Specifications and Variation

Six Sigma – iSixSigma Forums Old Forums General Specifications and Variation

Viewing 25 posts - 1 through 25 (of 25 total)
• Author
Posts
• #39412

Raju
Participant

We have a certain quality characteristic whose nominals and tolerance limits are say 100 + / – 5
Does this mean that this characteristic if it falls between 95 and 105, is it ok, or else the 6 Sigma ( + / – 3 Sigma) should be 10 with Cp = 1,

if we are asking a supplier to supply with this sepcifications, and drawing is released with these requirements then the drawing should say 100 + / – 5 with Cp=1 or just 100 + / – 5 is sufficient implying Cp=1, any thoughts

Let us say the data is

98,104,98,103,96,98,102,96,104,101: all these samples are acceptable according to the supplier, but if you see general statistics associated with this we have

Average =100

Range = 8

6 Sigma ( + / – 3 Sigma) = 18.97366596

and if has lot of variation

0
#119741

Titu John
Member

Hi Anjali,
Your problem is quite simple when ever you are stating your requirement yo should mention your specification limits i.e. USL & LSL , a supplier having  SPC experiece has to fix his production within the spec limits which you have mentioned i.e. his UCL & LCL so to have a good capable process he should have his Cp & Cpk  values of 1.33 or higher .
Cp is a ratio of the customer specification tolerance to six standard deviation of the short term process variation.
Cpk is a ratio of the diatance between the process average and the closest specification limit , to 3 standard deviations of a short term process variation , it is often refered to as process performance.

0
#119746

Raju
Participant

Thanks Titu.
You mentioned that we should give USL, LSL
but with the data I posted eventhough all the samples
are within the limits the Specifications, but still have large 6S
and hence Cp value also not good, in this case if we state drawing requirements as nominal 100 +/- 5.00 is it sufficient or we have to mention Cp=1, please clarify with your expertise Thanks

0
#119757

Mike Carnell
Participant

Anajali,
You seem to be making this more difficult than it needs to be. The Cpk value is telling you how the process is performing right now. Cp tells you that this is the best the process could perform. The current process is not capable (Cpk) and does not have potential (Cp).
Your data says that the process is centered (which is not always desireable) so your issue is size of the standard deviation. If you divide the specification width of 10 by 12 (+/- 6 standard deviations it will tell you how much of a reduction you need in the standard deviation to get to a six sigma capability.
Good luck.

0
#119760

Raju
Participant

Thanks Mike, my question is when we put requirements in drawing
100+/-5 means does it imply +/- 3sigma is +/- 5 or the samples
can lie between 95 and 105 and +/- 3 Sigma doesn’t matter,
are these things totally different

0
#119764

Mike Carnell
Participant

Anjali,
Stick with the number of std dev inside the specification limits.
Just imagine buying parts where 5 are at 95 and 5 are at 105. You still have an average of 100 everything is inside the spec limits but check out that std dev (it should be greater than 5 so just a +/- 3 sd is 3X the width of the spec). Sooner or later a process that is running rail to rail is going to spin out of control (actually it already is).
Just my opinion. Stick with the standard deviation.
Good luck.

0
#119766

Raju
Participant

Hi Mike, please specify what kind of requirements do I need
to put, because supplier is saying he can run is process post to
post, that is the reason I brought the issue of putting on the drawing
as 100+/-5 with Cp and Cpk =1 is that correct.

0
#119768

Bob Barone
Participant

“100 +/- 5%” means any value within those limits is equally acceptable.
“100 +/- 5% and Ppk >=1.33” requires that the mean value of the lot be distanced at least 4 std devs inside the nearest specified limit (i.e., std devs of the lot distribution – including mu drift over time – not point in time capability).  Ppk forces simultaneous attention to specified limits, standard deviation, and mean placement.  Suppliers can easily demonstrate compliance by supplying with each lot a histogram with Ppk calculations.

0
#119770

Issa
Participant

Bob BaroneI can tell that you are not very much acquainted with Dr Taguchi if you believe that “”100 +/- 5%” means any value within those limits is equally acceptable.”
This is a wrong statement. All the parts produced within these specified limits are not identical, do not operate the same way and therefore are not all “equally acceptable”. If one part is 95 and the other one is 105, how can you conclude that they operate the same way?
If the target is 100, anything within the specified limits that is not 100 is just “good enough” not perfect and not “equally acceptable”.
Issa

0
#119771

Raju
Participant

Thanks Bob,
So in the drawing if we don’t mention anything other than
just 100+/-5 (Nominal +/- Specs) means, supplier can send
samples in the 95 to 100 range without bothering about the spread,
(the spread could be +/- 3 sigma or +/- 6sigma), which is what
supplier is contending that we need to specify what is our Cp and Cpk
please throw some light in this aspects.
Thanks once again

0
#119772

Bob Barone
Participant

“100 +/- 5%” appearing on a specification or drawing means literally that any value within that range is acceptable.  It is a documented declaration of what you will accept.  That interpretation has been true for the entire history of specifications.  It is WHY such a specification stated as such is wholly inadequate.  It is precisely why Dr. Genechi Taguchi developed the Loss Function Equation.  His concept was conceived to oppose specified limits stated simply as “100 +/- 5%.”  If you require distribution centering within your stated limits and you do not specify that requirement in inescapable terms, then you deserve everything your supplier sends to you within “100 +/- 5%.”
Imagine how quickly your case would be tossed out of a courtroom if your spec said “100 +/- 5%”, the supplier then demonstrated that every piece shipped to you fell within “100 +/- 5%” and then you said to the judge, “But I wanted the distribution to be centered.”  That’s what the spec is for.  That’s why we write specs.  That’s where you get to say it and have it mean something.  Stating a Ppk requirement provides that.

0
#119774

Raju
Participant

Thanks Bob,
so to completely specify
we should do clearly state in the drawing, Nominal + /- (USL,LSL),
With Ppk (for Centering the Process and Process spread)
is that correct, thanks

0
#119775

BTDT
Participant

Anjali:
Yes, continue with some kind of continuous measurement of capability rather then the discrete measurement of within/outside of spec.
If the vendor performs within specs, then the conditions of the (poorly written) contract have been fulfilled. It may not be the best for a continuing relationship. Pushing the bounds of the specification is a tricky thing to do.
I am reminded about a story of plastic sheeting used for protecting buildings. The specs were relatively wide, one vendor was centered while the other was at the low end of the thickness. Both were ‘within spec’. A decent sized storm came up from the Gulf and all the sheeting from one vendor was destroyed, while the other remained intact.
BTDT

0
#119776

Bob Barone
Participant

Yes.
However…I have to say…since they are balking at your discussion regarding distribution centering, it is possible (even likely) that they already know they are delivering parts to you from an out-of-control or incapable, hodgepodge process (they may even be sorting to meet your current spec limits).  Don’t be surprised if they tell you they just plain can’t meet a Ppk requirement.

0
#119786

magesh
Participant

Anjali,
It is like this….
– If you want a 6 sigma process from the supplier, find out what tolerance (control limits and not the specification limits) should be given to the supplier and put that in the drawing itself.
– I mean, if your drawing dimension calls for 100+/-5mm which is very vague… tell the supplier that his control limit should be +/-2mm (i just gave this figure as an example… in fact you can find out exactly what this control limit should be …if u require a Six sigma process). By this way, all these kind of Cp, Cpk jorgans can be avoided (since small suppliers won’t understand these things very easily).
– On the other hand; if the supplier is knowledgeable, tell him that my Spec. is 100+/-5mm. But I need a Z-score of 6 which is nothing but a 6 Six Sigma capability. Hope you know how to calculate this Z-score.
– Another one method is, if there are 100 dimns. in your drawing, 10 may be critical, 20 may be major and the remaining 70 may not be that much important.
– You can declare that Critical dimensions require a Z-score of 6, Major require a Z-score of 5 and others 4.3 etc….. and the other way is to say that Critical dimensions need Cp, Cpk >2 ; Major dimns need Cp, Cpk > 1.67 etc…
Hope it is clear now…
If you need further inputs please mail me to…..
[email protected]

0
#119788

Anonymous
Guest

Anjali,
I don’t want to confuse you, but I should like to draw the attention of the forum to an ‘alternative view’
Let’s say that I have a supplier, and he uses low cost equipment to achieve a ‘contractual specification’ using poka yoke. If the process does not have sufficient ‘natural process capability’ then the 100% inspection will obviously give rise to some losses or muda.
Now a Six Sigma guy comes along as says .. no inspection is full-proof and I would like you to improve the equipment capability to achieve a Cpk = 2.
OK they say … but this will quadruple the price of the component.
What should you do? My answer would be to minimise the loss to society …. supplier and customer!
Good luck!
Andy

0
#119800

Raju
Participant

How do you reduce the loss to the society,
is it Taguchi loss function, I also see
Cpm and Ppm concept like
Cpm=Usl-Lsl/(6*sqrt(sigma square+(average-target)square)…
But how to estimate the loss,
please throw some light on this,
Thanks

0
#119802

Anonymous
Guest

Anjali,
I am going to stick my neck out again and advise you not to use Taguchi’s Loss function. Although I support the use of his signal-to-noise ratio, I don’t like his loss function because I don’t believe losses can ever become infinite.
My advice would be to find out more about your suppliers process. If they are relying on human inspection then there is an obvious problem, and I wouldn’t blame you for wanting them to improve their process capability.
On the other hand, if they are using Jidoka, then I should consider their losses against the cost of any potential improvement, either an improvement project, or by purchasing equipment with improved performance. (Remember, most of the time new equipment can a new set of problems, so you should find a way to reduce the risk.)
Good luck!
Andy

0
#119811

FTSBB
Participant

Regarding what should be on the drawing, the companies I have worked with had a small note telling whether the dimension was critical or not.  The notes vary by company: some put a small “boat” by the dimension, other label it as a “CF”, indicating a critical feature.  If the dimension was deemed critical by the customer, they would require some predetermined quality metric, usually a Cpk value.  I’ve worked with required Cpk’s between 1.33 and 2.0.  Just depends on the customer.

0
#119841

Anonymous
Guest

FTSSB,
You’ve made an important point ….
In other words, not all values within the contractual tolerance have the same quality, which is why some sub-assemblies, such as gear boxes, work better if components have a central tendency.
A supplier using poka yoke would not be able to achieve this central tendency, and lower gear box noise, unless they used a number of poka yoke devices to match mating parts!
Andy

0
#119846

Mike Carnell
Participant

Anjali,
I tend to agree with Bob. I have a problem with your statement about Ppk centering the process. It will not guarantee centering. If you specify a certain Ppk it will guarantee you a certain distance in terms of Std dev from the closest specification limit.The smaller the standard deviation  the closer to the specification the supplier is permitted to move if there is an advantage to them i.e. plating a precious metal – the supplier will want to run as close as they can to the lower specification. With good control (small Std Dev) they can run extremely close and still comply with the specification and Ppk requirements.
BTW – a Cpk of 1 as you stated in an earlier post would be pretty loose. You might want to run some tolerance stackups with some Monti Carlo simulation and see what you can live with.
Good luck

0
#119867

Raju
Participant

why did u say centering of the process with the Ppk is incorrect,
ie doen’t that indicate where mean value is distanced with respect to
spec limits, please explian in lay man terms,
Thanks

0
#119869

Raju
Participant

Question from this :
Titu JohnPosted on: Thursday, 19th May 2005, 11:37 AM.
Hi Anjali,

Your problem is quite simple when ever you are stating your requirement yo should mention your specification limits i.e. USL & LSL , a supplier having  SPC experiece has to fix his production within the spec limits which you have mentioned i.e. his UCL & LCL so to have a good capable process he should have his Cp & Cpk  values of 1.33 or higher .

Cp is a ratio of the customer specification tolerance to six standard deviation of the short term process variation.

Cpk is a ratio of the diatance between the process average and the closest specification limit , to 3 standard deviations of a short term process variation , it is often refered to as process performance.

Hi All:

I was wondering why do we need to give UCL,LCL limits to the supplier
And how to calculate the UCL and LCL limits,

0
#119872

Mike Carnell
Participant

Anjali,
Ppk has to do with position with regards to the spec limits but that position doesn’t necessarily have to be in the middle. If we use your example of 100 +/- 5 you have a specification width of 10. If I have a std dev of 0.83 that will give me about 6 std dev to each spec limit and position the mean in the center.
If that number is say a weight of a part that is some exotic metal such as inconel and it is expensive then it makes sense for me to run to the lower side of the specification and take out some cost. If I reduce my standard deviation to 0.4 I can produce parts with an average weight of 97.4 and not put my customer at a significantly higher risk than the process that was centered with a std dev of 0.83. Both processes have six std dev between the mean and the closest specification ( there is actually a difference but it isn’t of any practical significance). The point is the lower std dev allows me to run in a more optimum position for my business (the mean shifted lower by 2.6) which allows me a higher margin and the customer gets what they asked for (parts at 100+/- 5 and a Ppk of 2).
Good luck

0
#119874

Mike Carnell
Participant

anjali,
I have never seen specifiying control limits on a drawing we have produced or one we got from a customer. You can accomplish the same level of understanding betwwen yourself and a customer by specifying a minimum level of Cpk/Ppk.
You need to decide what the purpose of the drawing/spec is. You want to convey what you need for your process. You understand better what you need than they do. When you specify control limits you are on the edge of telling them how to run their process. Who knows more about their process you or them? (they should and probably do)
The best move you could make is have your supplier in, show them your process, show them your parts and explain specifically what you want then both of you agree how you will specify your need on the drawing.
Just my opinion.
Good luck.

0
Viewing 25 posts - 1 through 25 (of 25 total)

The forum ‘General’ is closed to new topics and replies.