Demonstrating Capability in 'Goalpost' World

S

SteelWoman

Demonstrating "capability" in "goalpost" world

Question (and if you reply, please talk to me like I'm a moron, my SPC skills are "decent" but not strong :rolleyes: ) In our world of cutting steel to a certain length or width, our customers supply us with what in our world we call "goalposts" - the high and low tolerances for their cuts. As long as we hit it ANYWHERE within those goalposts the customer is happy, we are happy, all is good.

My problem is this: We have a new line and I have been dutifully gatherning data points from various production runs over a 6 month period (we tend to do very brief runs, move on to something else, then hit that same order again later when the customer needs it). Anyway, I have over 1400 data points representing a 6 month period in which the vast majority of the data points are at 31.75", with the customer specification of 31.75 to 31.875. Of course when I plug this into my statistical software, because it all hugs the min, I'm "incapable", based on it's estimation of how many times we will fail - though the reality is we have not ever failed, customer is happy, we can almost produce this stuff in our sleep now.

Our registrar is coming in a few months to add this new line to the scope of our certification, and of course she'll be looking for "capability studies." Any advice on how to present this, since we ARE capable but "statistically speaking" we aren't?

Hope that explanation makes sense and any help you can offer will keep my gray hair from going white! :(
 

SteelMaiden

Super Moderator
Trusted Information Resource
To the other Steel -

Would that make you a mini me? sorry could not resist;)

There are formulas you can use for targets at some determined minimum or maximum values. That being said, I don't have that info available now, but there are several folks here that appear to my only semi-stat-literate mind pretty darn good at statistics, I'll be they will chime in eventually, and I'll be waiting for those formulas myself.

Unfortunately, I left my stat-pak at my previous div.
 
S

SteelWoman

"Mini Me"! I love that - lol... I needed a good chuckle, thanks!:biglaugh:
 
M

M Greenaway

Interesting that you say you are capable, but statistically you arent !!

I am confused by this statement, you are either capable or not.

My hunch is that you are saying you have determined the performance of your process, and some parts produced would be outside the specification limits, yet the customer has accepted all parts. I would simply suggest you change the tolerance to suit what you know is actually acceptable. Failing this simply adjust your process so that the mean is centred in the middle of the tolerance band.
 
A

Atul Khandekar

Since a 'vast majority' of your readings are the same, I suspect the resolution of your measurement method. What's the least count? Total tolerance you are working with is 0.125" (31.875-31.75) Can the measurement system distinguish parts within 0.01" (1-10 rule)?
 
C

Cristi?nC

I have the same feeling that Atul Khandekar has.

If I remember well, in the MSA manual there is an index named % of Contribution due to resolution (or discrimination).

It is calculated dividing the instrument resolution by the process variation (6 * process std. dev.) (usually the number 5.15 is used instead of 6).

For example: %C = 1 [mm] / 6*Process Std. Dev. [mm]

If this % contribution (%C) is greather than 30, then the measurement instrument must be replaced by other with greather resolution. If the %C lies between 10 and 30, it is acceptable, depending on the criticality of the measured characteristic. Finally, if it is lower than 10, then the resolution of the instrument is optimum for such a process.

An interesting characteristic of such an index is that if you continuously reduce process variation, some day your actual measurement instruments will be useless...

Which is the %C of your measurement instrument? (I am asking the same question as Atul, by using other words...)

Hope this help!
 
S

SteelWoman

Chris and Ata, I will have to perform that calculation - like I said, my SPC skills aren't Herculean - I was hired for my other talents! :D We actually use a tape measure (because the parts produced off this line are often in excess of 12 feet long each) which is the preferred instrument of our customer. Nope, we can't distinguish .01 on a tape, obviously, but neither can our customer who uses the same instrument. We actually tried, at first, using a VERY large caliper, but there were insane amounts of variation that we tracked to the simple fact that a caliper of that size was just too **** difficult to use (cumbersome, heavy, required two users for every measurement).

No, there are NEVER any parts out of spec - not a single one of the readings has been outside the customer's tolerances. It's not that the customer has accepted out of spec parts, it's that they've never gotten any.
 
M

M Greenaway

Steel

How can you say that you have never supplied out of spec lengths when your measurement system is clearly incapable ?
 
S

SteelWoman

This is one of those things where the world of Widget makers and the world of Raw Material makers don't speak the same languages. Like I said in my original posting, in our world (and by "our" I mean HERE and at OUR CUSTOMER) as long as the material is between those goalposts it is "good" for the customer and we are "capable" of producing it to their specs. I was happy to see the latest revision of the PPAP manual where provisions were made for raw material suppliers - we are a "different beast" from the world of parts makers and sometimes what is cut and dry in one world isn't so in another. I guess bottom line here is what does the CUSTOMER want? Here they want material that measures on a tape between A and B. That's what we provide.
 
Top Bottom