iSixSigma

Choosing Parts for an MSA

Six Sigma – iSixSigma Forums General Forums General Choosing Parts for an MSA

This topic contains 14 replies, has 6 voices, and was last updated by  Mike Carnell 1 year, 1 month ago.

Viewing 15 posts - 1 through 15 (of 15 total)
  • Author
    Posts
  • #55827

    sam
    Participant

    Hello,

    We are preparing for MSAs and the topic of choosing the correct parts for the MSA came up multiple times. The general rule I have heard is to choose parts representative of your process. The parts we are receiving in from our supplier vary from 15-30 millionths on roundness while our spec is 100 millionths. How will this affect the outcome of the MSA if our parts don’t “stick out like a sore thumb” as my old mentor would say (XBar chart by op).

    Thanks.

    0
    #201843

    Chris Seider
    Participant

    well if the parts come from the supplier, you can always tell them what you’re doing and wonder if they have any poorer quality parts from startup or incomplete processing etc.

    get what you can….see how the gage does for the process variation and see if you see any semblance of MSA degradation for those closer to the mid range of your spec.

    i’d say do with what you have. however, one idea….if you calibrate your instruments, do you have any standards to measure in the range also?

    0
    #201844

    Alchemist
    Participant

    Hello.

    I am having a similar dilemma regarding Gage R&R. Im confused about the sampling. Ive heard that the samples you choose should be random, but here is my problem with that.
    This is a hypothetical situation:
    You work at a plant that manufactures steel cubes, you control your process by measuring the dimension of every ~100th cube you produce with a micrometer.

    You decide to do a GRR to determine whether your measuring method/equipment/appraisers are good enough. You choose 10 cubes randomly from 10 different batches. Here is the part that confuses me. You make different sized cubes. You make ones that have the dimensions 10mm x 10mm x 10mm and ones that are 15mm x 15mm x 15mm or 50mm x 50mm x 50mm all the way to 100mm x 100mm x 100mm. The specified tolerances are the same for each sized cube.

    If you choose randomly, your part variation% can be huge to almost non existent and in turn your GRR% can vary greatly too.
    Should you choose the samples “randomly” for each specification, and make a GRR for each spec? The problem with the latter is that if you have over 100 different types of measuring equipments and you have about a dozen different specification for each type then you would have over 1200 GRRs… If you were to do 4-5 GRRs each workday than you could finish all 1200 in a year (not taking into account the time spent for sampling, finding appraisers and evaluating the data).

    Can anyone help me with this? Any experiences about what customers or auditors want?
    Appreciate your input!

    0
    #201845

    Chris Seider
    Participant

    yes, pick your “parts” for each product. You can’t run a gage R&R and expect good understanding of % of process tolerance or % process variation–even if same tolerance width.

    0
    #201846

    Chris Seider
    Participant

    @alchemist444

    fyi, your question isn’t uncommon.

    no worries.

    0
    #201847

    Bob Doering
    Participant

    Mark your parts, and have everyone measure the same spot. The roundness of the part is not the gage’s “fault”, so you do not want it influencing the outcome of the gage r&r. However, you do want to capture the roundness in your measurement process with that gage (measure around the part and report the highest and lowest value. If you ignore the roundness in your measurement technique, that is not gage error, that is measurement error – when you take a perfectly good gage (passed gage r&r) and use it wrong!

    0
    #201848

    Bob Doering
    Participant

    As far as “picking your parts” – the parts should represent the variation over the life of the process – not just one lot. The best way is to use historical standard deviation for process variation in the gage R&R (see bottom of page 121 AIAG MSA 4th edition). It is much more accurate than any 10 pc sample.

    0
    #201849

    D Mckean

    While randomly selecting samples for regular testing and monitoring a process is the correct way to approach a sampling regime, the Gauge R&R is not executed to accomplish this. The purpose of the gauge R&R is to validate your testers and the test process therefore you should select samples for the gauge that represent the range of variation that the testers will encounter. This will allow you to also see if your test process is failing when the parts are in a specific range.

    0
    #201851

    Chris Seider
    Participant

    @bobdoering
    I don’t understand your input above. If one KNEW that the roundness of the part would affect results, this is part of the measurement system variation that would be captured. To “mark” where to measure defeats the purpose of a gage R&R.

    The gage ought to measure the process “as-is”. I wouldn’t advise them how to improve the gage as part of the study even before measuring.

    D McKean, you worded it very well. One might attempt to randomly sample parts produced but one typically makes sure to capture parts across the typical spectrum.

    0
    #201852

    Chris Butterworth

    I always take a very large sample, measure them all and then select my GRR sample from these. In this way, you select the highest, lowest and some in-between. The problem with random samples is that you can get two or three that are nearly identical in the measurement under study.

    0
    #201853

    Iain Hastings

    In response to the original poster: Are you looking for product control or process control? My interpretation is that you receive parts from a supplier and you are inspecting them for roundness and comparing to specification limits.
    If this is the case, then you are concerned about product control and the variation in roundness between parts is not important as you would be comparing precision to tolerance.
    If you are comparing to process variation then yes, you would like to select parts that represent the process variation, or compare your GRR results to the historical variation of the process.

    0
    #201856

    Bob Doering
    Participant

    Chris Seider: First, you find out if the intrinsic gage error is small enough by doing the gage r&r on specific locations on the part. Then, you use the gage correctly by measuring and reporting the diameter high and low values so that the measurement reflects the roundness. If the roundness is insignificant to the process variation, you may choose to ignore it as a part of measurement. You don’t simply ignore roundness and let it roll into total gage error, unless – I guess – you intend to use to gage incorrectly. But, I certainly would not recommend it. That is not “as is” condition of the gage, that is simply poor measurement. Gage error and measurement error are two very specific different things. Gage error – measured by gage R&R – is the intrinsic error within the gage in its use. Measurement error is when you take a perfectly good gage and use it wrong – as in ignore roundness or taper.

    0
    #201858

    Chris Seider
    Participant

    We disagree @bobdoering

    I am stating that one measures the gage variability as it’s used–not as it’s ideally used.

    I don’t discount your thinking if the process is found unacceptable (% process variation or % tolerance) using the “as-is” process. Solve the problem by breaking the measurement system down to various parts of the measurement process.

    0
    #202845

    Urajan
    Participant

    From AIAG MSA Manual 4th ed pg.74 stated as ” the sample parts must be selected from the process and represent the entire production operating range

    Until today many of us has this dilemma on sample selection. The definition is not well defined by AIAG. Many of the people fail to address the definition precisely and quite vague. Hope someone can really help to clarify this. I would like to know what does it mean “represent the entire production operating range” as in the context of AIAG procedure.

    0
    #202849

    Mike Carnell
    Participant

    @urajan Regardless of what AIAG says you need to spread the samples out. You need to be across the spec limit and outside the spec limit.

    If you have a random number generator make some sets of numbers and run them so you can see the effect. Try running a bunch that are the same size and in the middle of the spec. See how that one turns out.

    I am not sure I would be as worried about AIAG compliance as I was about having a good measurement system.

    0
Viewing 15 posts - 1 through 15 (of 15 total)

You must be logged in to reply to this topic.