Have you done the math to show the lack of correlation?
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.
Again, have you done the math and calculated these quantities?
Whenever I did it obviously, one of the main things you can do is trade accuracy for time (sin(x)=x). This tends to be perfectly fine in fitting in most cases because the noise in the data is larger than the approximation, you can calculate these as well of course.
Actually, I'm not getting PAID anything for this work.
My mistake I assumed you worked for the company that was developing this process. So you have no mometary connection to the friction forged process?
Sorry, in my world the name of an author doesn't constitute a reference.
When someone mentioned that Verhoeven had performed sharpness experiments I did a literature search and it only took a few minutes to find the relevant papers. I cited a number of authors who have done significant research in a field directly related to work you are involved in and you ignore it because you don't have the exact paper references, don't ever look them up, your loss.
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?
They are not significantly different, either normalized, inverted or without the initial constant depending on what is being fitted. I tried a number of plots to see which ones were most easily understood by other people. Normalized sharpness was the one I found to work best so far.
I might want to, but I don't.
Yeah because such as disclaimer would be counterproductive to selling the product obviously.
Since you can disprove any assertion by a single counterexample, a link or reference here would strengthen your position.
The sharpness article I cited earlier and someone else cited right after again contains several examples of both. The older reviews are filled with notes where I updated methods of comparison and noted the early methods were problematic. Your arguement here shows a complete ignorance of anything discussed beyond this specific two threads. Now this is ok as everyone is new to a forum at one time but to make generalizations of that type is pretty absurd.
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.
That was my point, all you needed to know was that I did a random normal spread, the code is not relevant. Just like the code for the cut ratio algorithm doesn't matter either.
How would I tell when "the fits just go undefined"? What is the behavior of your algorithm when the fits go undefined?
The fits are undefined when they don't converge or the parameters are undefined (uncertainty is too large). You can calculate the probability of significance as well by looking at the chi-square difference for the introduction of additional parameters. It isn't "my" algorithm either, this is all standard nonlinear least squares.
...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.
This is a completely undefined test, all steels have an elastic region obviously. I showed years ago how the same edge would pass or fail (both ways) this test simply by adjusting the angle. THis is basic metallurgy and shown clearly by any stress/strain graph.
As can be seen in the image, the curves fit the data quite nicely, even though there is noise.
No they don't, there is a math definition of that statement and it failed there as I described clearly.
Since Cliff prefers to calculate a cut ratio, and the non-linear optimization routine he uses can't handle the amount of noise in this data, I decided to try a worst-case analysis.
I don't prefer to use non-linear methods, the curve is non-linear so you have to if you are going to model it.
[/quote]... because we weren't using Cliff's algorithm. So the data is not well suited for Cliff's algorithm. Probably the next time, we'll take more data to do a better job of matching his algorithm.[/quote]
THis is just absurd again, it isn't a personal issue. Again, the data is perfectly fitted by the model I developed, it is only when you manipulated it that the noise grew and it became undefined.
No, I can't put confidence limits on the data.
This isn't difficult at all, do it the same way I did for the cut ratio.
But this is the data that Cliff said he couldn't draw any conclusions from.
You can draw conclusions from anything, the point is are they supported by the data and do the transformations you have performed make sense. If you have two functions f1,f2, the difference or ratio of these functions will be effected by any calculations on them. You can make this difference or ratio as large as you want by the appropiate transformations on the functions. You could easily transform the functions so the advantage of the FF D2 was a million percent. Don't tell me this isn't obvious to you.
is it just loss, or other deformation?
Wear, deformation and chipping. As noted in the above you can model all of these separately but you would need very precise data. Chipping would just be a statistical modeling. I have interest in looking at this on Landes data because that is exactly what he measures in his edge stability tests where for example AEB-L is far ahead of RWL34. But on a CATRA the performance would be very different. Which one actually measures the edge retention?
-Cliff
