I don’t think there’s a need to come up with a GR&R study with an automated system since this will be the basis already by itself and it even has high resolution capturing time components.
What you need to focus on is the maintenance of the system itself such a mean time before failure, mean time to repair, among others.
If your data is not normal and it cannot be transformed anymore, try getting the AD test results of each distribution such as Weibull, among others then the distribution which has the best AD should be considered in getting the Cpk.
Take note that using percentile in a non-normal data can only be applied if basing on the graph, the…[Read more]
Computing for Cp and Cpk is already in relation to the target, which in return is usually defined by your customers. If your target was already defined by the customers, then use those.
If not, baseline your process, be sure everything is under control, then get the limits and average of your process or confidence interval then use that as your…[Read more]