PDA

View Full Version : Balancing a grinder with accelerometers



Erik Brewster
08-01-2012, 10:52 PM
I have been trying to sharpen drills on my bench grinder lately and have been irritated by the drill chattering all over the place because of the vibration due to out of balance. I remember having fun using the balancing rig in my vibrations class in college, so I went to reproduce it.

There are great tutorials and products for doing static balance:
- Oneway grinder balancer
- Various homebrew equivalents (or better) like this one (http://applescottysscrapbook.blogspot.com/2008/06/grinding-wheel-balancer.html)

I did it with accelerometers because:
1 - It sounds fun
2 - I would like to do this with a microcontroller mounted on the grinder at some point (not today), so it is a simple process, like a tire balancer
3 - From a theoretical perspective, it should be more accurate than static balance because there is no error of mounting the wheel after balancing because it is balanced in place
4 - I will learn something

Let's be clear -- #1 above is the real reason I am doing this :) #4 is also a nice fringe benefit.

The setup:
- One of my grinders is a on old Craftsman grinder I like that I inherited from my uncle. It's dear to my heart, hence no Baldor.
- I'm using a grinding wheel on one side and a Scotchbrite wheel on the other side for deburring.
- An accelerometer on the back of the grinder to read acceleration. I used this accelerometer (https://www.sparkfun.com/products/9652)
- I made an aluminum flange to mount the balance weights and also to sense the 0 degree position of the grinding wheel. I used this sensor (https://www.sparkfun.com/products/9454).
- Hot glue gun
- Fasteners in the shop for balance weights
- Digital scale from the kitchen to measure balance weights
- Oscilloscope to read signals
- Excel spreadsheet I made to calculate balance weights
- Towels to put under the grinder to keep it from skittering around on the workbench
- Netduino - used only for supplying power to the sensors right now

http://i46.tinypic.com/dyp1fn.jpg
Overall setup

http://i46.tinypic.com/rvatyc.jpg
Accelerometer mounted on back of grinder

http://i47.tinypic.com/30cq7u1.jpg
Balancing flange and index sensor. The index sensor is hot glued on the end of the block of aluminum for prototyping. The blue tape is to minimize the c-clamp handle from clanging around. It was surprisingly clear to see on the measurement.

Erik Brewster
08-01-2012, 11:06 PM
The general procedure is to:
- Take a baseline vibration measurement
- Add a trial weight and measure the resulting vibration
- Calculate the effect of the trial weight and then calculate the needed weight to eliminate the vibration

This is a simple, single plane balancing scheme and I think it should be fine for bench grinder use.

http://i46.tinypic.com/30jn14n.jpg
The picture above shows a yellow trace of the accelerometer output and a blue trace of the index pulse (sharpie mark on the flange, read by the optical sensor).
I get a 360 mV reading at 5.8 msec (117 deg)

http://i48.tinypic.com/2afbl04.jpg
Then I added a trial weight at 0 degrees and got an amplitude of 232 mV at 5 msec (101 deg).

I calculate the needed weight to be 1.24 units of mass relative to the first mass at -36 deg. Here is the result.
http://i46.tinypic.com/30saxvk.jpg
The resulting vibration is in the noise. I think I can see a vibration in there at 10 mV. In any case, this is 2.8% of the original value, which I think is pretty good.

Too bad these numbers are a bit optimistic...

ed_h
08-01-2012, 11:11 PM
Erik--

So how did the grinder's sound or feel change?

Also, which axis were you reading?

Erik Brewster
08-01-2012, 11:17 PM
The biggest problem I ran into is that there was a big 2 times per revolution vibration. A pure out of balance condition should be a once per revolution sine wave. Here is a nice example from before I balanced the grinder:

http://i46.tinypic.com/30jn14n.jpg
Note that there is one sine period per index pulse. Perfect. The problem is that I took this measurement with the motor off. I ran the motor up, turned the switch off and immediately took the measurement. The motor is basically going full speed at this point.

http://i46.tinypic.com/20zptvm.jpg
Here is a picture of the measurement right before I turned off the motor. Note that there is a significant 2 times per revolution vibration. It goes away completely the instant I turn off the power. It goes from a pretty good balance with the power on to absolutely spookily silent and smooth as soon as the power is off.

Any ideas? I'm thinking I'll take the motor apart to find the problem, but honestly, I don't know what I'd be looking for.

ed_h
08-01-2012, 11:30 PM
Erik--

It might be a resonance. Try damping different part of the machine by grabbing them tightly.

Erik Brewster
08-01-2012, 11:30 PM
Erik--So how did the grinder's sound or feel change?


Like I said in a following post, it's super smooth and silent now that it's balanced. It had a pretty noticeable vibration before. Enough that it was tough to grind with great precision. The Scotchbrite wheel was the big offender, though both had some out of balance. All the pictures here have the Scotchbrite wheel balanced and I am focusing only on the grinding wheel. Now that it is smooth, it is mostly a whooshing of air sound, when the motor is off.


which axis were you reading?
I am reading the vertical axis. For some reason, the axis front to back (z on my accelerometer) had strangely low gain that made it tough to measure. I understand that the up and down axis (X axis on my board) can pick up a torque vibration. I suppose I should go and remount the board to use the X axis in the front to back configuration to avoid picking up any torque vibration.

Erik Brewster
08-01-2012, 11:40 PM
It might be a resonance. Try damping different part of the machine by grabbing them tightly.

Interesting point. One thing I forgot to mention is that is clearly not a second harmonic of the out of balance -- it is just slightly a different higher frequency than twice the rotor imbalance. The rotor imbalance locks exactly to the index trigger. The "twice rotor speed" is just a bit faster than twice the rotor speed. I take that to mean that it is some vibration from the electrical lines, which should be a bit faster than the rotor speed, due to induction motor slip.

I did go and grab the machine. If I held it in the air, the twice per rev vibration reduced about 30%.If I pressed down, the vibration increased until I pushed with 30 lbs.-ish, where it started reducing. That seems to make some sort of intuitive sense.

ed_h
08-01-2012, 11:44 PM
If I'm reading your screen right, you have about a 120 Hz signal. Certainly could be a torque reaction. May not be much you can do. Add mass, maybe?

Erik Brewster
08-01-2012, 11:55 PM
To make reading the plots easier, the blue index pulse is at 58.something Hz. That's 3500 RPM or so. I remounted the accelerometer so that it would read front to back through the axis of the shaft. This should remove the torque couple. It read about like the vertical axis when I held it in the air, so that seems consistent.

J. R. Williams
08-02-2012, 12:15 AM
How does the unit run without the grinding wheels? I have almost the same grinder and had a balance problem and it was a cheap wheel. Install one wheel and run the unit and determine which wheel is the problem. Interesting project.

dp
08-02-2012, 12:15 AM
To make reading the plots easier, the blue index pulse is at 58.something Hz. That's 3500 RPM or so. I remounted the accelerometer so that it would read front to back through the axis of the shaft. This should remove the torque couple. It read about like the vertical axis when I held it in the air, so that seems consistent.

How many axes are you sensing? What does it look like on other planes?

interrupted_cut
08-02-2012, 12:48 AM
2/rev on motors is often "oil canning" of the stator from the rotating magnetic field, as you get a peak from both the positive and negative phase of the current waveform. That would explain why it goes away during coasting.

Davis

Erik Brewster
08-02-2012, 01:45 AM
2/rev on motors is often "oil canning" of the stator from the rotating magnetic field, as you get a peak from both the positive and negative phase of the current waveform. That would explain why it goes away during coasting.
Davis
I'm not familiar with "oil canning." Is that runout of the rotor? Bending of the shaft under speed? Other? It seems like some things could be fixed...

Erik Brewster
08-02-2012, 01:50 AM
I found a B&K paper on balancing that seems really useful:
http://www.bksv.com/doc/bo0269.pdf

Paul Alciatore
08-02-2012, 03:02 AM
Before you assume that your 120 Hz signal is an indication from the accelerometer, consider this:

Your sensing circuit is not shielded. That is an ELECTRIC motor that operates by generating an electromagnetic field and your sensor circuit board is certainly inside of the outskirts of that field. It could be that it is just picking up that field.

Try taking the sensor board off of the grinder but HOLDING it in roughly the same position WITHOUT TOUCHING THE GRINDER. Turn the motor on and see if that 120 Hz signal is still present. If it is, it is not from the accelerometer, but is instead just picking up the field in the air. Notice I said "HOLDING" it. Keep it isolated from the grinder and from the bench.

I have often noticed that an oscilloscope probe can be used to detect AC fields just by connecting the end of the ground lead to the tip of the probe and using the supposedly grounded loop formed in this manner to find a field. Certainly your circuit can do it also.

philbur
08-02-2012, 03:41 AM
Don't single phase motors produce a pulsed power cycle. It's not continuous X Kw but varies for 0 to a some value higher than the RMS a number of times per rev depending on the number of poles. 3 phase is much smoother because the phases overlap.

Just a thought from a mechanical engineer.

"Nonsense" would be sufficient reply from a qualified electrical engineer.

Phil:)

Erik Brewster
08-02-2012, 03:51 AM
Try taking the sensor board off of the grinder but HOLDING it in roughly the same position WITHOUT TOUCHING THE GRINDER. Turn the motor on and see if that 120 Hz signal is still present.

Interesting point. I made sure that my power circuit for the sensors was grounded to the grinder and the scope grounds on the probes were connected to the same ground. Ground is through the ground line on the index sensor. That is connected to the aluminum block I clamp to the grinder for index sensor mounting.

I then did your test: dismount accelerometer (was hot glued before, so no ground through the sensor previously). When I touch the sensor to the grinder frame lightly via the hot glue, I get a clear 120 Hz signal. If I don't touch it and hold it ~ 1/8" away, no 120 Hz signal, just a typical no vibration noise background.

Good idea and thanks. It's good to check these things.

Erik Brewster
08-02-2012, 03:57 AM
How many axes are you sensing? What does it look like on other planes?

I checked all three axes with no grinding wheels, washers, nuts, or balancing flanges mounted. All phase information is relative to X axis. I did not make any attempt to reference any phase information to an absolute reference. All data is 120 Hz dominant. Accelerometer was not mounted in the same orientation to the previous measurements. I saw no 90 degree relationships.

X (back front) axis: 40 mV 0 deg
Y (left right) axis: 60 mV 180 deg
Z (up down) axis: 100 mV 180 deg

Erik Brewster
08-02-2012, 04:00 AM
I also took the rotor out and trued it up. I used the bearing outer races to indicate it in. I supported the shaft on one side in a four jaw on the grinder mounting surface. On the other, I supported it by the ball bearing. I took a few thousandths of an inch off the rotor to true it up. It was out a few thousandths (~0.003") before. I could not notice any significant change in the 120 Hz vibration after I did this. I was able to get the bearings and rotor to spin true within a few tenths.

I also noted that the 120 Hz vibration is locked to the AC line frequency, not the rotor frequency. I triggered the scope on the AC line and it stayed precisely locked into the vibration.

alanganes
08-02-2012, 06:22 AM
This is a totally uninformed comment, so I may be way off base, but is it possible that the tool rest that the accelerometer is mounted on is flexing and introducing (or mechanically amplifying) some signal that is either small or not otherwise there?

This is a very cool experiment, thanks for posting this. I have thought of doing exactly this, I even got as far as making an accelerometer out of piezo buzzer, (there were not cheap, easily available MEMS accelerometers at that point)but never actually got around to trying it all out. I may need to dig that out and try it now.

Please do keep us posted. This is the essence of the home shop guy, I love threads like this!

J Tiers
08-02-2012, 07:56 AM
The biggest problem I ran into is that there was a big 2 times per revolution vibration. A pure out of balance condition should be a o
Here is a picture of the measurement right before I turned off the motor. Note that there is a significant 2 times per revolution vibration. It goes away completely the instant I turn off the power. It goes from a pretty good balance with the power on to absolutely spookily silent and smooth as soon as the power is off.

Any ideas? I'm thinking I'll take the motor apart to find the problem, but honestly, I don't know what I'd be looking for.

As Phil Burman stated, there is a 120 Hz torque ripple with 60 Hz single phase.... and the motor likely has other movements as the magnetic field goes to max and min, so there absolutely SHOULD be vibration at that frequency. I would imagine that your hand on the grinder would detect that also, with no suspicion of electromagnetic interference.

of course, interference is a possibility. Those cutaway grinder motors do have a weird magnetic path which is not so well enclosed as with a round motor.

You can determine which it is by taking off the sensor, and holding it in the same position just barely not touching the motor or table. If it is electromagnetic, you should see the same signal, a bit attenuated.

Rosco-P
08-02-2012, 09:15 AM
How does the unit run without the grinding wheels? I have almost the same grinder and had a balance problem and it was a cheap wheel. Install one wheel and run the unit and determine which wheel is the problem. Interesting project.

Wouldn't the first step be to balance the rotor and check that the bearing in the grinder haven't gone T.U.? Then compensate for any imbalance added by the grinding wheel, Scotchbrite wheel, wheel flanges, etc.? Did you check the motor shaft for straightness? A slight bend in one end of the unsupported shaft would introduce vibration as well.

Thruthefence
08-02-2012, 10:06 AM
I did a similar balancing project on my grinder using this:

Chadwick-Helmuth 192A / Strobex

http://i1204.photobucket.com/albums/bb405/thruthefence/Untitled.jpg

Old school, but effective. I found that by trying to balance the grinding wheel AND the wire brush together, I was unable to make the balance moves work as the "clock" angle would expect. Ended up removing one wheel, & balancing the other, one at a time. I made a disc from sheet metal, drilled #10 holes at 15 degree intervals at the outer edge. This was installed coaxially with the working wheel to accept lead rivets as balancing weights, and as a reference for reflective tape. I mounted the velometer to the wheel guard, the axis of which intersected the rotating shaft axis. The Plotter locks into the problem RPM, hit it with the strobe, and note the position of the target. Shut everything down, move the "target" to where the strobex "stopped" it, and install weight OPPOSITE the velometer (in this case, 6:00 o'clock). Once it dawned on me about the interference between the two rotating wheels, It fell right in, I ended up with .09 IPS on the grinder, .2 IPS on the wire brush. I used the "this looks right" technique" to choose weight, but after a couple of tries, I could estimate pretty well how much mass, equaled "this" much movement on the chart. In my case, I noted no 120 hz spikes in the plot, but the cabling is shielded. The "balancing disc" is just pinched between the wheel & the shaft, so I suppose it could shift at some time, but this was just a make work training exercise anyway.

lazlo
08-02-2012, 10:09 AM
Very impressive work guys!

Erik Brewster
08-02-2012, 11:56 AM
This is a totally uninformed comment, so I may be way off base, but is it possible that the tool rest that the accelerometer is mounted on is flexing and introducing (or mechanically amplifying) some signal that is either small or not otherwise there?
The tool rest is holding the index mark sensor. The accelerometer is hot glued to the grinder bod, so there is little chance of significant resonance in the accelerometer mounting. In any case, I have pushed and prodded the accelerometer while hot glued and not seen any noticeable changes. Since the accelerometer has a very small mass, if I were to touch it with my finger, I would expect to see any resonance behavior change, as my finger is many times the mass of the accelerometer.

Thanks for the kind words.

Erik Brewster
08-02-2012, 12:00 PM
...Old school, but effective...

That's a great project! I figure modern technology just makes things more convenient. Balancing has been around a long time and it's great you conquered it.

Evan
08-02-2012, 12:10 PM
This is a great project. Single phase torque ripple is approximately zero at no load. That isn't it. The power on vibration is caused by the magnetostriction effect in the armature pole pieces. That is the same effect that makes a transformer hum. The magnetic materials actually change shape under the influence of the magnetic field.

I have been considering doing this too so now all I need is an accelerometer.

Erik Brewster
08-02-2012, 12:18 PM
From taking the motor apart to true up the rotor, I realized that the alignment of all the part is not well guaranteed. It's mostly a bolt it up and hope for good alignment situation. Specifically, I am thinking (without proof) that the magnetic gap is conconsistent because the whole motor isn't assembled perfectly. It seems to me that this would be consistent with "oil canning" suggested by interrupted_cut. It would also be consistent with "Misalignment (parallel)" in the B&K paper I linked to.

I'll see what I can do to change the alignment and detect a change in vibration. If I can cause a change, then I can reduce the problem.

Erik Brewster
08-02-2012, 12:52 PM
I found a Siemens paper on motor vibration and one of the 2x line frequency causes is "Non Symetrical Air Gap". I think it is worth chasing a bit later today.
http://www.industry.usa.siemens.com/drives/us/en/electric-motor/anema-motors/specification/Documents/approach-to-solving-motor-vibration-prob.pdf

Paul Alciatore
08-02-2012, 04:00 PM
Well, Evan beat me to it. It is almost totally impossible to prevent any equipment that has steel in it and that uses a fair amount of AC power from humming. Many people think it is a 60 Hz hum, but it is most often twice that or 120 Hz.

In addition to the magnetostriction effect that Evan mentions, other effects are also possible, including simple attraction of a less than totally rigid steel part in the field. This is often the reason why the boxes that are used for power distribution will hum. No current passes through the steel sides of the box, but the heavy currents in the conductors inside them will cause them to vibrate.

I don't think you have any practical ways of preventing this in your grinder, but I may be wrong so do try. You can't eliminate the fields except by turning power off. Tighten everything you can. Add dampening weights if you can. Six inches of well adhered concrete all around the motor may help. Or a jacket of soft lead shot. Etc.

Of course, the torque ripple may also be at work here, but that should have only a minimum effect on grinding if the wheels are balanced. This is because it is a torque or rotational force and not a radial or axial force. If you could somehow mount your sensor on the rotational axis, you may not see it. Yes, I know that is hard to do.

Paul A


This is a great project. Single phase torque ripple is approximately zero at no load. That isn't it. The power on vibration is caused by the magnetostriction effect in the armature pole pieces. That is the same effect that makes a transformer hum. The magnetic materials actually change shape under the influence of the magnetic field.

I have been considering doing this too so now all I need is an accelerometer.

interrupted_cut
08-02-2012, 06:39 PM
"Oil Canning" is the stator taking an oval shape under the influence of the magnetic field. It is similar to magnetostrictive effects, but is more macroscopic in nature. Cheaper motors with less robust structure will likely be worse. I forgot to congratulate you on the clever setup. I've gotten spoiled at work, and now I can't measure and analyze vibration without a $25,000 16 channel analyzer. The reality is, it doesn't work much better than your setup..it just prints out better pictures! You have demonstrated a simple and elegant solution to the problem.

J Tiers
08-02-2012, 09:45 PM
This is a great project. Single phase torque ripple is approximately zero at no load. That isn't it. The power on vibration is caused by the magnetostriction effect in the armature pole pieces. That is the same effect that makes a transformer hum. The magnetic materials actually change shape under the influence of the magnetic field.

I have been considering doing this too so now all I need is an accelerometer.

Absolutely. The basic magnetizing current is MOST at no load..... so *field-induced* effects are MOST at no load. The difference vs max load is small, but still there.

The weaker flat-sided motor structure common to many grinders makes field-induced vibrations worse than they might be in a normal round motor. So magnetostriction may not be the dominant cause of vibrations.

As for torque ripple, there is least torque at no load, indeed, but against that is the fact that in a grinder, there is much more rotor inertia than usual (large grinding wheels), and likely more "windage" as well, since the wheels usually move some air. Obviously there is SOME torque, and whatever is there operates against a more 'solid" counterpoise with the more massive rotor structure including the grinding wheels.

Anyone doubting the inertia of the wheels can time the acceleration and spin-down with, and without the wheels..... Mine take a minute or so, the father-in-law's larger one takes nearly 3 to come to a stop. And they surely do not "snap" to full speed, either.



As Phil Burman stated, there is a 120 Hz torque ripple with 60 Hz single phase.... and the motor likely has other movements as the magnetic field goes to max and min, so there absolutely SHOULD be vibration at that frequency.

Mcgyver
08-02-2012, 09:53 PM
How does the unit run without the grinding wheels? I have almost the same grinder and had a balance problem and it was a cheap wheel. Install one wheel and run the unit and determine which wheel is the problem. Interesting project.

It is an interesting project in an area that has interested me for awhile. Those photos in one of your links early on of the static balancing were mine; I was able to make an improvement, but really without dynamic balancing (one of my 500 year long list of projects), it is limited.

I agree with JR, balance the rotor as best you can. Then balance the wheel by abrading a little off with a dressing stick- using it very gently sort of like a file to remove a bit of material from the wheel itself. Reason is every time you change wheels or dress you will need to rebalance ...but since it is pretty much a disk you could do a bang up job with one accelerometer (with the rotor first dynamically balanced). This is why on the surface grinder & T&CG you can leave the wheel mounted and swap the hub.

imo JR's is a revealing and interesting question. if for example the rotor was in good balance with no wheels, it's not as challenging a situation; the out of balance part (wheel if rotor is balanced) is essentially a disk and could be well balanced without dynamic balancing. The more challenging situation as we all know is the off centre CG's along the shaft creating an axis of inertia not the same as the axis of rotation....requiring dynamic balancing.

I think one could add a second accelerometer and you would have a dynamic balancer. There'd still be a bunch of math to do or putty trial and error, but you would know the relative magnitudes of the two, and at what angle they were

very neat stuff, thanks for posting it

Erik Brewster
08-02-2012, 10:41 PM
Some further info:
- The imbalance with no wheels is very small at 1/rev. The 2/rev imbalance is about the same. I'll take some proper data tonight or tomorrow.
- I bought the FFT add on for my scope, so expect to see some gratuitous FFT shots from now on. I doubt they will make a difference to the end result :P
- I took the grinder apart again to chase the stator alignment, but found that, without major surgery, it is not adjustable and was more positive in assembly location than I recalled.
- I have another 6" grinder of the cheapo variety that I'll do the same routine on at some point and will post the results.
- I couldn't help myself and bought a 6" Baldor on eBay, so I'll measure that one, too, when it gets in

Thanks for all the input!

ed_h
08-02-2012, 10:45 PM
I read in a paper recently about vibration sources in single phase motors. Some stem from magnetic sources (both magnetostrictive and "Maxwell"), and have a frequency related to 2 x line frequency. Other sources of vibration are mechanical and have frequencies related to the rotational speed. Since in a 2-pole induction motor, the rotation frequency differs by some (usually) small amount from the line frequency, there is a low frequency beat between the two. This is why the sound and/or vibration of an induction motor sometimes seems to be amplitude modulated at a low frequency that varies with load (since the slip varies with load).

Maybe everyone else in the world knew that already, but it was an Aha! for me.

Erik Brewster
08-03-2012, 01:44 AM
I read in a paper recently about vibration sources in single phase motors. Some stem from magnetic sources (both magnetostrictive and "Maxwell"), and have a frequency related to 2 x line frequency. Other sources of vibration are mechanical and have frequencies related to the rotational speed. Since in a 2-pole induction motor, the rotation frequency differs by some (usually) small amount from the line frequency, there is a low frequency beat between the two. This is why the sound and/or vibration of an induction motor sometimes seems to be amplitude modulated at a low frequency that varies with load (since the slip varies with load).

Maybe everyone else in the world knew that already, but it was an Aha! for me.
This has become very clear for me doing all the measurements here. With the triggering on the rotor and also on the AC line, it's really easy to see that a signal is locked in or not.

Evan
08-03-2012, 01:54 AM
For some reason, the axis front to back (z on my accelerometer) had strangely low gain that made it tough to measure.

I looked up the specs on a couple of 3 axis accelerometers and they both have a different spec for the Z axis than for the X and Y axis. It is very likely due to the physical geometry of the MEMs device.