Moving Range charts and capability?

[SteelMaiden]

OK, I'm out of my area of expertise...like way out!

Can a process capability be figured on a moving range with a lower specification limit but no upper?

The situation:
Once again, instead of finding a root cause, finger pointing is happening. We are producing nonconforming material for yield results during physical property testing (tensile tests). The metallurgy dept. states that our tensile tester must be goofed up, but looking at a moving range chart really does not show anything alarming other than the fact that the specification is 44000 min. and our Xbar is 46000. Coupled with the fact that we just performed calibration on the machine and extensometers, I tend to say the problem lies with our processing not our testing.

Can anyone give me any other thoughts on how to prove that we either need to adjust our chemistries or our rolling processes and forget about the tensile tester as being the cause? Or if I take my charts to the next product quality team meeting, will they be sufficient?

My only in-depth SPC experience deals only with sizes using subgroups and that have both an upper and lower specification, so I am not really confident with my knowledge of single reading, single specification statistics.

Thanks guys and gals, I know someone out there can point me in the right direction.

 

[Mike S.]

Steel,

I'm not sure I understand your problem completely, but here's a shot at it anyway since no one else has jumped in. Cp can still be calculated on a unilateral tolerance as (mean - LSL)/(3*std. dev.) or for you (46000 - 44000)/ 3 sigma.

I'm very familiar with the test being blamed when the product starts to look iffy, so I can appreciate your dilemma. I assume you don't have samples from a previous "good" batch that you could test today to see if the results repeat? I learned long ago (the hard way) to keep some "good" samples available for just such a possibility whenever possible. Until the issue of the measurement being bad is off the table, little progress will be made in other areas.

How about sending a sample quickly to someone else who has a tensile tester?

 

[Darius]

The process capability can not be figured on a XR chart, at least if you follow to the gurus of SPC, because they don't like specification limits into the SPC chart, they say that SPC has nothing to do with specification limits.

The traditional way to use the SPC chart is to know the within subgroup variation and the between subgroup variation of the variable and to show patterns into the data that help to understand the process variation and show how to control it.

The traditional way to show capability is with histograms, with the specs into it, there are other ways but almost the same, I like box plot's across the time (each box for each period of time) so it's like having histograms one in front of the other to show the trends across the time (I have seen box plot charts that show the highest values at the beginning of the month and lowest at the end because of the calibration of the measurement equipment).

I don't know about tensile tests, but it may be that the distribution is not normal, Wheeler say that the translation between cpk and ppm, the better must not be done because of the lack of normality and assumplions of the data, so the way to obtain ppm is just count the data points outside of the spec limit and calculate the % of non conforming units, and if you like capability index and you have a single specification, my recomendation is to use Cpmk (because Cpk don't behave fine on one spec conditions), you can chart it (the index) and see if any significant changes in the capability happen.

If "the metallurgy dept. states that our tensile tester must be goofed up", the control charts could be of help if you show that the behaviur is the same across the time (using control charts from different periods or a long period of time including the period that the tester suppouse to be working ok), but I preffer the box plot.

I hope this may help

 

[Chris May]

Steel,

not much time left today, but you could do worse than have a look at (broken link removed)

Lots of useful stuff about control charts in general.

Sorry this is brief, but I have to dash.

Hope it helps...a bit.

Regards,

Chris

 

[SteelMaiden]

thanks ya'll!

Just needed to get the input so I could be the "ASSERTIVE" evil twin this week

 

[Mike S.]

Steel,

Let us know how it all works out, huh?

 

[SteelMaiden]

OK, latest update. Number one, I impressed a few people by my knowledgeable presentation Maybe they thought I was just another pretty face?

By the time I got done plotting, graphing, analysing I pretty much was able to pinpoint exactly where we shifted the process. (not that the capability was great before then ) And once I started mentioning the lot numbers I found out that we stopped checking finishing temperatures about that time as "someone" felt it was not important what temp we finished at as long as we met dimensional tolerances.

Gosh I love it when a plan comes together! Thanks for all the assistance.