SPC (Statistical Process Control) Overview
- Thread starter Steve Prevette
- Start date
A
artichoke
Before jumping in with solutions, the first step should be to gain an in depth understanding of SPC. My recommendation is to buy or borrow:
Understanding SPC - Wheeler (this is the SPC "bible")
Advanced Topics in SPC - Wheeler
Out of the Crisis - Deming
A
artichoke
Take care with your selection of reading material. The EFCOG document makes the very common mistake these days, of failing to distinguish between what Deming calls enumerative and analytical studies. This is what Deming described as "poor teaching of statistical methods in industry" ( page 131, "Out of the Crisis" ). This failure has taken much of industry backwards. Demings lessons have been poorly learned. Perhaps Deming's book should be your first read, before getting into the deeper stuff with Wheeler.
P
pearsonow
Artichoke - I cannot agree more, i am re-reading Out Of The Crisis at the moment, as we are introducing SPC to some of our suppliers, and i want to ensure i teach it t them in a way that has meaning - not just gathering data for it's own sake.
I will have to look into Wheeler's work also. Has anyone got any suggestions for suitable reference material for short run based data?
I have read most of the Injection moulding related stuff in the SPC forum (guess where I've been told to look first). I am spending 2 day next week, before returning home, with one of our moulders - the longest time period of my visit - and am now wondering if what we orgininally set out to do is the correct method.
I will post in more detail shortly but will probably start a new thread as this will be a long and ongoing discussion I feel. I also intend to get some time off over the weekend, 15.5 hrs flying time, I am going to do some sightseeing!
I will have to look into Wheeler's work also. Has anyone got any suggestions for suitable reference material for short run based data?
I have read most of the Injection moulding related stuff in the SPC forum (guess where I've been told to look first). I am spending 2 day next week, before returning home, with one of our moulders - the longest time period of my visit - and am now wondering if what we orgininally set out to do is the correct method.
I will post in more detail shortly but will probably start a new thread as this will be a long and ongoing discussion I feel. I also intend to get some time off over the weekend, 15.5 hrs flying time, I am going to do some sightseeing!
A
artichoke
I don't know what your particular problem is but you might find Ch 6.2, Wheeler "Advanced Topics in SPC", on Rational Subgrouping, with an example of injection mouldings with 5 consecutive press cycles, to be of interest anyway.
A
artichoke
Bob,
I checked your link and came to this "... run the data through a distribution fit analysis, and find the best fit ..."
If you read Dr Wheeler's book "Normality and the Process Behaviour Chart", Wheeler describes very elegantly and in great detail, how this statement is utter nonsense. It is just one small part of the Six Sigma scam.
Wheeler also published an article in the Quality Digest recently on the subject.
The above statement is the sort of nonsense that has sent quality backwards to the days before Shewhart. It seems that in almost a century, little has been learned.
I'm happy to discuss if you need but I'm sure that reading Wheeler's material will bring you to an understanding more quickly.
Please, I posted that article on the forum. I own most Dr. Wheeler's work, including the one you cited. Most of it is very good and handy. But his statement - if read closely, mentions that you do not need to know the distribution "to make SPC work", but by no means does that suggest you do not need to know the distribution to understand capability or if any of the Western Electric rules are applicable to a particular process output among many other valuable pieces of information it provides. His point is as long as you chart you will find the signals of variation you need.
The notion of understanding the process distribution has nothing to do with six sigma - since that area is befuddled with normal-centric thinking. They rarely care about finding the correct model for their data.
You statement of "nonsense" is really unsubstantiated, but that is understandable. SPC takes more than regurgitating text, it takes thinking. Keep reading...it may sharpen your thinking. You have a good start, but have a ways to go.
A
artichoke
Wheeler explains on page 34 "Normality and the Process Behaviour Chart" that it takes at least 3200 points to test a distribution to 2.95 sigma ... and by the time you do, the process and its distribution will most likely have changed. For that reason, trying to determine the distribution of data for a process is a waste of time, that is, "nonsense".
Six Sigma adherents are concerned about "normalising" the data so that control charts may be used. This is also "nonsense". The "correct" distibution is not necessary for Shewhart charts. Wheeler proves this with an analysis of 1143 different distributions and shows, as Shewhart stated, that a knowledge of the distribution is not needed.
Similar threads
