DIY linear bearings (no PLA)

So, I have been doing some research in the last some weeks and learned a little bit about tribology and polymers after investigating plastic linear bearings.

One rather interesting material I found was UHMWPE (ultra high molecular weight Polyethylene). Basically, it is PE with extremely long molecule chains (which amplifies its positive properties, to put it simply).
In short, what we have got here is a self-lubricating low-friction (2nd or 3rd lowest after PTFE [and POM, according to one source]) polymer that is also quite resistant to abrasion and reasonably easy to machine.
Apparently, this material is being used in guardrails and slides in production lines, knee prostheses (which last up to 10 years) and also linear bearings - which is what this topic will be about.

The reasoning behind trying to manufacture bearings using UHMWPE is that - with the caveat that one would have to include subtractive tools in the RepRap setup - one could reduce the imported vitamin count of the base machine without adding excessive manufacturing complexity, thus coming closer to the original goal of a self replicating machine (from which the community has - not necessarily negatively - strayed a little bit in my opinion). IF these bearings prove to be feasible in terms of wear and ease of implementation into machine designs, users may hand-craft or let a stable RepRap mill them.

I ordered one 250x100x25mm sized piece for 12€ and built three bearings to test the feasibility of implementing such a contraption in a printer.
Also, I constructed a testing setup that is supposed to slide a carriage equipped with these bearings back and forth on the standard 8mm hardened steel shafts and report back to my computer about the number of runs it made.
Wear values obtained from these tests will be evaluated with the total distance run, derived from the saved number of runs.

Bearing, opened. Tension adjustable with screws


Bearing on shaft


Test carriage


Test setup

Please ignore the ears and grey corner...

The way the linear bearings are mounted right now is not optimal. The upper half is not tightly affixed to the carriage so that it requires quite a bit of force to overcome the potentially cantilevered bearings. I will try to remedy that with double sided tape tomorrow (the final design should be less hacked together). The bore is also quite tight as of now.

So far, this setup seems to be quite promising: As long as the bearings are straight to the shafts, friction is reasonably low even without lubrication (but still noticeably higher than the standard LM8UU's) and the carriage sits quite firmly without much wobble on the rails.
When I have completed the wear tests with the blank bearings, I will also test the performance with PTFE dry lubrication or grease etc.

Edited 1 time(s). Last edit at 01/17/2013 04:51PM by uGen.

Nice writeup! I'll be interested to 'ear how they hold up.

But seriously, what's with the ears and the grey corner ?!

Also, where did you buy the UHMWPE? Sounds like an interesting material.

Cheers,
Gary

---

------------------------------------------
garyhodgson.com/reprap | reprap.development-tracker.info | thingtracker.net

Thank you! I am currently writing a simple logger in python, so usable results should be in in about a week or so.

haha, I was bored one day and printed the ears for fun (found them on thingiverse). Just put them on the carriage for fun. And the grey corner is because I rotated the photo, so my image editor added a grey background to fill the new image.

I bought the UHMWPE here: Technoplast Kunststoffe, but there are a lot of sources including ebay where you can get it. They also sell it under the name "PE 1000". I was actually quite amazed how easy it was to get some UHMWPE as I did not hear about it before and as the specifications made it sound like some special high performance material like PEEK. In fact, it is listed among the "standard" plastics!

That's a great resource! Thanks for sharing!

---

------------------------------------------
garyhodgson.com/reprap | reprap.development-tracker.info | thingtracker.net

Update:
- After initial running-in, the bearings are much smoother now and require less force to move.
- Compared to the ball bearings, noise levels are lower and motion "slicker" - as in you don't feel balls circulating.
- Running the bearings without lubrication seems to work well (as I stated before, UHMWPE is self-lubricating; I did not expect such good results, though).
- The motors have enough torque to pull the carriage (you probably may have to set them a little bit higher in some circumstances).
- Play is minimal - though I don't have measured it (yet), it seems to be on par or better than with cheap linear ball bearings. Will confirm this tomorrow as it could also be an artifact of my test setup being shorter and better supported than the X-axis of the Mendel90 with which I did a quick comparison.


The advantages of plastic bearings in general are:
- low inertia, allowing for extremely high accelerations (probably way more than our printers can handle). At some point, the balls of the regular bearings slip when accelerated too quickly, leading to increased wear and decreased performance.
- dirt and debris on the shaft can be embedded into the plastic, removing wear-increasing elements from the system (as compared to plain metal bearings)
- self-lubrication and low maintenance

The disadvantages are:
- deformation under load, more pronounced dimensional changes due to temperature fluctuations and intake of air humidity (the effects of latter on UHMWPE and some other plastics like POM are negligible, though)
- wear. Compared to ball bearings, wear and thus service life values are worse. Depending on the application, you might have to change plastic bearings every few years, although some can also last for very long.
- increased play over time - obviously due to wear. I tried to counteract this by making the bearings adjustable. A three-point contact design like that with the bearings of the Sells Mendel might be better than my current design.

Thanks for the follow-up.

One thing worth keeping in mind when comparing with bearings is how good are the rods? After less than a year with bearings, my rods are developing grooves where the balls run, so I have to either rotate the rods slightly or replace them (with harder, and hence more expensive, ones). If the UHMWPE does last a few years it may be comparable (taking into consideration the other pros and cons).

Cheers,
Gary

---

------------------------------------------
garyhodgson.com/reprap | reprap.development-tracker.info | thingtracker.net

Oh, did not expect the rods to be worn that much as I have not seen the wear on our workshop's machine (about 3/4 year old). I don't know the exact composition of the rods, though, as the printer was built by the workshop instructor.

Thank you for reminding me: I forgot to mention that plastic bearings have a run-in time in which friction and wear will generally be a little bit higher than in normal service duty afterwards. That is because they take off the ridges of the shaft and fill indents with plastic debris. After this, the shaft should be covered in a very thin layer of plastic abrasions that provide lubrication, so theoretically, the shaft should last longer. Also, due to the larger interface between bearing and shaft, pressure is distributed more evenly.
However, different plastics have different optimum surface roughness. UHMWPE is better suited for shafts with 0.3µm average roughness whereas PA and POM (also a contender for linear bearings) prefer 2µm.
Cf53 ground steel rods like these should be suitable and reasonably priced for UHMWPE.
Another problem I forgot to mention is thermal build-up in the plastic bearings. The disadvantage here is that UHMWPE has got a relatively low maximum usable temperature point (around 60-80°C IIRC). That combined with plastic being a good insulator might lead to problems in a big print or with shafts heated by the hot air raising from the heated bed. I will try to examine frictional heating in my tests, but I think it should be generally safe to assume that there won't be that dramatic a heat build-up.

I'd be interested in how UHMWPE compares to the other "self lubricating" plastics notably Acetal and Nylon.

I have done a little bit of research on that:
Nylon (PA) in some variants seems to be excellent for linear bearings (with some being better and some performing worse than UHMWPE in regards of friction), but a big drawback is that it is way more susceptible to intake of air humidity, complete with dimensional change whereas UHMWPE has near-zero humidity intake. POM is somewhere in between, with a comparable coefficient of friction, better thermal dimensional stability than UHMWPE (which performs worst of all three materials in this field; just read that for continuous operation, maximum bearing temperature is 50°C for UHMWPE and 80-90° for PA and POM).
PA has lower wear than UHMWPE, which in turn has significantly lower wear than POM, but I would be cautious with these values because they depend on the material on which the plastic glides. In my research, I have seen a lot of people speaking very positively about POM for these applications and less for PA or UHMWPE, but I mainly concentrated on information about UHMWPE (and to a lesser extend, POM).
Conclusively, both UHMWPE and POM seem to be very suitable, with POM being popular because it is supposed to be easier to machine (you might need sharp tools for UHMWPE, but I found it reasonably easy to machine) and probably also because it is more well-known.
POM seems to be more expensive than UHMWPE, but I will try to get some POM samples in the future (might also be useful for plastic nuts).
Another interesting material is PET, which also has got nice tribological (science of interacting surfaces in motion) properties. It definitely feels a little bit harder and more brittle than UHMWPE.

Hi Guys, I am using graphite impregnated nylon bushes. They are running on silver steel shafts. I have them installed on my Y axis and will probably fit them to the other axes' as well in due course.

Sounds interesting. Do you notice the graphite on the shafts at all? How is wear and play?

I have some acetal LF13 which has 13%PTFE powder mixed in, I should probably run some tests on it, it's supposed to have a lower coefficient of friction than straight acetal, and wear slightly better.

Actually, I was thinking of using PTFE dry lubrication if the bearings prove to be difficult to move / wear too quickly, but a bearing with PTFE included further simplifies the whole thing.
But then, I went on to search for LF13, just to see how expensive it is and how to source it:
All sheets are 0.25'' x 12'' x 12'', other materials thrown in as comparison

PA: 29.00$
PEEK: 405.58$
PET: 53.81$
PTFE: 62.25$
POM-C: 25.20$
POM-?: 20.85$ (no description which acetal variant this is, but I guess it's POM-H)
POM LF13: 155.19$
UHMWPE: 8.98$

source: OnlineMetals.com

POM has one big advantage over UHMWPE for our applications in that it seems to be melt-processable. Some quick googling suggests that it is hot weldable, so one could actually print bearings from a suitable material. The drawback here is that when heated over 230°C, it releases formaldehyde, which is carcinogenic. Luckily, it melts at 168°C, but still, I would only process it with adequate ventilation and / or air filtering.
UHMWPE is extrudable (although it will be extremely viscous when molten), but to weld it, you have to apply 20 bar of pressure IIRC.
MSDS for POM-C
MSDS for UHMWPE
Generally, POM seems to be more of a potential health hazard when the limits of the material are not observed, but the advantage of being melt-processable seems very promising.

Update:

Today, I got my whole setup running in logging mode for the first time after programming in more features than initially planned. It held up quite well until after about 1000 actuations, the stepper either lost a step or the pololu temporarily shutdown (could not determine the exact reason because I immediately stopped the test to prevent the carriage from slamming into the end stops). Will try to exclude pololus running hot tomorrow.

Another thing I noticed: when accelerating by hand quickly, there was some noticeable stuttering. I first attributed it to some stick-slip effects, but after readjusting the floating bearing, the effect is gone. Maybe it was caused by some minor misalignment. Still running without lubrication, by the way.

I used to design conveyor systems a few years back and I can support using UHMWPE for bearings as it is highly suitable for what you are using it for. In fact I would suggest it is the best available. One key use is for chain guides and tensioners. It is also used for this in automotive engines for timing chain guides.

Another material to consider would be Oilon. This is basically oil filled nylon and is specifically designed for bearings. Its easily obtainable and is cheap. I use the green stuff, but it is also available in red which I think has slightly different properties. I am not sure you would be able to print this though which is what you are aiming to do, but it might be a cheap alternative if you were to drill and cross cut some bearings.

Using Oil on plain nylon is a really bad idea as nylon swells up when exposed to oil and can cause binding. Oil can also attract dirt which can reduce bearing life. Its all a balancing act really.

One thing that did occur to me was that you are using cylindrical shafts. Have you considered using a square or rectangular bar and flat pads of UHMWPE. You might only need one rail and also due to the larger surface area you will probably get better wear characteristics. You could print two vee blocks and clamp them diagonally. In addition printing the bearing surface on the heated bed makes it very flat which would help.

- martinprice2004: Nice to hear that someone else has positive experience with UHMWPE! Did not know that it is used in engines, though as I thought it would not stand up to the high temperatures there.

Thank you for the tip about Oilon. I just checked the price and found both quite comparable. The only thing I am unsure about is the higher humidity absorption of PA. While thermal expansion is in my eyes more predictable, changing the location of the printer might be the difference between binding bearings and good motion. Of course, one could leave a little bit more play to compensate for that, but I have to read up on PA generally first.

I used the cylindrical shafts because they were leftovers of another project. I already considered other solutions similar to what you propose, but I would like to test the bearings further before opening up too many things at once.

Unfortunately, UHMWPE does not seem to be printable. The pressure for it to weld to previous layers is way too high (20 bar IIRC). I have higher hopes for POM to be printable, though.

Edit: Looked at the Oilon datasheet. Moisture absorption seems to be a non-issue and the maximum service temperature is also 2.5 - 3 times higher than UHMWPE. I will try to get my hands on this...
Nice find!

Edited 2 time(s). Last edit at 01/23/2013 08:57AM by uGen.

uGen: How did you machine the UHMWPE, just sawing and drilling? Did you do anything special to smooth out the walls of the bearing holes?

From your links to Technoplast Kunststoffe and ebay.de i guess that you are from Germany.
About sourcing sheets of oil filled nylon:
Kern GmbH
Der Kunststoffladen
ebay

I ordered a sheet of oil filled nylon from ebay and will try to make some bearings out of it.

Edited 1 time(s). Last edit at 02/01/2013 06:18AM by Marshall Banana.

-Marshall Banana: Well, to be honest, the instructor of our plastics workshop helped me out a lot with his machines. His mindset is more "traditional", so he over-engineered the bearings by CNC-milling and reaming the holes. Afterwards, I split the bearing in his band saw.
Theoretically, one should be able to just drill the main hole with a hand-held drill - or better: a drill stand. As play is adjusted by compressing the bearing against the rails, tight tolerances should not matter that much. I would change the design of the screw holes, though, as one bearing half has to be affixed to the carriage while the other has to stay adjustable, which is not really possible with this design.

Yes, I am from Germany. Thank you very much for the links!

Also: UPDATE:
After one week of running the test carriage about 6 hours a day (amounts to nearly 10 km) with an additional weight (standard stepper motor), I finally got some really promising results.
Contrary to my initial concerns, even after long running periods, the shafts did not get hot; you could feel them being warmer than before, but insignificantly so.
Wear is ridiculously low. I could still see the ridges from the machining, so these bearings - also given their size - will probably outlast many other printer components. After all, you have got about 2mm of adjustability until both bearing halves meet. And even then, you could probably file away some mm on these surfaces until the bearings become unusable.
I could see some very fine UHMWPE powder at the ends of travel, though, so there is wear. Also, some kind of sharp debris left some marks on the bearing surface, but I could not determine what exactly did that. The upside of this is that this very debris could have been embedded into the bearing itself, preventing it from riding all over the rail and ruining it, as long as wear does not expose it again.
No lubrication was necessary throughout the tests, so your rails stay clean and dry.

Here is a picture of the upper side of one of the fixed bearings. The weight was directly above and in between the fixed bearings. As you can see, visible wear is only due to hard debris that caused the longitudinal scratches.
It was rather hard to get the lighting right for the machining grooves to be visible.



The next step: I will talk to the plastics workshop instructor about the wear of UHMWPE vs UHMWPE, POM or Oilamide. The idea is to print out a linear rail with ABS and attach strips of UHMWPE to that and run a carriage with UHMWPE, POM or Oilamide. The rectangular profile should minimize wear since the pressure per area unit is lower on a flat surface than on a round one. The play adjustment should also be easier since you have surfaces orthogonal to each other which you can easily shift with set screws. To give you an idea how this might look like:


Just printed and milled.

Actual friction force is friction coef of materials (which we cant change) *(multiplied by) perpedincular force exerted. So for a high perpendicular force, you can end up with a high friction, even if materials coefficient is low. By using 2 screws only, when the carriage is tightened, the friction is proportional to how much the carriage is tightened. At extremes you ether have a loose carriage + low friction, or a tight carriage with high friction.

To escape this common trap perhaps 4 screws are needed, 2 for tightening the bearing, and 2 for tigtening the carriage. Those for carriage would have to be tigtened to the upper plate alone, without engaging the bottom plate. Then you can have both a tight carriage and other 2 screws to independently adjust the friction force (regardless to how tight the carriage is).

Yes, I noticed this shortcoming of my prototypes already and fixed it with makeshift double sided tape. Guess this information got drowned out by my extra long posts.
The current approach would still involve two screws, though. By reversing the screws and cutting out nut traps or cavities in the carriage side of the bearing, one could use nuts to tighten this half. Then, using a second set of nuts, one could adjust the play of the second half.
Sort of like this (simplified for one screw):

HCBN-G-BN

H = screw head
C = carriage
B = first bearing half
N = nut
G = gap
B = second bearing half
N = nut

Your configuration of double nuts looke like a great way to do it. Further I would only be worried about the heat from extruder or heated bed, if and how it would affect the bearing material.

Yes, heat still is one of my worries, especially on the X axis.
UHMWPE has higher thermal expansion than PA and POM IIRC, and judging from Igus datasheets, plastic bearings in general tend to wear faster as they get hotter. I might also try the Oilon that martinprice2004 suggested and a POM variant the next time I order plastics.
Theoretically, through the carriage movements, the airflow might help with cooling the bearings, but the heat from the hotend might get a little bit more troublesome. A vertical X where the carriage itself acts as a heat shield to protect the bearings (maybe with the additional help of some aluminium foil) may be better suited than the horizontal layout where the rising heat gets trapped by the carriage underside. Alternatively, one can also flip the horizontal carriage.

I've been used UHMW bearings and notice the play increases with time very quickly. The material is wearing away. I wouldn't use them in a place where you need more than a single downward compressive force on the bearing, i.e., a bending moment that confers accuracy to a print bed some distance away. Other then that, they look promising at first.

Hi Simba,

could you please tell me in what applications/configurations you have used UHMWPE? This sounds very interesting to me as I only tested my bearing design by letting a carriage run ~10km, so long term results would be nice.

This was a test for a cnc i.e. overloaded application.

I bought 1/2" cold rolled steel from the hardware store and ran .5" (.498") UHMW 1 inch long bearings on it with a stepper for just 2 hours 3cm/s. Thats only 0.2KM. The weight was 5 LB and imparted a bending moment to the piece, so it wore in the outer edges first. Over 2 hours it lost almost .004" of ID. (note: this is just off memory) (note2: no lubriation was used to prevent bias) (note3: there is little if any heat build up.)

Edited 2 time(s). Last edit at 02/28/2013 01:21PM by Simba.

Ah, I see. Unfortunately, I wasn't able to find the typical surface roughness of cold rolled steel, but I assume that it isn't linear shaft-grade since you brought it in the hardware store?
The problem here is that UHMWPE and some other plastics like PET have lowest wear with very fine surface finishes like on ground Cf53 shafts made for linear motion. Other polymers like POM and PA have increased wear with the same surface roughness and actually prefer rougher finishes on their gliding partner. Maybe if you were to replace the UHMWPE with either of both materials I mentioned, you will get better results.
I conducted my tests with relatively smooth V2A rods over longer distances without noticeable wear, so I guess what I have read about surface preferences of plastics is congruent with our findings here.

Okay, good point!! I didn't think about roughness contributing to it that much, but you have a strong case there. Now imagine I used true smooth shafting but had also used lubricant. I could see these things working... 500 hours or more..

I just don't want to have to buy expensive shafting : D
Is this also true for bearings? Will a rougher (rougher meaning .008" tolerance instead of .0005") cause the bearings to somehow wear more/ I would think then the bearing will wear the shaft. I'm trying some stuff with hollow 12mm tubing that is blackened cold rolled steel. IT seems to roll well.

Edited 2 time(s). Last edit at 02/28/2013 07:36PM by Simba.

The strange thing is that as soon as I added PTFE dry lubricant to my test setup, performance worsened. I assume that the PTFE particles clogged my bearing rather than lubricating it. Probably, I just sprayed too much onto my parts. UHMWPE itself is rather slick and self-lubricating, though, and initial microscopic wear from residual shaft roughness will abrade some fine particles that smooth out the surface and provide lubrication. After I got rid of most left over PTFE so that only a very fine layer was left over, I could not make out any difference from the unlubricated state. I have not tried oil or grease, by the way.
Igus plastic bearings run marginally better with oil as lubricant than with grease; no idea whether this also applies to UHMWPE.

I don't know that much about the wear on the bearings, but assuming the shafts that you have got are not hardened, I expect them to wear faster than you get any noticeable wear in the bearings. In fact, the V2A shafts on our workshop's Mendel90 exhibit some grooving from the ball bearing races after about 3/4 of a year.
But in case you were talking about using radial ball bearings and not the linear bearings, I have calculated the service life of a V-groove rail (with typical data for 3D printing applications inserted into the variables) according to Bishop-Wisecarver datasheets - the thing would only fail after ~22+ AU's. That is from here to a little bit further than Uranus!
Given that the surfaces are hardened and the area of contact is rather large compared to a bearing running on a tube, you would probably still get impressive results with a DIY solution.

AS UHMW plastic is unusually solvent resistance, it would absorb only the tiniest petroleum based (ideal thinned smelly lubricant), giving it ideal performance and saving a nanoscopic amount of lubricant embedded in the plastics surface. I would not have recommended PTFE lube for this.

That is interesting to hear. I knew that UHMWPE doesn't absorb air humidity like PA does, but did not have any information about other fluids. Any reason why I should not use PTFE (besides obviously not working better than the unlubricated bearing)?
One of my goals was to have a completely wet lubrication-free printer for maintenance's and portability's sake and PTFE dry lube was sort of a borderline consideration in case performance is not as expected.