Edge Retention Testing and Analysis

[cds4byu]

There's been a lot of discussion on the Friction Forged Blades thread about testing and analysis. The discussion all started because there were questions raised about the testing of the FFD2 blades and whether or not the claims of improved performance were defensible.

We've now had enough discussion to start to get into the details of testing methodologies, data modeling, data analysis, and deciding whether or not results are significant.

In order to separate this discussion from the FFD2 blade discussion, I decided to start a new thread, and I'll invite Cliff over to participate.

Thanks,

Carl
-------------------------------
It is not necessary to believe things in order to reason about them
It is not necessary to understand things in order to argue about them.
- P.A. Caron de Beaumarchais, French Author, 1732-1799

 

[Wayne G.]

Carl,
Thanks for the new thread. I posted this on the original CATRA thread before I realized you had this one started and I hope you don't mind me postiing it here too.

Legitimate tests are not possible without starting the knives on an equal basis. The method that follows has been followed from the start with my testing. The test knife that did 300 cuts here had a factory edge on it but the wire or burr as some like to call it hadn't been removed. It was removed with the fine India and that put it on an equal basis with my other tests.

<<<<>>>>
Sharpening friction forged blades will not require anything other than a good combination stone like the India/Crystalon by Norton. The medium grit Crystalon (silicon carbide) side of the stone will remove steel from the hardest blade and the fine India (aluminum oxide) side will bring the edge to hair shaving sharp. The fine side is also used to take the wire edge off. The FFD2 at 67 Rc takes some fine cuts at a steeper angle to remove the wire, just as with softer blades.
The wire edge is ignored by many sharpen systems. However, to get the full potential out of any blade it has to be true sharp (wire edge removed) not false sharp with the wire edge lined up. Meat cutters and cooks work with the wire edge on the knife and keep it perfectly lined up with a steel. While this edge cuts veggies and meat like they were butter it is a defective edge for a hunting, or working knife for utility use.
I use the household strength Simple Green as a sharpening fluid, it gives an excellent cutting action with the harder blades and keep the stones clean and free cutting.
I’ll be using the India/Crystalon to sharpen a dull FFD2 blade at the sharpening class this Sunday at BLADE Show in Atlanta. If you’re in the neighborhood drop in and see what’s going on.

 

[Broos]

Carl,
Just an idea - do you think it would be helpful & possible for you to post an outline of the procedure/methodology you used - it would be helpful to look at here, and any improvements could be added to the existing methodology.
Thanks

 

[cds4byu]

You would do that in the initial stages to verify the robustness of the of the fitting, I do that as well, but I was speaking of using monte carlo methods to generate pseudo data sets to show the results that the noise will have on the inferences.

OK, let me ask the question a different way. How are you generating your 100 data sets? Are you using your fit model and adding normally distributed noise to the data, then rerunning the fit?

If so, how do you decide the standard deviation of the noise?

If not, would you please explain to me the algorithm to generate the 100 data sets?

You generally only do this when it is difficult to calculate the uncertainties directly from the correlation matrix.

What is "the correlation matrix"? What data are you doing the correlation with, and what statistical techniques do you use for the correlation?

You clearly don't understand my style of coding, that would be a cruel thing to force on someone.

I've been debugging and fixing other people's code since 1981. I'm perfectly willing to look at your code. Of course, if you don't want to share your source code, I'll understand. But I can't properly evaluate your method if I don't understand the algorithm.

The algorithm is actually quite simple, it is just brute force methods on curve intersection. For each point on one curve you find the point on the other curve which bound it, i.e. y1(x1) < y2(x) < y1(x2). Start off assuming the same point and then just go up or down as required to find the bounds. Now do a linear/spine approximation to get the approximate intersect value, i.e., y1(x')=y2(x).

I don't think you ever explained the curve you're doing this on, but I think I figured it out. I think you're working on a plot with the y axis being the CATRA ERT sharpness, or thickness of media cut per stroke. The x axis is either number of strokes or total media cut. I've seen comments from you that say I should only use the number of strokes on the x axis, because to use total media cut on the x axis I'm "putting a dependent variable on the x axis" which "makes a linear problem non-linear" and leads to "difficulties in curve-fitting" (my quotes are approximations, not literal quotes). But in your next paragraph, you call your x ratio the cut ratio, which means it's total media cut, not number of strokes. So I'll go with a working hypothesis that it's total media cut.

So my assumption is that y2(x) is the media cut per stroke (inches/stroke) for blade 2 after x inches of total media cut. And x1 is the smallest total media cut for blade 1 when the cut per stroke of blade 1 is less than y2. And x2 is the largest total media cut for which the cut per stroke is greater than y2(x). Is this correct?

But I'm not clear on how you do the "approximate intersect value" to get x', which is apparently an estimate of the total media cut for blade 1 when its cut per stroke (sharpness) is equal to y2(x). Is it the average x1 and x2?

This ratio x/x' is then the cut ratio, which means quite simply how much more material one blade can cut than another to a given sharpness. That is what most people would be interested in (the sharpness ratio at a cut length is of course trivial to calculate y2(x)/y1(x)).

Again, just to clarify, all of the plots on your model paper http://www.cutleryscience.com/reviews/model.html
show number of strokes on the x axis. But I have seen plots in some of your posts on bladeforums that show total media cut. Do you use total media cut for your x axis?

Now to take into account the noise in the data you do a montecarlo simulation on the raw data to thus produce a set of cut ratio values and you can present the mean values of these.

Could you clarify the details of your monte carlo simulation here?


Thanks,

Carl
-------------------------------
It is not necessary to believe things in order to reason about them
It is not necessary to understand things in order to argue about them.
- P.A. Caron de Beaumarchais, French Author, 1732-1799

 

[db]

CDS4 good luck in getting answers to your questions. How about good old spray paint for a blade coating? A good color would be hot pink for the test knives.

 

[cds4byu]

Carl,
Just an idea - do you think it would be helpful & possible for you to post an outline of the procedure/methodology you used - it would be helpful to look at here, and any improvements could be added to the existing methodology.
Thanks

Sure, I'd be happy to.

I think the methodology is clearly explained in our paper, available at TMS 2007 Paper

In summary, here's what we did:

1. Blade Preparation:
a) Heat treat ordinary blades, FF FFD2 blades
b) Grind blade geomety with either surface grinder (early) or Seipmann grinder (later)
c) check hardness to make sure it's where we wanted it
d) sharpen blades with 600 grit diamond belt, using fixture to maintain edge geometry. Same number of passes, same pressure.
e) Remove wire edge (burr) with cardboard wheel and CrO compound

Testing.

1. Measure blade sharpness, using REST tester to measure push cutting force. REST peak force was measured twice, and a third time if an outlier was found in the two original data points. The presence of an outlier was determined using subjective criteria. Data on all REST tests reported in the paper are available on the website.
TMS 2007 Data

2. Wear blade by placing it on a machine similar to a CATRA ERT tester, but with hemp rope rather than silica impregnated card stock. Wear was caused by a slicing stroke with a fixed amplitude and a fixed velocity. Total media cut was recorded on each stroke. A pre-determined number of strokes (20 in the early stages of testing, 40 in the later stages of testing) was performed on the blade wear machine.

3. Measure sharpness on REST machine, as in step 1.

4. Test to see if blade will still shave. (This was early in our testing process; lafter obtaining data we used a specified REST value, rather than ability to shave as our stopping criteria.)

5. Repeat steps 1-4 until we had at least three consecutive shaving tests that failed (or three consecutive average REST measurements above 3.0).

6. Plot the data according to CATRA ERT standards, as well as according to a modified CATRA ERT standard, with sharpness determined by REST force, rather than by cut per stroke (which is the CATRA measurement).

Hope this helps,

Carl
-------------------------------
It is not necessary to believe things in order to reason about them
It is not necessary to understand things in order to argue about them.
- P.A. Caron de Beaumarchais, French Author, 1732-1799

 

[Ravaillac]

d) sharpen blades with 600 grit diamond belt, using fixture to maintain edge geometry. Same number of passes, same pressure.

Actually, that sounds "conceptually wrong" since you want to test "edge retention", not "edge retention after X passes".

Not sure how this should be done but you should start with blades of equal sharpness.
On a more general note, rather than "equal sharpening", you should try something like "competitive sharpening": each blade sharpened in the way it's steel works best: you would have something like "FF at its best is better than S90v at its best", not "FF is better than S90V... in a X degrees edge after Y passes...".

 

[cds4byu]

CDS4 good luck in getting answers to your questions. How about good old spray paint for a blade coating? A good color would be hot pink for the test knives.

I'm afraid hot pink would put a total negative bias on our testers. Too feminine for real outdoorsmen. Now, if we could paint a camo coating ...

 

[sodak]

CDS4 good luck in getting answers to your questions. How about good old spray paint for a blade coating? A good color would be hot pink for the test knives.

A nice mauve might be nice...

 

[cds4byu]

Actually, that sounds "conceptually wrong" since you want to test "edge retention", not "edge retention after X passes".

Let me clarify. The passes were SHARPENING passes, and were used to raise the wire edge, which indicated that the sharpening was sufficient. The wire edge was then removed. No abrasives were used to dull the blades.

Not sure how this should be done but you should start with blades of equal sharpness.

Our sharpness test was a PUSH sharpness test, not a slicing sharpness test (the CATRA REST test). All the tested blades had approximately equal REST sharpness values at the beginning of the test.

I've gone back and forth over whether blade performance (speaking in the generic term, not the specific term "edge retention") should be evaluated by starting with equivalent sharpness. It's my observation that different steels have different ultimate sharpness limits. If steel A can be sharpened to a sharper edge than steel B, why shouldn't it get credit? On the other hand, getting blades really sharp takes as much skill as knowledge, as anybody who's tried to copy Wayne Goddard's sharpening techniques will appreciate. And having test results depend on the skill of the operator is not good practice. That's why we settled for a constant sharpening medium, a sharpening fixture, and a quite repeatable process, as evidenced by consistent sharpness of test blades after sharpening.

On a more general note, rather than "equal sharpening", you should try something like "competitive sharpening": each blade sharpened in the way it's steel works best: you would have something like "FF at its best is better than S90v at its best", not "FF is better than S90V... in a X degrees edge after Y passes...".

I agree with this statement in principle, but think it's unwieldy in practice. (And I won't refer to FFD2 or S90V in this thread, because I want this thread to be about testing practices, not blade materials.) It seems ideal to test the best possible performance of steel A compared to the best possible performance of steel B.

The hitch is that nobody knows the best possible performance of either steel A or steel B. And if I'm a knife maker who believes steel B is better than steel A, any time I spend trying to find the best possible performance of steel A is wasted. And given the number of different steels, and different heat treating processes, etc., I'd spend all my time trying to determine the optimal performance of some steel I'm not interested in.

I can think of two ways around that. The first is that, when you're comparing steels, you do a range of experiments. In Diamond Blade's case, it was a bunch of steels at the same geometry. In other cases, it might be one steel at a bunch of different geometries. Then you evaluate the results, and explain them as clearly as possible. If someone disagrees with your results, they're free to run the same tests, and see if they get the same results. That's how science works.

The second way is to have a standard reference blade. If it were up to me, I'd choose a razor blade produced by a reputable manufacturer, such as Olfa. It should have constant geometry and constant steel properties. There would need to be proof that such a blade was, in fact, consistent. Then, once there is a consistent reference blade, every manufacturer could compare their blade to the reference blade. And somebody who was interested could compare the blades from different manufacturers by comparing their performance relative to the reference blade.

Right now, I don't think either one of these methods is practical. So I think the best approach is for testers just to be completely open about their test methods, data, and analysis.

Carl
-------------------------------
It is not necessary to believe things in order to reason about them
It is not necessary to understand things in order to argue about them.
- P.A. Caron de Beaumarchais, French Author, 1732-1799

 

[Broos]

Actually, that sounds "conceptually wrong" since you want to test "edge retention", not "edge retention after X passes".

Not sure how this should be done but you should start with blades of equal sharpness.
On a more general note, rather than "equal sharpening", you should try something like "competitive sharpening": each blade sharpened in the way it's steel works best: you would have something like "FF at its best is better than S90v at its best", not "FF is better than S90V... in a X degrees edge after Y passes...".

To my knowledge there is not an absolute definition of edge retention without it being defined by a test method. In other words, our understanding of edge retention is limited & defined by what test we are going to use to measure it. The "after x passes" limitation is placed on that statement because of this, just to be absolutely correct.

And if we run the test again, but have different geometry and/or edge angle for each steel, who is going to provide the information on what angle and what sharpening procedure to use for each steel? There is no cutting model or solid understanding available that can quantify this to the extent of putting numbers to it and still have any assurance the numbers aren't getting pulled from somebody's posterior. Maybe Cliff can elaborate on this...

Better (I think) is to do what I suggested earlier in this thread. Perform the tests with all steels having the same geometries and same sharpening procedures for the different steels, then do again but vary the baseline edge angle and sharpening procedure.

If we do this, we can come up with some real data on the steels compared at varying geometries and edge finishes, and at which geometry and edge finish each different steel performs best (for the given tests). I would love this, just to see whether the results confirm the massive amount of conjecture that goes on here about it.

 

[Ravaillac]

First, thanks for answers.

Our sharpness test was a PUSH sharpness test, not a slicing sharpness test (the CATRA REST test). All the tested blades had approximately equal REST sharpness values at the beginning of the test.

What only astonishes me is that blades with same geometry (plus you added roughly same REST sharpness value), have so different "amount of material cut" on their first stroke, where edge retention shouldn't (probably) show yet.
-sorry to stay focused on that ERT graph, but I want to fill the blanks.

That's why we settled for a constant sharpening medium, a sharpening fixture, and a quite repeatable process, as evidenced by consistent sharpness of test blades after sharpening.

This seems actually a wise choice, but reasonable suspicion (or at least hypothesis that should be taken into account) is that method chosen might inherently favor one of the blades over the others.

I agree with this statement in principle, but think it's unwieldy in practice.

Of course that was kind of utopic.

The hitch is that nobody knows the best possible performance of either steel A or steel B.

While, this is true "in theory", knowledgeable people often good hints, many of them being quite consensual. Alternatively, once your "actual test" (the cutting part) is protocol is fixed, you could also try various standard method, and keep best result.

 

[cds4byu]

To my knowledge there is not an absolute definition of edge retention without it being defined by a test method. In other words, our understanding of edge retention is limited & defined by what test we are going to use to measure it. The "after x passes" limitation is placed on that statement because of this, just to be absolutely correct.

I think you're right. But I think it's even worse than that. We don't have a standard definition of sharpness that everyone agrees on.

Some people use CATRA Edge Retention sharpness, media cut per stroke. This metric is apparently embedded in ISO 8442.5 (If I remembered my ISO number correctly). I'm fine with this metric early on in the ERT test, when the blade is still sharp. But later on, when the edge is gone but high-carbide steels have little hard carbide teeth poking out of the edge of the blade, I don't think this definition is meaningful.

I like the CATRA REST tester. In my opinion, it's closer to my perception of sharpness than anything else I've seen. But I also understand that it's just a measure of push cut sharpness, not slicing sharpness. And I understand that an edge with "microteeth" will perform better in a slicing contest than a polished edge, but a polished edge will have a lower REST force (or higher REST sharpness).

I'd love it if we could come up with a definition of sharpness, but I don't think we will. So the next best thing is to come up with accepted standard methods of performing different tests, whether it be CATRA ERT, CATRA REST, Wayne G. rope slicing, or any other proposed test. Then the user of the knife can decide which test best matches his intended use, and make an informed decision.

I don't think we're completely there, but I do see some people documenting their test procedures, which is a great start, IMO.

Carl

-------------------------------
It is not necessary to believe things in order to reason about them
It is not necessary to understand things in order to argue about them.
- P.A. Caron de Beaumarchais, French Author, 1732-1799


P.S. Broos: Don't worry about stepping on my toes. I love a good technical discussion where I come away having learned something because I found out my previous knowledge was wrong!

 

[cds4byu]

What only astonishes me is that blades with same geometry (plus you added roughly same REST sharpness value), have so different "amount of material cut" on their first stroke, where edge retention shouldn't (probably) show yet.
-sorry to stay focused on that ERT graph, but I want to fill the blanks.

I agree. That was the biggest surprise for me, too. It appears to be repeatable, however. I think that area is one of the most promising areas for future research to learn more about what's going on. I hope to investigate more in the future, e.g. with before and after micrographs of both blades, etc.

Carl
-------------------------------
It is not necessary to believe things in order to reason about them
It is not necessary to understand things in order to argue about them.
- P.A. Caron de Beaumarchais, French Author, 1732-1799

 

[Wayne G.]

The reason for the fine India edge used on my endurance cutting tests is that it is the best all around working type edge for a knife. It is an edge that the average user can duplicate. Factory edges set up with fine belts and then polished can rarely be duplicated by the knife buyer.
The micro tooth edge from the fine India slices through a 1” rope yet won’t push through he same rope. Buff the edge until all the micro teeth are gone and it will no longer slice through the rope but will push through it.
A strop is made by gluing a strip of leather onto a strip of wood. The leather is loaded up with green or white buffing compound. Two strops on each side will refine the micro tooth edge and create a good compromise edge. Strop it too much and the result is the same as buffing…. The edge loses its slicing ability but pushes real good through the stuff being cut.

 

[cds4byu]

The reason for the fine India edge used on my endurance cutting tests is that it is the best all around working type edge for a knife.

I certainly wouldn't disagree with you, Wayne.

What ideas do you have about how to quantitatively measure this sharpness?

Carl
-------------------------------
It is not necessary to believe things in order to reason about them
It is not necessary to understand things in order to argue about them.
- P.A. Caron de Beaumarchais, French Author, 1732-1799

 

[Broos]

Carl, I think an step #7 is needed in your methodology outline - add that the test results were then compared to real world testing done by Wayne Goddard. This is an important step.

I still think to run tests using anecdotal evidence (hints) to establish each steel's optimum geometry, edge finish, and sharpening procedure is inherently flawed, and would yield results you could not defend.

I would like to have Carl running tests at different geometries and different finishes to satisfy our curiousity for years, but unfortunately it would probably not help sell one of those knives he invented, not to mention taking time away from him thinking of better ways to make an armored vehicle.

Is it possible to reach an agreement that having the same geometries and same edge finishes for the baseline of the testing is correct?

 

[sodak]

I certainly wouldn't disagree with you, Wayne.

What ideas do you have about how to quantitatively measure this sharpness?

Carl
-------------------------------
It is not necessary to believe things in order to reason about them
It is not necessary to understand things in order to argue about them.
- P.A. Caron de Beaumarchais, French Author, 1732-1799

I seem to recall in John Verhoeven's paper that he measured the quality of the edge by measuring the actual thickness head on. Anything under a micron (IIRC) was considered very sharp. I wonder if you could correlate sharpness to the edge width.

mse.iastate.edu/fileadmin/www.mse.iastate.edu/static/files/verhoeven/KnifeShExps.pdf

If that link doesn't work, just google verhoeven knife sharpening, it's the first hit. It's also at the bottom of this page:

http://www.mse.iastate.edu/who-we-are/people/emeritus-professors/john-verhoeven.html

A little light reading before bedtime!

 

[cds4byu]

I wonder if you could correlate sharpness to the edge width.

A little light reading before bedtime!

Nice article -- thanks for sharing it with me.

Verhoeven looks at push sharpness, which would be measured by CATRA REST testing. He mentions that his studies don't look at "sawing" action sharpness, which we would most likely call "slicing" on bladeforums.

I'm sure we could calibrate edge width to REST force, but there's also probably an effect due to edge roughness.

I think a follow-up article on this work would be very helpful.

Thanks for the tip.

Carl
-------------------------------
It is not necessary to believe things in order to reason about them
It is not necessary to understand things in order to argue about them.
- P.A. Caron de Beaumarchais, French Author, 1732-1799

 

[cds4byu]

Carl, I think an step #7 is needed in your methodology outline - add that the test results were then compared to real world testing done by Wayne Goddard. This is an important step.

In all honesty, I need to point out that step #7 was not part of the experimental methodology. Wayne did not test all of our blades for comparison purposes. He only tested a finished FFD2 knife, which is different from our test blades. And he compared it not with our other test blades, but with other knives he has tested over the years. So, although Wayne's test was confirmatory of our results, it's not part of our comparative test method.

In fact, I think that Wayne's testing method is an entirely different method. And Wayne has lots more experience and data than I do, so I expect that his test method is also valid, even though it's somewhat operator dependent.

I still think to run tests using anecdotal evidence (hints) to establish each steel's optimum geometry, edge finish, and sharpening procedure is inherently flawed, and would yield results you could not defend.

I think I agree. And I suspect that there is a substantial amount of disagreement over these parameters among the BladeForums participants. I'm toying with the idea of opening up a thread where I ask people to list what they consider to be optimum edge geometry for various steels, just to see how much agreement there is. But I'm a bit afraid of that Pandora's box.

Is it possible to reach an agreement that having the same geometries and same edge finishes for the baseline of the testing is correct?

That's a question I can't answer -- it'll take others to agree.

Carl
-------------------------------
It is not necessary to believe things in order to reason about them
It is not necessary to understand things in order to argue about them.
- P.A. Caron de Beaumarchais, French Author, 1732-1799