Dear All:
Im in the final stages of validating an automatic inspection machine. After running a Knapp test, I demonstrated that the automatic method is more efficient than the manual method, being the efficiency of the automatic method of around 95%.
However, questions arised around how that estimation of efficiency could apply to a real batch. I tried to prove it by estimating the amount of defectives that enter the process and the amount that “exits” from it, and then evaluate efficiency from those numbers.
The problem is that by calculating efficiency that way, it gets nowhere the 95% obtained from the Knapp test. My guess is that the statistical estimation (or the sampling method used to gain information) isnt powerful enough to give accurate estimations, since it shows that the automatic method is more efficient when there are more defectives at the start, unless this is a common behavior. Even then, the AQL of the outgoing product is still under acceptance criteria.
Moreover, there are new questions. Is there a point in proving the efficiency of the machine in that way? from my point of view, if the automatic method is more efficient than the manual method, even when it is not enough to eliminate all the defective product, as long as it works in a consistent way, it should be already validated.
So, how had you been validating this kind of processes?.
Thanks in advance.