iSixSigma

Range Chart in Control, Average Chart Out of Control

Six Sigma – iSixSigma Forums General Forums Tools & Templates Range Chart in Control, Average Chart Out of Control

Viewing 12 posts - 1 through 12 (of 12 total)
  • Author
    Posts
  • #55449

    Ed
    Participant

    Hi all,
    I’m having challenges with my pharmaceutical product. 5 samples per batch were sampled from the same machine and pH values were plotted on SPC chart. The range chart is in control but the average chart is out of control. The data set is attached below.

    Batch pH 1 pH 2 pH 3 pH 4 pH 5
    1 5.6 5.6 5.6 5.59 5.61
    2 5.57 5.56 5.56 5.56 5.57
    3 5.58 5.58 5.57 5.57 5.57
    4 5.56 5.57 5.58 5.58 5.58
    5 5.54 5.54 5.55 5.53 5.56
    6 5.5 5.51 5.52 5.49 5.48
    7 5.6 5.59 5.58 5.59 5.57
    8 5.56 5.56 5.58 5.54 5.55
    9 5.6 5.58 5.58 5.57 5.59
    10 5.6 5.61 5.58 5.59 5.57
    11 5.588 5.588 5.612 5.604 5.588
    12 5.557 5.571 5.557 5.557 5.559
    13 5.573 5.571 5.575 5.579 5.584
    14 5.573 5.575 5.575 5.586 5.575
    15 5.546 5.56 5.569 5.574 5.547
    16 5.559 5.544 5.544 5.543 5.557
    17 5.587 5.564 5.587 5.565 5.586
    18 5.574 5.576 5.581 5.574 5.575
    19 5.552 5.552 5.559 5.568 5.553
    20 5.569 5.569 5.568 5.578 5.568
    21 5.556 5.558 5.556 5.571 5.561

    I would appreciate it if anyone could comment on the attached analysis (whether it is an appropriate analysis), and suggestions on how to move forward to monitor the pH values.

    Thanks.
    Ed

    0
    #200046

    Lew Yerian
    Participant

    I’m OK with your choice of tools. Not sure what the dot plot gives you, but that’s OK.

    Looking at your data: Your process mean is shifting too much. Your batch-to-batch variation is too large. Eliminate what the problem cannot be, and whatever is left must be the source of your variation.

    Questions you need to ask yourself before moving forward:
    – Are you using a validated measurement system? If not, then none of your results are valid. I would address this question first.
    – Are your sampling methods sound (is the within-batch sampling homogeneous? For the most part, the values look pretty much the same)
    – What is different about the batches? Especially Batch 6. Why is the mean shifting from batch-to-batch?
    – What is the statistical vs. practical significance of what you are seeing? Are you measuring tiny variations in pH when it doesn’t really matter?

    0
    #200047

    Chris Seider
    Participant

    Confirm you’ve kept the data in time ordered sequence…if you look, it seems your within variation is typically more than your between variation.

    Have you done a gage R&R on the pH? Make sure where the process is being sampled isn’t impacting the pH result.

    A few aspects to consider.

    0
    #200048

    Ed
    Participant

    Thanks Lew and Chris.
    GR&R completed – OK
    Sampling method – 5 consecutive products from the same batch
    Data are in time ordered sequence.

    Chris,
    By within variation do you mean the range chart?
    Between variation the moving range chart?
    Range UCL = 0.045
    Moving range UCL = 0.069
    Please refer to attachment.

    Thanks.
    Ed

    0
    #200049

    Chris Seider
    Participant

    Within variation is the Range chart you originally posted. The between variation is your Xbar chart you originally posted.

    0
    #200051

    Ed
    Participant

    Thanks Chris for clarifying. Under what circumstances would it be appropriate to use the I-mR-R/S (between/within) chart?

    Lew’s comment “For the most part, the values look pretty much the same.” Does it suggest autocorrelation?

    Thanks.
    Ed

    0
    #200054

    Lew Yerian
    Participant

    I wasn’t thinking of autocorrelation. I was wondering if there is an issue with either the resolution of the gauge or maybe with the sampling. If I draw my sample from the same location in a vessel, will I get a very similar pH result?

    0
    #200055

    Ed
    Guest

    By gage resolution, do you mean too many decimal places?
    By sampling at the same location, you will get similar pH result.

    Thanks.
    Ed

    0
    #200057

    For a subgroup size of 5 you need at least 5 units of discrimination in your range chart, in your case you have (0,0.01…0.04).
    The range chart is in control so the measurement system is predictable and repeatable.
    The xbar chart has 7/21 points OOC – this means that your measurement varaition is quite large compared to your process variation (batch to batch), it would be better to have more points out of control (say 50+%). Remember the control limits come from the average range between repeat measures – the wider the control limits, the more measurement variation and the more likely the xbar points will be within the limits.
    As your batch to batch variation improves – when you improve your manufacturing process you’ll likely need to improve your measurement system.

    0
    #200061

    WIlliam Howell
    Guest

    Here are a couple comments.

    From my experience with pH meters, this is a very uniform across batch variability. This seems mostly like measurement variability.
    There is a lack of measurement consistency, with 1, 2, or 3 significant places shown.
    What is the target value and specifications for this process?

    0
    #200067

    Ed
    Participant

    The pH spec is 5.5±0.5

    0
    #200069

    Sergey
    Participant

    I would not take care about process of Cpk about 15 :) I believe there are a lot of other processes and measurement systems that require improvement

    0
Viewing 12 posts - 1 through 12 (of 12 total)

You must be logged in to reply to this topic.