AlanLeeKnots pull tests on bowlines

[mcjtom]

Alan Lee has several videos of pull tests of various loops.  I'm only interested in regular (#1010) and Scott's bowlines.

e.g. https://youtu.be/FPqvrFKpZrU

There are at least 3 ropes used: blue, white, and yellow.  What is the type/diameter of each and have Allan pulled them to failure without knots, just wrapped on larger bollards at both ends?

What is the unit of the tension force displayed?  I'm pretty sure it's in kilogram-force, but want to double check.

[mcjtom]

Alan Lee did a series of 12 pull tests where he tied the regular (#1010) bowline at one end and the Scott's lock bowline at the other end of a length of rope and pulled it to failure (recorded how the knots behaved when stressed and then looked at the remains).

First test:
https://youtu.be/0RViCfR1_GM

I took the info off the videos and did a little prediction magic to guess what the next, similar yet untested rope could do in the next test and what it likely means for either of those bowlines when tied on similar material.

Out of 12 observations in the sample, the rope failed 7 times (58%) at the regular bowline and 5 times (42%) at Scott's bowline. 

Whether the rope broke at one knot or the other, all ropes failed at roughly similar tensions / kN, n = 12:

m = 21.7, sd = 0.63
MIN = 21.0, MED = 21.5, MAX = 22.8

That makes their relative ultimate strength indistinguishable with this sample size (i.e. for all we know, the knots are equally strong).

There is some 8% chance that the next rope tested will fail at less than 21 kN.

There is some 1% chance that the next rope tested will fail at less than about 20 kN and likely less than 0.1% chance that it will fail at less than around 19 kN.  It could break equally likely at either knot.

There is no difference between the ease of untying the surviving knots - all easy.

To me it means that the regular bowline backed up with double overhand may be less compact, use more tail and look uglier but is equally strong and practical and at least as safe as the Scott's bowline. 

It probably also means that the idea that packing more rope strands in the choking loop makes the bowline stronger, safer, or easier to untie may not hold much water.

The tests don't show that Scott's bowline is any stronger or easier to untie than the regular bowline.  As to security, l suspect that the regular bowline backed by double overhand has (even) better chances of surviving accidental ring/cross loads, cyclical loads, or shaking/pulling lose owing to the stability of its backup dbl overhand knot.

I just want to add that I still love Scott's bowline :⁠-⁠)

[mcjtom]

And the real question with all these is: if you were to anchor a rope, expecting a drunk bear coming to try to untie it, which of the two would you use? :⁠-⁠)

[Dan_Lehman]

Allan Lee did a series of 12 pull tests where he tied the regular (#1010) bowline at one end and the Scott's lock bowline at the other end of a length of rope and pulled it to failure (recorded how the knots behaved when stressed and then looked at the remains).

?!  Seems a wrong-headed way to go --series of A-vs-B testing,
rather than (six each of) A-vs-A, B-vs-B (or just a single knot vs.
anchorage, 2 B *pure*).  One is assured of six survivors of each,
and six broken specimens (possibly the survivors are injured).

Note that with so short a specimen, it's possible the results
are tainted (or particular to) by uneven tensions in the S.Parts
where with longer runs such uneveness would have chance to
be ameliorated in effect.  --but then one is burning more rope
per test!  <sigh>

Whether the rope broke at one knot or the other, all ropes failed at roughly similar tensions / kN, n = 12:

In some testing of the BWL vs. Fig.8 EK,
EStar (Evans Starzinger) tested two sets of ten
specimen also in A-vs-B (BWL-vs-F8) arrangement.
(didn't ask, but I wonder if after initially thinking
he'd get a range of BWL failures --i.e., presuming
the F8 always to be stronger--, and getting some
early BWL victories, he THEN decided to continue
with the A-v-B set-up?)
Interestingly, in the two sets-of-ten I saw posted,
both were exactly 5:5; and the break forces
for the BWL just slightly lower than for the F8, around
82% (he had at some point tested pure rope, I think,
but had found manufacturer quoted figures to be
pretty accurate).

To me it means that the regular bowline backed up with [STRANGLE is the knot]
 may be less compact, use more tail and look uglier but is equally strong and practical
and at least as safe as the Scott's bowline.

Then you're being too quick & low-resolution.
Scott's Lock might have advantage on keeping
it all *together*, whereas just having tied off
the Tail doesn't ensure that the BWL itself
stays in shape (vs. loosening).  YMMV.
One can tie a dbl.OH in form not of Strangle
but Anchor Bend and orient it so that its Tail
can be tucked back through the BWL's nipping
turn, for a 3rd diameter.

It probably also means that the idea that packing more rope strands in the choking loop makes the bowline stronger, safer, or easier to untie may not hold much water.

I might agree (foolishly) except that that's my theory
you're seeing challenge to !!    )-:< !!
Seriously, the point is there, BUT LOOK AT THE EXACT
GEOMETRY, nevermind merely counting diameters.
(Consider : around the 3 dia of ulility poles aligned vs. 3 in a triangle --> ... vs. .:. )

In both of these knots, the S.Part's turn runs pretty
hard into the Returning Eye Leg.  Whereas it CAN
be made (in some knot variations) to compress hard
into a *limp* part such as the Tail :: better curvature,
perhaps, AND more of a heat sink, maybe?!

As to security, l suspect that the regular bowline backed by [a Strangle knot]
has (even) better chances of surviving accidental ring/cross loads, cyclical loads,
or shaking/pulling lose owing to the stability of its backup [Strangle] knot.

Depends how you put the strangle.  Ring-loading should pull
much S.Part through even though the BWL Tail is Strangle-ing
the Returning Eye Leg; such feed of material could be a concern.
Well, then ditto for Scott's; both stem the bad-Lapp-Bend spill
of U-fold tail (Strangle'd or interwoven as that is, resp.).


--dl*
====

[mcjtom]

Depends how you put the strangle.  Ring-loading should pullmuch S.Part through even though the BWL Tail is Strangle-ing the Returning Eye Leg

Which one is 'The Returning Eye Leg'?

I imagine the side of the loop closest (along the rope) to the tail?  Or the same way the HowNot2 fellow has tied it in the knots video?

[mcjtom]

Attached are the Tukey's boxplots to illustrate better the distributions of Allan's tests and their subsets. 

A boxplot elements show 5-number summary of a dataset: MIN, Q25, MED, Q75, MAX.
It's a fantastic visual tool to rapidly assess what story is being told by the data with little calculations.  A number of distribution parameters and diagnostics can be directly read off a boxplot and some inferences can be easily constructed.

In this case the plots further illustrate that those two knots can't be told apart by their breaking strengths (and also that even quite larger sample size is unlikely going to change this statement).

[mcjtom]

Interestingly, in the two sets-of-ten I saw posted,
both were exactly 5:5

Actually, if those knots are indeed no different in strength, it's almost twice as likely to observe 5:7 or 7:5 splits than 6:6 in 12 samples.  i.e. observing 7:5 splits (either way) in 12 tests is super consistent with the hypothesis that the knots are no different

You'll see it if you keep tossing a fair coin 12 times in a row and record ratios of Heads to Tails for each run.  It follows the binomial distribution.

[mcjtom]

?!  Seems a wrong-headed way to go --series of A-vs-B testing,
rather than (six each of) A-vs-A, B-vs-B (or just a single knot vs.
anchorage, 2 B *pure*).

This may not be such a bad idea if you want to detect small differences in the breaking strength of one knot vs. the other.  It is mostly because knots at two ends of one length of rope in a single test are likely subjected to more similar conditions than two knots, on two lengths of rope, subjected to a separate test each.

[mcjtom]

In some testing of the BWL vs. Fig.8, EStar (Evans Starzinger) tested two sets of tenspecimen also in A-vs-B (BWL-vs-F8) arrangement

.
These are probably the tests you're quoting:

https://forums.sailinganarchy.com/threads/rope-knot-splice-load-testing.154025/

[SS369]

Alan Lee has done some wonderful testing and investing and we should be grateful and appreciative for his time and efforts!

I think the pull to destruction testing has been done enough and we can glean sufficient data from the myriad of testers to be able to choose and size our knots and ropes for their intended uses. How many times have we actually brought our rope to destruction in use?

Drop testing, cyclic loading for security and ease of untying represents real world use, IMO.

Length of test specimens could shed some light on some things, but that will be dependent on rope materials and construction and similar dressing of the knots to be tested.

Knots and ropes used In-the-Wild see more abuse, therefore won't be represented in Pull-till-rupture tests. So, are these types of tests proving much to the layperson?

Real world use is dependent on so many factors; too many to describe! And not many are going to always use a brand new rope each time.

SS

[mcjtom]

And the real question with all these is: if you were to anchor a rope, expecting a drunk bear coming to try to untie it, which of the two would you use? :⁠-⁠)

The bear is coming - which one will you tie? :⁠-⁠)

[mcjtom]

The best appreciation of Alan's work may be using his results to tell the story and try to answer a practical question: is there a reason to avoid one of those two backed-up bowlines?  There is nothing to prove, just rationally explore.

The story seems to be that the two loops are no different in several important departments and thus the choice is yours - but one wouldn't know it beforehand.

There seem to be no difference in the ease of untying them after heavy loads.  That's good to know.

They are both strong enough, and equally at that.  That's also good to know.  Calling for a moratorium on breaking strength tests may be premature as those results are also quite useful in predicting how rope and knots will cope with e.g. drop loads.

https://itrsonline.org/tproduct/1-634887464371-analysis-of-impact-force-equations

http://web.mit.edu/sp255/www/reference_vault/second_order_rope_fit.pdf

Both bowlines don't seem to slip much under heavy regular loads.  That's reassuring.

As for abusing the knots by subjecting them to cyclical loads, accidental ring/cross loads, flogging, or a general mayhem (think of a bear) on new or old ropes, there is little data, but it seems reasonable to assume that a regular bowline with a strangle backup (especially with the strangle tied on the far side of the bowline loop - thank you Dan) is at least as likely to survive the abuse as is the Scott's lock variant.

Knowing all that, let's go back to the bear question...

[Dan_Lehman]

Of an eye knot, in the flow along material from
outside of the knot, first into ... ::

S.Part delivers full force into knot;
which is opposed by the continuation
of it (ultimately) into the
Outgoing Eye Leg,
and that 1st part of eye's return as
Returning Eye Leg,
leading ultimately to the Tail.

?!  Seems a wrong-headed way to go --series of A-vs-B testing,
rather than (six each of) A-vs-A, B-vs-B (or just a single knot vs.
anchorage, 2 B *pure*).

This may not be such a bad idea if you want to detect small differences
in the breaking strength of one knot vs. the other.
It is mostly because knots at two ends of one length of rope in a single test
are likely subjected to more similar conditions than
two knots, on two lengths of rope, subjected to a separate test each.

But it's at risk of misc. knot changes affecting the other,
same knots or different, vs. single-knot specimens.
(I got a long rejection of this hoped-for testing_cost-saving scheme
from the Czech superstatistics fellow (whose equations I cannot even
put into order they are so complex --not to say doing so would at all
help me!  (BUT I have figured that bit about odds of 50/50 for
likely output --that, no, a slight imbalance is more likely). )

Still, I'd like to see the CzechSlovaks do a thorough testing in the
*pure* way and in the 2-specimens way --which, yes, SHOULD
have a (predictable, I'd think) bias to posting lower values,
since it's the weaker one's break that occurs and is recorded
(one would need both to be strong for a strong knot to register,
increasingly less likely moving away from the mean).

'twas my hope that such a lower-force bias would be able to be
statistically *rounded* towards the *pure* values; and it
otherwise is understating strengths which is a safer thing
to do for use than overstating.


OTOH, as I put to him (unanswered), which of these ways
more resembles actual usage?  --where one DOES have knots
at both ends.  How is one to interpret *pure* results in doing
*real* (impure!) rigging?!
But, irrespective of right-on force measures,
there is the benefit of seeing behavior, of where
the knot breaks.


--dl*
====

[mcjtom]

I got a long rejection of this hoped-for testing_cost-saving scheme
from the Czech superstatistics fellow...

If this is the paper, they may be nicer to you if you call them CzechoSlovaks... :⁠-⁠)

[mcjtom]

Still, I'd like to see the Czechs do a thorough testing in the
*pure* way and in the 2-specimens way --which, yes, SHOULD
have a (predictable, I'd think) bias to posting lower values,
since it's the weaker one's break that occurs and is recorded
(one would need both to strong for a strong knot to register,
increasingly less likely moving away from the mean).

I think you are correct: taking a minimum of two random observations from a distribution will result in a distribution with lower mean, lower variance, and different shape than the original.  You can see this by running a Monte Carlo simulation of the situation (analytical solutions are complicated...).  The difference may not be great, but it's not trivial.


p.s. Figure 1 in this paper illustrates the point and shows the distribution of maxima of two observations from normal distributions with the same mean of zero but different spread.  The distribution of minima will be their mirror image.


https://www.semanticscholar.org/paper/Exact-Distribution-of-the-Max-Min-of-Two-Gaussian-Nadarajah-Kotz/956f307c3b63528c88f5858aef1716cd402348fd