1. Hello and Welcome to The Quality Forum Online...Continuing in the spirit of People Helping People !
    Dismiss Notice
Dismiss Notice
You must be a registered member in order to post messages and view/download attached files in this forum.
Click here to register.

Control Limit Adjustment Rule

Discussion in 'SPC - Statistical Process Control' started by Pongsakorn, Oct 5, 2016.

  1. Pongsakorn

    Pongsakorn Member

    Joined:
    Aug 10, 2015
    Messages:
    32
    Likes Received:
    5
    Trophy Points:
    7
    If we found out that the "New" Control Limit from calculation is "Wider" than the current one, please advise if the Control Limit should be adjusted to be the new wider control limit or maintain the current one.
    This cannot be found in AIAG SPC Manual.
     
  2. Miner

    Miner Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    269
    Likes Received:
    193
    Trophy Points:
    42
    Location:
    Greater Milwaukee USA
    There are a few reasons where it would be appropriate to revise the limits:
    1. The existing control limits were preliminary and based on insufficient data, whereas the new limits are to be the permanent limits based on sufficient data.
    2. The process was deliberately changed and the change will be permanent.
    Otherwise, the appropriate response would be to determine why the common cause variation has increased and address that increase to return it to normal levels, not to revise the control limits.
     
    Pongsakorn and Bev D like this.
  3. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    366
    Likes Received:
    369
    Trophy Points:
    62
    Location:
    Maine
    Miner is correct. In general we should not accept more variation unless there is a solid rational reason for it.
     
    Chris Glover and Pongsakorn like this.
  4. Pongsakorn

    Pongsakorn Member

    Joined:
    Aug 10, 2015
    Messages:
    32
    Likes Received:
    5
    Trophy Points:
    7
    Hello Miner and Bev,
    Really thanks your comment and advice.
     
  5. ncwalker

    ncwalker Well-Known Member

    Joined:
    Sep 21, 2015
    Messages:
    248
    Likes Received:
    148
    Trophy Points:
    42
    Location:
    North Carolina
    I will throw in my $.02 .... remember that "drift" is not necessarily variation OR bad. I often see people use diamond reamers to cut holes in soft metals. These things take FOREVER to wear out. They launch the process towards the large end of the hole knowing that the reamer will wear over time and make smaller and smaller holes. Depending on volumes, this can take YEARS. So initially, compared to the tolerance, they get very little variability. Cp north of 10. When they calculate control limits based on this, they are very, very small compared to the tolerance. They run the job, the reamer wears and the hole gets smaller. They violate their LCL because it was established with a very low standard deviation, yet, they are MILES from the lower limit. Does this mean they have to replace the (expensive) reamer? No. You re-evaluate the control limits and continue. Should you establish control limits much wider at the start? No. Because other things can still go wrong. Example - you can have a chip get in your tool changer and when you pick the reamer up, it will wobble. Loose control limits may allow this, the tighter ones are more apt to catch it.

    So your first step is to determine WHY the control limits have gotten wider. If your cutter is wearing (or the equivalent, you didn't specify a process) you can be seeing more variation as the process runs because it is drifting long term, but short term is OK. (And in this case, the appropriate reaction would be to shift the control limits, not widen them). But if last month between X parts you got Y variation and now between the same interval X you are getting Y++ variation, something has changed. Your reaction to both of these root causes should be different. And like Bev says, your decision should have solid, rational reasoning.
     
    Pongsakorn likes this.
  6. Miner

    Miner Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    269
    Likes Received:
    193
    Trophy Points:
    42
    Location:
    Greater Milwaukee USA
    You are quite correct in that scenario, though I would then question the need for SPC at all.
     
    Pongsakorn and ncwalker like this.
  7. ncwalker

    ncwalker Well-Known Member

    Joined:
    Sep 21, 2015
    Messages:
    248
    Likes Received:
    148
    Trophy Points:
    42
    Location:
    North Carolina
    You are clouding the issue with logic. :) I also would totally question the need for SPC. My customers on the other hand ... they sometimes ask for SPC on something we have robustly poka-yoked.
     
    Pongsakorn likes this.
  8. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    366
    Likes Received:
    369
    Trophy Points:
    62
    Location:
    Maine
    Well tool wear is a common issue for SPC (See anything Bod Doerring has written) BUT the approach is to have sloping limits (Wheeler's approach) OR Bob's approach which isn't statistical but based on the tolerances. The scenario of continually 'widening' or changing the control limits for tool wear isn't a very effective approach. I prefer sloping limits - in which case the limits don't widen - but Bob's approach also works very well.
     
    Pongsakorn and ncwalker like this.
  9. ncwalker

    ncwalker Well-Known Member

    Joined:
    Sep 21, 2015
    Messages:
    248
    Likes Received:
    148
    Trophy Points:
    42
    Location:
    North Carolina
    I do like Bob's approach as well. It handles variability and mean shift very simply in a way the shop floor can understand. My experience with SPC is that while the quality engineers typically get it, the knowledge and understanding isn't out there with the operators and the decision makers on the floor. I mean, they see "control limits" and tend to treat them like specification limits. Sloping limits are also good. I like to plot the mean and the standard deviations with a rolling method and monitor that against spec limits. Kind of like sloping limits the other way. And, easy to do now that we have computers. Back when Shewart came up with the things, it was all slide rules and tables. But I find even that sometimes gets lost on the floor because they aren't interested in understanding the stats. But they eat Bob's approach up. What I find "best" is something like Bob's method or Shainin's Precontrol to "run the floor" and make the immediate "continue or call for help" decisions. And the SPC is used for more macro analysis like "which machine runs the job more successfully" or root causing more external type effects like a change in coolant, weather effects on circ water, change in cutter brand, etc.
     
    Pongsakorn likes this.

Share This Page