|
boku
Jul 11, 2007, 10:08 PM
Post #51 of 78
(3760 views)
Shortcut
Registered: Jun 11, 2004
Posts: 278
|
tinnchris wrote: STDEV = .938 Yay, I have something to contribute! 
Hmmm... I wonder where I went wrong?
First, I calculated mean average:
mean average=(10 + 8.8 + 10.4 + 9.9 + 8.15)/5
= 47.25/5
= 9.45
Then I used the mean average to calculate sigma, the standard deviation:
sigma=sqrt(1/5*[(10-9.45)^2+(8.8-9.45)^2+(10.4-9.45)^2+(9.9-9.45)^2+(8.15-9.45)^2])
= sqrt(1/5*[(0.55)^2+(-0.65)^2+(0.95)^2+(0.45)^2+(-1.3)^2]
= sqrt(1/5*[0.3025+0.4225+0.9025+0.2025+1.69]
= sqrt(1/5*[3.52])
= sqrt(0.704)
= 0.839
(This post was edited by boku on Jul 11, 2007, 10:11 PM)
|
|
|
|
|
forkliftdaddy
Jul 11, 2007, 11:51 PM
Post #52 of 78
(3710 views)
Shortcut
Registered: Jun 3, 2003
Posts: 408
|
An awesome attitude:
maldaly wrote: I don't like failure so I use the minimum breaking strength. If I ever test a unit and it fails below the rated strength I'll lower the rating on the whole bunch.
A bit of history I didn't know:
maldaly wrote: Back in the good old days, we used average breaking strength.
Gulp.
|
|
|
|
|
yokese
Jul 12, 2007, 12:25 AM
Post #53 of 78
(3700 views)
Shortcut
Registered: Jan 18, 2006
Posts: 672
|
boku wrote: tinnchris wrote: STDEV = .938 Yay, I have something to contribute! Hmmm... I wonder where I went wrong? First, I calculated mean average: mean average=(10 + 8.8 + 10.4 + 9.9 + 8.15)/5 = 47.25/5 = 9.45 Then I used the mean average to calculate sigma, the standard deviation: sigma=sqrt(1/5*[(10-9.45)^2+(8.8-9.45)^2+(10.4-9.45)^2+(9.9-9.45)^2+(8.15-9.45)^2]) = sqrt(1/5*[(0.55)^2+(-0.65)^2+(0.95)^2+(0.45)^2+(-1.3)^2] = sqrt(1/5*[0.3025+0.4225+0.9025+0.2025+1.69] = sqrt(1/5*[3.52]) = sqrt(0.704) = 0.839
That would be the standard deviation of that particular subset of 5 units, but if you want to estimate the standard deviation of the whole "population" you'll lose one degree of freedom:
stdev = sqrt ( SUM[(x-average)^2] / (n-1) )
basically:
= sqrt(1/(5-1)*[3.52]) = sqrt(3.52/4) = 0.938
But of course, for a 3sigma test to be meaninful way more than 5 units should be tested.
|
|
|
|
|
boku
Jul 12, 2007, 12:32 AM
Post #54 of 78
(3695 views)
Shortcut
Registered: Jun 11, 2004
Posts: 278
|
yokese wrote: That would be the standard deviation of that particular subset of 5 units, but if you want to estimate the standard deviation of the whole "population" you'll lose one degree of freedom: stdev = sqrt ( SUM[(x-average)^2] / (n-1) ) basically: = sqrt(1/(5-1)*[3.52]) = sqrt(3.52/4) = 0.938
Ah, thanks, I'll look into that method.
yokese wrote: But of course, for a 3sigma test to be meaninful way more than 5 units should be tested.
Just so, as I noted in my earlier post on the topic.
Thanks, Bob "BoKu" K.
(This post was edited by boku on Jul 12, 2007, 12:36 AM)
|
|
|
|
|
billcoe_
Jul 12, 2007, 3:55 AM
Post #55 of 78
(3639 views)
Shortcut
Registered: Jun 30, 2002
Posts: 4694
|
Malcom - you rule. I wish you were making aliens.
|
|
|
|
|
justthemaid
Jul 12, 2007, 2:57 PM
Post #56 of 78
(3584 views)
Shortcut
Registered: Sep 11, 2004
Posts: 777
|
...and the # 5 ballnut lived happily ever after as it rode off into the sunset.
It's wires were frayed but it's love for Trango was strong.
After all... tomorrow is another day.
*(fade to silhouette of ballnut standing in the sunset)*
(This post was edited by justthemaid on Jul 12, 2007, 2:59 PM)
|
|
|
|
|
papounet
Jul 12, 2007, 11:25 PM
Post #57 of 78
(3510 views)
Shortcut
Registered: May 28, 2003
Posts: 471
|
maldaly wrote: boku, There are a lot of different ways to rate things, 3-sigma being a good one. If you want to use it go ahead. I don't like it because with most sample groups all it does is allow a specified percentage of the sample group (about 11 in 10,000 units) to fail below the rated strength. I don't like failure so I use the minimum breaking strength. If I ever test a unit and it fails below the rated strength I'll lower the rating on the whole bunch. Again, using a 3-sigma analysis with most sample groups a single unit that fails below the rating would be unlikely to change the rating. Most industries use a "working load" rating of 25% of average failure rate. I don't like that either. The working load rating for most carabiners would be 6 kN. Back in the good old days, we used average breaking strength. 3-sigma is a huge improvement over that and I like to think that minimum breaking strength is an improvement over 3-sigma. Mal
Common misconception that is worth challenging.
choosing the minimum breaking strength from a sample as the boundary of the whole population is not good or bad per se. It is dowright dangerous if the samplesize is small relative to the population. It could be superior to sigma based method if the ratio sample size/population size is large.
There is a definite probability that if the total population follow a normal distribution, that there will be a certain number x biners in THE TOTAL POPULATION failing below the minimum breaking strength observed in a sample.
This probability is dependant on the size of the population and of the sample.
with values 10; 8,8; 10,4; 9,9. 8,15
the stddev of the sample is 9,84 (it is only an estimate of the standard dev of the population)
the average of the sample is 9,45
the 3-sigma value is 6,64
99,85% of biner picked randomly from the population are at least stronger than 6,64kn
but only
91% of biners would be at least stronger than 8kn.
we are far from the minimum breaking strength of 8 kn you found from the sample and that you quote
If we enlarge the sample to 50 with the same values from 10,4 to 8,15, the 3-sigma lower limit is 6,91
if we enlarge the sample size to 500 with the same values from 10,4 to 8,15, the 3-sigma lower limit is 6,93
it is true that on large sample size relative to the population, it is possible to have one outlier in the sample such as one out the 500 which fails at a value lower than the 3 sigma
if the total sample size is = to the population size, the probability is near certainty that that there will be between 0,1% and 0,2% below average - 3stdev
Unless you are willing to share the population and sample size, i dare say that i would take the minimum breaking strength rating with a grain of salt.
Because it is not economically sound to have large sample size in destructive testing, in the absence of complete knowledge, 3-sigma and 6-sigma would appear to be safer than the observed minimum breaking strength.
in fact, I am almost certain that having the actual population and distribution characteristics, under the hypothesis of normal distribution, one can compute the equilibrium sample size for which there is a high confidence (99,5% for example) that observed minimum breaking strength is equal to the average-3estiamted standard deviation
On a related note, it is often the law that defines a Safety factor. In Life-protection-Equipment, the Safety factor can be 4 or 5.
Meaning that equipments are designed to sustain a Maximum Load 4 or 5 times larger than the Working Load (a load in normal operation).
Thus the use of steel biners in some rescue settings.
|
|
|
|
|
the_climber
Jul 13, 2007, 4:52 PM
Post #58 of 78
(3455 views)
Shortcut
Registered: Oct 9, 2003
Posts: 6142
|
Papounet, I think you missed much of what Mal was saying. Perhaps lost somewhere is the translation? I don't know. I would think Mal is mentioning a simplified/dumbed down (ie. stating without unnecessarily complicating the issue with to much detail) rundown of different types of safety ratings.
Ie. the construction and trucking industries require the Chain be rated by a safe working load of aprox. 25% of the breaking strength. I have however seen 70proof chain break at surprisingly low forces in 4x4 recovery. His reference to biners seems only to use an arbitrary piece of gear which, we as a climbing community, can relate to.
Ratings based on both standard deviation and minimum breaking strength are simply other ways of rating safety equipment. They happen to be the two which likely have the most relevance to real world climbing situations. But that’s my opinion. I have used standard deviation calculations in many situations, but relating to ore body concentrations and Paleoclimate calculations. So I’m not unfamiliar with the way sigma ratings work. And I’ll admit it works good with rating climbing gear. If it didn’t I’m sure a reputable company such as BD wouldn’t use it. I do have to side with Mal on the minimum breaking strength stand potentially having greater real world relevance to climbing. Yes it requires a large sample base, so does 3-sigma, to get accurate data on a specific batch. However, it does appear that Mal has no issued with destroying gear for the sake of safety. Quite simply the way he has handled this issue with the Ballnuts is an example one of the most responsible manners with a potential gear problem/issue has been handled with in recently (I’ll add OP with the link-cam to that list too).
The fact that should the minimum failure on a batch of gear be lower, but still within reason for climbing gear, Mal rates the gear at a lower strength shows great responsibility on the part of Trango. (I think this is what you were saying Mal, correct me if I’m wrong about the lower rating thing please)
When it comes down to it, it becomes a personal preference as to what ratings you trust in. I trust the safety of Trango gear just as much as any other brand of gear on my rack. It is amazing gear, if it wasn’t I wouldn’t use it. The same can be said for every piece of gear on my rack.
I guess this is where I stop rambling on and get back to work.
|
|
|
|
|
shogun
Jul 13, 2007, 5:23 PM
Post #59 of 78
(3435 views)
Shortcut
Registered: Dec 11, 2002
Posts: 107
|
billcoe_ wrote: Malcom - you rule. I wish you were making aliens.
what bill said!
thanks malcom!
-=g=-
|
|
|
|
|
the_climber
Jul 13, 2007, 5:31 PM
Post #60 of 78
(3431 views)
Shortcut
Registered: Oct 9, 2003
Posts: 6142
|
shogun wrote: billcoe_ wrote: Malcom - you rule. I wish you were making aliens. what bill said! thanks malcom! -=g=-
One more vote for Trango Alien-like cams!
C
mon Mal PLEASE!
|
|
|
|
|
stymingersfink
Jul 13, 2007, 6:23 PM
Post #61 of 78
(3414 views)
Shortcut
Registered: Aug 12, 2003
Posts: 7250
|
the_climber wrote: shogun wrote: billcoe_ wrote: Malcom - you rule. I wish you were making aliens. what bill said! thanks malcom! -=g=- One more vote for Trango Alien-like cams! C mon Mal PLEASE!
A TrAli-en-ago cam?
Yeah, I'd be down with that... especially the hybrid version
|
|
|
|
|
ktb
Jul 13, 2007, 7:02 PM
Post #62 of 78
(3397 views)
Shortcut
Registered: Jul 30, 2005
Posts: 8
|
Wow - good stuff to read in this thread! I will definitely work Trango into my next gear purchases. Hopefully someone like Mal will buy CCH and improve the production process. Could we start a petition! Pass the hat to help fund the buyout :) ?
|
|
|
|
|
maldaly
Jul 13, 2007, 9:50 PM
Post #63 of 78
(3377 views)
Shortcut
Registered: Oct 31, 2002
Posts: 1208
|
I LOVE you guys!
|
|
|
|
|
fear
Jul 14, 2007, 2:24 AM
Post #64 of 78
(3364 views)
Shortcut
Registered: Jun 16, 2003
Posts: 475
|
ktb wrote: Wow - good stuff to read in this thread! I will definitely work Trango into my next gear purchases. Hopefully someone like Mal will buy CCH and improve the production process. Could we start a petition! Pass the hat to help fund the buyout :) ?
You mean actually HAVE a production process. And no more smoking weed whilst brazing.
|
|
|
|
|
papounet
Jul 14, 2007, 2:29 PM
Post #65 of 78
(3336 views)
Shortcut
Registered: May 28, 2003
Posts: 471
|
the_climber wrote: Papounet, I think you missed much of what Mal was saying. Perhaps lost somewhere is the translation? I don't know. I would think Mal is mentioning a simplified/dumbed down (ie. stating without unnecessarily complicating the issue with to much detail) rundown of different types of safety ratings. ... When it comes down to it, it becomes a personal preference as to what ratings you trust in.
This is where I jumped in with Mal and i jumped in again with you.
The relative merits of the minimum observed breaking strength and 3sigma based values is not a place for OPINIONS.
Provided that the population's characteristic follows a normal distribution, there is a definite mathematical function that describe the probability that one item has a specification lower than a arbitrary value (and especially one computed on a sample.).
let n be the sample size
let p, be the population size
let v1= minimum breaking strength observed
let v2= observed mean - 3 observed standard deviation
for a small sample size, the probablibity that the value of a random item v =< v1 is higher than the probability that v =< v2,
for a very large sample size, the reverse.
one can compute the equilibrium n for a given p, that will give identical probability to both events
(it is equivalent to compute the Probability that v1=v2)
My objection is with OPINIONS on the merits of which rating to use. There is little room for OPINIONS when statistics are considered. The 2 ratings are only equivalent for a certain range of n and p. in other cases, one is superior to the other, which can be computed.
Due to the costs of destrructive testing, I am doubtful that a manufacturer would have the large sample size that would make minimum observed a better predictor than mean-3stddev.
I doubt also that the climbing items from a small scale manufacturing process follow a normal distribution.
I am happy with ratings on 3sigma combined with exhaustive non destructive testing to 50% as this later test would most likely catch the outliers.
ignoring the relevant data from Trango, i can only accept the minimum breaking strength on their products with a grain of salt (and I would suggest that Trango double check the legalese around the statement).
|
|
|
|
|
frodolf
Mar 4, 2008, 1:11 PM
Post #66 of 78
(3071 views)
Shortcut
Registered: Oct 1, 2006
Posts: 81
|
A bit of thread revival here, but I think this is relevant to the discussion.
I got a set of Trango Ballnutz a few weeks ago and being an aid climber, and having heard so much good about them, I joyfully went out to test them on the rock. It took me two minutes to break one.
What happened was that I bounce tested the smallest one. It wasn't a too hard a bounce test and my weight is in the neighbourhood of 75kg (150lbs). It broke at the "clamp" that keeps the two wires attached to each other (I don't know the word for it in english), which the spring rests on, above the coloured bar of the clip-in/thumb hole. The "clamp" simply blew apart. And I wasn't using too much force. I've bounced micronuts with half the rating just as hard. My estimate is that I created between 1,5 and 2kn of force.
It wasn't a lifethreatening breakage: the wire didn't brake or anything. But that "clamp" is what keeps the spring from being torn apart. Next time I hang from it will be the last. And that sucks. And what's even more stupid is that the clamps on all the five sizes has the same width, so they all have this same weakness. I still keep them on my rack, but I dare not bounce test them at all hard, and I don't care much for falling onto them. I regard them as bodyweight placements for now, sort of.
Thing is I ordered them from the States to Sweden and the shipment was a hazzle and took weeks, so I don't feel like returning it - it'll take mounths! And as I said, I don't trust any of the ones in my set, so in that case I'd like to/had to return the whole bunch. I'm not sure the retailer is up to that. I would't suprise me if they're not.
Has anyone had the same problem? Or am I simply unlucky? (or both)
And how could that peice get through security!?
F.
|
|
|
|
|
the_climber
Mar 4, 2008, 3:46 PM
Post #67 of 78
(3014 views)
Shortcut
Registered: Oct 9, 2003
Posts: 6142
|
frodolf, I would contact Trango directly on this one. Mal is great to deal with as are the rest of the Trango staff.
I've bounce tested the smallest two ballnuts HARD, and never had an issue I also took a short daisy chain fall on one without any incident like your describing. I'm much heavier than you too... about 75lbs heavier.
You don't have to send them back right away then wait, but I would start by calling trango directly and talking to them. Let them know what's up and perhaps get some photo's and let them know to expect them via email or something.
|
|
|
|
|
moof
Mar 4, 2008, 5:54 PM
Post #68 of 78
(2944 views)
Shortcut
Registered: Oct 17, 2003
Posts: 400
|
Yeah, what the_climber said, contact Trango. The owner, Mal, posts here as maldaly. He is a good guy and will treat you right.
|
|
|
|
|
maldaly
Mar 5, 2008, 5:27 AM
Post #69 of 78
(2872 views)
Shortcut
Registered: Oct 31, 2002
Posts: 1208
|
Hey Frodolf, that nicropress shouldn't have broken. PM me and I'll take care of you.
Mal
|
|
|
|
|
CCEbeling
May 26, 2008, 5:44 AM
Post #70 of 78
(2759 views)
Shortcut
Registered: May 26, 2008
Posts: 2
|
The folks at Trango are really great to deal with. I had a cinch fail to lock out last year when i was climbing in a gym, resulting in a 30-ft ground out. Being on the sharp end of the rope i didnt know if the fall was a result of device failure or belayer error. I called trango and they were happy to inspect and pull test the device and send me a new one. After inspection and pull testing it was determined it must have been belayer error as there was nothing wrong with the device.
I bring this up now because i believe I have discovered the way this accident happened. My climbing partner and i found a way of holding a Cinch at your right hip that prevents lockout, you may already know about this Mal but i think it is something that should be included in the warnings that accompany the product. I will take some pictures and post them as well as PM Mal.
I also found that lockout can be delayed if the belayer grips the rope above the device (a stupid thing to do anyway but in a panicky situation it happens) so it is best to keep hands away from the rope above the device.
|
|
|
|
|
maldaly
May 26, 2008, 3:13 PM
Post #71 of 78
(2681 views)
Shortcut
Registered: Oct 31, 2002
Posts: 1208
|
Hey CCEBeling, in part because of your near miss, we posted video instructions for the Cinch and many other of our products on the website. Check them out at:
http://www.trango.com/videos.php
Climb Safe,
Mal
|
|
|
|
|
billcoe_
May 26, 2008, 3:43 PM
Post #72 of 78
(2664 views)
Shortcut
Registered: Jun 30, 2002
Posts: 4694
|
frodolf wrote: .... I joyfully went out to test them on the rock. It took me two minutes to break one. What happened was that I bounce tested the smallest one. It wasn't a too hard a bounce test and my weight is in the neighbourhood of 75kg (150lbs). It broke at the "clamp" that keeps the two wires attached to each other (I don't know the word for it in english), which the spring rests on, above the coloured bar of the clip-in/thumb hole. The "clamp" simply blew apart. And I wasn't using too much force. I've bounced micronuts with half the rating just as hard. My estimate is that I created between 1,5 and 2kn of force.
First of all, and this might sound like bagging on you, but I'm not trying to disrespect you: however, in this country, it "wasn't too hard of a bounce test" isn't a scientific term and isn't really even repeatable by you. I've had one of these pull out on bodyweight that I thought should have held, but that conveys very little to the reader. What does that mean, "I pulled one with body weight"? I weighed @150 lbs back then. It really means very little.
Furthermore: I don't know how you can "..estimate is that I created between 1,5 and 2kn of force" unless you've tryed doing something similar on a calibrated tester in the same location. Falling on gear even a short distance can easily create very high forces. I'm not saying that you didn't create a low force, but that a higher force than you quoted would be easy to get to with even a short pull test.
Rock and Ice tested toprope falls and MEASURED forces as high as 1300 lbs, which surprised them and everyone else. I use my ball nuts ocasionally, and if I am forced to use the little ones, I make an attempt to back them up as soon as possible, as we all should. I just used the #1 last weekend on a short new route that went free, but if thats all that will fit, it is a great tool to have. My only compunction on using them on a FA instead of fixing a pin or using a bolt, is that now pretty much everyone who follows will need that piece.
I appreciated that you could break one of these and that you shared it, but just to let you know, you are not the first and I suspect you will be far from the last. I'd be interested if someone could find and show some of those short fall force tests that have been done, which may surprise you at how high a force you can get with a short daisy fall.
Anyway, take care all
Bill
|
|
|
|
|
frodolf
May 26, 2008, 7:50 PM
Post #73 of 78
(2602 views)
Shortcut
Registered: Oct 1, 2006
Posts: 81
|
A bounce test isn't a scientific term in my country either, duh. Tell me of one place where it is. And it wasn't my intention to be scientific.
My estimates of the force I created are, as you point out, dubious. I didn't have any intruments on me at the time to register the exact force. Sorry. But it was pretty much the kind of bounce test that breakes the wire of a #1 copperhead, a wire that is rated to the vicinity of 1,5-2,0 kn. Hence my estimation.
There is a difference between ripping a piece and breaking a piece. Mine broke, no question about it, in a way it wasn't supposed to. But I must say – and it has been my intention to post it here a while – that the service from Trango was great! I sent them some pictures and they sent my a new bullnut, totally free. Sweet!
F
|
|
|
|
|
stymingersfink
May 26, 2008, 7:51 PM
Post #74 of 78
(2602 views)
Shortcut
Registered: Aug 12, 2003
Posts: 7250
|
maldaly wrote: Hey CCEBeling, in part because of your near miss, we posted video instructions for the Cinch and many other of our products on the website. Check them out at: http://www.trango.com/videos.php Climb Safe, Mal
|
|
|
|
|
jt512
May 26, 2008, 8:12 PM
Post #75 of 78
(2594 views)
Shortcut
Registered: Apr 12, 2001
Posts: 21904
|
maldaly wrote: There are a lot of different ways to rate things, 3-sigma being a good one. If you want to use it go ahead. I don't like it because with most sample groups all it does is allow a specified percentage of the sample group (about 11 in 10,000 units) to fail below the rated strength. I don't like failure so I use the minimum breaking strength. If I ever test a unit and it fails below the rated strength I'll lower the rating on the whole bunch....I like to think that minimum breaking strength is an improvement over 3-sigma.
The minimum observed strength is not an improvement over 3-sigma. I suspect that by using the minimum observed strength your intention is to be more conservative than using the 3-sigma strength. However, if your sample size is not large enough, then the observed minimum strength will likely be greater than three sigma. Conceptually, this is easy to see: the smaller the sample of units tested, the less likely it will be that the sample will include units with unusually low strengths. In particular, since less than 1 unit in 700 will be weaker than -3 sigma, it is clear that all the observations in a small sample will likely be greater than -3 sigma. Hence, a rating equal to the minimum observed strength will likely be greater than -3 sigma.
Of course, if your sample size is large enough, the situation reverses: the larger the sample, the more likely that the sample will include a unit that is weaker than -3 sigma, and hence, the observed minimum strength will likely be the more conservative rating.
So, how big a sample size do you need to be confident that you are being more conservative than 3 sigma? Since the units tested are a random sample, there is no sample size that guarantees that the sample will include a unit weaker than -3 sigma; hence, you can never know for sure whether you are being more conservative or not. However, we can easily calculate the probability that a sample of size n includes a unit weaker than -3 sigma.
The probability of a randomly selected unit's strength being less than -3 sigma is approx. .00135. Hence, the probability p that a sample of size n includes at least one unit weaker than -3 sigma is given by the following formula:
p = 1 - (1 - .00135) ^ n.
Using the above formula we see that if you have tested n = 100 units, then the probability p that at least one of the units in the sample is weaker than -3 sigma is .126; that is, there is only a 12.6% chance that a sample of 100 units includes a unit at least as weak as -3 sigma. If you've tested 200 units, you still only have a 23.6% chance of having a minimum observed strength greater than -3 sigma. Want a 50/50 chance? Test 513 units. 75% chance? Test 1026 units. 95% chance? Test 2218 units. So, to be reasonably confident that your minimum-strength rating is less than -3 sigma, you have to test a pretty big sample.
Mal, you say that the 3-sigma system "allows" [13.5] units per 10,000 to be weaker than the rating. The fact is, though, that any reasonable rating you give to a piece of gear, there will always be a non-zero probability that a piece of gear will be weaker than the rating. What the 3-sigma rating does is admit this, and quantify what that probability is. Thus, when we buy a 3-sigma rated product we know exactly how reliable the rating is. In contrast, when you use a minimum observed strength rating, the customer does not know how reliable the rating is. As we see from he above math, the rating may be higher or lower than what the 3-sigma rating would be, depending on the number of units tested. By using this method you are actually giving the customer less information about the the minimum breaking strength he can expect than you would by using a 3-sigma rating, unless, of course, your sample sizes are quite large.
Finally, the expected value of the minimum observed strength decreases as the sample size tested increases. The more you test, the lower your ratings will have to be. No matter how many units you have tested, if you keep testing long enough, you will eventually find a new weakest unit, and have to lower the rating. Truncating the rating down to the next whole kN mitigates this problem. If you're lucky, your weakest observed unit breaks at "something".9 kN and you may never in the future observe one that breaks at less than "something".0 kN so your rating lasts forever. On the other hand, if your weakest unit was observed to break at "something".1 kN, you will likely have to drop the rating by a full kN eventually.
If you don't think that a 3-sigma rating is conservative enough, a more effective approach than using the minimum observed strength would be to simply use an n-sigma rating, where n is a number greater than 3. A 4-sigma rating would mean that 1 unit in 32,000 would be less than the rating. For 5-sigma, the figure jumps to 1 in 3.5 million; and for 6-sigma, it's a whopping 1 in 1 billion.
Jay
(This post was edited by jt512 on May 26, 2008, 10:42 PM)
|
|
|
|
|
|
|
|
|
