We are trying to PQ our tablet presses and are having trouble determining acceptance criteria.
We test 10 tablets every 30 minutes for weight, hardness, thickness using a TTS system.
We have action limits and lower and upper specifications. This is during normal production.
For our PQ it was suggested that in addition to capturing RPM and other settings we also perform a CPK analysis of a sample set of data.
I am no six sigma expert, but from what I have read PPK analysis would be more appropriate over a course of a batch
What I was thinking was doing the following.
Use batch size as total population
derive sample population from a .99 confidence interval and a 5% sample error.
Determine significant sample population and gather TTS data over the course of the run to satisfy the sample requirements.
Perform PPK analysis for each parameter (weight, hardness, thickness) and shoot for a number of 1.33 which would be 4 sigma. We would consider this passing because a 1.33 PKP would be 99.99367% and 68ppm defects. Pending approval from quality.
My question is this. Does this make sense? is this proper use of all of these statistical principles? Is it reasonable to derive a conclusion that our process is under control and effective if we acheive a 1.33 pkp.