Steel Edge Retention testing

[hardheart]

start from cut 8, initial sharpness is dependent on who is sharpening, and on what media, so just start at the earliest point where the measurements are equal.

 

[noss4]

nozh2002: I watched your video and saw you dropped the knife on the thread and cut it and your doing it by hand. If your using different blades of different weight how does this affect the out come of the test ?

How do you know you have the exact same amount of tension on the thread every time ?

And how do you limit the angle of the knife so the edge is being placed on the thread is completely the same every time and not cutting the thread at an angle ?

An will this affect the outcome of the test with many resets. and human, blade weight, and the exact same portion of the edge being used every single time ?

I have been making a machine to cut rope and have been struggling with a design so different blade weighs do not go against the scale.

The photo is an early prototype the new one is about 200 parts at the moment.

In the photo of the machine I have a counter weight system to set the knife
to zero weight. The knife is loaded under very light spring weight against the rope. About 1 pound. then the digital scale is zeroed out. The cutting stroke is set at whatever length I desire and is continuously repleted and controlled
by a computer for however many cuts I chose to make and the exact same edge portion is run over the rope.

I'm trying to remove all human error out of the equation

Thanks

 

[nozh2002]

nozh2002: I watched your video and saw you dropped the knife on the thread and cut it and your doing it by hand. If your using different blades of different weight how does this affect the out come of the test ?

How do you know you have the exact same amount of tension on the thread every time ?

And how do you limit the angle of the knife so the edge is being placed on the thread is completely the same every time and not cutting the thread at an angle ?

An will this affect the outcome of the test with many resets. and human, blade weight, and the exact same portion of the edge being used every single time ?

I have been making a machine to cut rope and have been struggling with a design so different blade weighs do not go against the scale.

The photo is an early prototype the new one is about 200 parts at the moment.

In the photo of the machine I have a counter weight system to set the knife
to zero weight. The knife is loaded under very light spring weight against the rope. About 1 pound. then the digital scale is zeroed out. The cutting stroke is set at whatever length I desire and is continuously repleted and controlled
by a computer for however many cuts I chose to make and the exact same edge portion is run over the rope.

I'm trying to remove all human error out of the equation

Thanks

Well, I have mathematical education, you probably have engineering or physics degree. You are making perfect tool to have precise measurements eliminating human error and I instead deal with it with increased quantity of tests. In result you may be will have to do one cut and I am doing 21.

I do not drop blade but very gently move it down by hand trying to eliminate any impact not to give to the scales any momentum. For everything else - I am not eliminating all this factors but dealing with them statistically.

There are too many factors to eliminates - rope itself not perfect, edge is not perfect and who knows what else. You have 200 parts already, and may end up wit your own CATRA macine for $15000 your own media etc:

http://www.youtube.com/watch?v=rNLLjkVwp7E

But test I have easy to reproduce by anybody and not expensive at all, easy to set up etc...

Thanks, Vassili.

 

[noss4]

nozh2002: I see... Thanks and yes tests that people can recreate
themselves cheap and easy is all ways good.

 

[nozh2002]

nozh2002: I see... Thanks and yes tests that people can recreate
themselves cheap and easy is all ways good.

Not only this but consider factors which are not under your control - edge are not ideal and equal - it is very well seen under microscope, it is more or less randm, also steel itself contains random inclusion of carbides, random grain distribution, media you are cutting - rope also random mix of fibers.So more or less but it will require some statistical analysis.

And ultimately any measurements at high precision will fall into Gauss distribution. Sometimes it is no seen because of grid/scale, but this is just law of Nature.

Thanks, Vassili.

 

[noss4]

nozh2002: I understand there is alway going to be things I can not control. There is always a certin amount of error plus or minus. I just want to eliminate as much as possible and test different blades and see how they compare.

Thanks for the feedback you have giving me a few things to think about.

 

[db]

There is no such thing as a perfect test. However almost all tests show useful info to some degree. Nozh great work. SODAK I also hope you are able to post your results and tests here. I understand if Cliff doesn't want you to though. That would be kind of ironic if he is now putting on any restrictions on his testers.

 

[PghMitchS02]

I think it is great what PghMitchS02 did. This is some point of view. I respect it. He may be wrong, may be not - we may discuss it.

Thanks, but it was Excel that did the work I knew that stats class I took was going to come in handy some day! What you did, Vassily, in collecting the data was the hard part.

All I was after was another way to look at the data. I think showing the increase in force over time makes it easier to compare edge holding ability by giving a common starting point to everything. Looking at the raw data, pressure required to make the cut, shows how sharp a given blade is, not how well it is wearing.

 

[Cobalt]

HH: Perhaps some don't also realize how crazy sharp 30-40g really is.

Nozh: I do not think that it is correct to say that 50g edge "twice sharper" then 100g edge.

I think these two statements really show the heart of the issue here. I tried to address this in my previous post. In order to give percentages you need an absolute point of reference and the initial sharpness is not one. It is the same mistake to say that 50g is twice as sharp as 100g as it is to say that the temperature increased by 100% when the temperature was raised from 50F to 100F. Let's examine this a bit more Cobalt's argument would be that you take a pot of water at 40F and heat it to 120F and a second at 50F and heat it to 100F. You then proceed to claim that the first pot is 300% hotter than it was before and the second is 200% hotter, a difference of 100%. Without wanting to step on anybodies toes but the truth is that this line of argument is complete and utter nonsense as the Fahrenheit scale happens to be about as arbitrary a scale as one could envision. In truth, the first pot was raised from 278K (K stands for Kelvin which is measured to the absolute 0 temperature) to 320K, a mere 15%, and the second from 283K to 310K, by 10%. Hence the difference of the temperature increase between the two pots is a mere 5%.

For sharpness, value of absolute sharpness is practically not really possible to define, as towards 0g the scale would get hopelessly nonlinear, but in principle, it would be possible to use 0g as point of reference. Far more practical is it to use absolute bluntness as reference. My guess is that the thread, that Vassili is using, breaks at about 1.5kg (this corresponds to the strength of the polythread that I started testing on) on a completely unsharpened edge. So 30g is actually 2% bluntness while 40g represents 2.7% bluntness. This is a much more illustrative number, as shows, that both are bloody damn sharp with little difference between them. Now, after so-and-so many cuts the 30g blade dulls to 120g while the 40g blade dulls to 100g, or 8% and 6.7% of complete bluntness. So the first dulled by 6% while the second dulled by 4%....[B]a difference of wopping 2%[/B]!!!!!

These numbers represent much more closely how the dulling is experienced in practice. They also explain why this test has changed Nozh's steel-snob attitude....(I still have my doubts.....about this change of attitude, I mean, not about his numbers; I would bet a good deal of money that he will continue to lust after the latest and greatest steel ....but then, who doesn't ).

Hob, what you compared what I did to, is way off and makes no sense, but whatever.

So go ahead and take those results and do an analysis yourself. There are a few more who will as well. The numbers are there. I did that analysis quickly in my head in a few minutes from a hotel room. That is how easy it is. But when I can back home I can get more detailed and esily show numbers. The results show everything.

oh, and the portion of your post that I underlined is way more senseless than my quick analysis. I will show you why when I get back to civilization

better yet, go ahead and graph the results and what pattern you get, I think you will see what I saw at a glance.

 

[PghMitchS02]

Trimmed to get at the heart of HoB's statement:

So the first dulled by 6% while the second dulled by 4%....a difference of wopping 2%!!!!!

Actually the difference is 33%, not 2%. In computing the percentage difference, you subtract the second data point from the first and divide that difference by the first data point. In your example 6 minus 4 is 2, 2 divided by 6 is one third or 33.33%, so there was 33.33% less dulling in your example.

Now to find out if either number is really meaningful, you need to go for some really higher, and boring, math called Analysis of Variance where by you look at all the differences, find the statistical mean and then see how far outside this figure your data point lies to determine if it is a meaningful difference . . .:barf:

Please understand, I'm not trying to fan the flames of a numbers war. I will be the first to admit, Disraeli was right (I think it was him) when he said "There are plain liars, damned liars and statisticians".

Bottom line is, we all have our favorite steel and favorite knife. If I need a knife NOW I really don't care what the steel is or who made it, I'll grab whatever I can reach.

What I find really interesting about Vassili's project is how quickly cutting something as fine as a piece of thread produces a measurable difference in sharpness. There are a lot of different ways to look at that data but the only conclusion we can all reach is "If you use a knife repeatedly, it needs to be resharpened".

 

[hardheart]

It's been shown several times before by many testers, both amateur and those in the industry, that high initial sharpness does fall off very quickly. I may do a regression this afternoon, if no one else gets it first. The results are gonna look familiar to anyone who has read through the myriad threads/flamewars we've seen already on this board and the toolshed/maintenance.

 

[db]

Hardheart is correct. The real questions many of the knife, and mainly sharpening nuts struggle with is. Do we resharpen at the point that the high initial sharpness is gone? If so is edge holding or sharpness more important? As often as I do touchup my blades they never get very dull.

 

[Cobalt]

Taking my own suggestion in the last post I made, here's the table:
__________________________________________________________________

% Increase Over Cut 0
Cut INFI 420 ATS INFI 420 ATS
0 40 30 40
1 60 50 60
2 70 50 70
3 70 60 70
4 80 50 70
5 80 60 70
6 80 60 70
7 --- 70 80
8 80 80 80
9 --- 70 80
10 70 70 80 75.00% 133.33% 100.00%
12 80 70 80
15 80 80 90
20 80 70 90 100.00% 133.33% 125.00%
25 80 80 90 100.00% 166.67% 125.00%
30 90 80 90
35 90 80 90
40 90 90 90
45 90 70 90
50 80 80 90 100.00% 166.67% 125.00%
60 90 70 90
70 --- 80 90 166.67% 125.00%
80 100 80 90 150.00% 166.67% 125.00%
90 110 80 ---
100 110 80 90 175.00% 166.67% 125.00%
110 110 80 90
120 110 90 90
130 110 90
140 100 80
150 110 90 175.00% 200.00%
160 110 100
170 120 110
180 120 110
190 120 110
200 130 100 225.00% 233.33%
210 120 110
220 130 110
230 110 ---
240 110 130
250 110 130 175.00% 333.33%
260 110 130
270 110 140
280 110 130
300 110 140 175.00% 366.67%
320 110 140
340 120 150 200.00% 400.00%
360 120 140 200.00% 366.67%
380 120 140
400 120 140 200.00% 366.67%
_____________________________________________________________________

I did not bother with comparing all the cut data, but I think this shows the trends well enough.

So how would compare the results of these two steels.

If I understand your data correctly, INFI sharpness degraded by 200% by the end. 420 sharpness degraded by 367%. Since the test media is the same and should be fairly consistent from thread to thread, there should be an easy way to compare the two.

by the way, my quick analysis was taking the average of the first 5-6 cuts and the average of the last 5-6 cuts. I chose more than one, since sharpness probably does not change much over 6 cuts and it represents an average sharpness.

IF you use this average your numbers will be closer to mine I think.

 

[Blue Sky]

Actually the difference is 33%, not 2%. In computing the percentage difference, you subtract the second data point from the first and divide that difference by the first data point. In your example 6 minus 4 is 2, 2 divided by 6 is one third or 33.33%, so there was 33.33% less dulling in your example.

You've got it a little backwards. The percentage change is the second point minus the first, divided by the first, times 100.

(4 - 6)/6 = -.33 x 100 or -33%. 6 to 4 is a negative change of 33%.

http://wiki.answers.com/Q/How_do_you_calculate_a_percentage_change_from_one_number_to_another

Regardless, I think you are making a mistake by putting the values 6 and 4 in a specific order in the first place. They are calculated from two separate, independent tests and their order relative to each other means nothing. I could just as easily say:

(6 - 4)/4 =.50 x 100 or +50%. 4 to 6 is a positive change of 50%

Since the order is arbitrary, your statement about the relative change in sharpness between the two tests is not valid.

 

[PghMitchS02]

I sit corrected, Blue Sky. I guess the important part is that it is not just a simple subtraction.

 

[nozh2002]

Almost finished with ATS-43. As I sad I just updating table in that post in this thread - you better check it time to time - I am putting data there right after I got it.

What a smooth steel! All test curves as well as result graph are almost ideal - no bumps no surprises, probably it is due to ATS-34 fine grain structure.

From other hand I was sooooo bored testing it - very predictable.

I think it will be good sampling curve to compare other steel with it.

I need more rope now 50 feet was enoght to test 3 steels. I finish ATS-34 and do additional testing for INFI and 420HC and ATS-34 over 400 cuts, may be up to 600... If my wrist will stand this load...

Thanks, Vassili.

 

[HoB]

Actually the difference is 33%, not 2%. In computing the percentage difference, you subtract the second data point from the first and divide that difference by the first data point. In your example 6 minus 4 is 2, 2 divided by 6 is one third or 33.33%, so there was 33.33% less dulling in your example.

Well, I am not invested in this enough to get in the midst of a flame war which is clearly already brewing so I will probably sit this one out at the sidelines. But the above I will address. No the difference is not 33%. By dividing by six you are making the 6% the new point of reference which is just as arbitray as using the 40g in the first place. By using the 1.5 kg as point of reference you are generating with the ratio a scale between 0 and 1 that is absolute. 0 is ideally sharp 1 is completely blunt. On that scale the difference is 0.02. Since the new scale is a ratio, you can write the 0.02 as 2%, which is still 2% of the completely dull value. The real question is whether the 1.5 kg is a reasonable point of reference. Instead of using a completely unsharpened edge as point of reference it would probably be more sensible to use the point at which the blade becomes unacceptably blunt which is of course somewhat up to personal judgment.

Blue Sky: Not sure this is really my original idea, Cliff has always tested the strength of the thread.

Cobalt: Yawn :yawn:. How about YOU do some graphing, instead of seeing things at one glance. Or better yet, you do some of your own testing. If you did, maybe you would realize that there are no landslides in Nozh's data set.

P.S. Ok, coming back to this one more time: I think all the talk about numbers, ratios and analysis hide a few very fundamental facts that what ever analysis you decide to do, will have to accomodate. And maybe they are not obvious, if you haven't done some thread cutting yourself: The INFI blade and the 420HC are essentially equally sharp at the beginning, because the perceived difference between 30 and 40g is very small and in practice hardly noticable (I am not making this up. Try it yourself!). They are both very sharp. After 200 cuts the two blades are still essentially equally sharp. The 420HC blade is a tad sharper but again a difference that is probably noticable but still nothing to get all excited about because the difference between 100g and 130g or whatever is just not earthshaking (again try it out). After 400 cuts the two blades are still about equally sharp. Yes, now the INFI blade pulls ahead, but the difference is still marginal. Again, you might notice the difference between 120g and 140g without a scale but the difference is simply not exciting. So after a significant amount of work, 200 and 400 cuts the two blades still show a very similar performance. Which is exactly what Nozh said in his first posts. Talking about landslides when looking at Nozh's numbers is just nonsense to anybody, who has ever tried threadcutting. If your analysis shows a performance difference of 200% (which would indicate that one blade performed twices as well as the other) and more you have simply used clever analysis to misrepresent the original data. To anybody without a scale, without numbers to analyze, the two blades will have performed very similar.

P.P.S: I think I will try and give one more comparison: Percentages are a very dangerous tool if the point of reference is meaningless. Envision two men, one earning $1 per day, the other $1.5. The first gets a pay raise of $2 and the second of $0.5. Now one makes 300% of what he did before, while the second only 133%. Mathematically this is correct and it surely looks like a landslide....but only to a mathematician. To anyone with common sense and to the two men the difference is insignificant. At a respective $3 and $2 per day both are still piss-poor and will have trouble feeding themselves. The pay raise could have been referenced for example to the minimum wage. At $7 per hour and 8 hours (I am using arbitrary numbers) the minimum wage is $56 per day. Now the first man received a payraise of 3.5% of the minium wage the second one a pay raise of 1%. Now the percentages actually represent what was already clear to anyone with common sense from the start: Both were earning well below minimum wage and the pay raise didn't change a thing and the difference between the pay raises is insignificant. Again, back to the blades. At the start both blades were very sharp. After significant amount of work both were still surprisingly sharp, and the difference in sharpness between the two blades after 400 cuts was hardly enough to make anybody yell halleluja. Now, that is what the rawdata tells us (again, try it if you don't believe it, but hardheart and Larrin have for example already confirmed this). If the analysis doesn't represent the these facts then there is a problem with the analysis.

 

[nozh2002]

To have completely statistically reliable data all this test must be repeated several times. I mean that may be I need to test INFI and 420HC an ATS-34 again with same procedure.

However I am not seeing myself doinig it again and again in nearest future. It will be really nice if someone will follow up with it. The beauty of this test is that it can be repeated by everyone and I did everything to disclose all my testing. If someone need more info let me know - I will be happy to see someone else results - too many steels around for me alone!

Thanks, Vassili.

 

[HoB]

probably it is due to ATS-34 fine grain structure.

Ummm, huh? ATS-34 isn't exactly famous for a fine grain structure....?

 

[nozh2002]

Ummm, huh? ATS-34 isn't exactly famous for a fine grain structure....?

I ha impression that it is better in this terms then D2 and CPM S30V and VG10. But I may not know for sure - it is my impression based on some posts here (Japansteel if I remember right mentioned this before).

Thanks, Vassili.