Geared 5:1 on 3-in Colsons for 14fps. Uses a BAG motor on a 90:1 VP to turn, and a 1:1 pulley allows me to use the VP absolute encoder.
Ideally, everything would be printed, but it would be easy to machine everything except the wheel blocks.
Gears and pulleys are printed again. Bevel gear is 15:25 10DP. Weight is under 7lbs easy, probably around 6.25 if we don’t use 100% infill on everything.
Center-bevel makes the turning module super short, at the expense of width. I’m trying to figure out how to get it slimmer, but you need very thin wheels and a small bevel pinion.
I’m also a big fan of center bevel. I love how it dramatically reduces the height and complexity of the swerve modules.
However, I have some questions about your plan to 3D print this.
What material are you planning to print from, and have you done any testing about its ability to endure the loads of a swerve drive? Especially in a design like this where the wheel blocks of the swerve are essentially cantilevered beneath the rotational plate. I’d have serious reservations about deflections in those pieces, and the impact of deflections upon the drive shaft and bevel gears.
Unless I’m missing a captive nut that cannot be seen, it also appears you’re planning on threading those (1/4"-20?) bolts into the 3D printed module side plates. I’d strongly caution against threading into plastics in an application like this if it can be avoided.
If you look at 192’s plastic/3D printed swerve module, they avoided both of the issues I laid out above. They used the clamping force of nuts/bolts to keep the pieces together, but did not thread into the plastics. Additionally, they had their rotation bearing on the bottom of the module and had structural walls on all four sides of their modules to mitigate the forces experienced.
But even 192’s design didn’t 3D print any of the gears. I haven’t bothered with any force calculations, but spitballing it seems like 3D printed bevel gears for this application would be a disaster.
I would watch scrub forces that are associated with having multiple wheels that are (relatively) far away from the steering axis. You did reduce this a bit with thinner wheels.
The torques experienced at the end of the transmission would be cutting it very close with 3D printed parts. If anything I would use 3D printed parts at the beginning of the transmission (near motors) since torques experienced are much lower. For example I think the AndyMark PG71 uses a plastic first stage, but metal for the others. That isn’t to say it won’t work, especially since you used larger 10DP teeth.
Please use 100% infill on the bevel gears. Less than that is surely asking for problems. Otherwise components that don’t see as much force are fine.
Overall, great design. Looking forward to seeing how it turns out!
Two comments: (1) I agree with the previous posters’ views about 3D printed parts in high stress areas. Competitive robots must be designed to play 80+ matches, plus a lot of practice driving. 16 showed pit visitors a 3D printed mock-up of their 2017 swerve module, which they made to study initial assembly and repair. Their final version has some printed parts, but more are machined. Even the best swerve drivers cannot dance around every threat – your module WILL take a few hits.
(2) The rotation bearing needs more diameter.
Overall this is a great start, and it is based on a very sound concept – center-bevel and dual Colsons. I became a fan soon after seeing its evolution on 2767.
Stratasys is a great sponsor of ours, and I’ve gotten an offer to use their advanced composite materials for this project (unofficially, from ehochstein, on my previous design.) Hopefully we can get a design review session done with them as well.
I posted in a rush and forgot to mention the captive nut They are 1/4-20.
I opted not to have a lower rotation bearing because I was trying to get more of the wheel “accessible” for driving on things that are not perfectly flat terrain. I felt this was a weakness of my previous design and would limit its effectiveness in any other game than Steamworks, not to mention as a demo robot.
Again I was thinking that stronger material/more controlled printing processes would help here (e.g. Stratasys). If not, we could possibly do Nylon or aluminum casting.
Thanks for your feedback. 3D printing makes designing easier, but pretty much every step after that will likely be met with challenges.
Let’s do some math…
Assume a full weighted robot is 150 lbs. ~1/4th of that will be on each module, so a normal force of 37.5 lbs. There are two wheels, but each takes half the weight, so we can act like it’s one wheel. The distance from the rotational axis is 1.5431 inches, and the coefficient of friction (mu) of a Colson wheel is 1.0.
Frictional torque should be (correct me if I’m wrong) t_f = Nmur. Then the max frictional force is t_f = 37.5 * 1 * 1.5431 = 57.87 lb-in. The stall torque of a 90:1 BAG motor is 342.5 lb-in. Even at the 90% efficiency that the JVN calculator has at default, it is well over the required force. I feel like steer motors are generally way overpowered for the application…
We’re in regionals- we got 28 (official) this season and that was a good year
That debate aside, this would definitely need a lot of stress testing before the season if we were to use it. It would be possible to move to a metal steer module (with printed steer pulley) with minor redesigns.
By “rotation bearing” I assume you mean the thrust bearing separating the steer module from the main plate? It’s currently a 1/8th oilite bearing running on two radial bearings, which is definitely kinda skimpy.
Agreed. Your center plate is going to flex tremendously and your bevels will lock up and your robot won’t move. I know how strong the stronger composites are, and it will not save your plate from flexing to death. I suggest printing the cavity as a single piece (maybe even attaching the wheels on the outside of the print rather than the inside) with very strong side fillets. I don’t think your current module will even last being put on the floor.
Thank you all for the feedback. I think it’s finally gotten through my head that 3D printing will probably not work for this module design. If this design ever comes to fruition, I’d still like to see if the bevel gears work. If not, one of my fellow students was saving pop cans from the shop, intending to do some casting.
Edit: Then again, it may just be easier/cheaper to switch to commercially available metal bevel gears.
Edit 2: Where can I buy bevel gears? Vex has the 15T set, AndyMark has the 2:1, SDP-SI has a variety. Anywhere else I can look?
Edit 3: McMaster-Carr has some as well, but these are all kind of expensive…
http://i.imgur.com/riM6rbk.jpg
Converted the module (except steer pulley) to aluminum. Wheel blocks are attached via threading instead of a captive nut. Pulley is attached with a captive nut. Replaced the Oilite module bearing with a roller bearing.
Agreed, agreed. We had a side impact/load severe enough to push the axle into the bevels causing the vertical shaft to bend. This happened on the 2014 design (picture on pg. 3 of the white paper). To be fair, the shaft wasn’t hardened but still illustrates that there is movement down there. We attempted to print the saddle using nylon with carbon fiber fill with carbon fiber overlay (pg. 4). They worked but the bearing flanges still moved too much for our comfort and the weight advantage wasn’t there so the composite saddle idea was scrapped. IMO the only thing that can be printed is the pulleys but if you have the ability to try… In the end, printing structural stuff is high risk so test, test, test.
Why print all those blocks and plates? Take greater advantage of the printing process to fill in high stress corners, reduce part counts, and make beautiful organic stress-reducing shapes!
(If I were designing this, I’d be really worried about scrub forces when turning this thing, to the point of considering running a miniature differential across the wheel shaft… but that would take us well into ‘science project’ territory)
EDIT: Oops, looks like you already are pushing towards Aluminum.
As far as bevel gears go… try to find a local shop that does custom gears and ask really really nicely. When I did Baja SAE we had a local gear shop that did our custom profile bevel gears for a miniature open differential design, and who taught us a lot about gear design in the process. (Associated Gear in Los Angeles, they’re fantastic)
I think you may need to reconsider your approach the the swerve wheel module. Simply changing to aluminum will not solve some of the weaknesses related to design geometry. I still expect your top plate to bend under load, regardless of what standard FRC material you choose.
I recommend studying some existing COTS swerve modules to understand how they achieve rigidity in the wheel module. You’ll probably notice designs tend to converge to a “box” shape, rather than the “c-channel” shape you have modeled.
Alternatively, you could build this module as-designed, and apply load to the frame (aka stand on the robot) to see how much the top plate deflects. I recommend counting the jumps until non-elastic deformation occurs.
Plan for the full life of the product, not just the season through Champs. 2 regionals + champs + States could be as many as 40 matches. Add in week-0 and summer/fall off-season events, and you’ll easily be up in the 60-70 match range. Add to that anticipated demo’s and driver practice. It all takes a toll on a robot over time, which is especially true for a complex drive train.
And once you look at all of that, ask yourself how long you want to keep this robot around. Perhaps the best mecanum robot my team built lasted 5 years as a practice and demo robot before we finally broke some critical parts of the frame, making the drive train non-functional. Funny enough, the student that built that drive train in the first place graduated college and returned as a mentor just in time to see it thrown out!
Thanks for your feedback. I’m not sure if I completely understand your difference between the “box” shape of other swerves and a “C.” Can you please elaborate?
My module design was inspired by the Revolution Pro 2 and by 2767’s white paper. 2767 used a half a piece of 1/4" wall 4" square tubing as their module. Because this design was intended to be 3D-printed, I wanted to optimize print orientation and couldn’t print the module as a single piece. The Revolution Pro 2 has discrete plates for the top and each side, but they are held together by tapped holes in the top plate. So I went down, and held the wheel blocks in with a captive nut.
So really, most of this design was based around 3D printing. I still like the center-bevel, but I’ll have to redesign the module to be machined instead of printed. Maybe I’ll just replace this steer module with a piece of square tubing (a la 2767) for now.
Of course. That was an off-color joke about regionals vs. districts, and I’ll be sure to keep durability in mind.
One big difference between your module and the Revolution Pro 2 is the top plate thickness. From the picture, your top plate looks to be about 1/8". The Revolution Pro 2’s top plate is at least 1/4". It also has the bolts facing sideways, which helps by supporting more of the top plate vertically (the length of the bolt vs the width of the side plate).
If you want to make a 3D printed swerve module, more power to you. There are a few designs that you can use to help you design yours. But there are big differences in the way you build a 3D printed and an aluminum swerve. Basing your 3D printed model on aluminum modules without fully taking into account the differences will leave you with lots of snapped parts. If you decide that you want to start with an aluminum module, that’s fine too. But design it like it’s made out aluminum, not like you wanted to 3D print then just switched the material to aluminum.
The Revolution Pro 2 isn’t a center-bevel module, and thus is much narrower. Consider how the forces on your center plate will change as the module is made wider. What lever arms are now longer? How will that impact material deformation? How will that deformation impact your bearings, bevel gears, and drive shafts?
If you go the way of a differential, not only have you added gobs of mechanical complexity, you’re also opening up the potential for traction loss; the maximum tractive force you can generate will be that of the smallest wheel. Unless you throw a fancy limited slip differential in there, which even further complicates the design!
You’re better off figuring out how to reduce scrub radius by pulling the wheels in. My thoughts are counterbore the wheels so the bevel can sit inside with the shaft just clearing the wheel. Of course, this introduces issues with your bevel gear not being adequately supported. But I think there’s scrub radius you could shrink up by moving components closer. This will also help packaging.
That and of course rigidity. Making that bevel mesh happy is important.
They are 1/4-20.