Skip to content
Rektron

Search

    Situation in production

    Values creep towards the limit

    Every part passes, but the measured values drift in one direction as the tool wears.

    What happens

    The measured values for a characteristic slowly drift in one direction, for example towards the upper tolerance limit, while the tool wears. Each individual part still passes.

    What should apply

    The process should run stably around the nominal value, with a spread that fits well within the tolerance.

    Why it is difficult

    • A single measurement looks normal. The pattern over time is the signal.
    • The difference between normal variation and a real change does not show in a table of values.
    • The person who sees the value at the machine does not always have the history, and the person with the history is not at the machine.

    The context that is needed

    • Earlier values for the same characteristic
    • Nominal value, tolerances and control limits
    • Which tool is in use and when it was changed
    • Machine, order and shift

    When does a slow change become a decision, and what happens afterwards?

    Illustration · Example data
    USLUCLNomLCLLSL3457

    The values pass all the time. The signal is the pattern: six values in a row rise (Nelson rule 3), and it comes before the control limit and the tolerance limit are reached.

    Step by step

    1. What should apply: the values should lie around nominal, within the control limits and far from the tolerance limits.
    2. Normal variation: the values spread around nominal.
    3. Signal: six values in a row rise. All still pass.
    4. Context: which tool, how many parts since the change, which shift.
    5. Decision: the operator decides whether to adjust the process, change the tool or keep running.
    6. Action: the tool is changed.
    7. Follow-up: the values are back around nominal. If the pattern recurs, the change interval is reviewed.

    Same situation, different questions

    Who needs to know what

    The same event looks different depending on which decision you have to make.

    The same evidence for everyone

    • Earlier values for the same characteristic
    • Nominal value, tolerances and control limits
    • Which tool is in use and when it was changed
    • Machine, order and shift
    • Operator

      Asks

      "Does the part pass, and should I adjust now?"

      Decides

      Adjust, change the tool or keep running.

    • Quality

      Asks

      "Is the process still stable and capable?"

      Decides

      Whether the process needs action and whether parts already made are affected.

    • Production

      Asks

      "Will this lead to a stop or scrap if nobody acts?"

      Decides

      Whether to schedule a tool change.

    • Engineering and planning

      Asks

      "Why is the tool behaving differently, and should the change interval be adjusted?"

      Decides

      Whether the tool change standard should change.

    • Management

      Asks

      "Does this recur on more machines, and what does it cost in scrap and stops?"

      Decides

      Whether the tooling strategy should be reviewed more widely.

    What may follow

    1. 01Compensate or change the tool
    2. 02Schedule the tool change
    3. 03Change the interval if the pattern recurs
    4. 04Check that the values return towards the centre

    Where Rektron can help

    What is available today

    In Eyesight the measured values are shown in trend charts and Xbar-R with control limits set per characteristic, and Nelson or Western Electric trend rules mark patterns with the rule number at the point before the value reaches the limit. The rules apply to manually entered values. Cp and Cpk are tracked per characteristic. In-depth capability studies are available in Rektron SQC.

    Do you recognise the situation?

    Tell us what it looks like where you are, and we will go through what is needed to handle it.