Sizing bearings for elevators and telescopes

FIRST General Forum
#2

Expected loads
Fudge room for impacts and competition
Packaging space
What you used the last time that worked
Overlap length in the sections
If you don’t know about the effects of L/D ratio on sliders, you -really- want to look it up.
Radial load rating for the bearings
Number of bearings used
Load sharing between multiple bearings
Acceptable friction (in case you use sliders rather than bearings).

The contact stresses between the bearings and the rails they run on are probably not a major factor in FRC; Aluminum vs steel spreads the load a LOT more than hard steel on hard steel. Nylon tires on Aluminum spreads it even better!

Keep in mind that your factor of safety needs to be related to how much you know about the system and how its used. In the case of FRC, our knowledge of what’s going to happen in competition is poor. We can design around the known loads, but we need a lot of room for “oops I just ran into the scale at top speed” or “I ran the elevator up into the bottom of something”.

4 Likes
#3

Honestly, we went with skateboard bearings because they are cheap and easy to get. You can get a 100 pack on Amazon for $23 or if you want to splurge you can pay a little more and get some in any color you want and it can be delivered the next day. They come in several standard sizes (unfortunately all metric, so it does make it a little tricky to get bolts to hold them).

1 Like
#4

I will add that uncoated aluminum can wear away pretty badly if you’ve got bearings rolling on it all season. 1072’s 2018 elevator rails saw a lot of aluminum flakes by the end of the season and needed to be wiped off occasionally. We had a pretty heavy and fast carriage that contributed to that.

For me, anything super low load gets R188s, anything decently high load that gets run a lot like the front and back of a carriage gets R4A bearings. You can calculate the load on the bearings by taking the moment that the carriage puts on the bearings and dividing by the distance between the bearings (drawing a force diagram helps with this calculation) and use that to spec the bearings as well, with a healthy factor of safety.

Metric bearings from China are very cheap (as are skateboard bearings) but metric screws are risky to keep in stock. Maybe if I had more of the blue ones from McMaster, or put the bearings on a turned shaft instead of a screw.

1 Like
#5

These are good reasons to go with 8mm i.d. bearings: 608, 698, 688 to name a few.

5/16" is just slightly smaller than 8mm (~0.002" difference) so for many FRC applications of these bearings, you won’t need to stock metric fasteners.

However, I wholeheartedly recommend that students get used to designing with metric bearings and fasteners. Some key industries went that route long ago (automotive is a big one), while others (like aerospace) lag behind. Someday, the inch will be as quaint as the cubit. Of course, my high school physics teacher told me exactly that about fifty years ago, so I am glad I didn’t hold my breath.

11 Likes
#6

definitely agree on the wear aspect: our 2015 robot had serious issues with the elevator carriage skipping out of its tracks. it was running vee bearings in rev extrusion, and after a couple comps we just could not keep it tensioned well enough due to the both extreme and uneven wear. in hindsight, plastic v bearing wheels would have been the better choice…remembering the hierarchy of materials is sometimes important :stuck_out_tongue:

#7

“Americans will measure with anything except the metric system”.

5 Likes
#8

To give a better answer than “idunno, we get away with R4s”… You should actually be able to analyze the required bearing size. The particular load case which should be important would be elevator fully extended, with payload, 2G (?) of acceleration (or slamming into something with some odd amount of force, idunno). Perform the force analysis to determine the load in the bearings (pretty straightforward lever analysis).

Rating bearing life is a little trickier - do some homework here. SKF

=============================================

Something to chew on: there are some pretty cheap needle-roller bearings out there. (McMaster-Carr 5905K73 with .473" OD has a similar load capacity to an R8 bearing with 1.125" OD) . Construction > size.

You’d need to use them on a shoulder screw (and maybe put an extra sleeve on the outside, the outer race is usually sheetmetal) but could be an interesting option if you need high-load in a low-space situation.

Big fan of using 608 bearings, or metric bearings in general. It is much easier to source metric bearings, and there are many more options. Amazon is a surprisingly decent place to get bearings. As for stocking hardware, IMHO it’s just as easy to get metric as english hardware anymore. Especially if I want something like a shoulder screw. Then again, I live in an area where the hardware stores’ selection of even english fasteners is rubbish - if I want hardware, I practically have to order it. COTS compatibility is the only real drawback.

#9

Almost more important than bearing size is what your minimum bearing spread/spacing is when fully extended. As this number approaches zero, the bearing forces go to infinity as well as a fixed amount of slop/deflection will lead to more angular error for next link.

It’s nice to stay in the 5"+ range for elevator carriages, 6-7"+ for the intermediate stage to robot frame, and as low as 1.5" for telescoping climbers (but I’d advise more if there are high lateral shock loads). You should always do more spacing if your design allows, and can certainly get away with less than these rough anecdotes but they’re good starting ponts based on fairly large dataset.

15 Likes
#10

shhhh don’t tell them the american system has been metric under the hood since the 1890s

#11

When applying these rules of thumb, do you base these on hard distances, or as percentages of range of motion?

#12

I didn’t even know these were a thing. I’ll certainly be placing an order for some of these M4 bolts for our motors. We’re constantly searching for the few we have in season when working with BAG or 775 motors. I never wanted to stock them heavily out of fear of students (or even myself) mistaking them for an 8-32. Having the metric ones be colour coded would be a game changer!

#13

+1 on AdamHeard’s comments! The spans are CRITICAL!!! If you don’t have enough overlap on your sections the forces spiral to infinity, and the coefficient of friction required to allow motion -at all- gets lower and lower. You can’t make that number negative unless you have powered wheels on your rails.

Here’s a good whitepaper talking about the ratio between length and width of bearings and friction. Pay special attention to the fun graph of coefficient of friction vs ratio of lever arm to bearing length! For most of our elevators, one of the dimensions is VERY wide relative to the span of the bearings. We REALLY need to be planning in ways to avoid racking and binding in that dimension! The thin part of the elevator always has a better ratio.

Here’s a page that clarifies the concepts with cantilevered loads:
https://www.pbclinear.com/Blog/2012/February/Cantilevered-Loads-and-the-2-1-Ratio

Hmm… that’s leading me to an interesting (and probably stupid) thought. What about an elevator where you had chain or belt synchronized wheels running up and down two rails? Even better would be belt sprockets running on a tensioned belt that was strapped to the rail. That could combine the actuation and bearings into one terrifying system! Complete racking control in the bad direction!

2 Likes
#14

Also, for a center-lifted carriage, a low height:width ratio on the carriage can lead to catastrophic failures if the carriage goes crooked. This happened to 1072 in 2018, and led us to change to a two-point lift in the offseason.

3 Likes
#15

This one has more discussion about wide carriages and setting up a fixed/floating system to improve this ratio. I’ve never seen a FRC carriage that wasn’t massively over-constrained… That means there are two or more bearings defining any given axis or rotation. This leads to systems that need to be -perfectly- parallel to work at all, and bearings fighting each other. Timing both sides of a wide system to force them to move identical distances can compensate for this problem -some-.

Sorry, its one of my windmills :wink: See my almost-exactly constrained 3D printer design: WeldingRod Bot: Exactly Constrained 3D Printer by Weldingrod1 - Thingiverse I funded a Kickstarter and later realized that the designers were completely ignorant of the concept of constraint. They were trying to have four v-wheels in grooves on all four vertical posts and were surprised that the resulting system was sticky and didn’t want to move smoothly, even with extreme levels of careful adjusting. I offered to help them, since they were also in Houston, but was rebuffed. I completely re-designed the printer and ended up sending plate sets out to 10+ other owners.

A key idea on this unit is having two types of V-bearing systems. One side works in the V so that it can’t translate perpendicularly relative to the extrusion. The far side has wheels working on the FLAT parts of the extrusion, so that the two extrusions don’t need to be perfectly parallel.
Here’s one side:

image
image936×1011 204 KB

The other side runs on the flats. Note that both sides are driven to limit the need for rotational constraint around the Z axis. Geometry dictated that the flat side couldn’t have just two wheels on flats (there’s a bar in the middle and vertical load to carry). The far end of the left and right extrusions use a two wheel on flats system to only constrain one rotation.
image
image793×1005 129 KB

#16

These numbers are ballpark good for 20-40" of travel per stage.

Lower load or stroke systems could deviate.

#17

688 flanged bearing (received same day from Amazon, $13.99 for a 10-pack) shown mounted on a 5/16-18 x 0.75" BHCS with a standard nut. Bearing diameter 16mm, flange diameter 18mm.

These bearings are usually rated for about 100 pounds static load, about 1/3 the rating of a 608. Probably not a good choice for a tall elevator (think 2019 or 2018), but could work well for a smaller sliding mechanism or shorter elevator. Can you tell I have been thinking about tiny FRC robots?

image
image818×685 120 KB

1 Like
#18

Following this setup on an FRC robot is a guaranteed “carriage falls out” situation. While technically we do over constrain elevators, clearances and flex create a low friction mechanism that’s less likely to pop out.

I did try to constrain an elevator with flanged bearings once and ended up with worn and sharp corners on the tube. But maybe it works better in certain situations?

#19

The method we’ve taken over the years to size and make elevators/telescopes.

  1. Use the largest bearing possible. Which for us has been .25" x .75". If that doesn’t fit use a .25" x .625" → .25" x .500". We haven’t found a need to step up to .375" x .875" yet…

  2. Do the largest span/gap possible.

  3. Make your mechanisms easy to maintenance, which for us has meant to let items be bigger than we want. Instead of fitting a telescope in a 2" tube with small gaps (cough 2020 hanger), go to a 2.5 or 3" tube with larger gaps (greyT). Many times we catch ourselves thinking we need to make something super tiny or fit into a small space, but often realize we had way more room IRL.

3 Likes
#20

Well yeah, that follows logically. The automotive industry just needs things to be cheap, so they use metric parts. In aerospace, we need things to be reliable, so we use standard parts. :wink:

3 Likes
#21

As one who left the automotive industry to join an aerospace company a couple of years ago, I resemble that remark.

Seriously, both industries have their standards. National Aerospace Standard (NAS) parts meet very stringent quality requirements, and their pricing reflects that.

Quick example: 1/4-20 x 1" Alloy Steel SHCS, corrosion resistant

Aerospace part (NAS1352-4-16P): 92562A271, $3.67/EA
Non-aerospace part (ASTM B117): 91274A172, $0.24/EA ($11.98/box of 50)

Oh, and metric NAS parts actually are a thing. :wink:

2 Likes