Ppk Issue - Some Products have Unilateral Tolerances

[bobdoering]

should i be adjusting the process after every 10th piece measurement? I mean if it is going out of tolerance I would be adjusting it back to the LCL...today i faced a bit of an issue as the 1st 10 odd pieces started trending out below the lower specs...

You need to adjust when you hit your 75% of tolerance - no matter when that occurs.

If your tool wear should indicate the parts get bigger as the tool wears, but the parts got smaller, then what you are seeing is start up/warmup variation. This is why start up/warm up is a special cause. It is not stable until the process comes to equilibrium, and requires special attention during that time - including more frequent checks. You may need some patience to get to that point of steady-state process.

Did the process ever "settle down" during the run?

Have a strong belief there are a lot of special causes which need to be looked into before i carry this excercise again..agreed?

It is hard to say from a distance, but at least you are paying attention to the things that count now!!

my question though is - should i select line with max variance or min variance for this?

I suggest using the technique on a line with less variance to get used to it and learn how it works, then go the the line with more variance and use the tool to sort it out.

 

[abhipatel]

i would send you a data set i have taken from before...but i dont think it would be much help as they as mentioned would have a lot of spcl causes...

anyways i believe you would be able to help me eye it better and also would be able to help me understand what the underlying distribution is....

if i run it with the current dataset in the distribution analyzer would it be able to tell me if it is uniform distribution? or would i have to do it after special causes are removed?

 

[bobdoering]

if i run it with the current dataset in the distribution analyzer would it be able to tell me if it is uniform distribution? or would i have to do it after special causes are removed?

You would need to not have the special causes, for sure.

Here is another problem. Look at the effect of "sampling error" on trying to use curve fitting. In order to correctly identify the uniform distribution, you can only use data from complete "cycles". If your run did not generate a complete cycle, or if you do not omit portions of cycles, you will distort the true distribution. For all practical purposes you simply need to know that the sawtooth curve you expect from adjusting to tool wear does generate a uniform distribution.

 

[abhipatel]

jeezz......process currently circles so much up & down...really need to c if it is actually overcontrol or it is special causes...am having a big fight on my hands with the traditionalists who believe that any data can be converted to normal..

simply say if it is not normal - convert it and use cp and cpk ......

guess by the time i remove the special causes and run a distribution analyzer they would have driven the heck out of my mind......

anyways their question to me now is "what can be termed as precision machining"?

bob, do u have any info @ insert/tool selection for facing (turning) operation?

 

[bobdoering]

jeezz......process currently circles so much up & down...really need to c if it is actually overcontrol or it is special causes...am having a big fight on my hands with the traditionalists who believe that any data can be converted to normal..

simply say if it is not normal - convert it and use cp and cpk ......

Darn normalcentrics. It is a pitiful way of saying "I don't understand any other distribution but normal, so if it isn't normal you must transform it." Even Shewhart said: "The total information is given by the observed distribution.” -Economic Control of Quality of Manufactured Product

I take it these guys are non-automotive, because the AIAG books cover this issue nicely...or at least start to cover it.

anyways their question to me now is "what can be termed as precision machining"?

That's an easy one. From my presentation:

DEFINITION OF PRECISION MACHINING
-A process where material is removed by a cutting surface – such as grinding, honing, turning, milling, etc.
-The process must be controlled in a manner that all variation (vibration, bearings, gage error) is statistically insignificant except tool wear.

bob, do u have any info @ insert/tool selection for facing (turning) operation?

No - every process is unique (material, feeds, speeds, etc.), but the key is the tool wear slope. The shallower the slope (less wear per part) the more stable the process.