SPC (Statistical Process Control) Overview

[Jim Wynne]

In Reply to Parent Post by Steve Prevette

As a practical example, let us say that I am working on a construction project that is killing one worker per month on the average, and that is stable on a control chart.

The rate of deaths of workers is STABLE and PREDICTABLE. I can project forward and say how many deaths will occurr. This may sound morbid, but in the early 20th century there were rules of thumb of how many deaths to expect per million dollars of project cost.

I hope everyone will agree here that a stable non-zero rate of killing workers in not ACCEPTABLE. The stable rate is thus NOT CAPABLE.

Stability and capability are concepts that apply to processes, not to the means of counting them. It makes no sense to say that a count of accidental deaths isn't capable--it's the process that causes the deaths that's not capable, assuming zero deaths is the criterion. Whether or not deaths (or any other undesirable outcomes) is "acceptable" is a red herring. The is math concerned only with what is, and not what should be. We can predict that there will be x airline crash fatalities per y miles traveled, but the prediction (and its statistical foundation) have nothing to do with whether or not any number of deaths is considered acceptable.

[sekaran]

how do create spc Specification Limits?

[Steve Prevette]

I assume you are asking how to create SPC Control Limits (not specification limits). The specification limits, if any, come from the customer requirement (cut a board to a 2 foot plus or minus 1/8 inch length).

I have a writeup on the internet on how to conduct SPC, including generation of the control limits from the data at http://www.hanford.gov/rl/?page=1148&parent=1144. There are also plenty of other writeups in stats textbooks, and other postings here on the Cove.

[artichoke]

In Reply to Parent Post by sekaran

how do create spc Specification Limits?

Your question is a little unclear but this may help: "How to Establish
Manufacturing Specifications"

http://www.spcpress.com/pdf/Manufact...cification.pdf

[bobdoering]

In Reply to Parent Post by Steve Prevette

I assume you are asking how to create SPC Control Limits (not specification limits). The specification limits, if any, come from the customer requirement (cut a board to a 2 foot plus or minus 1/8 inch length).

I have a writeup on the internet on how to conduct SPC, including generation of the control limits from the data at http://www.hanford.gov/rl/?page=1148&parent=1144. There are also plenty of other writeups in stats textbooks, and other postings here on the Cove.

If you are doing precision machining - and you are using correct process control, the calculations for control limits are straight forward: 75% of the specification. To see why, and for more information you can read: Statistical Process Control for Precision Machining

[artichoke]

bobdoering,

I'm curious as to why you have a logo "Stop X-bar/R Madness" ?

Machining, where constant tool wear effects results (Machine shops, SIC 3599, one of over 400 industry codes) is obviously a special case outside the general discussion. Amongst the hundreds of industries there may also be other special cases, such as those involving PID control, which fall outside the general discussion of control charts.

[bobdoering]

In Reply to Parent Post by artichoke

bobdoering,

I'm curious as to why you have a logo "Stop X-bar/R Madness" ?

Machining, where constant tool wear effects results (Machine shops, SIC 3599, one of over 400 industry codes) is obviously a special case outside the general discussion.

I always find it amusing that precision machining is considered such an insignificant "special case".

Even if so, so what?

It exists, and its practitioners need to understand that they have been sold a bill of goods with traditional charting systems. The madness is specifically for those that have been driven to use that chart when it does not apply - specifically precision machining. It has driven people in precision machining mad - with ridiculously compressed control limits and over and under reported capabilities. Most have no idea what to do or where to go, and it has been my goal to correct that problem.

I do not doubt that there are truly naturally occurring normal processes, that meet the requirement of being set to a nominal value (mean), and stay at that value (some values a little lower, some values a little higher) without any interaction by an operator. There may be a bunch of them in the other SIC codes - good for them. They can use the myriad of textbooks to get the information they need. They were written just for them. I have even stronger belief that there are other exceptions - but that others affected by them have, too, been drilled into the normal fallacy, and the associated charting methods.

My point (and the point of my avatar) is to rid the madness in precision machining. I find it a very important cause, no matter how minute as a ratio of industry codes (as if they are equally distributed, and as if 200 or 300 more SIC codes might not have captured machining in their industry) or by any other "measure". This excuse of "being an exception" has been used by the SPC software suppliers, which do not provide the correct tool for precision machining to use - leaving them to fend for themselves with paper charts or Excel spreadsheets - or worst, the wrong control because they paid for garbage software sold by people who have no idea about how to control machining.

Others may have the opportunity to correct the rest of the world's statistical problems - at least for now.

Often when marketing a solution to a problem, you have to catch someone's attention to the fact that they have a problem. Apparently, your attention was caught - so there is clearly a level of success.

Besides, I was told I couldn't use my old avatar any longer.

[artichoke]

In Reply to Parent Post by bobdoering

I always find it amusing that precision machining is considered such an insignificant "special case".

Please read my post. I did not say that machining is "insignificant". I stated that machining is one of more than 400 industry codes.

Mis-using any chart may "drive people mad" as you suggest.

... but that others affected by them have, too, been drilled into the normal fallacy, and the associated charting methods.

The belief in the normality fallacy widespead. Processes do not need to be normal for the use of Shewhart control charts. I strongly recommend Don Wheeler's excellent book on the topic: "Normality and the Process Behaviour Chart".