Bimodal data and Ppk?


Hoping someone here can help me with a query.
I completed destructive testing analysis from a manual process today.

The product i'm looking at has several features which must meet acceptance criteria of Ppk >1.33

When I ran the data for one feature it resulted in a bimodal model. The data is showing across people there is a distinctive difference due to technique. (It's a process heavily dependent on operator input so not unrealistic to expect this result)

My query is, all specifications are met but due to the bimodal model the Ppk criteria cannot be achieved. Is there another way to calculate Ppk based on a bimodal model with P value >0.05 ?

Thanks in advance!


TS Contributor
no, this is a clear sign that your process is not standardized and hence not capable. The point of the Ppk measure is to show that your process stays reliably within spec - and due to operator dependence you cannot do that.

It is not clear from your post whether the problem comes from the measurement error or the process itself - but this does not really matter, unfortunately.



TS Contributor
rogojel is correct. You have a mixture of different process streams. The correct approach is to adjust the means of the different streams to a common target to eliminate the bimodal effect. Once that is done, you may then calculate the appropriate capability index.