Rotating Torque Sensing

[Marty]

So the human powered submarine project is in its early stages, we have almost a year to play. Right now we're setting up to play with propeller designs. Available power is about 1/2 horsepower (max, in their dreams). Shaft speed tbd. We're trying to build a test cell to test various propellers for thrust at various speeds. We have the mechanics of it figured out. A battery drill delivers about the right power, is variable speed. Harbor Freight 4.5" RAgrinders ($12!) for right angle ~3:1 gearboxes, etc.

Once we get this in the water it's going to be obvious (these are ~middleschoolers w/o a lot of experience yet) that we need to measure shaft speed and torque, i.e. have a dynamometer. I haven't looked into measuring shaft speed yet but imagine a bike computer setup will work fine. I don't mind them doing the math (power = torque x speed) by hand until we have a lot of experience to imprint it a bit better in their minds. At that point we can probably use the same magnet/sensors, maybe w/o even any signal conditioning to feed the Teensy.

But torque has me wrapped around the axle :- ) . One idea is to measure the static torque on the drill as it's driving the whole mess. A simple load cell in that case, but the controlling hand might accidentally apply some torque as it squeezes the trigger. I'd actually love to have an actual rotating torque measurement. A purely/largely mechanical solution would also be fine. High accuracy is not needed, good repeatability is a good thing, tho. Shaft speed is limited to that of the drill (less than 600 rpm, 10 cps). I could have a whole rotating Teensy torque sensor with wireless talking to another wireless Teensy. That seems awkward. Any ideas? Thanks in advance, btw.

 

[skimask]

Purely mechanical...a few...first two using a fish weighing scale
1) Lock the R/A grinder into some sort of fixture that allows it to rotate a bit around the plane of rotation of the grinding wheel (in this case, the connection to the prop). Hook up the end of the grinder to the fish scale. Attach the fish scale to the floor or ceiling as appropriate. Turn on the power, grinder tries to rotate, torque applied will pull on the fish scale and give you a relative measurement.
2) Grinder goes over here with a pulley on the end of it. Prop goes over there with a pulley on it. Both are connected with a loose fitting belt. In the middle is an idler pulley attached to the fish scale on an arm that allows the idler pulley to be pushed down on the belt hard enough to get it to make the belt NOT slip. Push down on the idler to the point of no slippage, read the fish scale.
3) Dirt cheap, simple, probably not accurate, probably not repeatable. Make a flappy board behind the prop. Crank up the grinder. The farther the board moves, the more torque is applied to the prop, the more water is being moved.

Getting a bit more complicated...
Make an RPM sensor for the prop, hall effect, optical, guy counting really fast, doesn't matter. Use a heavy duty bench power supply to apply power to the grinder. Set a target rpm for the prop. Turn up the voltage on the bench power supply until the target rpm is reached. Read the volts and amps off the bench power supply. Multiply the two, you get watts, which is power. Accurate and repeatable...ish...assuming you've got halfway decent pieces/parts involved.

 

[Marty]

Thank you, SkiMask. Some more detail:

We're only using the gearboxes from the grinders. The power source is a hand drill which has a more reasonable speed, 0-10 turns per second vs ~200 for an RAgrinder. I'm not sure but I'll bet that the low speed torque is much better with a drill than the grinder, too.

Thrust from the propeller under test would be the OTHER quantity to measure, in addition to powerIn (to compute a propeller's efficiency). For thrust, I think a simple (waterproof) load cell being squeezed between the prop and the wall of the pool will do. No problem.

For powerIn we need the function of a dynamometer, measuring rotational speed (no problem) and spinning torque (hmm).

Your 2 fish (luggage) scale ideas are spot on. Measuring deflection of the drive side of a belt drive actually sounds maybe best (except that now I've got a group of kids thinking and building a shaft transmission instead of a belt). Measuring torque on the motor is likely the way we're going to go. Because we're using a drill, there'll be the torque 'noise' of the trigger finger wiggling around to set and maintain a certain speed, but it's super do-able and EASY.

I recognize it would be best to put the dynamometer right at the propeller blade, but that'd be wet and less accessible during testing. Your idea of measuring torque at the motor simple means that we have to measure the no-propeller (no-load) power consumption of the system before measuring the with-propeller power. No biggie, good conceptually for the kids.

Gee, this isn't turning into a very Teensy project. Yet. Load cell is there, bike computer speedometer is there. Actually, if we get a load cell in the place of the luggage scale then we're all set to integrate into a Teensy. That'll be SO COOL!

Hmm. Eventually we're going to need to have an on-board (wet) system. ______Marty

 

[jonr]

You can put a mark on each end of a shaft and then use optical sensors to measuring the timing. This provides RPM and, because the shift will twist and the timing between the two ends will change, an indication of torque.

 

[Epyon]

For dynamic torque sensing you could use strain gauges. You place the strain gauges, amplifier and Teensy on the shaft and make your own rudimentary rotary power transformer (not too hard for those low power levels). An optical (e.g. leds) or wireless (e.g. Xbee) data link reports the calculated torque to the supervisory system.

You could also build a magnetostrictive torque sensor, but maybe that would be overdoing things .

 

[skimask]

We're only using the gearboxes from the grinders. The power source is a hand drill which has a more reasonable speed, 0-10 turns per second vs ~200 for an RAgrinder. I'm not sure but I'll bet that the low speed torque is much better with a drill than the grinder, too.

Ahhh...I was under the impression you were going to use the whole grinder. Either way, same difference. Measuring power either way.

Thrust from the propeller under test would be the OTHER quantity to measure, in addition to powerIn (to compute a propeller's efficiency). For thrust, I think a simple (waterproof) load cell being squeezed between the prop and the wall of the pool will do. No problem.

Again, a boat type thing pulling a line hooked up to a fish scale will give you LBS of thrust. Cheap, simple, repeatable....ish...

This is more than likely way above what you want to do, but...
Back when I worked C-130's, there was an instrument called a "Torque Indicator" used to measure the amount of power put out by the engine on the propeller. Basically you had a shaft that would twist X amount under a certain amount of force. You had toothed hall effect sensors on either end of the shaft, one on the motor end, one on the prop end. With no load on the engine, teeth were perfectly lined up, no torque being applied. When you kicked up the engine rpm and angled the props, the props would get a load on them (eg. torque was applied), and the shaft would twist a bit. The hall effect sensors were there to detect that twist. The more the twist, the more the teeth would get out of alignment, the more the difference in the signal being produced between the 2 sensors, the more the torque was being applied.
Just throwing it out there.

 

[PaulStoffregen]

Can you get a reasonable estimate of torque by measuring the motor current, and perhaps subtracting an offset for the current needed for magnetic and static losses?

If you want to use Teensy for the tachometer, the FreqMeasure lib should help.

 

[Epyon]

Can you get a reasonable estimate of torque by measuring the motor current, and perhaps subtracting an offset for the current needed for magnetic and static losses?

That would require knowledge of the motor characteristic curve. It's not a linear relation, especially during low and high current conditions. Commercial VFDs try to make an estimate with a 'auto tune' to use torque control, but research in our lab has shown that current control is only really usable when having a detailed motor curve. When operating in the linear region it could be useful, but it would involve some trial and error .

 

[Marty]

It's an honor and a real pleasure playing with you guys!

Jonr, SkiMask, re your similar ideas (Torque Indicator) if I had or could contrive a shaft that'd twist a reasonable amount*, two bike computer-like sensors, or two light interrupters**, one on each end (well, spaced as far apart along the shaft as reasonable), would probably do great. I really like the idea--Once Teensy arrives in the game, of doing both RPM and torque with it.

*: So, what about a section of stiff hose as a twistable shaft? If I calibrate it periodically (like after disuse, high torque, significant temperature change...) it ought to work, ya think? Time to destroy my garden hose, maybe. ("Look, Honey, 24.5' of it still work perfectly, it's just that one 6" section that isn't so hot. That's, like 98% perfect, right?")

**: I don't have any experience with bike-like hall effect sensors and the ones used to home some stepper stages have a pretty long routine of super slow stepping and back and forthing, seemingly designed to overcome noise in the signal. I'm thinking an LED/photodiode with a interrupting disk could give me a nice sharp signal edge and so, better time resolution. Thoughts?

Paul, haven't looked at it but yes, now I think I'm headed in that direction, doing tachTork both. So, at risk of being too lamely a slacker, I ask is there a library with a handy function like a universal counter's "Time A->B" function? I kind of hate motors in part because nothing about them is very linear.

Epyon, I didn't know anything about magnetostriction, the Villari effect; Neat! I know too little to even hazard a guess on how to figure what level of signal is possible, quite how to rig it up, and what precision the machine would have to have. I know that the shaft would be part of the magnetic path. This would be a great science fair project for a kid who didn't mind not winning, or who even thought it'd be cool to not win, because the judges didn't know what they were looking at.

So let's see--First take on this will probably be a bike speedometer with its own readout (lame, yes but quick and do-able) along with the power drill handle torque-O-meter (SkiMask's first post idea 1) ). That will give me time to get the real thing built.

 

[Jake]

One commercial way to get torque using the twisting shaft idea is at http://www.binsfeld.com/torquetrak/torquetrak-10k/. This uses a ~$100 strain gauge and radio transmission of the strain signal to the receiver. Expensive but accurate. Slip rings are a way to get around the radio transmission technology, but I expect you want the amplification before the slip rings. Taking a look at the commercial solution might provide some food for thought though.

Electronic bathroom scales are quite inexpensive and I suspect incorporate strain gauge based load cells. It might be worth taking one apart to see what they have done. I believe quarter bridge strain gauges are in the $20 range.