D
We have a new customer that has asked for our specifications to be updated to include a measurement of the leakage rate of our parts. The parts open and close and the spec is related to how well sealed they are when closed. To perform the test a sealed part is placed in a chamber that is rapidly evacuated to some negative pressure, after a given period of time the pressure is recorded, if the parts leak the pressure in the chamber will rise. The increase in chamber pressure/time is the "leak" rate of the part.
The customer is allowing us to set the specification and we have agreed to set it such that we will achieve a Cpk (what Minitab reports as Ppk) of 1.33. However, I have two issues. First the possible measurements for this leak rate, barring equipment failure, are bounded, a perfect seal would give a leak rate of 0 but less than perfect seals always result in a positive leak rate. How do get Minitab to handle a one sided specification?
Second the data is not normally distributed, mostly due to the fact that the data is bounded and the target is the lower bound. I have found that a 3 parameter Weibull fits the data very nicely and sort of solves the issue of the bounded data. However, when I run the non-normal capability analysis with a spec of 0 (the boundary) and any upper spec higher than ~164 I always get the same Ppk value.
Any ideas what I can do? Is there a limit to the achievable Ppk for a Weibull distributed process?
The customer is allowing us to set the specification and we have agreed to set it such that we will achieve a Cpk (what Minitab reports as Ppk) of 1.33. However, I have two issues. First the possible measurements for this leak rate, barring equipment failure, are bounded, a perfect seal would give a leak rate of 0 but less than perfect seals always result in a positive leak rate. How do get Minitab to handle a one sided specification?
Second the data is not normally distributed, mostly due to the fact that the data is bounded and the target is the lower bound. I have found that a 3 parameter Weibull fits the data very nicely and sort of solves the issue of the bounded data. However, when I run the non-normal capability analysis with a spec of 0 (the boundary) and any upper spec higher than ~164 I always get the same Ppk value.
Any ideas what I can do? Is there a limit to the achievable Ppk for a Weibull distributed process?