10-08-2011 17:47

AdamHeard

Re: pic: 2011 "Nova" Swerve Drive (Front)

I'd be very worried about those bevel gears, what DP are they?

Also, the steering sprocket setup, is that one solid piece?

10-08-2011 18:31

Jedward45

Re: pic: 2011 "Nova" Swerve Drive (Front)

The DP is 32. They are a Boston Gear set. The Mounting Distances are up to spec with Boston Gear's technical drawings. Also, the hub of the Gear inserts into the wheel assembly.

The steering sprockets sit on a 1" OD Aluminum tube (3/4" ID). This tube is machined from a 2" diameter Aluminum rod because at the base of the 1" tube, a 2" mounting plate attaches to the rest of the chassis (Dark grey).

Hope that clears things up

10-08-2011 18:43

=Martin=Taylor=

Re: pic: 2011 "Nova" Swerve Drive (Front)

Wow. I once spent a long time designing a drive just like this. I have to say, this is a very good design!

I gave up on my design because I was worried about turning the modules. The very closest I was able to move the wheels was 4" wide. The effectiveness of a swerve is limited by how fast the modules can be rotated, and with only the window motors for rotation I figured this might be difficult with such wide wheels.

My next and final swerve design was a 45* module like in the 1625 teaser. In the end, this design wound up being more complicated to make then a conventional module and was only slightly smaller so I abandoned this too.

Its cool seeing all the new swerve designs on CD. I wonder how many teams will use swerve next year...

10-08-2011 23:45

Peyton Yeung

Re: pic: 2011 "Nova" Swerve Drive (Front)

I don't know if we will use it next year but team 45 has started thinking about adding swerve to the list of drive trains we can construct.

4wd
6wd
tread
ball drive
swerve?

11-08-2011 00:50

Tristan Lall

Re: pic: 2011 "Nova" Swerve Drive (Front)

Are they hardened? And is the gear's thickness drawn to scale? They seem rather undersized for a final drivetrain stage. I wouldn't proceed with those particular gears unless you've calculated the gear strength to verify the design—they're just too unlikely to be adequate.

Also, you didn't mean to include a differential, right? (I seem to recall a team having a design similar to this, but with differentials on all wheel pairs. Might have been 1140 around 2005.)

11-08-2011 01:14

Jedward45

Re: pic: 2011 "Nova" Swerve Drive (Front)

Currently the gears are baseline steel.

As you probably know, selecting the Bevel Gears was by far the trickiest part of this design. If the mounting distances are too large, the wheel separation makes rotation of the module impossible. On that end, if you happen to know where I might find a hardened, spiral bevel gear set that fits near the specifications for under $120, I would greatly appreciate it....

Also, how exactly do you calculate gear strength anyways?

11-08-2011 03:13

Tristan Lall

Re: pic: 2011 "Nova" Swerve Drive (Front)

My first instinct would be to check QTC (part of the Designatronics group familiar to many via their SDP/SI division). Although I realize it's not the ratio you were depicting above, you should probably look at this page, specifically SBS1-4020R and SBS1-2040L to get an idea of what's available, and what it might cost you. Those teeth are induction hardened to Rockwell C 48, which in steel, correlates well with a tensile strength of about 1 632 MPa (237 000 lb/in², or really strong). The teeth are 1.0 module (or 25.4 diametral pitch), which I'd say is about as small as you'd want to go for a final stage of a drivetrain—and even then, only with a hardened gear.

And by the way, if you need more clearance for the pinion (too wide), you might consider dishing out the wheels facing the centre of the gearbox (so that the pinion and gear sit in a depression in the wheel, sort of like a typical west-coast style wheel with the spokes on one face).

Typically, in FRC applications, gears fail because the teeth are overstressed. (Other failure modes are possible.) The important characteristic is therefore the bending strength of the teeth.

There are a few ways of evaluating this, with varying levels of complexity and accuracy. The Lewis-Barth method is traditional and conservative, and requires relatively few parameters. AGMA has another method that takes fatigue and contact stress (another failure mode) into account, as well as a whole slew of other design factors—but you probably won't know what values to assign to them without some sort of basis for comparison. Given that bevel gears are a bit of an unusual system for an FRC application, I'd avoid relying on rules of thumb alone (so don't just take my word for it).

For an introductory reference, see this. Check out the literature provided by the manufacturer. Also, there's this: a whole book about gears, with a good explanation of the Lewis method. For more about the AGMA method, I recall that any recent edition of Shigley's Mechanical Engineering Design should include a chapter (written for upper-year mechanical engineering students). You might find this a convenient resource for equations and examples. Definitely search the Chief Delphi forums for other resources.

11-08-2011 07:40

Hawiian Cadder

Re: pic: 2011 "Nova" Swerve Drive (Front)

I suggest looking into worm gears, they can get you a lot more reduction in that space, and I bet it will be easier to find a set that small. with a worm gear you may be able to remove the cimplebox entirely.

11-08-2011 23:18

Tristan Lall

Re: pic: 2011 "Nova" Swerve Drive (Front)

I'm not generally a fan of worm gears for drivetrains (see here, at page 30 for a discussion of low efficiency, especially as the reduction increases). However, seeing as you're only replacing a CIMple Box (4.67:1), and maybe part of the 4:1 bevel gear set (since I found a 2:1 set that might work for you), your total reduction is only around 10:1. With a well-designed structure and high-quality worm gear, you can probably achieve 90% efficiency on that reduction, which is quite competitive with spur gears. You'll probably need a stiffer structure to make it happen, though (alignment is critical with worm gears).

12-08-2011 13:40

Jedward45

Re: pic: 2011 "Nova" Swerve Drive (Front)

Out of curiosity, with those 2:1 Bevels, the 5:1 Cimple box, and 3" wheels, what would a good FPS estimate be?

12-08-2011 13:47

Hawiian Cadder

Re: pic: 2011 "Nova" Swerve Drive (Front)

According to the JVN mechanical design calculator;

7 feet per second after frictional losses.
196 LBS of pushing force with rough-top tread.

12-08-2011 13:52

Jedward45

Re: pic: 2011 "Nova" Swerve Drive (Front)

Is it possible to reduce the CIMple box further, I was hoping to get right around 10 FPS

12-08-2011 14:01

lemiant

Re: pic: 2011 "Nova" Swerve Drive (Front)

You mean reduce it less. And I would just decrease the reduction on the bevel gears.

12-08-2011 17:54

Chris is me

Re: pic: 2011 "Nova" Swerve Drive (Front)

With all due respect, if you can't figure out gear ratios, you really should not be designing a swerve drive.

12-08-2011 18:33

Jedward45

Re: pic: 2011 "Nova" Swerve Drive (Front)

With all due respect, I didn't post myself on this website, I posted my design. If you have a comment or criticism about that, I'd be happy to hear it. But if you only want to discourage me from experimenting and learning, I'd prefer if you kept that to yourself.

12-08-2011 18:51

Aren_Hill

Re: pic: 2011 "Nova" Swerve Drive (Front)

From looking at it, you have a fairly robust looking module with most of the issues attended too, aside from that bevel gear looking tiny.
What i'd be concerned with in this module is the amount of power necessary to steer those wheels (depends on the number of modules and what motors are steering).

Also depending on your steering arrangement some method of angle adjustment between the steering sprocket and the module may be necessary.

Do you have the machining capabilities to make that one piece bracket and securely mount it to the vertical support?

It also looks rather huge, but i like small compact packages if you look at our recent swerves.

Chris's statement has some merit, and I've learned to like when people are straight up blunt with me, as it makes things much easier to gauge and get a realistic evaluation of your current state.

I would suggest listening to his advice and getting a firm grasp on gear ratios and the efficiency of various methods of power transmission, this will only help your design. Exercises like this are a chance to develop further, and as has been pointed out you could use some work in the gearing arena.

Don't immediately think someone is attacking you when in reality they're just trying to bump you back to the path you've strayed from. He gains nothing from tearing you down.

12-08-2011 19:25

Jedward45

Re: pic: 2011 "Nova" Swerve Drive (Front)

Thanks for the advice. As far as my previous comment is concerned, I understand the need for blunt criticism. I understand that Chris was trying to explain I need to better understand gear ratios, and I agree with him. However, I'm slightly taken back that anyone in FIRST would discourage someone from trying something new, especially when those new and difficult things are a source of the knowledge so vital to their own creation. Moreover, the criticism wasn't followed by an attempt to teach me the skills I need, or even show me where I could find the materials to teach myself.

As far as the module is concerned:
-The module is designed to be tied (via 25 chain) to the other three modules. (All I know is a banebots motor was used to rotate, but I wasn't in charge of that system, so I don't know how much power was needed. We also used 1" wide wheels, and rotating wasn't a problem)
-following some earlier advice, I'm working on interfacing a more robust and larger set of bevel gears.
-We do have access to the machining necessary to create the lower chassis.
-The size may be deceiving, but the wheels are only 3" in Diameter and 1/2" wide. I don't know if that still constitutes a "large" design, but if it does, I'd sure like to see Winnovation's swerve design

Thanks!

12-08-2011 20:05

ProgramLuke

Re: pic: 2011 "Nova" Swerve Drive (Front)

Not everything you want is going to fall into your lap, sometimes you have to go get it. The world belongs to those who show up. In this case showing up might mean searching CD's White Papers for JVN's Design Calculator.

12-08-2011 21:59

Tristan Lall

Re: pic: 2011 "Nova" Swerve Drive (Front)

I might be concerned if this was week 2 of the build season, and you were only then figuring out how to do design work with gears. (That would be the time to drop everything and assemble the kitbot.) Instead, it's the summer, which is the perfect time to figure all of this stuff out.

12-08-2011 23:46

DonRotolo

Re: pic: 2011 "Nova" Swerve Drive (Front)

My concern with the dual wheels is that every turn guarantees one or the other will have to scrub, and hard - based on my assumption that the wheel axle is solid and common to both wheels.

13-08-2011 03:05

Tristan Lall

Re: pic: 2011 "Nova" Swerve Drive (Front)

And incidentally, it was posted above that the JVN gear calculator would give you 7 ft/s. That's probably a decent estimate, but it's worth discussing some of the theory behind that result.

First, you take your wheels, which are designed to be Ø3.00 in. (From examining that drawing by eye, based on the size of the CIM, which is Ø2.50 in, I'm guessing the wheels are actually slightly larger than that. Did you account for the thickness of the tread backing, and maybe some of the tread?) We'll stick with Ø3.00 in for simplicity.

Then find the circumference per revolution: C = π × d = Ø9.42 in/rev = Ø0.785 ft/rev. That's the distance the wheel travels in one revolution. Though I'd really rather work in SI, it would just confuse the Americans—so were going to calculate the speed of the robot in ft/s.

That means we need rev/s, or its familiar cousin, rev/min. This is determined by the speed of the motor, at its operating condition. Note that this is distinct from free speed. Most of the time, when we talk about robot speeds, we're either talking about the theoretical free speed (when calculating), or the actual speed (when measuring).

Since the free speed of the CIM motor is ω_free = 5 310 rev/min = 88.5 rev/s, and we've got a 10:1 gear ratio (Z = 0.1) a little multiplication gives:

ω_free × C × Z = v
88.5 rev/s × Ø0.785 ft/rev × 0.1 = 6.95 ft/s
v ≈ 7 ft/s

In JVN's calculator, there's usually an arbitrary factor of 81% corresponding to operating condition (graphed on a torque curve)—in other words, the motor is operating at 81% of free speed. It's arbitrary in the sense that there's no physical reason why it has to be 81%, but based on JVN's testing and experience, it was a reasonable value. (I've occasionally picked a slightly higher value for this—around 85% or even 90%. Again, it was just a guess, but I've designed several drivetrains with lots of motors, which tend to be less heavily loaded under ordinary driving, so the motors presumably run closer to their free speed.)

In the calculations presented by Isaak* above, I think he's either using a 100% (i.e. at the free speed) calculation, inputting the correct speed and rounding up to 7 ft/s—in other words, using JVN's calculator to do the math I showed above—or else he's using the rough free speed JVN preloads it with (5 500 rev/min) and a weird speed loss constant of 97.25%. In any event, remember that the speed loss constant is just a way of saying "when the wheels are off the ground and spinning under power, their speed is 81% of what it would be in a frictionless gearbox." That's different from, but related to efficiency.

Anyway, in my estimation, your wheels are probably more like Ø3.25 in (due to the tread thickness), so that might skew things a bit. Using speed loss of 81% and efficiency of 90%, using 1.2 as the coefficient of friction, substituting in the correct values for CIM motors instead of the rough/old ones in some versions of the JVN calculator, and taking robot weight to be a rather plump 167.5 lb_f (hostbot + minibot + battery + bumpers), I'm actually estimating that you'd see something closer to 6.10 ft/s and 201.0 lb_f pushing force.

By the way, remember that all of this is predicated on your battery actually being able to deliver that performance—which it won't. As the load on the battery increases, its voltage drops, basically shrinking your speed curve (and hence your power). The more advanced versions of the JVN calculator will let you model motor performance over time at some voltage, from which you can determine the (constant-voltage) load on the battery. If the load on the battery is significant for a long enough period (e.g. when accelerating), it may be time to look at a few cases where the applied voltage is lower, in order to estimate mid- and late-match performance.

For now, though, don't worry too much about the battery voltage and power consumption, because your design isn't crazy enough for it to matter too much. (I've had a few designs where these absolutely did matter.) Estimate a desirable speed, then try to find gears to match.

* I just realized he's not "Hawaiian" or even "Hawaiʻian", but rather "Hawiian Cadder"...I'm not sure what to make of that.