowdsourcing timing belt data was talked about a while ago here.

For mechanisms with specific loads, I saw issues with FOS around 1 for belts. After shooting for minimum FOS of 1.5, and usually ending up above a FOS of 2, all of the issues went away. Is it conservative? Probably, but never worrying about belts breaking or skipping was nice.

You can do the basic physics calculations for belts on a drivetrain as a starting point, but driving style is a huge factor. I would take the drive train belt feedback with caution and do your own testing to match the teams driving style. Since this is a dynamic problem, dynamic testing will provide the best results.

The rated torque numbers from the Gates catalog is for near infinite life. It has nothing to do with torque to ratchet or tensile strength of the belt. 9mm belts are marginal even with 2 CIMs. They will be even more so with falcons or Neos. Not to say they will not work. We have successfully used 9mm belt and have lived with the occasional ratcheting and belt failures.

It’s an interesting discussion regarding infinite life or a shorter life. I did not have good luck going under the infinite life timing belt load ratings. Getting the data on a shorter life cycle would be difficult. It would be nice to see the structural material properties of the timing belt composite materials and see where the load ratings land in the elastic range. They don’t specify (at least what data sheets I have seen) what FOS they set their load ratings at.

Some may view an occasional ratchet/skip and breakage acceptable, others may view that as unsuccessful. I’d be in the camp of a ratcheting/ skipping or breaking timing belt/chain as an unsuccessful implementation more often than not.

1 Like

If you design for stall torque of 2 (3?) Falcons and low gear of a two speed transmission using the Gates published rating–you are going to need bigger belts. It is also not a condition you are going to see very often. For the routine case industrial you want to design so you do not ratchet or break the belt. The fun fact about engineering design: Most cases are not routine. Another example: Rotation gear boxes for arms are routinely designed for fast rotation and low stall currents when holding the arm in position. If you design the arm to withstand full torque, you end up with something too heavy to put on the robot. You can implement driver training and software controls, but it also good to have a weak link designed in that is easy to fix. Better to ratchet than to bend the arm.