pic: 254/968 Gearbox - Destructive Testing Results

For everyone who was curious about the strength of the gearbox components, we had a third party testing laboratory verify strengths as part of the final analysis for our senior project.

Some results to note:

  • 170 foot pounds of torque before failure on the output shaft. That’s the equivalent of a higher powered 4 cylinder engine you’d find in your car!

-4,119 pounds bending force before failure on the output shaft. Again, very high and quite unexpected!

-155 foot pounds of torque on the 80 tooth gears web before the test rig failed! The gear was intact with no noticeable deformation!

Also, if you look closely at the components, you can see clear examples of the inherent stress risers caused by sharp corners. Almost all of the shafts failed where there is a sharp corner. This was avoided as much as possible, but without specialized tooling, elimination of these sharp corners was difficult. For our purposes it was not completely necessary either, as almost all components had factors of safety well above what a FIRST robot needs.

Feel free to ask any questions.

Good job! I think this is one of the first times strength testing has been performed on a FIRST robot powertrain?

Wow. I knew that it was strong when I played with it… But not THAT strong. Props on a seriously sweet design!

Not all teams have a gearbox that is also a senior project though. Its cool to see this information. I seem to remember travis saying that it wouldnt be to practical to manufacture at the same level/cost as an AM shifter. Thats too bad. I’d love to throw a couple of those on my robot.

Kind of…except that transmission output shafts are subject to the torque from the engine, times the gear reduction in low gear. Which is why automotive transmissions have larger output shafts than input shafts.

Interesting tests, thanks for sharing! It’s a nice transmission, but not nearly as cost effective as the AM shifters as noted. Unfortunately real life generally means economic considerations are top priority for a design.

Yes, automotive torques are measured at the wheels in a ~1:1 gear or at the crank… However, I will stand by my original point that 170 foot pounds of torque at the wheels is still more than a honda civics engine can output. Correction: (without a transmission)

I am not saying this will withstand that torque, as there is no factor of safety present, I am simply giving a basis for comparison for those who see these results as just numbers.

Also, with regards to cost effectiveness, that’s not always true. This project was based around a model of weight being the utmost concern, and cost being forth of fifth down the line. Also, weight can become a cost savings itself. When it costs ~$3,000 to $5,000 per pound (estimate) to send a sattelite into space, that weight savings sometimes can create an economic savings greater than the added cost of manufacturing.

Wow! That is one strong tranny.
Were these results close to what you were able to calculate?

More or less. We didn’t concentrate too much on these forces, as the gear teeth would strip long before any of the shafts or webs would fail. We spent considerable time calculating bending strengths on the gear teeth to make sure they would be okay.

As such, yes, these were very similar to what we calculated and predicted. Some components had an initial 10X factor of safety in them, but due to the design, the shape and size couldn’t be changed much and further weight wouldn’t have been easily removed. Since the components didn’t weight much at this point anyway, we didn’t feel that a design change was worth while or necessary.

How were you able to calculate the strength of the components?

Thanks, Travis

For gear tooth strength, we used the Barth revision of the Lewis bending forumula. For other strengths, we used other bending eqns, torque eqns, and shear strength eqns depending on the part we were analyzing.

We also did finite element analysis on all of the components to verify that our initial calcs were correct.

Here is an example of the FEA for the 80 tooth gear. These parts were subjected to two 20 KSI loads on the gear teeth, ~ 90 degrees apart.

Stress Distribution




So, how much are you guys selling it for next season? :slight_smile:

I think the real question is, where can we find the plans/instructions so we can make it better? (Is that even possible??)

Always! “Continuous Improvement” is the future! :wink:

(first part removed, I think we cleared that up)

In the real world, cost is a major factor affecting design. The 254/968 robot, and space missions, are notable exceptions. AndyMark sells a lot of transmissions, because they have a very cost effective design.

Yes, torque is increased as the gear reduction increases… of course. I think I added a little too much there at the end with “transmission or otherwise” My ASSumptions were based on engine output torque, neglecting that fact that you have the ability to increase that torque in lower gears. Sorry.

I think the confusion here stems from my failing to explain that this rating I quoted is the “specification” value. Usually these horsepower and torque ratings are measured on an engine or vehicle dyno in a gear close to 1:1, or simply at the flywheel if on an engine dyno.

So no, this shaft wouldnt handle the torque at the wheels in 1st or 2nd gear on a honda civic. However, most people wont say their honda has a torque of 400 ft lbs because that is entirely dependent on the transmission and gear reduction, as you mentioned. And it would in fact be strong enough to handle the torque coming out of the engine.

I should read more carefully next time. :wink:

Also, I should add as I was somewhat upset by the other comment… we did do a cost analysis and mass production process and material selection. As the quantities get higher (very high), these gearboxes actually do have the ability to be cheaper than a steel alternative. Material costs become negligible, and the decreased machine run time and tool wear begin to make up for the costs added by anodizing and additional lightening. As manufacturing engineering students, we were far more concerned with these issues than we were the mechanical engineering based calculations.

Interesing points about the reduced costs with very large production runs.

Many teams rely on others to shoulder the “up front” cost of buying/leasing CNC equipment, and end up buying parts that available off the shelf …it’s neat that you had the opportunity to design and make these transmissions!

Any chance you’ll do endurance (constant fluctuations in fatigue) testing with this transmission? Granted these things are built for (at most) 3 competitions and a bit of practice (about 120 minutes of run time if you make the finals at competitions) but what would happen if you tried to put the robot under its stresses for say, twice that amount of time in a competition-style environment?

My Material’s Science intro class went over a bit of sheer stress analysis but the details escape me at the moment. Is there a way to get a product lifetime estimate from the tests done already?

Excellent opportunity you have to be able to do such a disciplined and detailed analysis.

Funny you ask. We did do long term accelerated wear and fatigue testing.

We built a total of 16 gearboxes. 4 of them went on competition robots. 254’s comp bot saw ~8 hours total run time including practice at 4 events. 968’s comp bot saw only ~4 hours run time including practice at 2 events.

However, there were another 4 gearboxes built and used on two practice robots which each saw ~100 hours of driving. After these 100 hours of aggressive use, we saw little to no difference between the practice robot gearboxes and the competition gearboxes with less than 10% of the runtime the practice ones saw. It was determined that the anodize wears initially after 2 or so hours of use and remains at that worn, broken in point for the remaining time.

In addition to those 8, we used 2 more gearboxes on an accelerated wear test platform. We mounted the gearboxes onto a mock side rail and programmed a continuous loop program to cycle the motors and actuate the shifting cylinder: full forward, shift, full reverse, shift, for another 100 hours under load.

Here is a picture of that test platform after the 100 hours of testing. You can see the anodize wear, but after measurement, the involute tooth profile is intact. We also disassembled the gearboxes and checked for any fatigue issues and further damage or wear… while we didn’t have the ability to do a thorough lab test, our brief visual inspection showed no damage or signs for concern.



Also, here is a video showing the test platform with a wheel mounted for later use as a demonstration unit. As shown in the video and pictures, there is no load on the output shaft. This video is strictly to show the cycling procedure used. We may bring this to IRI if people want to see it?


Lastly, here are the components removed from the test platform after ~100 hours of continuous use. Some wear is noticible, other portions are just dirty from the anodize wear.


These are amazing results. Even better, Travis has done a great job explaining this testing and failure modes.

It is one thing to be a great designer, it is another thing to be able to explain what you did, how things work, and why things happen.

Personally, I appreicate this work. Needless to say, FIRST will see better designs from many people because of this effort. This design has indeed “raised the bar” in FIRST.

Most respectfully,
Andy Baker

Very nice, very strong, looks like a great piece of equipment.