OT partly: Motor balancing app challenge

Machine tool motors, hand power tools, anything with a rotating mass for that matter requires some minimum standard of balance to function properly. Power tools that leave your hand tingly slowly degrade your sense of touch. Machinery with dodgey motors telegraph vibration into cut surfaces, etc. Out of balance poses a real problem. The remedy is to balance the rotating parts but that requires a tear down, sending out the parts a day or three in limbo, then the assembly. PITA and in the meanwhile progress has halted for everything down stream of the equipment out of service.

We now have smart phones elaborately equipped with an array of sensosr including a accelerometer and a gyro. These are sensitive as hell, equal to the sensors on a balance machine. Load up a vibration app, set the phone on the table in a quiet room and the phone will register in the squiggly line fall of a green pea onto the table from an inch of height.

Quadracopter drone nuts balance their lift motors with a smart phone: There's some relevant videos on YouTube

What's needed is an app designed to meet the needs of on-machine balancing. Set the phone on the motor and the app determines the motor RPM from the cyclic nature of the vibration then goes on to quantify it and if the distance between bearing is entered diagnose the accelerations and phase relationship for each Add a known trial weight like the hub and enter its radius and location from a particular bearing and it will tell you the amount, radius, and, angle for correction for each re-position presuming 90 degree intervals.

Take it a step farther. Buyers of junky import power tools often complain of out of balance borne vibration. Take your handy smart phone app, remove the armature, cobble up wooden V's spaced on a board at the armature bearing centers, stick the phone on the board connecting the V's, spin the armature by some means, and the phone will tell you how much and where to correct.

The challenge therefore is to the geeks who frequent this forum: write an app to do the above. A grateful world will reward you with 32 virgins.

I'm old and no longer have the smarts or I'd do it and make everyone jealous.

Forrest Addy said:

The challenge therefore is to the geeks who frequent this forum: write an app to do the above. A grateful world will reward you with 32 virgins.

Click to expand...

I could do it, but I already have more women than is good for me, so I pass....

PDW

Problem is not detecting the vibration, problem is tying it to the rotation of the motor - shaft - spinney bit. Smart phone will happily detect the vibration, it won't be able to tell you the angular rotation on the sharft to add a given mass too to counter act it. thats the problem.

Using the accelerometers it should be easy to see the state of - magnitude of vibration and its rotation in relation to the smart phones position, thats ifffffff the smart phones sensors will react fast enough - can read at a high enough frequency? have no idea what frequency they read em at. Obviously that needs to be a significantly higher frequency than shaft rotational speed.

I'll add it to the list behind:
1. Electronic servo motorised, dividing head and precision digital protractor.
2. Multichannel thermocouple input, ramp-soak temperature controller.
3. Interferometer based, blue/violet laser, optical dial indicator measuring in increments of 200 nanometres and temperature compensated.
4. Over unity Farnsworth Fusor chamber.

Forrest Addy said:

The challenge therefore is to the geeks who frequent this forum: write an app to do the above. A grateful world will reward you with 32 virgins.

Click to expand...

Everyone : be careful - it's trick. He doesn't specify the gender or even the species.

As adama says, it's not going to be just that simple. The computer rotor balancer at work has TWO strain gauges on the tuning fork like frame that the rotors run on, one for each bearing fit in an electric motor. It also has a rotary encoder that acts as a dividing head on the rotor being balanced so you know where to place the weight or remove weight to balance each end of the rotating mass. You can do single plane, but you have no idea if the rotor is wobbling like a football at each end unless you had two phones so you could get dual plane balance. On a drone copter, the blades are indeed single plane and are going to have gobs more influence on overall balance than a tiny rotor inside the motor.

If you could accurately flash the mobile phone's bright LED as the vibration signal peaked, wouldn't that let you strobe the rotor position with the highest out of mass component?.

Remember the movable trial weight? It has two purposes: to provide a marker for sensing RPM and to force an out of balance that can be compared to previous readings. If the app is written to take advantage of a movable trial weight. (4 trials at 90 degree intervals on the same plane of rotation). Locating the trial weight on an easily correctable plane (cooling fan cast into each end of an induction motor armature for example) gives you two planes of rotation for which the correction can be determined therefore. If all goes well and the planets align, you can accurately dynamically balance the armature conditional on the sensitivity of the sensors. Other point of elaboration are a BlueTooth tach, remote sensing where LDVT heads with seismic mounted masses are installed at suitable locations, etc. But these are more advanced and expensive considerations therefore I suggest sticking with the smart phone's on board resource for now.

Modern balancers have two sensors which detect out of balance at two axially separate locations simultaneously. The computer crunches the sensors readings with the data input by the operator and displays where the correction is to be applied and how much.

My proposal draws from early balancing technique. The app designer will have to research the dawn of balancing technology. Look up "Yakimov" balancer. and "static roll balancing" just as background to see what can be done with time and some very simple equipment. There is lots of balancing video on YouTube but beware of BS and plain idiot hacks.

This proposal is not a trivial project. The fundamentals of practical balancing, the capabilities of the phone sensors, the algorithms employed, and the operator interface require careful integration. Failing to do so will result in an unworkable kludge. A good job will result in a valuable tool both diagnostic and remedial with application in industry. .

I have a hunch the phone sensors are not that consistent. They may require some form of calibration prior to first use. another point is some data may be required the operator has to accurately determine and enter.

This is not a "it cant be done because it's never been done before" project. It's a "we got these new tools and this old problem, what can we do to solve it" project.."

Yeah, if you can allow for the flash delay and the delay in reading the accelerometers. Gotta remember, a phone accelerometers are designed to pick up human movement not machine vibration, that’s a very different sensor read frequency requirement.

Now thats not to say you can not fast enough to do it, nor the sensor read, a phone may well have these running at a good few hz, problem is a typical 4 pole motors 1400 rpm here in the slower 50hz part of the world, thats 23 revolutions a secound, then you need to read the accelerometers at least every what 45 degrees of rotation? ie 8 times per rotation, thats, 8x23=184 events a secound, you need to not only measure, but record calculate the highest and strobe the phones light. A led will happily stobe fast enough, but remember, it will only work if the phones light is facing the part and you can not hold the phone as it has to be getting the vibration.

A quick google seams to bring up a max accelerometer sample rate of 100hz, even then androids not real time, so its not going to always be constant and theres going to be a delay between reading the value and getting that processed to flash the led light.

Roughly-of the top of my head you might if your lucky be able to get 90 degree resolution at typical line speeds for a 4 pole motor. But even then, few people worry too much about balance at that kinda shaft speed, its often more like 3K+rpm that balance on most things really starts to matter and a phone just is not fast enough to do this with any degree of accuracy.

This is one of those times a good old fashioned CRT oscilloscope comes into its own, simple single index pulse encoder which need be little more than a photo detector and shiny bit of tape on one channel as the trigger and then the other channel on the accelerometer. Multi planes easy, just add another accelerometer and use a scope with more channels, Most CRT scopes have a multi Mega hz frequency response, so you will get great definition even at the 100,000+Rpm range.

Theres more than a few small cheap digital scopes out there too now, often based on similar hardware to phones, hence with the correct sensor hook up, it should not be too hard to end up with say a device you can comfortably carry in one small case to do this. Dont know how much accelerometers are, but a small scope and a wall wart power supply and the leads to hook it up is easily sub $100 these days, just the accelerometers on top.

The phone also has a camera.
Turn it on and focus it on the shaft.
Put a dot of white out on the shaft so the phone can sense the rpm's and shaft position.

Why not sample for 200 revolutions? The mean Vs time (less random noise) will be more representative than a sample taken from a single revolution and actual sample rate will be less of a problem.

The whole idea of the movable trial weight is to provide a number for a max condition. Comparing trial weight readings from the 4 quadrants will determine amount radius, and angular position of the out of balance. Detecting our of balance from sensor data alone is a whole order of magnitude greater in difficulty.

Look at it pencil on paper graphically:

Run he trial weight at 4 quadrants noting the raw figure from the phone

Draw two lines crossing at right angles in the center of the page..Call the intersection "origin"

Pick a convenient scale and measure vectors from the origin proportionate to th eout of balance reading in order first to last. Label them in quadrant degrees

Connect the ends of opposite vectors to form new crossing lines. Call that crossing "intersection"

Draw a ray from the origin through the newly constructed intersection

Scale the distance between origin and intersection. That distance divided by the vector quantity represents the out of balance proportional to the trial weight.

The direction of the ray orients the angle of out of balance to the initial trial.

This is the old school descriptive geometry approach to solving out of balance polygons. Simple, elegant, even a second year apprentice can grasp it's possibilities.

I don;t think you need refined incremental data. A mean figure of merit drawn from a number of cycles over four quadrants is plenty. The results are massaged graphically by the app.

IOW, you don;t set the phone on the motor and get complete data. That's too easy.

If the app can take the wobble out of any of the several ceiling fans I've got, I'm all in!

CalG said:

If the app can take the wobble out of any of the several ceiling fans I've got, I'm all in!

Click to expand...

Brainstorm!

Mount a laser pointer on the stator hub of the fan pointing vertically down on the floor.

Turn the fan on and note how the out pf balance orbit is now magnified by the laser so it draws a circle on the floor.

Number the blades.

Tape a penny on the fan blades one at a time and run the fan Measure the circle drawn for each blade..

The diameter of the circle is the trial weight reading for each blade.

Work the descriptive geometry I out lined in Post #11.

Work the ratio and proportion math for the correction weight and its radius.

Chances are the correction won't be all on one blade.

Play with it.

When the circle diameters are all equal with the trial weight remove it and check the size of the circle after balance..

Lemme know how it works.

Another approach using dedicated hardware would be to use off the shelf Arduino hardware: The prototype below took one night to solder together the 5V to 3.3V logic level shifter to interface the rotary encoder to the on-board quadrature encoder circuitry. Because it's using a dedicated encoder reader circuit instead of software it's fast: up to 42 MHz speed or more than fast enough to keep up with the encoder's 7500 RPM maximum speed. The software I wrote to read the encoder register and convert to 0.05 degrees resolution is six lines in length and took about two hours to get running.


Costs were:
Arduino Due: $17
LCD adaptor board: $12
Colour graphic TFT LCD: $19
2048 Count encoder: $36
TOTAL: AUD $84

three axis accelerometer/gyro sensor boards to make this into a dynamic balancer are $2.50

Clip on guitar tuners may also have some useful technology in them, seems they have some way to pick up tiny frequencies/vibrations quite accurately.

If you've only got one sensor you're probably stuck with doing a 4-trial measurement with an added mass, one plane at a time. If you can get phase it's probably a bit more direct, but the calculations are worse. It doesn't matter if the sensor is in a fancy cell phone or just something hooked to a meter or scope, but filtering is essential. The vibration frequencies for normal speed motors are quite low, say 57.5 Hz fora 3750 RPM motor. I did a spreadsheet for this some time back that works with ceiling fans and grinder motors.

Forrest Addy said:

Machine tool motors, hand power tools, anything with a rotating mass for that matter requires some minimum standard of balance to function properly. Power tools that leave your hand tingly slowly degrade your sense of touch. Machinery with dodgey motors telegraph vibration into cut surfaces, etc. Out of balance poses a real problem. The remedy is to balance the rotating parts but that requires a tear down, sending out the parts a day or three in limbo, then the assembly. PITA and in the meanwhile progress has halted for everything down stream of the equipment out of service.

We now have smart phones elaborately equipped with an array of sensosr including a accelerometer and a gyro. These are sensitive as hell, equal to the sensors on a balance machine. Load up a vibration app, set the phone on the table in a quiet room and the phone will register in the squiggly line fall of a green pea onto the table from an inch of height.

Quadracopter drone nuts balance their lift motors with a smart phone: There's some relevant videos on YouTube

What's needed is an app designed to meet the needs of on-machine balancing. Set the phone on the motor and the app determines the motor RPM from the cyclic nature of the vibration then goes on to quantify it and if the distance between bearing is entered diagnose the accelerations and phase relationship for each Add a known trial weight like the hub and enter its radius and location from a particular bearing and it will tell you the amount, radius, and, angle for correction for each re-position presuming 90 degree intervals.

Take it a step farther. Buyers of junky import power tools often complain of out of balance borne vibration. Take your handy smart phone app, remove the armature, cobble up wooden V's spaced on a board at the armature bearing centers, stick the phone on the board connecting the V's, spin the armature by some means, and the phone will tell you how much and where to correct.

The challenge therefore is to the geeks who frequent this forum: write an app to do the above. A grateful world will reward you with 32 virgins.

I'm old and no longer have the smarts or I'd do it and make everyone jealous.

Click to expand...

This does not directly address your particular suggestion but it's close:
http://bbs.homeshopmachinist.net/threads/70934

BTW I had the rotor and pulleys from the motor on an import 20" drill press professionally balanced recently. It cost me $150 and took about an hour and they let me stand there and watch. They applied epoxy putty for weights. The cheap machine tool motors can be quite imbalanced and mine sure was. It took a fair size glob of putty to get it smooth. The result was night and day improvement. Now I can have a very tricky alignment on the table unclamped or very lightly clamped and start the spindle and nothing on the table will move at all. It's like an entirely different machine. IMO totally worth it in both the time and money.

There are older school ways of balancing things yourself. Here's one such:
http://www.antiqueautoranch.com/montana500/septnewsletter/page4.html

Finally, if the machine you want to balance is a bench grinder, see this:
http://www.nwnative.us/Grant/shop articles/sharpTable/balancing

metalmagpie

CalG said:

If the app can take the wobble out of any of the several ceiling fans I've got, I'm all in!

Click to expand...

Celing fan wobble is usually due to pitch of the blades rather than balance issues. Just twist one of the blades to increase it's pitch, and note if it's better or worse. make adjustments using logic until it runs right.