Hello everybody,
I would like to get some feedback from the ocmmunity - most likely somebody was solving a similar problem at some point.
I have a set of measurements/corrections that are done more or less at set intervals on number of tools. We see that the measurements are not very far from each other each time for one category of measurements and we would like to extend the lifetime of each measurement, so the intervals between measurements are longer. In other words we want to prove that it is still safe to some degree NOT to do those measurements as often.
So far what we want to do is to set 2 sigma control limits for the data set and then for each tool and for each point calculate time it needs to hit the first OOC point. Then just take interval right before that so that 95% of all points would have been measured BEFORE they hit OOC.
Does that sound reasonable? Or there are some other methods to determine this "lifetime" and numerically prove that it is safe to extend the "lifetime" of this measurement?
Any ideas would be highly appreciated.
Thanks,
M.
One thought for you is to take a look at a CUSUM approach to your practical problem. Essentially it sounds like you are trying to set up a process monitoring and alarm system for a given critical to quality characteristic's performance. The new CUSUM platform in JMP 14 might be of use? You'll have to establish what are called detection limits...not control limits. This is guided by some practical consideration for how much deviation from a target value are you willing to live with, with an assumed process variance. There is a very cool Profiler within the platform for you to use with different charting parameter conditions and a ARL profiler element as well. The new CUSUM platform can be found at Analyze -> Quality and Process -> CUSUM Control Chart. For more details in the JMP documentation I suggest visiting here:
https://www.jmp.com/support/help/14/cusum-control-charts.shtml#
