Measuring torsion in shaft

[tmoritz]

I need to measure the torque applied to the shaft of a 42" butterfly valve. Head against the valve is around 1000 ft. The valve was adjusted to eliminate leakage. I am concerned we may be torsionally overloading the valve shaft and so I would like to measure the torsion in the shaft as we close and seat the valve and again when we open the valve.

My plan is to use strain gauges glued to the shaft between the actuator and the packing gland. My experience with strain gauges consists of one lab back in school fifteen years ago. From what I recall I need to mount four strain gauges 45° and 135° off axis, two on each side of the shaft. The gauges should be wired into a full bridge configuration. I need to use a closely regulated voltage source to supply the bridge and a precision amplifier/filter to condition the signal. I plan to run the signal into an IOtech Chartscan. I will also record the valve angle using a string transducer with the cable wrapped around the shaft.

My questions are:

1) do I need to address temperature compensation?

2) can I accomplish the same goal with a quarter or half-bridge?

3) do I need resistance wire and, if so, what for?

Thank you in advance.

Tom Moritz
Mechanical Engineer
US Bureau of Reclamation

 

[BobM3]

1) If you use 4 active gages (2 in compression and 2 in tension) then, as long as each gage sees the same temperature, you should not need temperature compensation.

2) Four gages will give you 4x the sensitivity as 1 gage as well as limit temperature effects.

3) You don't need special wire. A 4 wire cable consisting of 2 sets of twisted pair, shielded wires is often used.

 

[israelkk]

Why not use a torque handle with gage such as used for torquing large bolts.

 

[IceStationZebra]

1) agree, shouldn't be needed as long as all gages are the same temp.
2) You can use 1/4 or 1/2 bridge, but if you have the room go with the full bridge. The full bridge has the benefit of canceling bending and axial loads.

http://www.vishay.com/brands/measurements_group/guide/ta/ftm/ftmg.htm

3) just make sure that all four leads are the same resistance (i.e. same type and size of wire, same length, same temp) And don't make the lead wires super long.

The two sets of gages need to be 180° from each other on the shaft to properly cancel bending loads. To make the job easier you can get gages made for this that have pairs preset at 90°, though I have used standard gages several times. If you are short on time definitely have spare gages - you will almost certainly destroy a couple getting back on this bicycle! You can also buy preleaded gages, but they are about 2-3x the cost.

http://www.vishay.com/strain-gages/gages-sensors/shear-torque/

The standard adhesive used is cyanoacrylate (super glue, Black Max works fine). The trick is not to polish the shaft too much, the glue needs a little roughness to grab on to. The final polish should be with 300-400 grit emery paper. I do lot of work on cold rolled steel so I tend toward 320, but if the shaft is more porous I would lean towards 400 grit. I roll my finger over the gage to help squeeze out any excess glue - laying your finger flat tends to trap the excess glue which will give you a weak joint. And once it sets for a minute pull the tape off at a 45° angle to the gage to help from lifting the edge of the gage.

I don't follow this to the letter, but it will work for a couple of gages.

http://www.vishay.com/docs/11127/11127_b1.pdf

ISZ

 

[rb1957]

i still think that a paint line would be sufficient to measure the twist on the shaft, which is directly proportional to torque.

 

[tmoritz]

BobM3 and IceStationZebra - Thanks for your input. This helps me feel more comfortable with the process.

Tom Moritz
Mechanical Engineer
US Bureau of Reclamation

 

[tmoritz]

israelkk,

Good suggestion but there are two issues that I didn't include in my original posting. Firts, the mechanical operator produces torque in excess of 80,000 lb-ft. I don't want to try to counter that with a mechanical torque gage. Second, we are concerned about this valve because it emits a loud bang when opening. I suspect we are over torqing the valve on closure and the seal/seat are seizing. I need to have torque vs. valve position data. Hard to do that with a mechanical torque gage.



Tom Moritz
Mechanical Engineer
US Bureau of Reclamation

 

[tmoritz]

rb1957,

We tried marking the shaft at two elevations and measuring the angular deflection. My calculations based on that data indicate the torque is an order of magnitude (10X) the manufacturers documented maximum torque value for the valve. I haven't looked at the gear box yet to determine its maximum torque.

I suspect measuring the angular deflection method lacks accuracy and resolution resulting in a large error. Also, as I mention above, I would like to relate shaft torque to valve position and that's best done with strain gages and some sort of shaft encoding.

Thanks,



Tom Moritz
Mechanical Engineer
US Bureau of Reclamation

 

[rb1957]

thx for the feed-back ... i'd be interested to see if the strain guages support your proposition (i'd've thought that you'd be able to measure the angle of twist very accurately) the problem i see with s/guages is that they measure the surface strains and you have to project these into the solid section. if your section is carrying a significant amount of shear, this'll upset the measurement of shear stress ... unless you measure the shear stress at (say) 3 or 4 locations around the perimeter.

don't like the sound of your "big bang" !

i guess it is possible to exceed the manufacturers allowable torque by a factor of 10 ... maybe they're limited by fatigue ? ... maybe you have a very large safety factor ??

good luck

 

[rob768]

We perform strain gauge torque measurements on a more or less daily basis (as part of shaft power measurements) and it is a very accurate and reliable method. Accuracy within a few percent, depending on correct positioning of the strain gauges, and knowledge of G (which varies a little depending on used material). One set of strain gauges suffices, just make sure the measurement is influenced by bending (if at all possible)