Cp / Cpk on position using multiple MMC bonuses

[Proud Liberal]

Does anyone have a good approach to including MMC bonuses on both feature and datum sizes into a Cp and Cpk calculation.

 

[Miner]

This is a difficult question to answer as there are numerous methods with no consensus on which is best. One approach uses the average bonus tolerance. Another uses the residual tolerance (bonus tolerance - position). And there are several, more complicated approaches. Each approach has its own shortcomings. However, the biggest shortcoming of all the approaches is failure to deal with violation of the normality assumption. Most position measurements are far from normal, being highly skewed, which invalidates the capability index.

 

[Bev D]

Here's where I channel my Mother: if all of the other kids are going to jump off a bridge, will you?

less cryptically: why do you need a Cp/Cpk number? If it's just to satisfy your Customer then ask them which method they endorse. In addition to the methods alluded to by Miner, you can always calculate the index from the categorical data on pass/fail. it's a heck of a lot easier and simpler and probably answers what the Customer probably really cares about: yield.

If you really care about process capability I find that a simple chart of values against the spec limits with the parts that are accepted using MMC in a different color and the failed parts in a third color to be the most insightful analysis