Andymark Churros

[llamadon]

Hi, our team is prototyping a wcd this summer, and we have an excess of andymark Churro extrusion (http://www.andymark.com/product-p/am-2595.htm) and plan to use it as a live axle. Does anyone have experience using this material? If so, is it strong enough?

[orangemoore]

If you are just building a drivebase they might work.

But I have never heard it being used as an axle before and I am skeptical that it would be strong enough for the job. Most hex axles don't have a hole going all the way through it, as well as a dimple on every side of the outside edge.

If you could I would suggest you use something else.

[Michael Hill]

The churros are 6063 aluminum. Hex shaft from Vex is 7075-T6 (black anodized for superior performance, of course). I'd say there's not a chance they hold up, but that's just my speculation.

[Akash Rastogi]

I wouldn't even use regular 6061 hex shaft for this application, let alone the 6063 churros. Bad experiences with bending and snapped axles. (Just ask 2791, among others).

[s_forbes]

Excess material you say? Build it and see what it takes to break it! You will learn more by testing the churro extrusions to the limit than you will by using the "right" material from the start. If they fail, you can replace them with stronger shafts later.

I'm a big fan of destructive testing. It's fun and educational. Every now and then you find something that you were sure wouldn't survive, but ends up working and saving you time/money/weight.

[Boe]

Excess material you say? Build it and see what it takes to break it! You will learn more by testing the churro extrusions to the limit than you will by using the "right" material from the start. If they fail, you can replace them with stronger shafts later.

I'm a big fan of destructive testing. It's fun and educational. Every now and then you find something that you were sure wouldn't survive, but ends up working and saving you time/money/weight.

I agree if you have it test it. Also since it already has the hole through it you can use that to put a bolt and washer on each end to hold the shaft in place, either by putting a bolt all the way through, or by tapping each end.

[AndreaV]

Sure, it's a 'learning experience" and you can use it, but I doubt it will last long. simple directional changes(with 120lb worth of robot) would probably be enough to twist them. The profile is probably the WORST for handling torsion loads, it's missing material from where it's needed the most for this application.

If you do use it make sure you have some solid hex(as high a grade as you can find) to have ready to go.

[MechEng83]

Spacer, yes. Live axle, no.

We tried this on this year's robot and it twisted it into a twizzler shape. I recommend using solid hex aluminum for your application. 6061 might suffice if the load isn't too high. We ended up using 2024 solid aluminum hex. That worked well for us.

[asid61]

6061 1/2" hex will not bend in drivetrain applications, as we had to substitute a few 7075 axles for 6061 during build and they held up through 2 competitions. This might have been due to to other drivetriain problems that stopped us from moving too much though.
6063 is considerably weaker than 6061, so I would be wary of using it. For prototyping, it might be fine, but it would be much safer to just purchase some 1/2" 7075 hex stock from Online Metals instead.

EDIT: This would be in our team's experience only.

[R.C.]

6061 1/2" hex will not bend in drivetrain applications, as we had to substitute a few 7075 axles for 6061 during build and they held up through 2 competitions. This might have been due to toher drivetriain problems that stopped us from moving too much though.


I'm not totally following. Just to clarify, your saying that 6061 1/2" hex would be more than fine in a drivetrain application such as a WCD (or other drivetrain using live axle)?

[asid61]

I'm not totally following. Just to clarify, your saying that 6061 1/2" hex would be more than fine in a drivetrain application such as a WCD (or other drivetrain using live axle)?

I'm sorry, it did not bend for us. Our drivetrain didn't see very heavy use this year, and that might be why it worked fine. Teams like 254 probably would need the extra durability afforded by 7075, but for a prototype low-load drivetrain 6061 should be fine.

[Chowmaster4695]

6061 1/2" hex will not bend in drivetrain applications

http://www.chiefdelphi.com/forums/showpost.php?p=1388120&postcount=16

You stated they would bend here. Please give us accurate information. Our students use chief delphi to help them make decisions where others have more experience. If they see contradictory posts like yours then they and many others will be confused. Please edit your post.

Chow Out

[Gary Dillard]

...twisted it into a twizzler shape.

Churros and twizzlers, this thread is making me hungry.

[DampRobot]

I'm not totally following. Just to clarify, your saying that 6061 1/2" hex would be more than fine in a drivetrain application such as a WCD (or other drivetrain using live axle)?

We've used 1/2" 6061 axles in a drive before, however, they were supported by bearings on both sides, and we were running 6" wheels instead of 4" wheels. We didn't see it twist up at all. YMMV.

Of course, with the 7075 VP stuff so cheap, easy, machinable, and strong, I would never order 6061 hex for drive shafts. They almost never come with the right dimensional tolerance on the OD anyway, whereas the VP stuff is good to go out of the box. So really, why use 6061?

OP, if you're fine with wrecking some material, go for the Churro tube. Otherwise, buy the real stuff from Vex and profit. I've seen Churro tube twisted up ("into a twizzler shape") multiple times at regionals, and I wouldn't bet much on it holding up.

[cbale2000]

We used some Churro for our collector on this years bot with no issues...

http://s2.postimg.org/56e4nahrp/DSCN1226b_Collector.jpg (http://s2.postimg.org/ft7xsppx5/DSCN1226b_Collector.jpg)

You can't really see it in that picture because the PVC spacers cover it, but it extends across the entire length of the collector, drives the wheels, and is only supported on the ends. Overall we've been very happy with it, especially because it's light and rigid (and because we already had some around when VEXPro ran out of their long hex stock during build season).

Now granted this application probably undergoes far less force than something like a drive system would, so keep that in mind when planning its use.

[philso]

I believe our drive train team used the churros for all the axles in our octanum modules (2 per module). I do not recall seeing any fail. The holes in the ends were handy for holding the modules together. Please keep in mind that these axles were not cantilevered. They were always supported on both sides of the wheels with hex bearings.

[Gray Adams]

http://www.chiefdelphi.com/forums/showpost.php?p=1388120&postcount=16

You stated they would bend here. Please give us accurate information. Our students use chief delphi to help them make decisions where others have more experience. If they see contradictory posts like yours then they and many others will be confused. Please edit your post.

Chow Out

I can't say that I recommend blindly following random posts on CD and expecting success. You could always either go through to math with the max torque seen on the shaft during a direction reversal at top speed, calculating the stress for a hex shaft, a churro, or a keyed shaft, and then selecting an appropriately strong material.

Or you could just test it and see what happens. Both are totally valid.

[EricH]

I can't say that I recommend blindly following random posts on CD and expecting success. You could always either go through to math with the max torque seen on the shaft during a direction reversal at top speed, calculating the stress for a hex shaft, a churro, or a keyed shaft, and then selecting an appropriately strong material.

Or you could just test it and see what happens. Both are totally valid.
Why not do both? Run the math, then run the tests, and see how they stack up. Then if (when?) the test doesn't match the math, run more math and more tests to see which one forgot to take reality into account. There've been cases where tests showed a bit of a reality that doesn't show up at all in high school physics (or college physics, for that matter).

[Alan Anderson]

I believe our drive train team used the churros for all the axles in our octanum modules (2 per module). I do not recall seeing any fail.

Were they live axles, actually transmitting torque? That's the twizzler-inducing situation that people are worried about.

[philso]

Were they live axles, actually transmitting torque? That's the twizzler-inducing situation that people are worried about.

Yes, they were live axles. The drive train team may have used the churros so they would not have to wait to get the proper hex shaft material. I don't recall them having the "twizzler" problem but I can understand that it could happen. We may just be lucky.

[AndreaV]

Yes, they were live axles. The drive train team may have used the churros so they would not have to wait to get the proper hex shaft material. I don't recall them having the "twizzler" problem but I can understand that it could happen. We may just be lucky.

They may have been live, but i am sure these live axles were supported by bearings on both sides, with the shaft section probably being around 3". In a butterfly drive whatever is driving the shaft is bolted directly to the wheel or driven by the hex in close proximity to the wheel. On a live axle WCD the sprocket/belt is on the opposite side so the whole axle is susceptible to torsional loading, which is the twisting action mentioned earlier.

With a WCD you have traction wheels being cantilevered, the shaft has to deal with axial and radial loading. Sure all shafts have to deal with this in some regard, but in a WCD the wheel(which is acting like a lever) transfers it directly to one side of the shaft, instead of evenly distributing it between both. This is when these forces become a problem.

I busted out my paint skills to illustrate what this looks like on a wcd with live axles.

http://i.imgur.com/3tbpmOZ.png (http://imgur.com/3tbpmOZ)

[JamesCH95]

They may have been live, but i am sure these live axles were supported by bearings on both sides, with the shaft section probably being around 3". In a butterfly drive whatever is driving the shaft is bolted directly to the wheel or driven by the hex in close proximity to the wheel. On a live axle WCD the sprocket/belt is on the opposite side so the whole axle is susceptible to torsional loading, which is the twisting action mentioned earlier.

With a WCD you have traction wheels being cantilevered, the shaft has to deal with axial and radial loading. Sure all shafts have to deal with this in some regard, but in a WCD the wheel(which is acting like a lever) transfers it directly to one side of the shaft, instead of evenly distributing it between both. This is when these forces become a problem.

I busted out my paint skills to illustrate what this looks like on a wcd with live axles.

http://i.imgur.com/3tbpmOZ.png (http://imgur.com/3tbpmOZ)

This is called a cantilever axle.

Simplistically, a cantilevered axle sees six times the bending stress compared to a axle (beam) fixed on both sides. I would surprised if AM churro survived as a cantilevered axle for very long because of its low XC area and relatively poor material properties compared to 7075T6.

I would also be surprised if the shaft failed in torsion instead of bending.

[Nate Laverdure]

Polar moments of inertia:

AndyMark 1/2" churro tube (http://www.andymark.com/product-p/am-churro.htm) .00535 in^4 [Solidworks]
VexPro 1/2" round tube (http://www.vexrobotics.com/vexpro/hardware/shaft-stock.html) .00570 in^4 [Shigley 8th ed. Table A-18]
1/2" round shaft w/ 1/8" keyway .00570 in^4 [Solidworks]
1/2" round bar .00614 in^4 [Shigley 8th ed. Table A-18]
1/2" hex bar .00752 in^4 [Wikipedia (http://en.wikipedia.org/wiki/List_of_area_moments_of_inertia)]

[AdamHeard]

Polar moments of inertia:

AndyMark 1/2" churro tube (http://www.andymark.com/product-p/am-churro.htm) .00535 in^4 [Solidworks]
1/2" hex bar .00752 in^4 [Wikipedia (http://en.wikipedia.org/wiki/List_of_area_moments_of_inertia)]
1/2" round bar .00614 in^4 [Shigley 8th ed. Table A-18]
VexPro 1/2" round tube (http://www.vexrobotics.com/vexpro/hardware/shaft-stock.html) .00570 in^4 [Shigley 8th ed. Table A-18]

It's worth noting that the 6063 that the Churros are made of has a substantially lower yield.

[Greg Woelki]

Polar moments of inertia:

AndyMark 1/2" churro tube (http://www.andymark.com/product-p/am-churro.htm) .00535 in^4 [Solidworks]
1/2" hex bar .00752 in^4 [Wikipedia (http://en.wikipedia.org/wiki/List_of_area_moments_of_inertia)]
1/2" round bar .00614 in^4 [Shigley 8th ed. Table A-18]
VexPro 1/2" round tube (http://www.vexrobotics.com/vexpro/hardware/shaft-stock.html) .00570 in^4 [Shigley 8th ed. Table A-18]
While the polar moment of inertia of the churros is decent, not only is it 6063, but also the dimples will cause it to perform poorly in torsion.

[MechEng83]

Polar moments of inertia:

AndyMark 1/2" churro tube (http://www.andymark.com/product-p/am-churro.htm) .00535 in^4 [Solidworks]
1/2" hex bar .00752 in^4 [Wikipedia (http://en.wikipedia.org/wiki/List_of_area_moments_of_inertia)]
1/2" round bar .00614 in^4 [Shigley 8th ed. Table A-18]
VexPro 1/2" round tube (http://www.vexrobotics.com/vexpro/hardware/shaft-stock.html) .00570 in^4 [Shigley 8th ed. Table A-18]

*WARNING* Summary of Graduate-Level Mechanics of Materials concept ahead

The polar moment of area is only useful in terms of torsional rigidity. The torsion constant requires a much more complex formulation (the Prandtl membrane analogy). It is only identical to polar moment of area for circles.

Then, you have to use the modulus of rigidity (G) of the material and the distance from the central axis to the outer-most point to determine the shear stress. and compare this to the maximum allowable shear stress of the material.

The dimpled sides of the churro profile actually make it incredibly weak in torsion compared to a solid section or even a full hexagon with a hole in the middle. The membrane or "soap bubble" analogy lets you have a bit of understanding as to how rigid something is in torsion. If you imagine a membrane or soap bubble is attached to the outside edges and the membrane is inflated the volume is analogous to the torsion constant of the section. If there are open sections that are completely contained, the membrane is "flat" in that area. This is why a thin walled tube is very strong in torsion compared to a 359 degree non-closed section.

tl;dr - science says churros are weak in torsion.

[Briansmithtown]

Nothing beats Andy Bakers Churros... Nothing!

[hrench]

it's great that someone has actually posted the equations for torsion, but I want to weigh-in and agree with the churros being the 'worst shape' for transmitting torsion.

In an engineering class somewhere along the line, I learned the 'membrane analogy' http://en.wikipedia.org/wiki/Membrane_analogy for torsion in a shaft.

Essentially, the calculus is the same for an imagined 'membrane' (think square of latex of a balloon) that's stretched over the end of the shaft. So if you could stretch a balloon over the end of the shaft and then air it up, if it holds alot of air then the shaft will withstand alot of torque. But if the shaft has cuts in it--like a churro--then it won't hold much air and hence not much torque. This is also true for shafts that have been through-drilled with a pin--if you do that, you cut the torsional limit to about half to a third of what it would've been. Also shows how keyways and/or splines are good because they transmit torque without hurting the torsional limit of the shaft very much.

Also, in the membrane analogy, if you air up the balloon-stretched over the end, the slope of the transition indicates stress. If it produces gentle-slow changes in slope, these indicate the stress will be even. If you stretched the balloon over a churro, the slope in the vee-grooves would be very steep, indicating high stress. This also shows how if you use a square-section for torsion, the failure will be in the middle of the sides, not at the corners.

Will it fail? I don't know--it would be difficult to estimate the loading. Is it a bad shape? yes.

[Chris is me]

Hi, our team is prototyping a wcd this summer, and we have an excess of andymark Churro extrusion (http://www.andymark.com/product-p/am-2595.htm) and plan to use it as a live axle. Does anyone have experience using this material? If so, is it strong enough?

I wouldn't even use regular 6061 hex shaft for this application, let alone the 6063 churros. Bad experiences with bending and snapped axles. (Just ask 2791, among others).

Akash is right. Do not use churro tube as West Coast Drive cantilevered axles!!! I would go so far as to say that churro tube shouldn't ever be used as a live axle in any drivetrain.

There are a multitude of reasons churro tube is bad for a cantilevered axle, most of which have been touched on already. The profile shape is atrocious in torsion, which means your shaft will twist and possibly snap under heavy drivetrain loads. The material is 6063 aluminum, which generally has a lower yield strength than even 6061 aluminum. I suspect you would also have problems with bending or transverse shear loads, especially since the shaft has a giant clearance hole going all the way through it.

6061 1/2" hex will not bend in drivetrain applications, as we had to substitute a few 7075 axles for 6061 during build and they held up through 2 competitions.

This is absolutely, positively not true. It's okay to post your experiences on CD, but don't use that information to make blanket statements about all drivetrains. At least not without solid evidence or reasoning.

This might have been due to to other drivetriain problems that stopped us from moving too much though.

This is exactly why you didn't have any failures. Failure is not always instantaneous.

To share a story, 2791 used 6061 hex stock to make drive axles in 2011, and we had numerous failures. Both center drive axles failed halfway into our first event. One center axle, if I recall correctly, actually stripped itself from hex to a rounded profile. Another split in half in an apparent torsion + shear failure. These were solid 6061 axles without intermediate snap ring grooves or other stress risers.

I'm sorry, it did not bend for us. Our drivetrain didn't see very heavy use this year, and that might be why it worked fine. Teams like 254 probably would need the extra durability afforded by 7075, but for a prototype low-load drivetrain 6061 should be fine.

What's the point of building a prototype drive if you're not going to test durability? If you're building a drivetrain before the season starts, you should build as close to how you would in build season as you can. That way, if and when you do have failures, you can directly apply them to your future robot. If you use parts you know will fail to cut corners, you won't know if the part you intended to use in build will work properly.

We used some Churro for our collector on this years bot with no issues...

You can't really see it in that picture because the PVC spacers cover it, but it extends across the entire length of the collector, drives the wheels, and is only supported on the ends. Overall we've been very happy with it, especially because it's light and rigid (and because we already had some around when VEXPro ran out of their long hex stock during build season).

Now granted this application probably undergoes far less force than something like a drive system would, so keep that in mind when planning its use.

We also used churro tube for our collector this year. This was partially because we were out of hex stock and partially because a collector is probably the least loaded part of our entire robot. The loads a collector shaft sees, unless exposed to a high speed collision, are extremely low.

[hrench]

I also want to weigh-in on the 'build it and see if it breaks' philosphy that's being encouraged here.


That is NOT SCIENCE. What Science IS is using previous observation to determine what will happen deductively. The reason we engineers make books of statistics about materials and books of equations about stresses is because science works. If we design something that will work because we've used science, we've taught the kids the value of STEM.

if we design something with the guess that it might break or might not, then we didn't teach math, we didn't teach use of historical empirical statistics and instead we've taught 'trial and error.'

Not a good way to be an engineer.

Don't get me wrong, I was a 'farmboy' and stuff on a farm is most always pushed till it's about to break--or does. Seeing how stuff breaks is a great learning experience and I recommend it.

But engineering --doing the math--works better in the real world and I think that's what we're supposed to be teaching the kids.

[asid61]

http://www.chiefdelphi.com/forums/showpost.php?p=1388120&postcount=16

You stated they would bend here. Please give us accurate information. Our students use chief delphi to help them make decisions where others have more experience. If they see contradictory posts like yours then they and many others will be confused. Please edit your post.

Chow Out

Excuse me. Please read the rest of my post:
"This might have been due to to other drivetriain problems that stopped us from moving too much though.
6063 is considerably weaker than 6061, so I would be wary of using it. For prototyping, it might be fine, but it would be much safer to just purchase some 1/2" 7075 hex stock from Online Metals instead."
Emphasis on my own post mine. If I could edit my post, I would, but I can't. I posted later in the thread to clarify this.
As a side note, please keep in mind that I am indeed a real person, and I have my own experiences, which might be different from yours. Different does not mean incorrect.

Now, to clarify my statement:
My team used some 6061 axles due to 7075 being unavailable late season. We experienced a multitude of drivetrain problems, leading to us only being able to move during half the matches. However, post-season while we were debugging the drivetrain, we discovered that axle bending was not one of the issues in our drivetrain. More likely it was due to incorrect tensioning of chain.
So, I can say this: 6061 is fine for axles, provided is is not under undue stress. It is cheap, which would be the main reason why you would use it. For a prototype drivetrain, it should work fine. However, once the season rolls around, switching 6061 for 7075 should not be a huge issue.

There have been threads in the past detailing shaft strength as compared to shaft diameter (for rounds). There is one here: http://www.chiefdelphi.com/forums/showthread.php?t=38737&page=2
It would be wise to run some calculations, regardless of material being used. It's just plugging in a few numbers.


What's the point of building a prototype drive if you're not going to test durability? If you're building a drivetrain before the season starts, you should build as close to how you would in build season as you can. That way, if and when you do have failures, you can directly apply them to your future robot. If you use parts you know will fail to cut corners, you won't know if the part you intended to use in build will work properly.


Good point. I was thinking of just a proof-of-concept drivebase, but for a proven design like WCD you're right, it would be better to design realistically instead. Might be a good idea to have some 7075 on hand anyway then...

[Andy Baker]

It's worth noting that the 6063 that the Churros are made of has a substantially lower yield.

We have made a mistake on the AndyMark website, incorrectly stating the wrong alloy of aluminum for these extruded churros. They are actually made of 6005A aluminum.

As for using them on an FRC-sized drivetrain as a cantilevered axle, I would not recommend this, for many reasons stated previously in this thread.

Sorry for this error in stating the wrong alloy! The website is now corrected.

Sincerely,
Andy Baker

[llamadon]

Wow this thread got a lot of responses, thank you all for your input!

In regards to the material: After seeing that the churro extrusion was probably not going to hold up in a high stress environment like a drive train (thanks to all of the input from you guys), we researched the different alloys of solid hex stock we have available to purchase. Now Vexpro has issues with being out of stock right when we need something, so we opted for McMaster. We chose the high strength 2024 alloy hex shaft, and will be using this for our drivetrain project. This shaft had the highest yield strength of all the shaft available: 47,000 psi.

[MrBasse]

C1018 steel yield is 54K and C12L14 steel yield is 70K if you need stronger. We've gotten some recommendations from local teams to look directly at tool steel for direct driven cantilevered wheels after this past seasons abuse to avoid bending.

[tim-tim]

Wow this thread got a lot of responses, thank you all for your input!

In regards to the material: After seeing that the churro extrusion was probably not going to hold up in a high stress environment like a drive train (thanks to all of the input from you guys), we researched the different alloys of solid hex stock we have available to purchase. Now Vexpro has issues with being out of stock right when we need something, so we opted for McMaster. We chose the high strength 2024 alloy hex shaft, and will be using this for our drivetrain project. This shaft had the highest yield strength of all the shaft available: 47,000 psi.

You also have the option of using AndyMark 1/2 Hex (http://www.andymark.com/product-p/am-2291.htm) that made from 7075, although not sure if it is -O, -T6, or something else.

[asid61]

C1018 steel yield is 54K and C12L14 steel yield is 70K if you need stronger. We've gotten some recommendations from local teams to look directly at tool steel for direct driven cantilevered wheels after this past seasons abuse to avoid bending.

I really can't think of why you would possibly need to use tool steel axles for anything in FRC. Online Metals has 7075 hex aluminum, which even 6 cim drivers like 254 use.

[magnets]

I really can't think of why you would possibly need to use tool steel axles for anything in FRC. Online Metals has 7075 hex aluminum, which even 6 cim drivers like 254 use.

All cantilevered wheel setups aren't the same. If you have the wheel far out from the bearings, and you're using a large diameter wheel (more torque needed to drive it), I might consider using something stronger than 7075.

[asid61]

All cantilevered wheel setups aren't the same. If you have the wheel far out from the bearings, and you're using a large diameter wheel (more torque needed to drive it), I might consider using something stronger than 7075.

It would have to be an absolutely massive wheel (like 8") and cantalievered out 2" from the chassis bearings to have a factor of safety of 2, according to my calculator. And tool steel axles is still overkill. 4140 steel would be my next choie.
Interestingly, the cantaliever only slightly affects the factor of safety.
Of course, if they want to run something lie 24" wheels, then I could see tool steel being a valid choice. But before doing that I would just increase the axle diameter.

EDIT: Actually, is tool steel even capable of taking axle type loads? It might be too brittle.

[MrBasse]

It would have to be an absolutely massive wheel (like 8") and cantalievered out 2" from the chassis bearings to have a factor of safety of 2, according to my calculator. And tool steel axles is still overkill. 4140 steel would be my next choie.
Interestingly, the cantaliever only slightly affects the factor of safety.
Of course, if they want to run something lie 24" wheels, then I could see tool steel being a valid choice. But before doing that I would just increase the axle diameter.

EDIT: Actually, is tool steel even capable of taking axle type loads? It might be too brittle.

I know teams that used steel axles this year that had bending issues with aluminum and mild steel. Progression of materials landed them at tool steel. Calculators have issues with being based in theory. The problem with theory is that it doesn't always jive with reality.

[asid61]

I know teams that used steel axles this year that had bending issues with aluminum and mild steel. Progression of materials landed them at tool steel. Calculators have issues with being based in theory. The problem with theory is that it doesn't always jive with reality.

What setup were they running?
What type of tool steel, out of curiosity? Some steels are labeled as such, even if they aren't technically tool steels.

[Mike Marandola]

EDIT: Actually, is tool steel even capable of taking axle type loads? It might be too brittle.

There are many varieties. Tool steels become brittle due to their heat treatment. Of course the composition of various tool steels can cause the hardness to be different from each other pre-heat treat. I have only worked with D2 and it bends before heat treating.

[ratdude747]

I also want to weigh-in on the 'build it and see if it breaks' philosphy that's being encouraged here.


That is NOT SCIENCE. What Science IS is using previous observation to determine what will happen deductively. The reason we engineers make books of statistics about materials and books of equations about stresses is because science works. If we design something that will work because we've used science, we've taught the kids the value of STEM.

if we design something with the guess that it might break or might not, then we didn't teach math, we didn't teach use of historical empirical statistics and instead we've taught 'trial and error.'

Not a good way to be an engineer.

Don't get me wrong, I was a 'farmboy' and stuff on a farm is most always pushed till it's about to break--or does. Seeing how stuff breaks is a great learning experience and I recommend it.

But engineering --doing the math--works better in the real world and I think that's what we're supposed to be teaching the kids.

As a senior in EET, I have to at least partially disagree.

More times than not, in EET at least, that IS how designs are implemented. Why? Because more times than not the math is, in practice, an ideal figure that is not as precise as one would like it to be. For example, this is why they still teach Smith charts in Electromagnetics classes (for those that don't know it's a graphical way of analyzing RF circuits), because while the modern math is easy given a good calculator (we used TI-86/89's), in reality, there are so many factors at play that the realistic accuracy is equivlent to using graphical methods.

Another example is with microcontrollers and DACs. While often times what the DAC is controlling (in my case, it was an VCO tuning an FM radio) can be modeled with an equation, in practice in a fair number of cases it works better to build the system and then through trial and error find the values needed. In my case, it was the 100 values corresponding to the 100 "channels" in the FM band. While, even with an optimized "scrolling" set of code to expidite the process, this took a couple of hours to find all the values (in the 12 bits the DAC would see), the result was a radio that despite being in an RF lab with poor reception, would pick up stations with reasonably clear sound. Using an equation to find the values probably would have worked, yes, but the results wouldn't have been as good.

Yes, math has a place, like initial designs (you have to stat somewhere), ballpark values (for situations that are hard to model), and feasibility analysis, but any good design is backed up with a tested and tuned prototype, which in some cases require some educated "guess and check" to get dialed in.

Relying on math alone may get a good grade on an exam but for actual design work, relying on math alone is a poor practice. To me, engineering is ultimately about good design work, not good test scores.

(yes, it's a long post, but I had something to say)

[EricH]

Yes, math has a place, like initial designs (you have to stat somewhere), ballpark values (for situations that are hard to model), and feasibility analysis, but any good design is backed up with a tested and tuned prototype, which in some cases require some educated "guess and check" to get dialed in.

Relying on math alone may get a good grade on an exam but for actual design work, relying on math alone is a poor practice. To me, engineering is ultimately about good design work, not good test scores.
I am going to disagree, and agree. (I know, can't lose, right?)

Disagree: Sometimes, you have only one chance to get something right. Or for whatever reason, you don't have the resources to build a prototype and tune it prior to the final design. In such cases, it's actually a lot better to run the best numbers you have available--factoring in edge cases, outside factors. Relying on math in this case is actually good design work, primarily because there is minimal chance of actually being able to test your work. (There are a number of cases like this--think of large Civil Engineering projects, where you don't get X number of retakes. You just get a case of "OK, we found a problem, how do we fix it before we continue?")

Agree: For most cases, particularly high-volume production cases, it's usually a good idea to build a prototype or other test just to make sure that X will actually work. Even on those one-shot cases, your best math is actually going to come from scale testing, where you build a nice, cheap model and run it through various simulations to figure out exactly what's going on. For more complicated systems, a multitude of tests will be conducted.

tl;dr: Sometimes, the math just needs to be done, and done RIGHT, because no prototype will serve the purpose.

[JamesCH95]

It would have to be an absolutely massive wheel (like 8") and cantalievered out 2" from the chassis bearings to have a factor of safety of 2, according to my calculator. And tool steel axles is still overkill. 4140 steel would be my next choie.
Interestingly, the cantaliever only slightly affects the factor of safety.
Of course, if they want to run something lie 24" wheels, then I could see tool steel being a valid choice. But before doing that I would just increase the axle diameter.

EDIT: Actually, is tool steel even capable of taking axle type loads? It might be too brittle.

Does your calculator include torsion? Or the combined stresses of bending and torsion? Do you make some allowance for the fatigue strength of aluminum?

[FrankJ]

tl;dr: Sometimes, the math just needs to be done, and done RIGHT, because no prototype will serve the purpose.
Like when you press the "fire" button on your lunar lander & hope it will take off and mate up with your command capsule somewhere in orbit?

[Nathan Streeter]

I also want to weigh-in on the 'build it and see if it breaks' philosphy that's being encouraged here.


That is NOT SCIENCE. What Science IS is using previous observation to determine what will happen deductively. The reason we engineers make books of statistics about materials and books of equations about stresses is because science works. If we design something that will work because we've used science, we've taught the kids the value of STEM.

if we design something with the guess that it might break or might not, then we didn't teach math, we didn't teach use of historical empirical statistics and instead we've taught 'trial and error.'

Not a good way to be an engineer.

I'm also going to agree and disagree...

I agree that good engineering inherently should involve calculations/FEA/CFD/'running the numbers,' and that if we shield our students from that entire side, we're giving a poor image of what good engineering is.

However, we'll also be very poorly teaching students if we hand students a single equation for torsion or bending and a table of material properties. In FRC, far more dangerous than testing something (being unsure of failure or success) is the attitude that the equation or FEA is a magic box that spits out a highly accurate solution. Quite frankly, Mechanics/Fatigue/Failure/Stress Analysis are complicated and tedious enough that if we took the time to truly 'run the numbers' for 5% of the bearings, shafts, gears, keyways, fasteners, and frame members we'd be entirely out of time! Much of my Mechanical Design class was spent doing just these calculations for a single loaded axle with gears, bearings, keyways, and fasteners... accounting for the impact of keyways, stress concentrations, cyclic loading, reliability, factor of safety, etc. is very tedious, generally requires iterative calculations, and even then fails to really include the effect of heavy impacts.

Quite simply, it's best to teach students some mechanics of materials, materials science, and mechanical design, but it's also good to teach them that Engineering is saving time with some simple calculations, understanding the significant short-comings to theoretical tests, setting up a good physical test, predicting the success of a component from prior experiences, and knowing when to 'just try it.'

[Nathan Streeter]

Like when you press the "fire" button on your lunar lander & hope it will take off and mate up with your command capsule somewhere in orbit?

Absolutely! However, I'd never climb into a Lunar Lander if it was based simply on the best calculations, FEA, and CFD without any prototypes or prior iterations.

How many times did we send stuff up into space, to orbit the moon, to even land on the moon, or to return to Earth before Apollo 11? Dozens, at least, and excluding those I'm sure there were many mock-ups, prototypes, and tests completed in advance!

All that said, NASA and FRC are at about opposite ends of the spectrum in terms of cost of failure vs Cost of Development. If you guess wrong on the shaft material of your FRC robot, at worst case you'll be dead in the water for a couple matches and may be ruled out from reaching your potential at an event. If you guess wrong on your manned trip in space, not only do you likely lose human life, but you may also shut down an entire program or limit future funding.

[BBray_T1296]

There is a reason that the Soviet Union was able to develop a closed cycle rocket engine deemed by American rocket scientists to be impossible. That reason? Iterative testing, failure, fixing that problem, discovering a new one. Rinse and repeat.

Sure their specific methods may be dirty or whatever, but they still produced an engine ~10% more efficient than anything we could come up with as of the early 90s.

Sure, doing the math and everything is a good way to get stuff done, but at some point it becomes more cost effective (time = money!) to simply do it twice then spend three times as long doing it once.

Lives are not on the line in, failure is always an option. It makes a great learning experience too

[llamadon]

You also have the option of using AndyMark 1/2 Hex (http://www.andymark.com/product-p/am-2291.htm) that made from 7075, although not sure if it is -O, -T6, or something else.

Really? I always thought AndyMark did not carry their own hex shaft for some reason, I would look on their site and never be able to find it. Obviously I did not look hard enough :o

[FrankJ]

All that said, NASA and FRC are at about opposite ends of the spectrum in terms of cost of failure vs Cost of Development. If you guess wrong on the shaft material of your FRC robot, at worst case you'll be dead in the water for a couple matches and may be ruled out from reaching your potential at an event. If you guess wrong on your manned trip in space, not only do you likely lose human life, but you may also shut down an entire program or limit future funding.
Like lets use this O-Ring stuff to seal a solid rocket joint even though experience shows joint rotates in an unpredictable manner under pressure & the O-ring takes a set when it is cold? (Challenger for you young people. :))

FEA really didn't exist in today's terms in the 60's. 640 K ram (kilobytes) was a lot back then. Apollo more so than Challenger.

[Michael Hill]

Like lets use this O-Ring stuff to seal a solid rocket joint even though experience shows joint rotates in an unpredictable manner under pressure & the O-ring takes a set when it is cold? (Challenger for you young people. :))

FEA really didn't exist in today's terms in the 60's. 640 K ram (kilobytes) was a lot back then.

Don't blame the O-ring. Blame the people that knew it was a problem but forced the launch anyway.

[FrankJ]

Don't blame the O-ring. Blame the people that knew it was a problem but forced the launch anyway.
Precisely.

[Oblarg]

Often (especially in FRC) we have complicated systems that you can mock up really easily, for which it is far easier to tune a prototype than it is to work it out the behavior theoretically. There is also the fact that FRC is a high-school competition, and tuning a prototype is something most high school students can do and understand a lot better than a lot of the math required to correctly model many of the things we deal with in FRC (for example, how many high school students are realistically going to understand continuum mechanics?).

Take, for example, the frisbee shooters in 2013. Those were complicated nonlinear systems that would be a nightmare to model. I don't know a single team that did any sort of theoretical modeling of the effects disk compression or motor speed or rail friction on the reliability of such a shooter. I don't know why anyone would even consider approaching the problem that way, when all of those things can be figured out empirically with a simple prototype. I don't think this is sloppy, nor do I think it builds bad habits.

[EricH]

I don't know why anyone would even consider approaching the problem that way, when all of those things can be figured out empirically with a simple prototype. I don't think this is sloppy, nor do I think it builds bad habits.

Correct. In FRC, prototyping is probably the best method you can come up with for quickly getting to the answer. The math done around the prototype should be doing one of two things: It should either be giving you a ballpark setting for the prototype BEFORE you build it ("Hey, the best guess is that the prototype oughta be able to throw 15 feet if we do thus-and-so"), or it should be in use AFTER the prototype is used to see if there is any optimization that can be done and will actually be worth the effort ("Uh... guys, it threw 20 feet, but we can get an extra 5 feet by adding a whodijingle and if we also add a whatsit we get another foot on top of that, but if we just play with this-that-and-the-other we can get 3 feet extra").

Where the bad habits set in are when nobody does even ballpark numbers. That's something that could cost a lot of time and money down the line for somebody.


True story: I've seen what happened when a prototype didn't work as planned, and then it was modified so that it would, but somebody forgot to re-run the numbers for a critical piece. That R/C aircraft was really, really squirrelly to fly. The critical piece? The control surfaces weren't resized after the wing area was increased.

Personally, I really like to set up MathCad (or Excel) with the equations, and see what happens if I monkey with one or two numbers. If I monkeyed with the right numbers in the right way, I get better output numbers. Otherwise... guess I gotta take 10 seconds and re-enter that number and see if my output did what I wanted it to this time.

[MechEng83]

I'll preface by saying I feel like this topic has strayed too far from the original post and may warrant a new thread over the value of prototyping vs doing theoretical calculations.

Now, my day job involves doing FEA (finite element analysis) all day, every day. So I have a personal stake and I feel it's a disservice to teach the kids that analysis is wasted time and you should just build and try it out. I also recognize that the analysis can become so complex and take too long to solve that you'd be better off getting empirical data from a prototype.
Analysis and calculations are tools. Some tools work better in certain situations than others.

One of my work leaders has a saying that I think provides good insight in to how design/analysis should be conducted: "All models are wrong. Some models are useful." The complexities of the physics involved in most of the systems are well beyond high school, undergraduate, and even some graduate courses. But there are simplified models that have the capability of giving you a ballpark estimate of "will this work?"

I often tell my kids to "Do the math. If it works in theory, it might work in reality. If it doesn't work in theory, it probably won't work in reality"
Before some of you jump on this and provide counter-examples about things that work even though theory says they shouldn't, know that this is useful as sorting tool. It helps in the decision making process.

Running numbers using college level mechanics and playing with the numbers allowed my team to come up with a "perfect" elastic counterbalance for a rotating arm. In implementation, it wasn't exact, but it worked well enough to make a system which was effective for the game challenge that year. Don't forget the I in FIRST. Showing my students what was possible with math inspired some of them to learn about it when they got to college.

Relating this back to the original post, If you pulled out the formula for torsional strength, plugged in the numbers and found out you needed 30 ft-lbs, but your system "theoretically" could only take 10 ft-lbs, then you go explore other options, rather than waste time doing the experiment.

[s_forbes]

It's neat that this thread is still alive, I guess "do the math" vs "just test it" can be a hot topic. Relevant:

http://imgs.xkcd.com/comics/efficiency.png

If we have the ability to test something quickly, I encourage it so we get something done. If it requires more analysis first, then we do that. Teams will vary depending on resources. Folks who manage to do both all the time are awesome.


Re: original post - did you test them? If so, what did you find?

[hrench]

[QUOTE=MechEng83;1408152]
One of my work leaders has a saying that I think provides good insight in to how design/analysis should be conducted: "All models are wrong. Some models are useful."
I often tell my kids to "Do the math. If it works in theory, it might work in reality. If it doesn't work in theory, it probably won't work in reality"
QUOTE]

Really like what this guy said.

I realize that axle loading can be a complex equation if you add the cantilever load, maybe some thrust loading, maybe even some torsional resonance (which can destroy systems). Actually, each axle sees different loading based on where on the 'bot it is. Not what I was talking about.

The equation for a churro in torsion has already been posted. That's the math that this thread started out talking about. Other posters have already shown with math that they don't take torsion well.

Simpler math that we can teach kids. So they can learn STEM, not trial and error.

[MechEng83]

It's neat that this thread is still alive, I guess "do the math" vs "just test it" can be a hot topic. Relevant:

http://imgs.xkcd.com/comics/efficiency.png


A bit creepy that this is actually the most recent XKCD...

[FrankJ]

A bit creepy that this is actually the most recent XKCD...

You do know that Randal is a Firster from way back?

[MechEng83]

You do know that Randal is a Firster from way back?

http://imgs.xkcd.com/comics/first_design.png

Yes.

[Oblarg]

Simpler math that we can teach kids. So they can learn STEM, not trial and error.

Trial and error is a integral part of STEM, though. It's always worth considering to what degree the problem at hand is worth modeling, and to what degree you should just try stuff and see what works. Reality is a really good simulator ;).

We should certainly teach math when it is relevant and useful, but it is foolish to write off trail and error as "unscientific." After all, empiricism is pretty much the fundamental idea behind all of science.

[Chief Hedgehog]

Trial and error is a integral part of STEM, though. It's always worth considering to what degree the problem at hand is worth modeling, and to what degree you should just try stuff and see what works. Reality is a really good simulator ;).

We should certainly teach math when it is relevant and useful, but it is foolish to write off trail and error as "unscientific." After all, empiricism is pretty much the fundamental idea behind all of science.

Agreed! Sometimes my shop detail can mock-up a working part before my strategy team can prove their math!

Which goes back to FRC 4607 CIS failing better than most teams...

[hrench]

Trial and error is a integral part of STEM,

I was going to start with "where in STEM is trail and error?", then I realized...

STEM: Science, Trial, Error, Math.

Thought that was funny.

[FrankJ]

A lot of great inventions are a result of an accidents or observation of an unplanned events. Vulcanized rubber, post it notes, Penicillin are good examples of this. The intelligent, documented observation of the event is what distinguishes from the "hey watch this" type of randomness.

Since the people with the money don't like the term trial & error, you can always call it experimentation or research. Both calculation & experimentation are useful tools to solve a problem. Engineers are really just applied problem solvers.

[JamesCH95]

A lot of great inventions are a result of an accidents or observation of an unplanned events. Vulcanized rubber, post it notes, Penicillin are good examples of this. The intelligent, documented observation of the event is what distinguishes from the "hey watch this" type of randomness.

Since the people with the money don't like the term trial & error, you can always call it experimentation or research. Both calculation & experimentation are useful tools to solve a problem. Engineers are really just applied problem solvers.

http://rubicon-foundation.org/wp-content/uploads/2014/02/Adam_Savage_Quote.jpg

Often the hardest part of this is fully understanding what parameters are actually affecting the situation, as well as to what degree they matter. This is where having a theoretical understanding of the mechanisms at play (i.e. a model) is very useful - it tells you where to look. And thus experimental variables were born...

At work recently a colleague of mine and I spent hours trying to work out the math for a particular problem. Couldn't figure it out to a satisfactory result. So, we just started testing and *poof* the answer revealed itself. However, had we not at least TRIED to do the math we wouldn't have known what parameters to vary in our experiments.