Debalancing machine - is SPC needed?

Debalancing machine is consisted from two parts:1 measurement instrument and debalancing part. ALL parts (rotors for DC motors) are measured (100% inspection). If balance is more than prespecified target (eg. > 40 gmm) then parts go through balancing operation that lowers the debalance below target (measured again). Parts, that go to the next operation, than comes from two distributions (supposed nonnormal):1. good parts from initial measurement and 2. balanced parts. How to measure the Ppk (I understand that Cpk can not be computed). Is it necessary to compute Ppk, since it seems all parts are good but parts from the first distribution can be near the target (but below!)

------------------
Vojko Krizman
SLOVENIA

[Laura M]

My initial thoughts are you have 2 separate calculations to do, because you essentially have 2 different populations. You would want to know how capable the rework process is.

Also wanted to mention that my Great grandparents were born in Slovenia!

Laura

Debalancing operation cannot be omitted.The result of this operation are parts that come from two distributions and we have 100% inspection. My main question is: If we have 100% inspection and all parts are good, do we have to compute Ppk, especially if all parts are good (but from distribution it comes out we have some fraction of nonconform parts!?)

[Laura M]

100% inspection is often the alternative if you do not meet process capability requirements. If you have a bimodal distribution due to 2 processes rolled in to one data set, then it is possible that each process is separatly capable, but when you combine them the standard deviation increases and the capabiiity index would indicate parts out of spec. Without knowing more about the prcocess, that's probably as much as I should say so I don't mislead you.

[Rick Goodson]

I agree with the advise Laura has given but wanted to add a thought. If the two separate processes are capable, why combine them and then and force yourself into a 100% inspection operation that may not be necessary? If you run both populations across the same secondary operation you could consider a 'short run' SPC technique such as a 'Nominal' or a 'Target' chart. Once again, as Laura stated, I would need to know more about the process before we could definitively say short run is the best option. If you want to investigate short run further try the International Quality Institute, http://www.i-q-i.com/index.htm or Dr Don Wheeler's website http://www.spcpress.com/

Adding to the Balancing Process Characterization Issue ...

How are you going about calculating the process capability for a polar oriented process (imbalance amount and angle) when the usual capability calculation assume a linear measurement? Of course, the polar coordinates can be converted, but you still have two dimensions instead of one.