iSixSigma

GageRR: Percent StudyVar versus Percent Tolerance for Non Statistians

Six Sigma – iSixSigma Forums General Forums Tools & Templates GageRR: Percent StudyVar versus Percent Tolerance for Non Statistians

Viewing 11 posts - 1 through 11 (of 11 total)
  • Author
    Posts
  • #53611

    van Helden
    Participant

    In our production process we measure the quality of a liquid product and have a specification target of 15 +/- 10%. Thus the spec.limimts are 13.5-16.5.

    I have done a GageR&R study on this measurement system and the results are:
    – %StudyVar = 5%.
    – %Tolerance = 49%.

    The question is how should I interpret these results with respect to StudyVar and Tolerance. Knowing that this has to be explained to people with no knowledge of statistics.

    In addition, we are able to measure 15 +/- 0.75 (5%) while the customer spec is 15 +/- 1.5 (10%). Now 0.75 is 50% of 1.5. Is this because the %Tolerance is almost 50% or does this relationship not exists.

    Thanks
    J.

    0
    #190866

    Quesnel
    Participant

    % tolerance is how much % of the available tolerance is gobbled up by the Measuring Systems Error.

    % Study Variation – is comprised of components (ie – Total GR&R, Repeatability, Reproducibility, Part-toPart) – I am assuming that the 5% is Total GR&R % (but I may be wrong??).

    If this is the case, you can say that the GR&R Variation (Total) / Total Variation * 100 = 5%.

    Or the Total GR&R – % Study variation is 5% of the Overall Variation.

    I hope this helps you out!

    Eric

    0
    #190867

    Mikel
    Member

    Assuming that you choose your samples correctly, you’ve got a very incapable process.

    The bad news is you have difficulty judging good/bad, but a MS capable of studying your process for improvement.

    0
    #190877

    van Helden
    Participant

    Thanks for your replies. I understand the StudyVAr, meaning that only 5% of the overall variation is taken up by the measurement system. I also understand that our process is uncapable – that’s what I figured out from the initial data analysis (and hence was the reason to start a project).

    But I am still trying to get a better understanding of %Tolerance. If I am correct %Tolerance is calculated as follows:

    %Tolerance = 6 x StdDev. / (USL-LSL) x 100%

    Can I translate this as follows:

    %Tolerance = Total MS variation / allowed variation x 100%

    And we want this to be a small number (<10%) to make sure that the MS variation is only a small part of the allowed variation in order to be able to detect on/off spec.

    Is this a correct way of stating (taking into account that English is not my native language)

    0
    #190878

    Mikel
    Member

    Yes

    0
    #190879

    van Helden
    Participant

    Ok, now I would like to put this into the context of the project (and try to sell it to senior management that the MS is not capable, while they already celebrated the success of this new process):

    The target is 15 and we have R&R of 5%, thus 0.75. Let’s assume this is constant over the whole measurement range. With respect to the USL of 16.5 is it safe to say that with any measurement between 17.25-15.75 we can not distinguish between on or off spec. Simular to the LSL of 13.5 with any measurement between 12.75-14.25.

    Thus only if we measure a value between 14.25-15.75 we can assume it is on spec in all other cases it is off spec or we don’t know.

    Is this correct…?

    0
    #190880

    usa
    Participant

    “Rat in the Wheel” approach !

    0
    #190881

    van Helden
    Participant

    Well bbsua, as you could read in one of my replies I am not a native English speaker. I don’t know what ‘Rat in the wheel’ means. I would prefer constructive feed back on my questions / statements. So if you think it is wrong what I am saying, then please educate me.

    The reason why I keep digging into this issue is because I have to convince senior management that we have a serious issue with the current MS (and the process). Whilst the same management already declared this process as a big success. Now the challenge is that I am working in a large company which does not have a six sigma program. In addition, the company is located in the Middle East (and in this part of the world failure does not exist and people don’t make mistakes). The management here is not interesteed in MSA, Measurement system error, %Tolerance etc at all. The only thing that talks is $$$$$. So somehow I need to translate the MS error into dollars.

    thanks
    J.

    0
    #190882

    Nick
    Participant

    I was once told that it costs $1 to fix a problem if it is detected in raw materials before production. It costs $10 to fix a problem if it is detected in finished goods before they leave the production line. It costs $1000 to fix a problem once it has been delivered to your customer. Very arbitrary but illustrates the point that the earlier you detect that something is outside of specification or ‘defective’ the cheaper/easier it is to correct.

    How was the 15 +/- 10% specification limit determined? Were these set up by someone in the production process or set by the customer? If set by production then they might be tighter than necessary and could be relaxed enough to make the measurement system more capable (and reduce the amount of product labeled out of spec). If the customer defined the specification limits then you could look at customer rejection rates, warranty claims against the product, etc. Perhaps you could determine the cost of product sent back to you (and all of the transportation costs, investigation costs, …) and compare it to the cost of detecting the defect on the production line.

    This would be the cost of not detecting a defect. This isn’t perfect as someone could argue that this cost is also dependent on how often you test production but it might be a start.

    0
    #190886

    van Helden
    Participant

    nmudd,

    Some clearification. The spec limits are defined by our customer, which is an internal customer. The spec limits are based on models. And will probable discussed when I present the MSA results.

    The process is not as you described that the customer sends back products. The product is deliverd at the location by pipeline and used instantaneously – regardless it is on/off spec. There are 40 pipelines and there are samples taken for QC from 4 pipelines a day. Therefore I need to calculate the cost implication of our incapable MS/process.

    0
    #190887

    Nick
    Participant

    That makes things more challenging. Do you think that, even if you could assign cost, you could get leadership to take any action? If your customer never rejects any product then there is no incentive for producing on spec product and no penalty for producing out of spec product. Why would leadership want to spend money to fix a process that their customers seem happy with?

    Can your internal customer make the connection between off spec material entering their process and defective finished goods? If so, this might be an opportunity to team with your internal customer to assign cost to on/off spec material. From your description they either cannot measure the incoming material well enough to detect the different between on/off spec, fail to act upon the information, or the specs are modeled improperly. Regardless, all affect the cost and quality of the finished goods and should be a concern.

    0
Viewing 11 posts - 1 through 11 (of 11 total)

You must be logged in to reply to this topic.