[WIP] Geartrain Testin Stand: horrible pictures and first results

[schraubedrin]

(Warning: lot of text, dry theory and zero nice looking pictures ahead)
Sorry for those horrible pictures - my phone from 2014 has a nice size but the camera is showing its age

Whenever i see big and/or powerful model i wonder about the geartrain design, efficiency and reliability.
I always assumed that it's better to transfer power with high rpm, rather than high torque.

It took some time to figure out how to put this assumption to the test, but here it finally is: my geartrain testing stand V3.1:


The idea is to measure the torque loss across the tested geartrain at a given rpm for a known resistance.
Those measurements can then be compared for different geartrains.

There are several challenges i had to solve and that still can be optimised - input is very welcome!

• Adjustable and reliable power source: i opted for a switching mode power supply on PF motors
• reliable torque measurement: originally i planned to get the input torque from electrical measurements on the motor and trust the brake's torque. Instead i built sensors based on differentials.
• rigid test-geartrain mounting: i think i used more 5x7 frames than 42055 but i'm still not satisfied (more on that in the results)
• consistent braking torque that doesnt change over time: i wanted to be able to run the testing stand for extended time, so no weight lifting or friction. Instead of an electric brake i opted for a fluid/air brake

The assembly consists of:

1. 2x PF L motors as power source with the added rpm sensor on a 1:3 ratio to get a better sensor resolution (from 20 rpm to 6,7 rpm)
2. A rotating set of weight blocks on a 1:3 ratio as high inertia rpm-buffer
3. The input torque sensor based on an inline planetary gearset and a lab scale
4. The test-geartrain
5. The output torque sensor
6. The aerodynamic brake as power sink

Let's take a closer look at the seperate modules:

The power sorce are these two PF-L motors, as the brake is powerful enough to drive one motor alone into thermal shutdown - even without a testing geartrain.
They are regulated by a switching mode power supply. Idealy i'd set the driving torque via the current (an electric motor's torque is proportional to its input current).  My current-dial isn't nearly acurate enough, but fortunately those armchair-engineer thoughts don't matter in the real world
As the rpm sensor only has a resolution of 20 rpm, i attached it with a 8t gear to to get a 1:3 ratio and therefore a 6,7rpm resolution (not visible in the picture).

The inertia rpm-buffer was added because i had rpm oscilations, especially at higher torques.
It's build from two weight blocks which spin at a ratio of 1:3 to store more energy (kinetic energy is proportional to rpm squared). This ratio is achived by 36t and 12t gears instead of the simpler 24t and 8t gears as bigger gears generally cause less bearing loads (longer levers with the same torque). They are hidden in the dual 5x7 frames to the left and right.

To minimize the bearing losses of the weight block assembly, i didn't mount it directly in stud holes but instead layed it on four black discs. This way the weight lies on 8 holes (2 per disc) instead of 2 and has a lower rpm. Another advantage of this bearing system would be a way lower breakaway-torque, but that's irrelevant in this case.

The torque sensors work by changing the direction of the roation and bracing the idler gears on lab scales.
I built my own planetary differentials because they are more efficient than bevel geared differentials like those from lego and i expected a lot of torque.
I had a hard time figuring out the angle between the gears so they don't have any preload. Then i realised that i could have made my live a lot easier by switching the 20t and 16t gears around
The input torque sensor (red) is mounted on small turntables which transfer the force from the high torques. As the output torque sensor (green) only has to transfer the torque from the air brake, i mounted it directly on the turning axles to get more precise readings (no stiction from the turntable)
The lab scales have a resultion of 0,05g up to 1kg, which is far to precise for this application.

Here's a cutaway of the gears in the sensor. The input (red) and output (green) turn in opposite directions through the idler gears which are mounted on the yellow sensor beam, which then presses onto the scale.

The tested gear train get's mounted in a big compromise of stiffness, space and removability. This part of the whole assembly has the most potential for optimization.
For the beginning i made tests with a 3:1 ratio (power transfer with high torque) or 1:3 ratio (power transfer with high rpm).

The power sink was the part of the project i experimented the most with, until i settled on this air brake.
It's surprising how much energy it takes to turn those dishes. I can change the resistance by moving the dishes further inside, changing the gear ratio to the air brake, or change the dishes for some different airfoil.
It has a permanent 1:3 ratio to provide some resistance even at low rpm.
The size of the airbrake is also the reason i had to mount all the other components higher

The connection between the modules is achieved with universal joints to prevent potential resistance from misaligned axles, the modules are braced against each other with two 5x7 frames: one immediately under the universal joint to assure correct allignment and one further below for torque transfer.

 

Testing is unfortunately a pretty involved process:
First i have to set the rpm via the voltage of the power supply (this can be a challenge even with the added inertia)
Then i have to note the voltage, rpm and the min. and max. values on the scales manually. This is necessary (and difficult) because the scales have a high readout frequency and there's a lot of vibration in the system leading to readout variations from 2 to 7g.
Then it's off to the next measuring point.
I do this once with rising and faling rpm to prevent measuring errors by hysteresis.

Automating this with a mindstorms system would be great but i have neither the sets nor any idea how to implement the torque sensors.

First results showed that the variations in the testing runs with the same gear assemblies are bigger than the differences between the different gear trains
I don't trust those results because of this (and because it goes against my intuition )

(green is the high torque transfer, violet the high rpm transfer)

 

Now i'm not sure how to proceed.
I could make more measurements in the hope that they'll even out eventually.
Or rebuild the modules to be more stiff?

I'm looking forward to your comments and helpfull insights

[TeamThrifty]

I love some theory..! This is basically a test of mechanical losses, the energy in is consistent, so you're measuring how much survives... so high rpm drive trains have more losses (friction) and less torque is available at the end. I think this sounds logically correct?

So the conclusion would be to drop the rpm as close to the motor as possible to slow the drive train down. This probably means planetary's aren't the most efficient option?!!

[allanp]

From my experience it is true that high rpm drive trains are a bit less efficient, but I generally prefer them because you get less jerky movement at the output.

[MinusAndy]

11 hours ago, TeamThrifty said:

I love some theory..! This is basically a test of mechanical losses, the energy in is consistent, so you're measuring how much survives... so high rpm drive trains have more losses (friction) and less torque is available at the end. I think this sounds logically correct?

So the conclusion would be to drop the rpm as close to the motor as possible to slow the drive train down. This probably means planetary's aren't the most efficient option?!!

Definitely, but sometimes the torque of running everything at low rpm high torque from the motor can put loads of stress on gears and cause wonky gearbox syndrome.

[JintaiZ]

Great idea!

[schraubedrin]

10 hours ago, TeamThrifty said:

high rpm drive trains have more losses

As i said, i wouldn't trust this data too much. The difference between the test runs is bigger than the difference between the geartrain concepts.

 

11 hours ago, TeamThrifty said:

This probably means planetary's aren't the most efficient option?!

I'm with @Allan and @MinusAndy on this one: Although higher torque is more efficient, it's much more difficult to brace it, especially through wheel hubs. And the bracing seems to influence the efficiency even more.

[2GodBDGlory]

13 hours ago, allanp said:

From my experience it is true that high rpm drive trains are a bit less efficient, but I generally prefer them because you get less jerky movement at the output.

Also, low torque in the drivetrain makes it less likely that your gears are going to skip--most important in complex gearboxes