Homemade dynamic balancer instrument

[JMarkWolf]

I plan to develope a simple vibration analyzer for a specific experimental helicopters rotors using a small dedicated micro-controller to capture accelerometer and phase angle data for a full revolution. I think I can perform a simple numerical integration, via Newtons method or similar, to generate velocity data at a given phase angle that will correspond to published polar charts describing corrective action.

The variety of MEMs accelerometers available seem to make the acquisition job fairly straight forward.

My background is electronics which makes the micro-controller part easy for me, but I have little idea what to expect as far as mechanical gotcha's. Can anyone on the forum offer re-assurances or opinions of likelyhood of success following the described approach?

 

[CESSNA1]

JMARKWOLF:

I think you are taking on an unecessarily complex (forget simple) project for what can be a fairly easy task. Model helicopter rotors can be statically balanced with the equipment available to modellers through the model aviation magazines. Dynamic balancing is really only used for rotating machinery that has some length like generators, turbine rotors, etc. The "dynamic" balance at tire shops is really a static balance done using centrifugal force. The integration may not be simple, noise is an ever present concern

There are a lot of complexities involved in your project such as: aliasing/noise/filtering/and others.

Hope this helps
Regards
Dave

 

[JMarkWolf]

Thanks Cessna1

But I'm not talking RC model helicopters. I own a 2-seat experimental class living breathing helicopter, for which I've borrowed handheld dynamic balancers, at least one of which has nothing more than a $5 PIC processor in it, and a $2200 price tag. I design small computers for a living so I have a leg up on this aspect of the project.

Static balancing as you describe is indeed part of the process but dynamic balancing is required for truly smooth in-flight operation.

There only two frequencies of interest, 8.7hz and 43hz. Everything above 50hz is attenuated by the accelerometer module I've selected, so I'm hoping noise and aliasing will not be an easily insurmountable issue.

But I'm soliciting input here so keep it coming.

 

[GregLocock]

Cassna1

No, a modern wheel balancing machine is a true dynamic balancer, that is, it specifies weights to be added at two planes, the inside and the outside rim.

JMarkWolf

Your problems are associated with getting good physical data, once you have the waveform captured the rest is just maths. I really doubt that a cheapo accelerometer will be much good at low accelerations. What sort of accelerations do you see, typically?

Cheers

Greg Locock

 

[JMarkWolf]

Cassna1

The polar charts I alluded to, in my first post, describe the corrective action for a given magnitude and phase angle of measured imbalance. The charts are available for various aircraft, mine included.

Once you have the captured the acceleration data and phase angle successfully, and integrated for velocity, you need only to "plot the dot" on the polar chart to determine the corrective action.

The typical corrective action for helicopter blades is to add/remove weight to the end of one blade, and/or adjust the angle of attack of the blade by a couple different means.

The accelerometer I plan to test first generates 1200mV / g. The expected g's will be approximately 0.3 g's with a high degree of imbalance, with 0.03 g's with a low degree of imbalance. I've included a programmable gain stage immediately after the accelerometer for when the degree of imbalance diminishes and I need bigger signal deflection.

 

[GregLocock]

You'll find that picking a 0.3g signal on a running vehicle is very difficult. I suggest you use a tone wheel as your sampling clock so that you can accurately pick up the relevant orders and get a good phase measurement.


Cheers

Greg Locock

 

[JMarkWolf]

Greg

Please advise what a tone wheel is.

The sample clock I was planning to use is based on the microprcocessor frequency clock. Taking one acceleration measurement for every degree of revolution of the rotor.

Starting and stopping the acquisition when the master rotor blade is at zero degrees phase angle (directly out front), triggered by a Hall effect sensor.

I can take multiple revolutions worth of data until I get a repeatable and believable data set, the integrate, analyze and average during the "off cycle" of rotor revolution then display the result in IPS (inches per second) @ phase angle.

"Plotting the dot" on a published paper polar chart with concentric rings corresponding to IPS, and radial lines corresponding to angle of rotation of the main rotor yields the corrective action.

I agree that digging a signal out of the mud might get a little tricky, but common accelerometers provide only 100mV / g of signal. The accelerometer I hope to use, the Motorola MM2260D, provides 1200mV / g of signal.

 

[GregLocock]

A tone wheel generates a clock pulse based on rotor position, which is what you propose.

Averaging in the time domain using a phase locked clock is a very powerful way of improving S/N, superior to almost any frequency domain technique in practice.

Sounds like you are on the right track, but I think your accelerometer may overload, depending on where you mount it.




Cheers

Greg Locock

 

[magnograil]

Your accelerometer manufacturer should provide a frequency-phase plot with it. As you get near the natural frequency there will be a phase shift in the output which you need to compensate for. Another effect you might consider is the tail boom blocks the downwash which will probably show up as a cyclic load variation on the rotor.
The attenuation has a rolloff not an infinite slope. With 50 Hz being so close to your desired measuring frequency (43 Hz) you will probably see a much lower signal that actual.

 

[GregLocock]

Yes, assuming that's an anti aliasing filter you'd be better off with 64 or 100 Hz.

Take a look at the unfiltered signal, it is very unlikely that acheap accelerometr has any effective cutoff, and it will overload due to signals outside teh frequency range of interest. This will ruin your measurements. It is very hard to detect.



Cheers

Greg Locock