Edge Retention Testing and Analysis

cliff is now using multivariate statistics?

i guess he has come a long way from the days when he didn't even seem to understand significant digits or p-values

i wonder what kind of academic appointment he now holds?
 
Isn't that essentially backwards? Initial sharpness should be the same, when you change steels to get different wear resistance, toughness, grain & carbide size, etc., shouldn't that be affecting cut ratio later in the test? I don't see how, if I hand sharpened a FFD2 blade next to an S90V blade, that one would be nearly twenty times sharper than the other right off the bat, and that the performanc gap would actually decrease by such a large margin.

I must be looking at this wrong, but I would expect a higher ratio later, meaning to me that the knife would cut longer. I understand the curve is moving right to left, I don't understand why the ratio is decreasing as more media is cut.
 
hardheart said:
Isn't that essentially backwards? Initial sharpness should be the same, when you change steels to get different wear resistance, toughness, grain & carbide size, etc., shouldn't that be affecting cut ratio later in the test? I don't see how, if I hand sharpened a FFD2 blade next to an S90V blade, that one would be nearly twenty times sharper than the other right off the bat, and that the performanc gap would actually decrease by such a large margin.

I must be looking at this wrong, but I would expect a higher ratio later, meaning to me that the knife would cut longer. I understand the curve is moving right to left, I don't understand why the ratio is decreasing as more media is cut.
Click to expand...
That's part of the mystery of this test. It's a very interesting result, but doesn't match "common sense".
 
You guys don't have knives that get sharper than others with the same sharpening methods?
 
yeah, but the test shouldn't be run that way. I could sharpen one knife as I normally do, and take another one and scrape it on the driveway. I'm starting at two different sharpnesses then, and results would show it. How the edge got to whatever state it is in at the start of the test isn't important, comparing the steels on equal grounds is. This just further restricts the results-it can at best be said that when sharpened the same way, but not to the same degree of sharpness, FFD2 outperforms S90V for slicing by a machine. Still a useful result, let's you get some idea of how much you would need to change habits when switching steels, but knowing what each steel does at different degrees of sharpness also helps. This is lacking in a lot of reviews, though.

I was thinking at first that maybe the problem was one knife was over sharpened, but really either or both could be over or under sharpened. So I don't even know if FFD2 is easier or harder to polish up or make toothy compared to the other steels tested. The numbers are still good, lots of rope cut, but tracking down the baseline is tough.
 
It is my personal opinion and experience that some steels get sharper than others, even when both are sharpened perfectly. Now of course things can be changed by different grit sizes and different sharpening methods for each but that is not what I'm talking about.
 
I've assumed all along that they did what they said - they made the blades the same geometry and same finish. I thought it was reasonable to believe that the number of passes on the machine guaranteed that the tested edges were fully developed at the stated angles, which goes along with "equal geometry" and equal REST values. I don't understand how this is so easily disregarded.

Sounds like multiple pics of each blade will be needed to defend against the burr or other edge problem issue.

Again, unless you think all steels have the same Ci(max), you need to decide whether a steel should get the benefit in testing of having a greater Ci(max).

Unless of course you want the blades/steels all at varying geometries and finishes to achieve "equal" testing.

If you only want to compare across equal ranges of C, I think this will give you a definition of edge retention that could be great for industrial cutters, but would suck for knife users.

Knifenuts want a steel that can be made sharper easier than other steels.
 
It's edge retention, so we have to know what sort of edge we are retaining. A few more runs would work for me. Something I would like to see in further testing is one where the same beginning values are measured, and then the edge retention compared from that point, if that's possible.

I don't know if these have the same max, or what that max is, they only tested for one finish at one angle. Again, I don't see how a difference as large as the one shown is possible if the edges are exactly the same, maybe it's because of waiting the first 20 cuts to measure the push cut again.

Is this the best FFD2 offers, or is there even better performance? I think that's a good question to ask.

We already have plenty of easy to sharpen steels. This test wasn't about that, it was about retention of the edge, not forming of it.
 
hardheart said:
Isn't that essentially backwards? Initial sharpness should be the same, when you change steels to get different wear resistance, toughness, grain & carbide size, etc., shouldn't that be affecting cut ratio later in the test? I don't see how, if I hand sharpened a FFD2 blade next to an S90V blade, that one would be nearly twenty times sharper than the other right off the bat, and that the performanc gap would actually decrease by such a large margin.
Click to expand...

The initial cut ratio is due to the repeatable but unexplained phenomenon (more to come later on further testing) that at an equal REST sharpness, the FFD2 cuts more rope per stroke. The early cut ratio advantage is an artifact of the testing startup.

The cut ratio reaches a minimum at intermediate test values, then increases again. The combination of the minimum and the continued increase shows to me that edge retention is improved.

Thanks,

Carl
 
Broos said:
Carl,
Does this indicate the data correlation is bad, or does it just indicate the results do not fit acceptably into Cliff's theoretical model?
Click to expand...

Probably neither -- I think it indicates that there is a wide range of correlated a and b values that could be used to fit the measured data. C is uniquely identified, but a and b can be traded off in a correlated way to get an acceptable fit. Since neither a nor b is a physical parameter of interest, but just a fit coefficient, it doesn't give me heartburn that there is a range of a and b values.

However, if in doing a monte carlo simulation with various a and b values, you ignore the correlation between the value of a and b that is identified by the fitting, you'll get garbage simulated curves. BECAUSE OF THE CORRELATION, YOU CAN'T VARY a AND b INDEPENDENTLY IN A MONTE CARLO SIMULATION.

I suspect, but don't know, that treating a and b as independent may be where Cliff had some trouble with the fit parameters. But that's a wild guess on my part, because I don't have any real information about how he does it.

Thanks,

Carl
 
Well, I'm off with the scouts to do some field testing of knives, rather than lab testing.

I'll be away from the computer for 2 weeks, so don't think I'm just running and hiding.

I'll be in contact when I get back.

Carl
 
db said:
It is my personal opinion and experience that some steels get sharper than others, even when both are sharpened perfectly. Now of course things can be changed by different grit sizes and different sharpening methods for each but that is not what I'm talking about.
Click to expand...

I believe (and will post a description of why I believe it when I get back) that the harder the steel, the smaller the burr formed at a given edge angle, and therefore the sharper the blade. However, I have not yet obtained physical proof of this in the form of measured burr widths.

Carl
 
cds4byu said:
Since neither a nor b is a physical parameter of interest, but just a fit coefficient, it doesn't give me heartburn that there is a range of a and b values.
Click to expand...

I think this is one of the main points of contention here. Cliff places much importance on these coefficients within his model and wants to use them to extrapolate performance.
 
How far is it of value to make inference of behavior as tested by machines on human behavior? The raw data generated in the inital experiment corresponds with the phyiscal model, so it is not of zero value.

The argument is that the force vectors upon the edge in human use is much more variable and would lead in faster edge degradation. If we accept that, is the difficulty then prescribing the relativity of faster degradation in comparing steels due to various factors that are largely missing in the machine test (ie edge stability to a high degree of sensitivity?)?
 
wow, this is a very intense thread, lol, but also VERY interesting! By far the best thread I read in a while!!!
 
kel_aa said:
How far is it of value to make inference of behavior as tested by machines on human behavior? The raw data generated in the inital experiment corresponds with the phyiscal model, so it is not of zero value.
Click to expand...

Hello,
I must admit that my introduction to the idea that machine testing is invalid was here - Can you point me towards any test results out there where human testing yielded greatly different results than machine (CATRA) testing? Didn't the human results in this case for the most part confirm the machine results? Don't forget the FFD2 blade had a thicker edge than the other blades in Phil Wilson's test.

If lateral forces play such an important part in how a knife dulls in normal human cutting, then couldn't one person achieve completely different results than another, depending on how they cut? If the lateral forces play that important a role in edge retention wouldn't we see more differences when comparing edge retention test results? When multiple humans and CATRA all get similar results doesn't it tend to diminish the role this lateral force plays?

And even if lateral forces do play a critical role in the test results, for a benchmark test I would rather test using a known lateral force than an unknown and transient lateral force. With lateral forces during human tests being unknown, how will you be able to defend your results from hand testing against claims the test was intentionally biased by uneven cutting? The claims of bias made against these tests is a good example of what you are going to have to defend against.

Another point is that by minimizing the variables in a test methodology you will be more likely to be able to explain the results & build a theoretical relationship based on results. I would rather develop a better theoretical understanding of the roles that the downward force and the speed of the slice play before adding additional forces to the test. Then we could also help identify push cutting and slicing differences and similarities - I generally do not agree with a total disconnect between slicing and push cutting that seems to be assumed by some here, and think that push cutting type behavior is an important part of slicing.

Regardless of all this, I am not against hand testing, & think Goddard's test may be the best test we have - it is easy to perform, provides a good comparison of knives, or a good comparison of steels if an equal test baseline is used, and with some practice can yield very repeatable results. I would also think that having a nice even and consistent cutting stroke is critical in getting repeatable results with this test.

I will say I'm just thinking out loud, so feel free to let me know where you disagree!
 
We wouldn't get test results within 5% if we had lateral forces taking place. It’s not hard to hold the knife at 90-degrees to the cutting platform.

I think a variable that isn't talked about very much is the condition of the edge. When I read of tests that blunted the edge and where steel pulled out of a blade because of large carbides I can only think that the test blade had a wire edge on it. We take great care to prepare the true sharp edges for our tests.

I think I know of what I write. I've been teaching my sharpening class for thirty years at knife shows, hammer-in's, and schools and on hunting trips. Most knife users don't know a wire edge from a carbide. Some get it lined up and think they have a sharp blade; however it will not have the endurance cutting ability that it will with a true sharp, wire-free edge.

Wayne G
 
Wayne,
Thanks for your reply. I agree with you, but was just trying to explore the forces that are behind the contention made here that any machine test results are invalid to use as a reference to gauge edge retention in normal (human) use.

I do not disagree that there are some lateral forces encountered when humans cut. The issue is how important they are to edge retention testing. I suspect when you and Wilson test that these forces are very small and insignificant, and do not play much of a role in the results - this belief is made stronger by your results confirming the CATRA results. The way I read their objection is that ANY human test will have these forces, and that CATRA has none. And that the lateral forces are critical in how an edge blunts in normal human use.

I can't find any other instance where we do have CATRA results and 2 experienced human tests to compare. To my mind there has been more theoretical debate than is warranted given the small amount of data available, and regardless I am satisfied in defining and testing for edge retention without any theoretical basis - the results will tell me what I want to know. Having more of similar test results to compare would make me feel a lot better about making any theoretical assumptions.

Just fyi, here are the posts where this objection to machine testing were made:

Cliff Stamp said:
Buck does, the engineers report to Chuck but at a lay level. He was simply noting the performance ratios off of the CATRA data did not represent what happened when the knives were used by people. He specifically cited the blunting mechanism ignored and this mechanism is basic physics, edges will deform and chip under lateral loads when used by people. I cited a simple example to prove this with the razor blade.Contact him and ask for the actual data if you are interested. You can also see the data reported by Landes in his book where he talks about edge stability which is ignored by CATRA type testing.
Click to expand...

Cliff Stamp said:
No Chuck Buck did himself directly (from memory) "We gre bored waiting for the knives to blunt on the CATRA machine, but in REAL LIFE it is otherwise." He specifically mentions the effect of forces on the edge which blunt by a mechanism other than wear. He even notes the exact cause (correctly) and makes it public. This was years ago on the forums.

A human hand will have a much higher variance in the stroke vector (magnitude and direction both) for the cut path. This induces significant forces on the edge laterally. This was why such machine testing was quickly noted to be problematic in germany and other methods used to predict the performance of blades when used by people. Ideally, and this is the obvious part, you just use people (with as noted the correct methods to prevent systematic bias).
Click to expand...
 
I don't have a direct response to your thinking out loud, Broos. I too will engage in some of that:

From Test for measuring cut resistance of yarns, Hyung-Seop Shin, D. C. Erlick, D. A. Shockey:



The blade (Stanley utility blade, 2 um mean radius, 30 deg included) made one perpendicular cut through a yarn (made of 330 12-μm-diameter polybenzobisoxazole (PBO) fibers) of Zylon (a cut resistant material they were testing) material. The cut energy was 0.105 joule. Before the cut is completed, the edge deforms and expands to a radius greater than 5 um (evidenced by the fact that a harder ceramic blade of 5 um radius was outperforming it).

I think you can imagine what work even on this scale (1/10 of a J) directed at the edge not in line with the axis of the blade can do.
 
You must log in or register to reply here.
Back
Top