M6 8.8 bolt fatigue failures (HELP!!!!) 4

[trikeflyer]

I am trying to improve a bolted connection (aluminum plate bolted to threaded aluminum case) which currently uses three M6 8.8 allen head cap screws. I have had all three bolts fail in fatigue, presumably due to vibration. I suspect that the problem may be that the bearing strength of the aluminum is lower than the tensile strength of the bolt, resulting in creep of the aluminum and loss of pre-load. Unfortunately, I don't have the option of using a different size of bolt or more of them.

I am looking for suggestions. I was considering replacing the 8.8 fasteners with 10.9 hex head cap crews and a stainless steel fender washer. I realize that the 10.9 fastener will require a higher torque to reach a working pre-load. The purpose of the washer was to spread this pre-load over a greater area in order not to exceed the bearing strength of the aluminum. I'm no expert (Obviously!), so I'd really appreciate an expert opinion.

Thanks for suggestions and direction!.

Dave C.

 

[MikeHalloran]

Uh, Trike, what carries the propeller thrust to the airframe?

The plate directly?

Or the plate through the motor?



Mike Halloran
Pembroke Pines, FL, USA

 

[trikeflyer]

Sorry about the confusion, looks like I used the term "fretting corrosion" rather than the correct term "fretting wear". I'm trying to be careful (still trying not to call a screw a bolt). I'm just a darned electrical engineer. That's my excuse, anyway.

Mike, if I am understanding your question correctly the plate is mounted only to the motor. The propeller thrust is carried through the plate to the motor, then through the motor to the airframe (via rubber mounts). This is a pusher type propeller, and other than the effect of belt tension on the top bolt, I believe all connections are in compression.

Dave C.

 

[dik]

Thanks Cory... Is this caused by the difference in the hardness of the materials? fasteners aren't my forte (I'm not sure what is at times <G>)... just adding my $.02... it's a learning process for me, too. It appears to be a real problem, then... I would think the proper solution would be to increase the size of the bolt or change the material into something that be of similar hardness. Is it possible to 'ultramachine' or try fitting several fasteners for the threads to better match and to reduce the clearances?

Dik

 

[MikeHalloran]

But the propeller axis is not coaxial with the motor shaft, right? In that case, the plate is stressed in bending, and because it acts like a lever, the propeller thrust is _magnified_ at that 'top' bolt.

Or, I could be completely wrong, as often happens.

Could you post a picture of the assembly?



Mike Halloran
Pembroke Pines, FL, USA

 

[trikeflyer]

Mike, I just went out to take a picture, then I remembered that I have it all torn apart to fix it! HA! Oh well, I found a picture that shows the thing folded up. If you zoom in to the propeller area you can see the motor and reduction drive. Anyhow, in normal "unfolded" mode the prop is in the back.

You seem to have a pretty good idea of how it is configured. I'm still impressed by your insight regarding the belt tension taking pre-load off of the top bolt location. For sure the thrust is contributing a moment about that top bolt which is opposite the moment from the belt. I'm having a bit of trouble figuring out how to calculate the bending moment from the belt, and I'm wondering what the static vs vibration contribution is of each. The other factor is that the belt tension is more or less static, but the propeller thrust varies quite a bit depending on mode of flight (climbing, cruise, etc.) This is just a SWAG, but I'm thinking that for most of the time the moment from the belt tension dominates.

Dave C.

 

[MikeHalloran]

Oooh. That's not a plate anymore, not with that big hole around the engine crankshaft. It's more like a pickle fork.

Okay, when I zoom in, it gets kind of fuzzy. A close-up including just both pulleys and the plate might help. If I'm seeing it right, the 'top' bolt is between the pulleys, and the other two bolts are nearer the engine cylinder, basically at the tips of the pickle fork.

There's no idler, so even if the belt is a little tight at assembly, it will lose static tension as it stretches (permanently) in the first few minutes of operation.

The good news is that the dynamic belt tension, the direct result of the power being transmitted, should produce a moment opposite to that produced by the propeller thrust, and because they both originate in torque transmitted, they should track each other.

The less good news is that the net moment is trying to rotate the metal being squeezed by the top bolt, and the legs of the pickle fork can't help much in resisting that, because they're very narrow relative to the sturdy remaining portion of the plate between the hole and whatever carries the prop, presumably some kind of flanged stub shaft.

The belt tension also tries to shear the bolts. The traditional (terrestrial) solution for a situation like that is a hollow dowel around two of the bolts (in matching counterbores) or small short dowels next to the bolts. Or a pilot diameter and counterbore around the big hole concentric to the crankshaft; I can't tell from the photo if a suitable feature in the crankcase is present, or available and unused.

It might be instructive to get a 'tenths' dial indicator and good adjustable base, and try to directly measure deflections of portions of the plate relative to the engine as you manually apply thrust to the propeller hub. I suspect you'll find the plate is a lot more flexible than you'd like.


Mike Halloran
Pembroke Pines, FL, USA

 

[hydtools]

The bending moment from the belt would be the product of the belt tension times the distance from the belt midline to the plate mounting surface.

My visualization of the plate is that it flaps back and forth as the thrust changes.

Ted

 

[trikeflyer]

Yeah, you are right, it is more of a pickle-fork.

I just pulled it apart today, and with the standard bolts and installation torque after 5 hours of running time there is noticeable fretting corrosion near the top bolt.

The good news is I have another option which I am in the process of testing/implementing. There is an alternative "plate" available that has the same offset, just down instead of up. By flipping the whole motor over and using this other plate the thrustline stays the same. It's a bit of a plumbing mess to do it, but the upshot is that the alternate arrangement then has the single bolt at the bottom and the two bolts on the sides sort of half way between the crankshaft and the propeller pulley. Overall the plate is stouter in the area between the two top bolts and the propeller pulley. It's still a pickle fork on the bottom part, but at least there are two bolts at the top where they are really needed.

I just realized that I can calculate the belt tension. It is speced at 2mm deflection with 5kg applied to the center of the span. I need to measure the span do the trig and convert that all to fig newtons, but it's getting late. I'll do that as well as take some pictures tomorrow morning. Oh, and for thrust, it's 27hp and my forward speed is around 35mph, so with a bunch of unit conversions I can do that too, I think. I'm thinking its about 100lbs, but I'll calculate it tomorrow.

BTW, the pickle fork is about 0.8" thick and the fork tines are about 0.4" wide.

Great stuff, thanks guys for taking an interest.

Dave C.

 

[desertfox]

hi trikeflyer

I realise I have entered this thread quite late on and not 100% sure I have the configuration correct in my head, maybe we can have a close up picture of the area concerned at some point.
Anyway I have sent a link showing a typical failure of a cap screw by fatigue.
In addition I agree MikeHalloran that the one bolt is seeing most of the load and I have uploaded a file on how you can calculate that based on the assumption that the plate doesn't bend and that it rotates about one edge.
This would confirm why the top bolt breaks first and then the other two bolts.
Incidently my calculation assumes all the bolts are in tension and that the bolt loads in the othr two bolts can also be calculated by multiplying the linear distance to the bolt centres from the pivot point.
If I have mis-understood your problem please let me know.

Regards

desertfox

http://books.google.co.uk/books?id=...esult&ct=result&resnum=8#v=onepage&q=&f=false

 

[trikeflyer]

Thanks, desertfox, and thanks for joining the thread.

I'm out of town for the weekend and having a hard time with my laptop, so I won't get a chance to carefully look at your analysis until I get home to my printer. From a first glance, though It looks like you have the configuration correct.

Going back to the question of the M6 screws and specifically choosing 10.9 vs 8.8 grade, I'd like to ask another question. Maybe corypad could provide the definitive answer. Here's the scenerio: Whenever I mention changing the 8.8 screws to 10.9, I invariably get the comment that although 10.9 are harder they are also therefore more brittle, and thus more likely to break.

So what is the real story? My simple minded thinking is that by going to 10.9 screws properly pre-loaded I am more likely to prevent relative motion than with the 8.8 screws tightened to their (lower) pre-load. If there is no relative motion, then there is no fatigue, and the screws don't fail. But what if the 10.9 screws do not provide enough pre-load to prevent motion? Are they in fact then more or less likely to fail than an 8.8 screw? Does that make sense? A slightly different angle on the question might be to ask this: If I can achieve a higher pre-load by either using a 10.9 screw or a larger diameter 8.8 screw which is the best approach?

Thanks again, everyone. It's going to take me a while to sit down and go through all the comments and calculations in detail, but I already can see that there are several insights that are immensely helpful. For example, I wouldn't have considered changing from the 'up" plate to the "down" plate without the insights about how belt tension and thrust impact the load on the screws.

I think I saw someone else on another thread comment that it was a bit like drinking from a firehose. I have to agree; I am feeling a bit drenched here! You guys are great!

Dave C.

 

[MikeHalloran]

Firehose? Nah. More like a Wood Screw Pump.

Wait'll the clutch hooks up...



Mike Halloran
Pembroke Pines, FL, USA

 

[CoryPad]

Property class 10.9 material has lower fracture strain than property class 8.8, but it isn't brittle. Class 10.9 material is stronger than 8.8 material both for static and fatigue condition, but property class 8.8 and 10.9 screws have essentially the same fatigue strength due to the high stress concentrations present in threaded fasteners. The fatigue improvement to the joint for 10.9 screws is that you can use a higher preload which reduces the oscillating load applied to the bolt. Usually, a smaller diameter, stronger fastener is better than a larger diameter, weaker screw in fatigue because the smaller screw is more compliant, which reduces the oscillating load applied to the bolt.

 

[desertfox]

Hi trikeflyer

Here's are a couple of sites to give you some info on blted joints:-

http://www.roymech.co.uk/Useful_Tables/Screws/Preloading.html

http://www.rbwmfg.com/images/HelpfulHints2.pdf

With reference to our question about 8.8 and 10.9 grade fastenings I came across a situation where one company prefered to use 8.8 over 10.9 wherever possible, their reason for it was explained to me as follows:-
"Most of our joints are tightened using a torque figure that generates 80% to 90% of the fastener proof stress (yield stress), given that torque settings can have an error of 20%-30% there is a very good chance that some of the bolts will be stressed over the yield figure.
Now once over the yield figure the stress/strain curves differ and as Cory stated the 10.9 as a lower fracture strain than grade 8.8 therefore they felt that when the bolt exceeded the yield stress at installation they had a little bit more margin against completely fracturing the bolt using a grade 8.8 as opposed to a grade 10.9.
Just looking at the Torques you are quoting for your application it is possible that you could be reaching or exceeding the yield stress of the 8.8 bolts particularly if there is any lubricant present on the threads at the assembly stage (whether lubricant is present delibrately or accidently)which could also contribute to the failure mode your seeing at present.

desertfox

 

[trikeflyer]

Mike, I had to look up a Wood Screw Pump. HA! I think you are right. It's not the pressure that's going to drown me, its the sheer volume. Thanks again everyone for taking an interest in this.

desertfox, I've had a chance to look at your analysis. Thanks very much for that. If I am interpreting your drawing correctly, it's not quite the situation that I have. In my case the plate bolt pattern is centered on the crankshaft, and the pulley that is mounted to the plate is offset (to the top in your drawing) I think the analysis is similar, except in my case I think the pivot point (assuming a ridgid plate) would be at the bottom of the plate rather than the top. I am assuming that the pully that you are showing is the one on the crankshaft and the one that is out of the picture is mounted to the plate.

I've attached a couple of pictures: One shows the motor and the mounting surface for the plate, the other shows the plate and the eccentric to which the propeller pulley is mounted. The plate has a rim approximately 3mm high at the top and near the mounting bosses. For a reason I'm not sure of, the manufacturer machines the motor and plate so that that this rim is an interference fit on the motor. I tried to get a decent picture showing the fretting wear: It appears to be on the edge of that rim perpendicular to the plate surface, and mostly on the rim near the top bolt.

I've tried to catch up on some of my calculations. Here's the results:


Bolt pre-load:

Assuming 0.15 to 0.25 coefficient of friction (screw lubricated with Loctite) a torque of 15Nm results in approximately 11kN of pre-load. This is 66% of the proof load of the property class 10.9 M6 screw. This should result in a maximally pre-loaded joint for the size of fastener and the strength of the crankcase threads.



Engine thrust and torque:

At full power:
Torque: 32.6Nm
RPM: 6000
Engine HP: 27

Based on 35mph forward speed, 500fpm max climb rate, and power off glide of 10:1 the actual thrust delivered to the airframe is 106lbs.



Drive belt:

Center-center distance between pulleys: 110mm
Center of belt to plate/motor surface: 50mm
Belt tension: 1.348kN
Bending moment on plate: 67Nm
Unloading of top bolt assuming ridgid plate (original plate): 515N
Unloading of top bolt assuming plate pivots on bottom bolts (original plate): 644N
Unloading of each of two top bolts assuming ridgid plate (reversed plate): 335N
Unloading of top bolt assuming plate pivots on bottom bolts (reversed plate): 322N


Thanks again,

Dave C.

 

[trikeflyer]

For some reason only one picture seems to have made it. Here's the other.

Dave C.

 

[desertfox]

Hi trikeflyer

I may have the drawing the wrong way round however I was really just drawing a pulley with a plate with 3 holes at 120 degrees and showing the belt tension, I believe that my analysis is right for your situation though even if my configuration is slightly out.
The step for the motor and plate which you say as an interference fit is probably to take the direct shear load off the bolts and possibly reduce some of the bending load that were discussing now.
Any chance of a picture with it all assembled and close up?
Those bolt loads you have stated in your last post are they using my calculation method and also what is the pitch circle diameter of the bolts? The preload of the bolts on assembly it seems are far higher than those you have listed as external loads, there are presumeably more external forces on those bolts than were aware of at present? Finally it appears that your saying "unloading bolts at top etc" can you expand on that please or even better upload your calculations.
Your bolt preload of 11kN with the friction coeff for loctite gives a torque figures of 9.9 to 13.2Nm is that what your using?

desertfox

 

[hydtools]

trikeflyer,
I would say the counterbore on the plate is to centralize it on the motor and take the shear load. Is the counterbore/recess deep enough that the surfaces of the ears for the bolts do come hard together? If the faces do not come together this may be your bolt loading problem. Just asking because of the fretting on the diameter of the motor. Is there fretting on the face of the bolt ear surface?

Ted

 

[trikeflyer]

desertfox, thanks for the clarification. It does seem that your drawing is an accurate schematic representation of what is going on, I just wanted to try and head off or clear up any possible confusion over the actual configuration. This thread has evolved well beyond my original question about screws, so my original simplistic explanation of the configuration may be misleading.

I am in the process of re-builing everything the other way up, so I don't have a complete assembly to take a photo of at the moment, but I will as soon as I can.

For the pre-load calculation I was using the FAQ that corypad linked to earlier (faq725-536 How do I calculate an assembly torque?).

Here are the numbers I am using:

Assembly torque - 15Nm
Pitch - 1mm
Pitch diameter - 5.35mm
thread fiction coeff. - 0.15 to 0.25
bearing fiction coeff. - 0.15 to 0.25
hole diameter - 6mm
bearing surface outer diameter - 10mm for original allen head capscrews, 13.3mm for proposed flanged screws

I calculate a pre-load tension of between 10.42kN and 11.11kN for the flanged screws (depending on the friction coefficient)
I am using the numbers corypad suggested for friction coeff of a screw lubricated with Loctite.
I will try and generate a (readable) fascimile of my other calculations.

When I say "unloading of top bolt" (dang it, I mean top screw; there I go again!) what I mean is that the belt tension has the effect of peeling the plate away from the motor in the top screw location. I am assuming that nothing moves and that the tension in the screw stays unchanged, therefore the "peeling" action would reduce the contact force between the plate and motor in the top location. Rather than "unloading of the top screw" I should have said "unloading of the joint in the top screw location" This may not be relevant if the interference of the counterbore is preventing the flat surfaces from coming together.

Finally, I would agree that the static loads seem small compared to the pre-loads. I think that the vibrational loads are the culprit here. The motor manufacturer has indicated that this joint is quite sensitive to proper propellor balancing. It also appears that in at least two and probably four of the failures that I know of rough idling of the motor has been implicated. Excessivly rich idle mixture is causing this two stroke motor to "four stroke" at low rpms resulting in significant impulse torque.

hydtools, I think that what you seem to be implying is correct. The flat surfaces do not come together, as I saw no fretting wear there, only on the perpendicular surface of the counterbore. The counterbore has a tapered wall thickness, approximately 4mm thick at the surface of the plate taperering to 0.5mm. I am wondering if the purpose of this counterbore is to amplify the clamping force. This would explain why it is machined to an interference fit. If so, then it would seem that any wear or relaxation of this material would quickly cause the joint to lose pre-load. On the original part in which the screw failures occurred this counterbore is almost completely worn away.

 

[hydtools]

The flat surfaces of the ears should come together. That is also how the machining would control the squareness of the plate to the motor. Otherwise, you would have to carefully tighten each bolt/capscrew to maintain plate squareness to the motor and parallel alignment of the pulleys.

This area would be worth checking for machining or design error. Just my opinion.

Ted

 

[trikeflyer]

Ted, that's a good catch. I agree that it seems the "ears" should come together. It would seem like the machining of the counter-bore would need to be very accurate in order to ensure that the ears come together while at the same time achieving the desired clamping action by the counter-bore. I'm having a hard time imagining a way to unequivocally verify that these surfaces have come together.

What I am going to say next is really ugly, so just a warning to the squeamish....

Is there some kind of joint filling material that could be applied to these flat surfaces to fill the gap? What I am thinking of is something that can be applied and then the whole assembly is clamped to some intermediate torque. The filler is then allowed to set and then the screws are tightened to their final torque.

OK, that should firmly establish that I am open to any suggestions regardless of how hairbrained or repulsive!

Thanks again for the great insights,

Dave C.