Edge Retention Testing and Analysis

[Broos]

You are free to show that his model does not account for the behavior you measured for example.

Isn't it the responsibility of the scientist to prove their theory, not on everyone else to disprove it?

why is the FFD2 blade so much sharper than the SV90 blade to start?

What is your definition of "sharp"? According to my definition, they were the same, unless I make the assumption that they were sharpened incorrectly.


Despite our differences in the definition of "sharp", here is previous explanation given by Carl:

As far as the metallurgy, and the cause for the extra performance, we're trying to determine that. The grain structure is so fine that we can't use optical metallurgy, and we haven't yet been able to develop the right specimen preparation for electron microscopy. We're working on it, and hope to have better reports out in the future.

In almost every technological advance, the advance comes before the science is fully understood. I believe there's a real advance here, and I hope to get the science there eventually.

Carl

There is a little further discussion on it if you read the other thread.

 

[Cliff Stamp]

Is this math?

A lay person's summary. The math would be the exact responce to the data on multiple knives (CATRA result and sharpening frequencies for the users) and the lack of correlation between these and the systematic error of a blunting mechanism which is supported by published papers and also quite immediately obvious from first principles of basic physics as I outlined in the above.

Until there are numbers attached, it's not math.

Obviously you do the math and calculate the determination times, correlation coefficients, etc. .

If I gave that description of the physics of any process in a paper for peer review

Congradulations, you have deduced that an article written for lay people to introduce a fundamental blunting mechanism and the means to model it is not written in the same style as would be for a formal paper for a physics journal.

No, because I don't know which papers you consider relevant.

As I told you, I have only a brief outline of what work was done, the basic concepts covered in the time period. Those two authors were involved and did some of the work. The review article is in progress. You are the one actually getting PAID in this area, I am doing it as a hobby. Once references were provided in any way it should have been obvious to you that you had a responsibility to investigate it in detail.

I'm still waiting for an answer on this question.

1. You perform edge retention tests to determine the sharpness as a function of the amount of media cut. You do this multiple times for a given blade.

2. Using the data from step 1, you calculate the mean and standard deviation for each x-axis value.

3. With the Levenberg-Marquardt method, you use a nonlinear curve-fitting procedure to fit your model (C(x) = Ci/(1+ax^b) to the data.

4. The fitting procedure gives you values for Ci, a, and b.

5. The covariance matrix (which I believe is the same thing you called the correlation matrix) contains data on the goodness of the fit.

6. If the goodness of the fit is adequate, you have a reasonable measurement of the cutting performance. If the goodness of the fit is inadequate, you don't have a reasonable measurement, and you conclude that the data isn't yet good enough to report.

Yes, with a few points. The covariance matrix contains information on how the parameters are correlated to each other you can use this to test the goodness of fit or just look at the residuals. The model will change depending on exactly what is being plotted, sharpness or cutting ability, or blunting, but the basic principles are the same. I have shown the variants for all of them

There is a lot of noise in this thread and when you encourage it then it is kind of absurd to complain when serious questions are answered. If you responded in one post it is also more likely that points would not be missed.

Would you please identify which calculations "exploded the noise"? Was it your calculations or was it our calculations?

As I noted AGAIN I showed the raw data, no calculations. As I noted clearly your difference equations exploded the noise they always do. This is basic math of error propogation.

The repeatability of the wear test has not yet been quantified.

You might want to kind of put that as a bold disclaimer after any conclusions. "I have no idea if this result is representative of the population." (that is a math statement by the way, it has an exact definition) or in similar terms "Feel free to ignore this conclusion as it is completly unknown if the data actually supports it."

So, why is it that when you “manipulate” data, it is science, math, physics, and ABSOLUTELY correct. If anyone else is "manipulating" data, it is ....

LIE 1:I have shown on many occasions how the work other people have done produces valuable results.

LIE 2:I have shown on many occasions that work I have done in the past was likely influenced by systematic errors.

However as for the base point, yes you can perform calculations on data which cause misleading conclusions and you can perform calculations which are not. As an example, you can perform a z-test on a small sample and get a perfectly nice probability which will have a large systematic bias and be misleading.

In addition, your lack of willingness to fully disclose your analysis method for peer review implies that you are the one ...

LIE 3: I have described it in great detail, showed the exact results and even put the raw data being fit on the website on many occasions. Of course the actual code I use to implement the algorithm isn't the issue but a straw man, you can code it in any language or even use a different algorithm if the result remain the same. But if you really want to see code, here is a random number generator in AWK :

function nrand( x1,x2,w,y1,y2) {

do {
x1 = 2.0 * rand() - 1.0;
x2 = 2.0 * rand() - 1.0;
w = x1 * x1 + x2 * x2;
} while ( w >= 1.0 );

w = sqrt( (-2.0 * log( w ) ) / w );
y1 = x1 * w;
y2 = x2 * w;
return y1
}

BEGIN {i=0
iL=4
diL=5
iU=2
diU=3
srand()
}
{
if ($1 !~ /#/) {
i++
x=$1
k1=$iU
d_k1=$diU
k2=$iL
d_k2=$diL
}
}
END { for (j=1;j<=i;j++) {
n_k1[j]=k1[j]+d_k1[j]*nrand()
n_k2[j]=k2[j]+d_k2[j]*nrand()
printf("%4.2f %4.2f %4.2f %4.2f %4.2f\n", x[j],n_k1[j],d_k1[j],n_k2[j],d_k2[j])
}

It's just not organized very well.

Yeah, it was never meant to become what it has, it was started as a hobby. Had I known I was going to end up doing nonlinear modeling I would have structured it much better from the start. Right now there is a huge inertial moment preventing the organization because of the massive amount of work I know it will take.

-Cliff

 

[gator68]

Isn't it the responsibility of the scientist to prove their theory, not on everyone else to disprove it?

I am talking about the apparent willingness of cds4byu to discuss what Cliff Stamp has done. Cliff has "proven" his model does a good job of fitting the data. (He has done lots of testing of knives.) As far as I can see the diamond blade people do not know why or how their knives work -- they are presenting data that the knives are so much better. This is a good first step, but it would be interesting to fit this into some sort of physical model to understand the behavior and see how this might translate to other steels.

What is your definition of "sharp"? According to my definition, they were the same, unless I make the assumption that they were sharpened incorrectly.

By this I mean, why did one blade cut so much better than the others on the first cut? Look at figure 1. Then look at figure 3. They all have the same REST value (except for the reference D2 blade -- why? sharpened incorrectly?) The discrepancy between these two results is curious and would be IMHO a good place to start looking at why the FFD2 blade is "better". Note that the FFD2 and SV90 blade dull with the same slope. This is why understanding the initial "sharpness" or edge state is important.

cds4byu's explanation does not address this, especially given that this test was done once.

I got no dog in this race -- I'll probably end up buying one of these knives if the price is right. That said, I'm still going to ask questions as long as they're putting up data.

 

[tnelson]

I have to disagree here. Cliff has derived a fairly simple model for how a knife blunts. This is derived from the idea that the force applied to the material is the force applied to the knife edge; and that the force is "used" to blunt the edge and push material aside during the cut. From this simple idea, he comes up with a model -- a proposed relationship between sharpness and amount of material cut. He then tested this with data on tons of knives, and did repeated testing on the same knife. The model actually fits the data. Using the model, one can make comparisons between the performance of different knives. This is valid as long as you believe your model is useful and you take appropriate data that allows you to make meaningful fits to your parameters.

This is a perfectly fine approach to a problem in physics. I've done similar things to model particle detector performance. Rather than do first principle calculations of phonon propagation through gold-germanium interfaces, simple create a model using heat capacities and a "weak-link" between two heat baths. This turns the problem into a question of simple algebra, and captures 90% of the behavior. If that's all you need to know, then stop there.

As far as I can tell, Cliff has lots of data etc on his website. It's just not organized very well. I doubt he is trying to hide anything or avoid "peer review." You are free to show that his model does not account for the behavior you measured for example.

Here's a question for you TN -- why is the FFD2 blade so much sharper than the SV90 blade to start?

gator68,

I agree with you completely. Don't get me wrong, I think Cliff has done a lot of work and derived a very nice simplistic model. This is a perfectly acceptable approach. However, from what I can find (yes, it is a bit unorganized and hard to find, but I am too) his approach is strictly empirical.

The thing I am trying to understand is the "physics, mechanics of blunting". Cliff repeatedly comments that his approach, or his model, is all 'science, physics, and mechanics' based. I do not see this in his methodology.

I understand principles of physics and mechanics. I am trying to better understand the "physics of slicing/cutting", and the "mechanics of blunting". Such knowledge will help us better understand our results, and how our approach may affect results. This knowledge would help all of us to improve our methods/approach in the future, e.g. further science. If Cliff has some knowledge in this area, teach us (I can't find it in the materials he has posted). If he doesn't, that's OK too, but just say so. Don't keep ignoring the question.

Now, regarding the question of FfD2 versus S90V: based on the REST test, they were the same sharpness (i.e. same radius at the cutting edge). Now, why the large difference in amount of material cut: I don't know. We are working to better understand this.

I have a hypothesis as to why the difference. This is support by some anecdotal evidence. If I have time later, I will put my hypothesis together and post it. However, this is only a hypothesis and I want to make it clear that it is such. Feedback and thoughts are welcome.

thanks,

TN

 

[Cliff Stamp]

Now, regarding the question of FfD2 versus S90V: based on the REST test, they were the same sharpness ...

No they were not, because that is not what the test measures. I clearly explained that in the above. The fact is that even CATRA knows this because I discussed it with them years ago.

-Cliff

 

[tnelson]

Like the annoying little kid who won't put his hand down and says "ooh, ooh, ooh" - sorry.

Broos,

That's OK, I was one of those annoying little kids (I think I still am?).

I appreciate your comments on the discussions. I can see you use your background and put some thinking into your comments.

thanks,

TN

 

[Kohai999]

LIE 1:I have shown on many occasions how the work other people have done produces valuable results.

LIE 2:I have shown on many occasions that work I have done in the past was likely influenced by systematic errors.


LIE 3: I have described it in great detail, showed the exact results and even put the raw data being fit on the website on many occasions. Of course the actual code I use to implement the algorithm isn't the issue but a straw man, you can code it in any language or even use a different algorithm if the result remain the same. But if you really want to see code, here is a random number generator in AWK

Cliff,

Webster's defines "lie" as "to make an untrue statement with intent to deceive".

Do you really think this is what the BYU boys are doing?

If so, why? Because they are humbled and trembling in fear of your clear as mud position?

Pick another verb to describe what it is that you feel they are doing. "lie" is about as inaccurate as you can get.

Best Regards,

STeven Garsson

 

[gator68]

...

Now, regarding the question of FfD2 versus S90V: based on the REST test, they were the same sharpness (i.e. same radius at the cutting edge). Now, why the large difference in amount of material cut: I don't know. We are working to better understand this.

I have a hypothesis as to why the difference. This is support by some anecdotal evidence. If I have time later, I will put my hypothesis together and post it. However, this is only a hypothesis and I want to make it clear that it is such. Feedback and thoughts are welcome.

thanks,

TN

Thanks for the response. I would be interested in hearing any hypothesis you have.

Another question. the press release mentioned a third test:

The third machine tests a blade’s strength by securing a
blade against a fixed stainless steel rod at a 20 degree angle.
Force is applied until a maximum of 65 lbs is realized. The
degree of flexing, bending or chipping is noted and recorded
for each blade tested.

No results were presented from this test. Anything you can share?

 

[Cliff Stamp]

In regards to the manner of blunting I gave you an article which refererences published work which describes the methods of blunting in detail from principles of metallurgy supported by experiment (Landes) plus direct experiment (Furi), and a bunch of other work showing various types (the chisel guys). Yeah, my model fits all their data too, it has to because the basic physics is the same. Like I said it even fits dental scrapers. In the webpage I didn't explain the physical laws in detail (such things like the rate of wear would be inversely proportioal to the amount of wear) because it was a lay arguement, but I find it absurd you don't follow it anyway because it is basic physics. Plus anyone who has sharpened a blade actually knows first hand this is the way it behaves and both are the same actually in regards to abrasive wear.

-Cliff

 

[Kohai999]

....... In the webpage I didn't explain the physical laws in detail (such things like the rate of wear would be inversely proportioal to the amount of wear) because it was a lay arguement, but I find it absurd you don't follow it anyway because it is basic physics. Plus anyone who has sharpened a blade actually knows first hand this is the way it behaves and both are the same actually in regards to abrasive wear.

-Cliff

I find your frequent use of the word absurd to be absurd, but it does rhyme with turd.

Best Regards,

STeven Garsson

 

[db]

Good to see Cliff reverting back to name calling personal attacks and insulting and accusatory posts. Pretty silly coming from Cliff who has shown himself to be a lire, distort facts, mislead, and refuses to answer simple questions. Talking about peer reviews and published papers, what about?
Ref.
http://www.bladeforums.com/forums/showpost.php?p=4532586&postcount=39

Heck it would just be nice if Cliff would simply answer questions as clear and directly as tnelson or cds4 does.

 

[hardheart]

dunno if these will help

http://www.bladeforums.com/forums/showthread.php?t=441164
http://mse.iastate.edu/fileadmin/www.mse.iastate.edu/static/files/verhoeven/KnifeShExps.pdf

I'm guessing everyone has taken a look at Cliff's work on cutleryscience by now.

http://www.cutleryscience.com/articles/sharpness_review.html
http://www.cutleryscience.com/reviews/model.html

 

[Cliff Stamp]

That one kel_aa posted and the thread have a number of excellent references. I was impressed until I found out that it was all done in germany about a hundred years before.

-Cliff

 

[cds4byu]

That one kel_aa posted and the thread have a number of excellent references. I was impressed until I found out that it was all done in germany about a hundred years before.

-Cliff

Yes, I have all of those papers. They all refer to a smooth blade. I'm working on an analysis of them.

Carl

 

[cds4byu]

This is a perfectly fine approach to a problem in physics. I've done similar things to model particle detector performance. Rather than do first principle calculations of phonon propagation through gold-germanium interfaces,

I agree. Particularly since you have to be certain whether phonon propagation or electron propagation is dominant in the particular material and temperature regime you are considering.

simple create a model using heat capacities and a "weak-link" between two heat baths. This turns the problem into a question of simple algebra, and captures 90% of the behavior. If that's all you need to know, then stop there.

And if you use a model that says q=hA(deltaT), and you're fitting data for h, then you can say your model is based in physics.

But if you are fitting a model that says q=hA(deltaT)^beta, and fitting for h and beta, then you're no longer fitting a physics-based model, but an empirical model that has been modified from the raw physics model.

What's the difference? The physics model is unique and correct, at least until a new physical explanation comes out (e.g. Newtonian mechanics vs. special relativity). On the other hand, the empirical model is only a convenient fit, and other convenient fits are also permissible.

I'm trying to understand whether Cliff's models are physics models, or empirical models based in physics. I can't get enough information to make the decision. The "lay" answers in cutlery-science.com are too vague, and I haven't seen anything more specific, even though this thread was started specifically to have geek-talk.

Carl

 

[db]

After reading this thread and the FFD2 thread I thought this quote of Cliffs from Hardhearts link was pretty funny.

Of course published papers will attempt to make something seem very complicated but at a basic level physics is very simple and you can summarize anything
in a few simple words. If you can't then you don't understand it.

 

[Cliff Stamp]

Empirical models are things like spines, polynomials, etc., there is no physics behind them, they will fit any curve. There are also semi-emperical models which are curves which are based on the physics of a simple or ideal system which are then applied to a real system by allowing some perturbation factor, in the model I use this perturbation factor would be x^(0.5+a). The a would be the perturbation factor representing the deviation from the ideal (simplistic behavior).

As I noted the underlying physics is quite simple. If you assume it is wear based then as the edge thickens, under the same force the pressure is reduced so the rate is reduced, thus the 0.5 law. However this is just the ideal behavior. In reality at the same time the edge is deforming and chipping which changes the effective thickness. As well the wear rate changes because the edge changes from wear around the matrix to the large carbides coming out. This is why you can see high scatter in wear, the blades will even get sharper and then blunter in a cyclic fashion. I have talked about this and cited examples in my work and others before.

Now to be rigerous you would want a model which had parameters for both wear, deformation and chipping, and included the nonlinearity, you could do this by probability calculations for the chipping and carbide influence on wear. But the problem I found when looking at this was that none of the data, even the CATRA results are anywhere near close enough to the noise level needed to give model parameters like that because the dependencies would be far too close, the fits just go undefined.

-Cliff

 

[cds4byu]

Quote:
Originally Posted by cds4byu
Is this math?

A lay person's summary. The math would be the exact responce to the data on multiple knives (CATRA result and sharpening frequencies for the users) and the lack of correlation between these and the systematic error of a blunting mechanism which is supported by published papers and also quite immediately obvious from first principles of basic physics as I outlined in the above.

Have you done the math to show the lack of correlation? If so, where can I see it?

Quote:
Until there are numbers attached, it's not math.

Obviously you do the math and calculate the determination times, correlation coefficients, etc. .

Again, have you done the math and calculated these quantities? If so, can I see it?

Quote:
If I gave that description of the physics of any process in a paper for peer review

Congradulations, you have deduced that an article written for lay people to introduce a fundamental blunting mechanism and the means to model it is not written in the same style as would be for a formal paper for a physics journal.

I've asked for a technical discussion, not a lay-person's discussion. You've claimed it's all there. I found the lay-person's discussion, but not the technical discussion. I'd still like to see the technical discussion.

You are the one actually getting PAID in this area, I am doing it as a hobby.

Actually, I'm not getting PAID anything for this work. It's just a hobby for me as well

Once references were provided in any way it should have been obvious to you that you had a responsibility to investigate it in detail.

Sorry, in my world the name of an author doesn't constitute a reference.

... Carl's summary of the modeling process was here ...

Yes, with a few points. The covariance matrix contains information on how the parameters are correlated to each other you can use this to test the goodness of fit or just look at the residuals. The model will change depending on exactly what is being plotted, sharpness or cutting ability, or blunting, but the basic principles are the same. I have shown the variants for all of them

Can you give me a list of the models you use, in addition to the one I quoted, or specific references to the other models?

There is a lot of noise in this thread and when you encourage it then it is kind of absurd to complain when serious questions are answered. If you responded in one post it is also more likely that points would not be missed.

I'm trying to avoid noise, which is why I keep trying to ask specific questions and get specific answers.

My personal preference for this kind of forum is one post per issue, which makes it easier for me to follow the specific issues. Since you apparently prefer to have the posts lumped together, I'll try to follow this procedure in the future.

As I noted AGAIN I showed the raw data, no calculations. As I noted clearly your difference equations exploded the noise they always do. This is basic math of error propogation.

I'll post a different analysis of your data later, because it would be too long to post here. Please note that it will be in a different post, which you don't like.

You might want to kind of put that as a bold disclaimer after any conclusions. "I have no idea if this result is representative of the population." (that is a math statement by the way, it has an exact definition) or in similar terms "Feel free to ignore this conclusion as it is completly unknown if the data actually supports it."

I might want to, but I don't.

LIE 1:I have shown on many occasions how the work other people have done produces valuable results.

LIE 2:I have shown on many occasions that work I have done in the past was likely influenced by systematic errors.

Since you can disprove any assertion by a single counterexample, a link or reference here would strengthen your position.

LIE 3: I have described it in great detail, showed the exact results and even put the raw data being fit on the website on many occasions.

It's not the data, it's the algorithm I've been asking about.

Of course the actual code I use to implement the algorithm isn't the issue but a straw man, you can code it in any language or even use a different algorithm if the result remain the same.

Again, it's the algorithm that's the issue. And different algorithms will almost certainly provide slightly different results.

But if you really want to see code, here is a random number generator in AWK

Cliff, you complain about noise, and then you add it. You know perfectly well that the code asked for was not random number generation code, but the code to perform your analysis. In order to minimize the noise in the thread, I won't go back to the earlier posts. Every question I've asked about your analysis method has been aimed at understanding what you do. Not one question about your method has been aimed at saying what you do is wrong. I can't make a judgment about whether I agree or disagree with your method until I understand your method, and that understanding is what I'm after.

Yeah, it was never meant to become what it has, it was started as a hobby. Had I known I was going to end up doing nonlinear modeling I would have structured it much better from the start. Right now there is a huge inertial moment preventing the organization because of the massive amount of work I know it will take.

If you're interested in help, I'd be willing to look at ways I could help you accomplish the task.

Carl

 

[cds4byu]

EBut the problem I found when looking at this was that none of the data, even the CATRA results are anywhere near close enough to the noise level needed to give model parameters like that because the dependencies would be far too close, the fits just go undefined.
-Cliff

How would I tell when "the fits just go undefined"? What is the behavior of your algorithm when the fits go undefined?

Carl

 

[cds4byu]

Another question. the press release mentioned a third test:

No results were presented from this test. Anything you can share?

This test is an attempt to make an automatic test of one of Wayne Goddard's subjective tests.

I'm not sure what this test means, or what properties of the edge it measures, in either Wayne's implementation or Charles's implementation, so haven't really done anything with the data from that test.

Wayne believes, IIRC, that a soft edge will bend, a too-hard edge will chip, and a just-right edge will flex, then return, without chipping or bending. I believe that all of the blades in the test set we provided passed the test. Some of the FF parameter sets we tried resulted in chipping, others resulted in bending. But I don't have the data available.

Carl