Unconventional DataLogging Question

[The Dark Side of Will]

I'm designing a "flywheel proof loading" apparatus. It will consist of an electric motor with a belt drive to a hub on which I can mount a bolt circle adapter and a flywheel. I will use some sort of crank trigger on the hub with an RPM switch. The RPM switch will operate a relay to reverse polarity of current to the motor. In this way, the motor will accelerate the flywheel to the set-point of the RPM switch, then reverse torque, decelerate the flywheel to zero and accelerate to the set point RPM in the opposite direction, then reverse torque again, etc..

The whole point of this is to be sure that the "home made" button flywheel I'm going to make for my Tilton clutch won't come apart at high RPM. The flywheel lines up on my right shoulder, of course, and my RPM goal is 8200-8500.

I think that it will be pretty easy to set up a counter to count the number of reversals.

However, another idea occurs to me. I can log the time required for each cycle, average over a large number of cycles and come up with a relative measure of the moment of inertia of different flywheel configurations. However, doing this is beyond my meager experience. What hardware/system should I be looking at to do this?

I think a *real* trick would be to detect a sudden decrease in cycle time (IE, flywheel came apart) and automatically stop the machine.

The way I expect to use the machine is to start it going and let it run for a week or two (accumulate tens of thousands of cycles) to proof a given design. After that I'd retire the test article and build another one just like it for actual use.

[Series8217]

Do you plan to bolt the clutch onto it for testing too?

At that RPM I'd put a scattershield on your transaxle regardless of what it does on the test rig..

[The Dark Side of Will]

Are you familiar with the difference between verification and validation?

"Validation" tests assess fitness for purpose. "Verification" tests assess compliance of the article with the design.

A "Qualification" test is a validation activity that stresses a test article beyond normal operating limits in order to validate that the design of the article is up to the task we want it to do, and then some.

An "Acceptance" test is a verification activity that stresses a test article to its normal operating limits in order to verify that the article was built correctly.

IE, the first article produced of a given design becomes the qualification test article. It is assumed to be stressed so much by the qualification test that it is not fit for duty afterward.

Subesquent articles are acceptance tested to verify that they were built correctly, then integrated into the system.

So a flywheel qualification test might be 10,000 cycles (zero to test RPM and back to zero) at a stress 25% over the normal operating stress. IE, for use up to 8500 RPM, the qualification test would be run to 9500 RPM. After that much abuse, the article would be retired and become a wall ornament.

The next unit would be subject to an acceptance test which might be 100 cycles at 8500 RPM. The intent of the acceptance test is to reveal a latent defect (such as a poor weld) that would contribute to the "infant mortality" side of the bathtube curve.

I wouldn't want to put the clutch through qualification testing. Use of Tilton clutches in racing applications is pretty darn good.

I would certainly put the clutch through acceptance testing.

I wonder if I can get my hands on GM's flywheel validation standards...

[The Dark Side of Will]

And I've pretty much decided I'll have to use an Ardruino to control the thing.

[Indy]

How will you be constructing the button?

There are factors here that make a home-brew fatigue test less effective than simpler methods. First, the material (fatigue) properties of 6061 and 7075 are already well known (just an assumption of your primary material). If you're using steel construction your maximum design stress should already be below the fatigue threshold. Fatigue tests are often used to find failure modes in complex assemblies rather than to give an exact quantification on failure. In a finely calibrated material testing machine hitting tension-compression loads within .01% and using very much identical test specimens we're often lucky to hit fatigue lifes within 15% of each other.

The way I see it is that there's conflicting design regimes here. If you try minimize weight, or MOI in this case, by increasing max stress and decreasing lifespan, you have no overhead for induced damage. If you maintain a decent safety factor you'll have so many cycle counts that the fatigue life won't mean anything on a test simulating inertial loads only. Something closer to a "static" test (max load, or failure) would probably tell you all that you need to know about the structure without the need to rack up a big electric bill.

If you still want to do a fatigue test, you should consider ways to make the input forces as realistic and brutal as possible, i.e. off-angle/off-axis resistance loading from the transmission end, or anything that forces the button out-of-plane due to loads or harmonics. You could probably make things more simple by eleminating reverse rotation with a simple stop-start in one direction.

[teamlseep13]

How will you be constructing the button?

There are factors here that make a home-brew fatigue test less effective than simpler methods. First, the material (fatigue) properties of 6061 and 7075 are already well known (just an assumption of your primary material). If you're using steel construction your maximum design stress should already be below the fatigue threshold. Fatigue tests are often used to find failure modes in complex assemblies rather than to give an exact quantification on failure. In a finely calibrated material testing machine hitting tension-compression loads within .01% and using very much identical test specimens we're often lucky to hit fatigue lifes within 15% of each other.

The way I see it is that there's conflicting design regimes here. If you try minimize weight, or MOI in this case, by increasing max stress and decreasing lifespan, you have no overhead for induced damage. If you maintain a decent safety factor you'll have so many cycle counts that the fatigue life won't mean anything on a test simulating inertial loads only. Something closer to a "static" test (max load, or failure) would probably tell you all that you need to know about the structure without the need to rack up a big electric bill.

If you still want to do a fatigue test, you should consider ways to make the input forces as realistic and brutal as possible, i.e. off-angle/off-axis resistance loading from the transmission end, or anything that forces the button out-of-plane due to loads or harmonics. You could probably make things more simple by eleminating reverse rotation with a simple stop-start in one direction.

I agree with Indy along the lines of stressing the flywheel to a max stress(plus a certain safety value) test to see if your construction is adequate

Fatigue testing would need to take into account(as well as can be estimated) trans axle deflection that would put a bending stress on the flywheel and of corse overload and shock loadings.

I do love the idea of actuall testing going on for parts you are creating. Good job!

[Series8217]

Use whatever microcontroller you're comfortable writing code for. Measurement and control for this apparatus are extremely simple. You just need to count some pulses out of a hall effect sensor and switch some relays for an electric motor. If you've used an Arduino development board before that should be perfect.

A PLC may be easier/simpler if you already have experience with one.

Are you familiar with the difference between verification and validation?

Yes.

I don't think the flywheel proof loading apparatus is close enough to the real use conditions to convince me it's safe to use. Does the test rig replicate crankshaft harmonics at high RPM, sudden clutch grab on an upshift from high RPM, etc.? A belt drive has a LOT more give than a sticky clutch; it's going to soak up the transient spikes. The inertia of the motor stator will also soften things up too. There are many other factors (such as off-axis load that Indy mentioned above). The results from the proposed test rig would not convince me to spin a "home-made" flywheel spin at 8500 RPM in line with my major arteries.

When you're designing such parts for a large corporation to sell hundreds of thousands of, a single very unlikely failure results some legal expenditures to settle the resulting civil action. When you're designing parts for you, that failure results in death.. so I'd go ahead and use a scattershield anyway.

[The Dark Side of Will]

These are pics of the Quartermaster flywheel for the 7.25" dual disk FWD application:





It is a single piece of steel with what are essentially rivnuts set into it. I was initially thinking that I'd make something like this, using a disk of 1/2" plate blanchard ground down on both sides down to about 0.450. The crank bolts would have to be countersunk, and a shoulder machined onto the outer perimeter. The packaging in general would be extremely tight in terms of stack height.
I would bore the center out of a flexplate and weld the remainder to the OD of the disk to produce something that looked a LOT like the QM piece. The basic question I'd like to answer is "How many of what size speed holes can I drill in the remainder of the flexplate and not have it come apart within a reasonable (or even unreasonable) service life turning 8500 RPM regularly?"
Packaging, temperature tolerance and wear tolerance--the flywheel being the friction surface for anything but a carbon clutch--are driving me to steel vice aluminum.

In expending further thought on the matter, I am starting to think it would not be too hard to mill away the right part of the bellhousing of a 282 (or even a 284, F23 or F40) and weld in a starter mounting plate in order to use a small diameter ring gear and a reverse rotation gear reduction starter.
That would dramatically reduce both the diameter (and the accompanying centrifugal loads) on the ring gear AND the amount of "new engineering" that would have to be done on the unit...

As far as the other loads suggested above...
I'm not sure what's mean by "off-axis" loads. If my transmission loading starts to become anything but axially symmetric, my transmission has already come apart.
The clutch apply and release loads are relatively tiny. The release load for the Tilton is in the range of 500# and the apply load is between 2500 and 3000. For "bell washer" loading through the thickness of a piece of .450 thick steel plate, I'm inclined to think that's not significant.
The drive loads are interesting... but they're loading in shear which is greatest at the hub. The stress from the centrifugal force is predominantly at the rim. The drive loads are thus at their greatest on a portion of the part that is essentially not loaded by the centrifugal force.
The thing that *does* intrigue me is temperature... It's probably no trick to get the friction surface of a low mass clutch to 1000 degrees or better with a couple of hard launches. This obviously affects the properties of the material, but ALSO causes effects from thermal distortion. As the hub experiences thermal growth, the balance between the radial stress on the spokes and the circumferential stress--"hoop stress"--on the ring gear and/or outer loop would change to increase the hoop stress and decrease the radial stress.

[The Dark Side of Will]

I do love the idea of actuall testing going on for parts you are creating. Good job!

Hellz Yeah!
One thing I could use the test apparatus for would be DE-structive testing of flywheels... I could build a scatter shield out of 3/8" thick steel (or thicker...) and just let the thing rip until a stock one comes apart. The controller would sense the sudden change in acceleration and stop the machine.
Blow some shit up!