I have now obtained a copy of Cannondale's report of the testing they did for the CPSC. Here is my analysis of their experiment.

---

The first error is in getting my nationality wrong, but I'll let that pass :-)

I will focus on their "Item #1" since this is where the greatest failings of their test are apparent.


In contrast to the extremely weak braking force (which acts to push the hub down relative to the fork), this is an extremely large load which acts to keep the hub seated snugly in the dropout. The angle is not specified but it looks close to 45 degrees in the photo (presumably it is supposed to simulate the forcing from the rider's arms) and I will use this value in the calculations. Even at 30 degrees from horizontal, the conclusions are qualitatively the same.

Obviously, a 275 pound (1210N) force on the handlebars is only realistic under very heavy braking, and even then it sounds optimistic. But maybe they have measurements to justify it (see below). And even if it is plausible in itself, it is certainly not plausible in combination with a 5 pound lever force.

Analysis of the combined effects of Errors 1 and 2

This test combines a 10-year-old with weak hands pulling the lever, and a 250-pound-plus gorilla with all his weight shifted to the front wheel under extreme braking!  The net effect of the weak braking and huge force on the bars is that there is a negligible force pushing the wheel down in the dropouts - but a precondition for the skewer loosening is that the hub actually moves.

Working back from the DIN test results, we can estimate that the 22N lever pull which they apply would generate a deceleration of 6.2*22/100 = 1.4m/s^2 for a load of 100kg, which means 140N of horizontal force at the tyre contact point (330mm from the hub). With an effective disk radius of 82.5mm (size not specified, but I am assuming a mid-size 185mm diameter rotor, and pad centre 10mm inboard of this) the force at the pad contact point will be given by 140*330/82.5 = 560N in a roughly downward direction. The vertical reaction force at the ground due to the "arm force" of 1210N at 45 degrees is 1210*cos(45) = 850N. Even when one assumes that this vertical reaction force is shared between both sides of the hub, and the disk force is generated at one side only, the net downward force on the left hand side is a trivial 560-850/2 = 135N, which will easily be resisted by the feeblest of QR skewers. This is a meaningless test that completely fails to address the basic premise - that the large downward force applied to the hub under heavy braking can cause the wheel to slip in the dropouts. As my calculation on the main page shows, the net downward component of force may easily be as much as 1825N, some 13 times greater than that generated by Cannondale's test!

In the face of this failing, the fact that Cannondale only performed a single test (rather than a range of diferent skewer tensions, disk diameters, braking forces etc) and did so on an essentially smooth roller (0.5" bumps? That's hardly off-road in my book) are relatively minor omissions - even though they are also in themselves significant failings.

Now, on to Cannondale's comment in the summary: "At this time there are no reasons to believe anything is mising or over constrained in this test". I believe I have shown in a few brief paragraphs that this test is wholly inadequate and does not begin to address the weakness of the design. Furthermore, I did so with the help of a few minutes googling (I didn't even know there was a DIN test when the report plopped through my letter box) and some simple back-of-the-envelope calculations. But we are expected to believe that no-one in Cannondale's Experimental Stress Analysis Lab was capable of this much. Although they claim to have read my website, they must have stopped before they even got to the "executive summary":

"There are two main aspects to the failure.

• The disk brake generates a massive force largely downwards in the direction of the open fork ends. The friction of a quick release skewer is often not sufficient to stop the axle slipping down in the dropout slot.  This is explained in more detail here,
• The QR is initially restrained by the retention lip on the fork (assuming it is present), however over time the slipping of the quick release leads it to unscrew, which is described here. Once it has unscrewed enough, it can be forced over the retention lip and the rider will crash. "
  

I thought the use of the term "massive force", together with supporting calculations, made it pretty clear!

Further information from Test #2

In fact, it gets worse when we look at "Item #2 does the disc brake force cause the wheel to come out of the dropouts?" This is a rather badly designed test which achieves little although it does highlight one failure mode which I will not discuss here but will refer to on the main page. What is most interesting about test #2 is not the test itself (which I do not intend to go over) but the measurements of an instrumented bicycle which was ridden briefly by a test rider. The only measurements which are referred to are those from a torque measuring device on the front wheel.


So, they obtained a measurement of a massive 235 foot-pounds of torque (which works out at 950N deceleration force). This is higher than the standard 0.6g deceleration that is generally considered maximal (to avoid the rider pitching over the handlebars), but this was an instantaneous peak measurement rather than a steady braking force. The torque meter was calibrated against direct load application to the wheel rim, and appeared to be accurate. This torque was by no means a freak value, wth 200 foot-pounds achieved several times in a very short test ride in a car park (real-life off-road riding was actually less stressful on the brake). At this torque, the downward force at the brake pad is a massive 3800N, far far greater than the estimates I used in my calculation, and almost 7 times greater than in Cannondale's test #1. So not only was their first test wholly inadequate, but they published data which proves this point just a few pages further on in the same report, apparently wthout realising the conflict.

Summary

Cannondale's test combines a very weak braking force with a very large downward push on the handlebars. The net effect is that there is a negligible downward force on the left hand side of the hub relative to the fork end, which is more than an order of magnitude smaller than the force that can reasonably expected in normal riding. By design, this test could not possibly have generated a failure.

A meaningful test of the skewer loosening would involve

1 A realistic braking force - to equal and exceed the braking force that Cannondale actually measured
2 A realistic weight on the front wheel (perhaps the 275 pounds figure is not so far off, when combined with the torque value of 235 foot-pounds, but it certainly requires that a very large braking force is applied)
3 A range of skewer tensions from tight to moderately tight to perhaps a little looser than recommended (I'll note in passing that Cannondale's procedure of 20 pounds of force on the lever end is not a 'light' closing force, it is a normal one).
4 A variety of skewers of different design
5 A range of disk rotor sizes
6 A number of different forks
7 A roller with significant bumps - say 3" high, with sloping tops to generate some out-of-plane forces
8 A check for skewer slippage as well as actual unscrewing

Cannondale's test doesn't begin to address the issues. I've asked them to justify their experimental design and to reconsider their statement that "At this time there are no reasons to believe anything is missing or over constrained in this test." Any response will be published here.