Bolted joint slippage and implications 1

[ivymike]

Suppose I have a pulley bolted to a shaft. Clamp load between the bolt and shaft shoulder provide torque capacity (via friction). I've performed static slip tests (clamp shaft, measure quasi-steady torque vs time while gradually increasing torque until the pulley slips) and found that the slip torque is very close to what I've calculated, and that it is reasonably well controlled (multiple units measured). I've match-marked the pulley to the shaft and used the assembly in its intended application. After use, it is apparent that the pulley has slipped significantly on its shaft. If I re-mark the bolt, loosen it, and measure torque while retightening to the new mark, I find that the torque required is the same as the installation torque within measurement error (ie bolt hasn't loosened significantly, if at all).

The question / argument is:
Position 1 (mine): Loads in use must be exceeding the quasi-static slip torque value often enough to cause the rotation observed (regardless of duration of high-torque events, they must exceed the quasi-static slip threshold)
Position 2 (coworker): The gear will slip at a lower torque if the load is applied suddenly, so the loads required to explain the slippage may be much (~50%) lower than I think if there is any kind of "impact" going on in the system

Anyone care to weigh in? Am I all wet - is the "dynamic slip torque" really much different from the "quasi-static slip torque" of a clamped joint? Is there a third position that one might take?

 

[handleman]

Is there any way you could measure actual torque in use? Of course, if loads are applied suddenly then you may get instantaneous torque loading that's far above your quasi-steady state. Is the load applied suddenly as a rule? If so, how is the load applied? Let's say you're driving this shaft with a motor attached to a clutch. Spin the motor at 1750 rpm and then drop the clutch. Your motor's inertia will hit the shaft with significantly greater torque than the steady-state load.

I know that's an extreme example, but it's all I can come up with given the information you've posted.

 

[ivymike]

I expect that we are hitting the shaft with sudden torque increases. I think the spikes are high. What I'm wondering is whether I can draw a "minimum spike height" line at the static slip torque, or if the joint might slip with a lower torque impulse if the impulse happened fast enough.

There is a great deal of disagreement currently about how high the spikes are - we've done shaft torque measurements, and they don't look that bad. We don't slip every joint, though, and some joints will run for many hundreds of hours before they begin slipping (no identified system changes to speak of). I suspect that there is something going wrong from time to time (which we haven't found) that increases the height of the spikes above what we've measured, and above the static slip torque level. Others believe that the spikes are always as low as they were in the one test, and that there is something-or-other that makes certain joints decide it's time to slip at a low torque level.

 

[iainuts]

Is there any chance of a thermal gradiant/transient being created during operation? If there was, the bolt might loosen during the transient due to differential thermal expansion and then go back to 'normal' afterwards.

I've had similar problems with torqued fasteners, especially ones under a cyclic load even when that load is easily calculable and is much less than the frictional load. Part of the problem might be explained due to thermal transients.

 

[handleman]

Would it be possible to change your clamp style pulley to a wedge-type keyless bushing? I believe you'd get quite a torque capacity increase.

 

[ivymike]

I'm not sure a torque increase is what I want - if I start breaking shafts instead of slipping pulleys, I'm in a worse spot. We're also sensitive to positiona and alignment, and taper joints are notoriously crappy on both counts.

 

[handleman]

So, basically you want to get torque enough to not slip but without potential shaft damage, right? Would some sort of shear pin arrangement be an option?

 

[ivymike]

the question is simpler than that:
is the torque "seen" by the joint the static slip amount, or something less?

 

[diamondjim]

Have you tried using loctite on the
interfaces? Sounds like a creep
phenomenon.

 

[GregLocock]

I can't think of a mechanism by which a 'sudden' torque of a given magnitude would be more likely to cause slip than a steady torque of the same magnitude, with normal engineering materials (guessing you don't have a silly putty shaft).





Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[CoryPad]

If I re-mark the bolt, loosen it, and measure torque

and

bolt hasn't loosened significantly, if at all

How did you conclude this when you loosened the joint? It would have been better to measure the new tightening torque before loosening the joint and compare to the original tightening torque. Isn't it possible that the joint experienced vibration loosening? This phenomenon can occur with torque inputs lower than the static slip torque.

Regards,

Cory

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips Fora.

 

[ivymike]

I measured the retightening torque after loosening (as stated above). Do you think that tightening farther would give a better answer than retorqueing to the same angle?

The only other thing I can think of is that if there is a substantial "tilting" load applied to the pulley, and the ratio of bolt stiffness to abutment stiffness is not low enough, I might reduce my friction capacity by separating the joing somewhat.

 

[GregLocock]

Maybe you could examine the true capacity of the joint by applying a tangential force at the rim of the pulley.

Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[rorschach]

Is milling a keyslot is out of the question?

 

[ivymike]

fixing the slip isn't the problem.

 

[ArchetypeJoint]

If you are getting slip immediately at start-up it is likely to be torque-rate dependant (effect of pully inertia causes slipping at lower loads than iniitally predicted). If slippage occationally happens over time it is likely a case of loosening. Is the pully located by a precision fit over the shaft or a preciaion diameter bolt? If the pully simply butts to end of the shaft secured by a standard bolt and through hole, small imbalances in the pully during rotation will create shear forces wanting to create relative motion in the joint. These forces are additive to the forces you initially predicted and is a prime cause of vibration loosening. Also, is the direction of shaft rotation such that bolt will tend to tighten or loosen? Bike pedals have right and left hand threads to get rotaion in a tighening direction - not the same joint design, but illustrates the point.

 

[ivymike]

Slip seems to occur after hundreds or thousands of hours of use. Slippage happens in either the loosening and tightening directions (both have been observed). Rotation is primarily in the loosening direction. Loading occurs alternately in both directions. The pulley is located via a light press, then clamped stoutly in place via a flange-head bolt. Total diametral runout is on the order of 0.020mm on this ~100mm diam pulley.

Again, the question was not in regard to stopping the slipping. I would have thought that bolt loosening would be obvious during retorque measurements (is that assumption incorrect?). I want to know "what I know" about the loads based on the slippage.

 

[ArchetypeJoint]

No, I don't think you can discount fastener loosening, because measuring residual torque with torque wrench (re-torque measurements) can be a very misleading test. It is very operator dependant and can lead to higher that actual results because the peak measured is due to overcoming static friction, often when the head moves relative to the bearing surface. The reading you want is the torque at the point when the male thread starts to rotate relative to the female threads(by the way this should be done in the tightening direction). This can be more accurately measured using a torque-angle transducer and a transient recorder. The better method is to measure bolt tension directly. It is tension, not torque, that is keeping the joint together. This can be done most accurately using ultrasonics or strain-gaged bolts.
Regarding understanding the loads, your quasi-static test is an ideal case, because it nearly static. If your pully loosens with time either the demand side of the equation (the loads applied to the joint increase)increases, or the supply side (clamp load and coeffcient of friction - and COF is constant)decreases . My guess is that it's the latter which is causing slip.
Dave

 

[ivymike]

as above, the bolts were marked, loosened, and re-torqued while torque was being measured. Torque vs time was plotted electronically as the bolt was brought back to alignment with its original (marked) position.

The bolt is too short for ultrasonic measurement (or so I'm told). Strain gages don't seem to survive long enough in use.

It sounds as though you agree with my position (position 1, above), that the slip torque is close to the qs slip torque if the joint hasn't loosened.

 

[GregLocock]

Sounds like you have a good handle on the life of the joint.

How did you measure the in-service torques? Where were they measured?

This is beginning to sound like a case for a classic 8D analysis, bad luck.



Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.