iSixSigma

Established Control Limits

Six Sigma – iSixSigma Forums Old Forums General Established Control Limits

Viewing 4 posts - 1 through 4 (of 4 total)
  • Author
    Posts
  • #28625

    woey
    Member

    How to establish control limits (UCL, LCL)? Do I use the UCL&LCL from the X-bar chart or individual moving average chart? How many sample do I need so that the UCL/LCL is statistically valid?

    0
    #71559

    Jim Johnson
    Participant

    The control limits are statistically defined as 3 standard deviations above and below the mean of the data sampled.  In the case of the X-bar chart, the UCL/LCL is 3 STD DEV above/below the average of the sub-groups of the data.
    To answer your question (I think) you would always use the UCL/LCL on the chart where it appears.  It exists to determine when a process is statistically in or out of control, not whether the process is meeting specification limits.  The specification limits are set by the customer.
    Bottom line, is you use the UCL/LCL to determine statistical control on a process.  Use the specification limits (customer requirements) to determine whether the process is meeting customer expectations.
    I hope that this helps. 
    Jim

    0
    #71571

    woey
    Member

    Jim: how many subgroup of data that I need so that I can use the UCL/LCL on the charts? The reason I ask because I need to establish UCL/LCL for a process on the production floor for in-process inspection. I want to be able to justify using UCL/LCL on the charts. Do I need to use different set-up at the machine, then collect more data, and create UCL/LCL and use it so that it is the representation of the process? Thanks!

    0
    #71593

    bob farey
    Participant

    The choice of chart is dependent on the quantity and type of data. Ten data points plotted on a chart (this may be ten sub-group means and ranges or ten single readings) give a reasonable indication of the state of the process. Fifteen to twenty give a strong indication and 25 to 50 is ideal. It gets more complicated if the process is, or appears to be, out of control, since data values due to special causes need to be eliminated before you can calculate the true underlying mean value and the degree of natural variation. Always remember that if you use control charts to monitor processes, you always know much more about the process than someone who doesn’t !

    0
Viewing 4 posts - 1 through 4 (of 4 total)

The forum ‘General’ is closed to new topics and replies.