Measuring mechanical torque of a generator while running

GetDeviceInfo

Joined Jun 7, 2009
1,844
You see where I'm going with this? Just because I have volts and amps doesn't mean I have the data necessary to declare xxHP.
Then again, volts and amps are the necessary values. When you measure an engines output, You could either apply a known load, or you could apply some load, and measure it by torque and speed. Volts and amps give you a known load.

I’m kind of thinking your original idea has some merit. I’d rig up a test fixture and try it
 

shortbus

Joined Sep 30, 2009
8,467
Sorry, didn't realize you were talking about engine dynos. I was talking about drive-on chassis dynos. I don't know anything about engine dynos.
Not just engine dynos use that type. The AC dyno is just an "absorber" as they are called. Wheter tis any of the type dyno there is an absorber, to put the load on what is being measured.

The really old and out dated type of dyno you were talking about, so lod I didn't know without looking them up even existed couldn't give what is commonly sought out from dyno testing. Accerating a weight pretty much would only give a total load, not what dynos today are after. The moden dyno is used to make a curve, a plot of values at different RPMs. They set a RPM and alppy a load until the RPM drops. That makes point of refference for that RPM then they increase the RPM(usually in 100 RPM steps) and do it over again until the preset redline is reached.


Ok let's say you're using a dyno which has an airboat propeller as its load.
Don't think they do it that way. They would tie the air boat to a load cell and apply power to find the prop was capable of, but that wouldn't be HP, just force of movement from the prop.

Or let's say you're using a dyno which has a brake disk as its load.
Same answer as above. Your not finding the power of the prime mover but the power of the brake, and it would be in force not HP. You know the HP of the prime mover.

Or let's say you're using a dyno which has a water pump as its load.
Again same answer, you are comparing oranges and apples.

You see where I'm going with this? Just because I have volts and amps doesn't mean I have the data necessary to declare xxHP.
And your gen/alternator is just being used as the absorber that was mentioned in the first response of this post. The more I thought about what I said about not using the PTO to find your motor HP, the more I wondered, is your PTO live? Does it work even if the transmission isn't engaged? A live PTO usually has it's own clutch, and is sized to not break things in the PTO drive train. So another thing against use in the PTO for engine performance data is you would only be reading the PTO clutch slip point.
 

shortbus

Joined Sep 30, 2009
8,467
Now I spent some time last night looking at DIY dynos and here are some links for you. Each of the Youtube ones also have many other you might be interested in.

http://dtec.net.au/Inertia & Brake ... - Kart, RC, Bike & Car- Engine & Chassis.htm

The next ones are about Absorbers, hydraulic type, using water. They are not as complicated as I was expecting.
How they work -

These are about a guy making his own from scratch -

 

GetDeviceInfo

Joined Jun 7, 2009
1,844
I had the pleasure of commissioning electric and hydraulic topdrives for a few years. These were dyno’d for 12 hours at full load as part of thier certification. Used both friction and water dyno. They applied a load only, no sensors. With that, we knew what the input power was, but no idea what the efficiencies were. That was solved with a calibrated strain fixture mounted on the output quills.

Being a bit of a motor head, Ive been looking for a used water dyno to test turbo applications on the Jeep 4L., just in case anyone knows of one kicking around.
 

hrs

Joined Jun 13, 2014
308
since the center distance between pulleys is exactly 1ft, that load cell should read out directly in ft*lbs! Right? of course? ...
The principle is correct. I don't know what kind out output a load cell gives. Volts per x units of force? The torque will be that of the generator shaft. Losses will be due to friction, maybe a few percent.

If the cardan shaft connects tractor and generator at an angle I think you will measure a cyclic torque proportional to the sine of the angle on top of the nominal torque. This apparent cyclic torque will not go "into" the generator, it's actually a cyclic force on the load cell. If you manage to trim the angle to 0 degrees the force will go away.
Code:
                /=== tractor
               /
              /
             /
generator===/
I think it could work.
 

Thread Starter

strantor

Joined Oct 3, 2010
5,532
I had the pleasure of commissioning electric and hydraulic topdrives for a few years. These were dyno’d for 12 hours at full load as part of thier certification. Used both friction and water dyno. They applied a load only, no sensors. With that, we knew what the input power was, but no idea what the efficiencies were. That was solved with a calibrated strain fixture mounted on the output quills.

Being a bit of a motor head, Ive been looking for a used water dyno to test turbo applications on the Jeep 4L., just in case anyone knows of one kicking around.

I also commissioned topdrives for a while. Didn't do any dyno testing but saw some pretty crazy stuff. One that really made my nerd knob swell was the NOV IDS900(?) with a coreless PM motor with like 9 phases. Mud went straight through the center of the motor. it was actually a giant P.O.S. of but it sure was cool.



I don't know where you could get a used water dyno, but I came up with this (i think) cool idea you could use with any kind of load. You might even be able to apply it to the driveshaft of your jeep and have a constant HP readout as you drive around. I thought it up when I was afraid my idea in the OP wouldn't work, and I may still pursue it, whether or not my OP idea works out. It would enable me to test HP using any attachment, not just my generator.

I was looking at how shaft torque is measured in industrial applications and found that they often measure actual twist of a shaft. Examples:







I tried to think of a simpler way to achieve that and I probably missed the mark (mechanically speaking) but in terms of electo-implementation, it's a single sensor which provides a PWM output containing the data for both RPM (pulse frequency) and torque (pulse with/duty cycle). You could wire this straight to an arduino or similar and have it working with just a few lines of simple code.

You would use a shaft slightly smaller than normal so you get a sufficient amount of twist out of it at max torque, and surround it with nested tubes that are coupled to slotted disks. And a photointerrupter.



Capture1.JPG
Capture2.JPG
Capture10.JPG
Capture3_9.jpg



Each tube is fastened to the shaft at the end opposite its slotted disk, but otherwise they are not fastened anywhere else. they are free to slide against each other.


Capture11.JPG



As the shaft twists, the tubes attached to its exterior don't. So the plates fan out sorta like a camera iris, narrowing the slot as seen by the photointerrupter. This will appear to the measurement device as a smaller PWM duty cycle



Capture12.JPG



To simplify the thing you could probably do away with the slotted disks and just slot the tubes. Move the photointerrupter 90 degrees to measure the slots in the tubes (I didn't bother to draw the slots in the tubes, I think you can get the idea)



Capture15.JPG
 
Last edited:

Sensacell

Joined Jun 19, 2012
2,774
Good solution - but why the multiple tubes??

All you need is one shaft, one tube, both with slotted disks.

The inner solid shaft sees torque, the outer is a reference.
 

Thread Starter

strantor

Joined Oct 3, 2010
5,532
Good solution - but why the multiple tubes??

All you need is one shaft, one tube, both with slotted disks.

The inner solid shaft sees torque, the outer is a reference.
The 2nd video in my post describes a system like that, one torque pulse, one reference pulse. It's simpler (probably better) but as I see it the "gotcha" is that (I think) you would need a very precise speed feedback in order for the time between the two pluses to mean anything concrete. And even if you had that, your torque value is still the result of a theory and an equation that you could only ever take on faith unless you had another calibrated Dynomometer to prove it against. I wanted something I could set up in a jig on my workbench, put a lever on it, hang some weights on, observe and measure the twist, and be totally confident that it will do the same thing while spinning as it did on my bench. Directly measuring the torque during operation as opposed to calculating it, gives me more of a warm fuzzy. And I thought it would look cooler. And I like PWM.
 

Thread Starter

strantor

Joined Oct 3, 2010
5,532
Good solution - but why the multiple tubes??

All you need is one shaft, one tube, both with slotted disks.

The inner solid shaft sees torque, the outer is a reference.
After some more thought I think I'm closer to your line of thinking. Two slotted disks should do the trick. You could construct a PWM wave by combining their two outputs, but it wouldn't be as simple as just adding them.

Here's how I envision the output at a low torque value. A leads B by just a few degrees. They overlap, so they could just be added (well, not added, but... overlaid?) OR. If A OR B = true, out = true.
Screenshot_20210319-023325_Samsung Notes.jpg

But when there is more significant twist, something more sophisticated will be needed. Or else we get this:
Screenshot_20210319-023816_Samsung Notes.jpg
When what we want is this:
Screenshot_20210319-023938_Samsung Notes.jpg

That's what my multiple slotted disks would do; fill the gap between the torque pulse and the reference pulse. But figuring out a digital solution is probably easier than making all those tubes and disks. What would you propose? I know how I would do it in a PLC but this isn't something I want to use a PLC for.

But even the PLC solution I have in mind would be confounded by change in direction of rotation where the slotted disks wouldn't.
 

LesJones

Joined Jan 8, 2017
3,127
My understanding is that one disk is fixed to the driven end of the shaft that twists due to the torque and the other disk is fixed to the driven end of the shaft. The tube just extends the position of one of the disks so it is close to the other disk so the light has to pass through the slot in both disks. when the disks are lined up (I.E No torque). As the shaft twists due to torque the percentage of time the light is allowed is reduced. If we assume that the slots are the same angular length as the space between them the the output wave form will have a 50% duty cycle. This will gradually decrease until the slot in one disk is fully covered by the gap between slots in the other disk. At this point the duty cycle would be 0%. If the torque increased above this level then light would start to get through again. You would have to ensure that the torque did not reach this level. So the duty cycle of the waveform would be a direct measure of torque. To avoid having to us the tube the disks could be mounted at each end of the shaft and the outputs of the two sensors logically ANDed together to generate the PWM signal. Using a low pass filter to convert the PWM to voltage may be a problem if the shaft speed was very low. Using a counter clocked with high speed pulses counting the number of pulses in high and low parts of the waveform may be a better solution.

Les.
 

Thread Starter

strantor

Joined Oct 3, 2010
5,532
My understanding is that one disk is fixed to the driven end of the shaft that twists due to the torque and the other disk is fixed to the driven end of the shaft. The tube just extends the position of one of the disks so it is close to the other disk so the light has to pass through the slot in both disks. when the disks are lined up (I.E No torque). As the shaft twists due to torque the percentage of time the light is allowed is reduced. If we assume that the slots are the same angular length as the space between them the the output wave form will have a 50% duty cycle. This will gradually decrease until the slot in one disk is fully covered by the gap between slots in the other disk. At this point the duty cycle would be 0%. If the torque increased above this level then light would start to get through again. You would have to ensure that the torque did not reach this level. So the duty cycle of the waveform would be a direct measure of torque. To avoid having to us the tube the disks could be mounted at each end of the shaft and the outputs of the two sensors logically ANDed together to generate the PWM signal. Using a low pass filter to convert the PWM to voltage may be a problem if the shaft speed was very low. Using a counter clocked with high speed pulses counting the number of pulses in high and low parts of the waveform may be a better solution.

Les.
The issue there is that the minimum duty cycle you could ever read is 50%. Half your measurement range is eaten up with dead space. When I was drawing the multi-tube model I had to find a medium amount of dead space that gives a good range of measurement without using too many disks. I probably still used too many disks.

I had another idea though this morning. The disks don't have to be identical. One of them could have a slot corresponding to 90% duty cycle and the other a slot corresponding 10%. So the output wave latches high when the 90% wave goes high, and unlatches when the 10% wave goes low. I think that would achieve the exact same objective as the multi-tube model but without all the tubes. But I only slept 3hrs so that might be stupid. I'll probably come back later and contradict myself again.
 

Sensacell

Joined Jun 19, 2012
2,774
With a microcontroller, it's really easy to measure timing with nanosecond precision.

You need to have overlapping flags that create 2 edges that can be tracked. The shaft speed must be calculated from the total period of the cycle, the phase shift (torque) can be measured by the timing variation between edges. It's best to have this idea repeat several times within one rotation, you need to be able to handle speed ripple within a single revolution. (typical of piston engines) I still don't see how all those tubes help in any way? just overlap the flags to avoid ambiguous time signatures, and limit the travel for overload conditions, so the flags never move too far.

For example- The leading edge is your reference edge, you time the RPM leading-edge-to-leading-edge. The falling edge is variable- (it's the moving edge) the time between rising and falling is your torque measurement. You calculate the phase using both measurements. If this setup had 8 teeth, you can get a measure every 45 degrees.
 

LesJones

Joined Jan 8, 2017
3,127
I was thinking of doing the timing over the period of one segment of the encoder.
190321.jpg

I was thinking of using a 16 bit counter with a capture register. The counter would be clocked at a suitable frequency.
This is the sequence of events.
Zero the timer
Wait for the leading end of the pulse for the driven end encoder. (Point A) and start the counter.
Wait for the leading edge of the load end encoder and capture the value of the counter. (Point B)
Wait for the trailing edge of the driven end encoder and capture the value of the counter. (Point C)

The count at point C is the length of the driven end encoder high state. ( This will be known in terms of angle.) So we know how many clock pulses represent this angle at the current rotational speed
The count at point B represents the angle of twist in the shaft.
So if we divide the number of pulses captured at point C by the number captured at point B we can calculate the angle of twist in the shaft.
The system would need to be set up so the leading edges at point A and point B line up with no twist in the shaft. This could be done with a dual channel scope with the shaft rotating and adjusting the position of one of the sensors.
Some code could be added to the microcontroller to display the difference in timing of these two points in setup mode.
The shaft would still have to be calibrated by measuring the angle of twist for a known torque applied with a torque wrench.

I used a very similar system for a tachometer I designed a few years ago. The software description on this web page gives you an idea of how it works.

Les.
 
Last edited:

Thread Starter

strantor

Joined Oct 3, 2010
5,532
Ok @LesJones & @Sensacell I'm sold on it. I'm going to need to get a new PTO shaft for this anyway. Now I just have to decide if I still want to try the belt force measurement thingamajig as well. I have a burning desire to know if that would work. I can't stand being this confounded by something. I would like to do both and compare them, but I... I probably... I don't know. We'll wee.
 
Last edited:
Top