iSixSigma

How to set QA specifications?

Six Sigma – iSixSigma Forums Old Forums General How to set QA specifications?

Viewing 8 posts - 1 through 8 (of 8 total)
  • Author
    Posts
  • #40427

    Jeremy
    Participant

    Hello,
    I have noticed that the QA specifications were set quite some time ago and that because of raw material, process change and especially a lack of monitoring, the process has shifted to what it is today.
    I wonder if someone could tell me how specifications should be set today depending ofΒ process characteristics (mean, stdev). Any advise would be appreciated or references.
    Many thanks
    Regards
    Jeremy

    0
    #125410

    eliz
    Participant

    Jeremy,
    If the process has changed (modified / redesigned etc) the customer would have been told?
    Customers usually set your specs for you, (eg – i want it this big / small / size / weight / volume etc etc) Speak to them, as they probably aren’t happy that they are receiving product that is out of spec (old spec)
    what happened regarding the monitoring? does the CEO not care about the product that the customer receives? – can you eloborate more please.
    Regards
    E

    0
    #125411

    betty
    Participant

    Eliz,
    what happened if we are receiving product that is out of spec or out of control limits? ” Receiving product that is out of spec” and ” receiving product out of control limits” are the same thing?
    B

    0
    #125413

    Akram
    Participant

    There is no statistical relationship between the control limits and the spcification linits.
    The specifications are set by the customer requirements while the control limits are determined by the process.
    try this http://www.sixsigmafirst.com/controlcharts.htm

    0
    #125414

    eliz
    Participant

    Hi Betty,
    no, not necessarily….
    If you are receiving product that is ‘out of spec’ that is, out of the specifications limits that are set to adhere to your requirements (as a customer) then that product is failing to satisfy you (as a customer) ‘out of control limits’ is different (but equally as damaging in some respects) but ‘out of control’ does not necessarily mean ‘out of spec’
    eg
    your spec is LSL = 20, USL = 40, if the product is out of these limits, you (as the customer) will be unhappy & most likely reject the product.
    however, if the same product / process (based on the same ‘spec’ of LSL = 20, USL = 40) has control limits of, eg..
    LCL = 26, UCL = 37 (based on a control chart)
    & you get various data points that are outside of these ‘control limits’ eg 24, or 38, the product will not be ‘out of spec’ i.e you (as the customer) would not reject the product (as its within your spec) but would be ‘out of control’ for them data points & would need aanalyzing & addressing.
    Hope that makes sense!
    E

    0
    #125416

    betty
    Participant

    Thanks for the answer Eliz. then i have another question.as we all know, if the mean and the st.deviation of a normally distributed sample are known, it is possible to predict the proportion of the population that will fall between any two limits. for instance ,any normal distribution will have %99,73 of the population in the area under the curve between three st.deviation below the mean and three st.deviation above the mean. Similarly, the area within +/-6s is 99,9999998 % of the total. In essence, this implies that almost 0,0000002 percent of the product would be outside the tolerance. all we talk about are the outside of the tolerance limits and the defects. well then, what happens the homogeneity between the products if we set the process width as +/-6 sigma about the mean instead of the +/-3sigma?
    In other terms, if we multiply the standart measure for process width by six, we accept/increase the product that is distant from the mean,doesnt?
    Betul

    0
    #125418

    eliz
    Participant

    Hi Betty,
    yes, but why would you want to do that?. That is why a six sigma process would still create 3.4PPM, +/- 6s & you would most likely not have any points out of control, as the limits (U + L) would be so vast!!! (which is your last point i believe) thus creating 0.2PPM (& everyone happy with the process!!) & the only thing that you would probably look into (via a control chart) were trends of the data points (3 in a row above / below / cusum ascending / descending etc)
    not sure that helps? ! but i think it sort of agrees with your last point!
    E

    0
    #125479

    betty
    Participant

    Hi Eliz,
    if we expand the limits to six sigma, we will sacrifice the homogenity of the products. yes i agree with you that the number of the products accepted as defected will decrease, but then the homogenity of the products will decrease also(because of the products added with specifications far from the mean). This is important for the statisticians but the lack of products accepted as defected is more important for the employers. My question is, how appropriate is it for a statistician to sacrifice from the homogenity of the products?
    B

    0
Viewing 8 posts - 1 through 8 (of 8 total)

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