There are a few known difficulties when measuring free body torsional natural frequency, which include means of torsional excitation and need for strain gage attached to the shaft in order to measure oscillations.

How about simplifying the test? Assume a system consisted of: motor rotor - flexible coupling - shaft - fan wheel.

Attach accelerometer to the wheel tangentially. Impact the wheel also tangentially.

Any suggestions?

Thanks,
David

This message has been edited. Last edited by: David_G,

Dave,
Yes, for a down and dirty check, if you mount the accel tangentially on a fan blade (as close as possible to the backplate of the wheel, and then impact the other end of the fan wheel, tangentially with a VERY soft hammer tip, you should be able to excite the torsional mode.
Now, separating out any lateral modes that were also excited is going to be difficult.
There really isn't any better method than using a torsional encoder on the end of the shaft, or strain gages.
Also, a 2D FEA model is excellent for identifying these modes and also helping separate out the lateral modes.
Good Luck!

Ron,

I did consider the potential problem you mentioned with the proposed test as being able to pick up also lateral vibrations. To overcome it I thought of locking up the wheel in order to see blade natural frequency and doing a reqular impact test in order to see shaft/structure directional lateral resonance.

Likely, the torcional resonance frequency is going to be the smallest one.

Well we did that by remote on a pulp screw press assy in China measuring tangentially and hitting tangentially by the standard 2"x4" and we got a peak. You could test having the tangential position either in horizontal or vertical direction and see if there is a peak that is common. Since the device is not running it will not be perfect but still give some answers or more normal, more questions. Olov

David,

I have done this test a few times on fans and large turochargers. With accelerometer at 12:00 or 6:00 then hit at 3:00 or 9:00 clock positions. For an additional check then move accelerometer (or rotate rotor 90-degrees) to 3:00 or 9:00 and then hit at 12:00 or 6:00. Accelerometer and hammer impacts are all in tangential direction. If possible, repeat test with accelerometer at one end of the rotor and impact at opposite end. This technique minimizes exciting lateral vibration modes, since impact is 90-degrees from response and primarily only torsional modes would be excited. Hammer tip hardness has to be selected based on F-max and not necessarily the softest tip available.

Walt

Walt,
The main reason I use the softest tip is that the frequency of interest, especially on a large fan,coupling,motor will be low. I always emphasize it to keep people away from using a large plastic mallet, etc.
I like your approach on testing in multiple directions.
Ron

90 degree idea is a good one, Walt. Thanks to the rest of responders as well.

One more concern... The flexible coupling is part of the power train. For torsional vibration purposes it is just another element with its moment of inertia and stiffness. The coupling, such as of gear or grid type, is not preloaded during the test and is rather loose in tangential directrion.

How to overcome this if one is looking for the whole power train natural frequency? Or you have to do this test on both sides of the coupling, thus determining two separate resonances, any of which could be excited during operation ?

Sounds like good practical suggestions. I don't have anything practical to offer. Just some thoughts:

There were several different types of torsional resonances discussed above. The one discussed in the original post was two relatively rigid machines connected by flexible coupling. That's what I'll focus my comments below on.

=====================================================================

1 If you have two rotors A and B both torsionally rigid with flexible coupling between, then how about:
Put the accelerometer on rotor in torsional direction
Use a riggers strap with hoist with load cell to put a pull on B in the CCW direction.
Restrain A so it can rotate CW but not CCW. (i.e. place a stationary barrier to prevent further rotation of A)
Now there are two things you can do:

• 1.1 - Try to measure the torsional spring constant = change in angle across the coupling divided by torque. Change in angle might be gathered from a wire wrapped around a shaft and securely fastened to dial indicator plunger (use circumference formula to convert wire movement to angle rotation). Torque would be force on the hoist times radius. If the system is two torsionally rigid rotors/shafts connected by one flexible coupling, then the only other thing you need to compute the torsional resonant frequency is the rotating inertia on each side of the coupling - not too tough to compute. Easy to calculate for that simple scenario. But if the rotors on each side of coupling are not torsionally rigid, it becomes a whole lot tougher to estimate torsional stiffness associated with the fits etc.
  
  
• 1.2 - Very quickly release the pull on B by cutting the strap (with review of safety aspectss before you do it). It seems like this might possibly be more effective at getting a high magnitude torsional excitation than bump test.

===========================================================================

2 We already know that for estimaging a lateral resonant frequency we can use a "shaker".....how about tryinging to apply the same concept for torsional resonant frequency. Two product ideas come to mind:
2.1 - Build a torsional shaker that clamps to the shaft near one end of the machine train (for example outboard fan shaft on TEFC motor). It would have to develop variable frequency torque oscillations by accelerating an internal mass.
2.2 - Build a special power supply which will apply voltage to a motor in a manner that will create variable speed torque oscillations, but has zero average torque (does not cause the motor to start rotating). I'll bet it is not such a challenge to figure out a voltage pattern that would do this.

=====================================================================

quote:

One more concern... The flexible coupling is part of the power train. For torsional vibration purposes it is just another element with its moment of inertia and stiffness. The coupling, such as of gear or grid type, is not preloaded during the test and is rather loose in tangential directrion.

How to overcome this if one is looking for the whole power train natural frequency? Or you have to do this test on both sides of the coupling, thus determining two separate resonances, any of which could be excited during operation ?

That would be a non-linearity - i.e.coupling stiffness depends on load. The resonant frequency for a nonlinear system can change depending on the magnitude of the excitation. If you're lucky, you can generate high enough excitation during impacting etc to match operating conditions, but probably not. If you don't match the excitation magnitude on a non-linear systems, you can get resulting errors in estimating the "resonant frequency" that will be seen during operation.

This message has been edited. Last edited by: electricpete,

So two types of torsional resonances were mentioned.

1 . Blade resonance. I can imagine a blade impact test would be fairly easy since the blade inertia is so much smaller than the rotor... if only the blade (not the rotor) moves during impact you still have the correct mode shape. Don't need to impact hard to excite it.

2 - As I interpretted the original question and what I discussed above - two rigid rotors connected by flexible coupling. Seems like it is more challenging to test since we need to get both rotors moving torsionally.

So which is more common? I haven't seen either myself but I don't work on turbines etc. I have heard #1 talked about on the forum. I can't say I have heard of any instances of 2... has anyone?

EP,
The one a field tech is usually trying to solve is a torsional mode between the driver and the driven that may be due to the wrong coupling stiffness.
Ron

Attached is torsional impact test on the driven machine only. The coupling and driver mass-spring may bring these low frequency resonance numbers even lower.

Impacting and sensing directions were perpendicular.

Mixer_impact.doc (43 KB, 25 downloads)

David,

I guess that I don't understand what you are trying to measure. The torsional natural frequencies of a machine train would have to be tested with all components in place, no? How can you arbitrarily test without the motor coupled? How can you interpret yor data?

Walt

Walt,

Of course testing just partial machine train makes no sense, but currently I have no access to the unit in question and just practiced on a spare driven machine.

Unfortunately I did have an instrumented hammer and therefore can't validate the results by not having the phase. But when taking normal background vibration tose peaks were not present.

I was surprized to find such low resonant frequencies.

David

David,

Not much info to work with. If your mixer has large paddles, then it could be a blade/paddle bending resonance. If your mixer is loosely/flexibly supported, then it could be a resonant support structure. Most machines do not permit access for an adequate impact test for torsional vibrations. I suggest measuring torsional vibrations directly with machine operating.

Walt