iSixSigma

Gage repeatability question

Six Sigma – iSixSigma Forums Old Forums General Gage repeatability question

Viewing 19 posts - 1 through 19 (of 19 total)
  • Author
    Posts
  • #47033

    Ray in L.A.
    Participant

    What I have is a hand-held X-ray metal sorter/analyzer, and I want to set up a repeatability study.  Every measurement has multiple outputs (i.e. element percentages), between 9 to 12, depending on the metal.
    I don’t think reproducability is an issue, given there isn’t much technique involved, besides squeezing a trigger.
    I’m struggling how to set this up in Minitab.  Any suggestions?

    0
    #156256

    Mike Carnell
    Participant

    Ray in L.A.,
    Set it up just like any other crossed design. You might find the reproducibility shows up based on the persons ability to select a different spot to read.
    Just my opinion

    0
    #156259

    CT
    Participant

    Minitab is really difficult with these kinds of studies when the operator is a none factor. I have an Excel spreadsheet if you would like. Just post an email and I will send to you.

    0
    #156261

    Mike Carnell
    Participant

    CT,
    It seems a little dangerous to drop the operator until you have data that says they are not a factor. Depending on how sensitive this thing is and how homogeneous the material is the spot they read will show up in the reproducibility number.
    Regards

    0
    #156263

    Ray in L.A.
    Participant

    When I’ve set up other crossed designs, it was a single output per measurement (i.e. micrometers, hardness testers, calipers, etc.)
    With this set up, I have columns for:
    Part No.
    Trial No.
    Values (9 per measurement)

    0
    #156266

    CT
    Participant

    Mike,
    Agreed, but with hand held device with a simple trigger not much the operator can do here and this should really be discounted in this specific case, or at least I think so. Reason is you really want to incorporate natural operator influence into the reproduceablity of the the gage. You want the R&R to be as natural as possible, obviously if there is huge operator difference then another problem exist and should be corrected prior the R&R test.

    0
    #156264

    Ray in L.A.
    Participant

    Hey, thanks.
    [email protected]

    0
    #156267

    Mike Carnell
    Participant

    Ray in L.A.,
    No difference. You have to analyze buy each value. You are probably going to find that the precision varies by element so you want to analyze 9 different ways.
    Regards

    0
    #156268

    CT
    Participant

    I’m sending now…one thing to keep in mind is you will probably need to run a horrible amount of test and actually set up an R&R for each data point on the gage, not that difficult but very time consuming
    Good Luck

    0
    #156269

    Mike Carnell
    Participant

    CT,
    Agreed you want to see the natural operator variation and although it is simply pulling the trigger it is also where it is pointed. If your run something like FTIR you can make it look like the precision is terrible just by using a large sample and moving it around between readings. If you leave it in the same spot and just read over and over again it does extremely well.
    That was why I said not to drop the reproducibility right away. If there isn’t a method to get the same spot over and over again there will probably be a reproducibility number. It will be a MSA issue but probably not a gage issue.
    Regards

    0
    #156270

    CT
    Participant

    Mike question for you
    Would it be possible to do a Ppk initial study on each element prior to R&R just to note the variation from element to element? This maybe enough to at least prove that the elements coming in are in control enough not to influence the R&R. Maybe? hell I don’t know!

    0
    #156271

    CT
    Participant

    Mike Agreed, and ignore last question, already answered!
    Thanks, been a while since I’ve done detailed R&R’s like this

    0
    #156272

    Ray in L.A.
    Participant

    Thanks both of you.
    I was trying to get around not having to do a separate study by each element. 
    Since this looks like a slower project than I originally thought, I might as well add operators to the study.
    So let’s see…3 operators, 10 parts, 3 trials, 9-12 elements, 6 different alloys = 4,860 – 6,480 values
    This should be fun

    0
    #156273

    CT
    Participant

    SIX SIGMA, Good Ole Fashion Way to Spend the Weekend!

    0
    #156274

    Mike Carnell
    Participant

    CT,
    Alright you are going to get me into one of those stats discussions that always ends up with one the stats guys logging on and telling us how stupid we are. Normally I just wait for Robert Butler to answer.
    What I normally use Ppk for is to decide what type of analysis I am interested in. If the Ppk is less than 1.0, I use Percent of Tolerance so I can decide if I want to work on the process or the gage first and if we need to guardband specifications to protect a customer. If the Ppk is over 1.0, I use percent of study variation. It tells me how much I can trust the Measurement System to work on the process.
    Initial thoughts on your suggestion are really pretty similar. I would think that if there isn’t a large difference in the range charts you would be fairly safe on that assumption.
    Regards

    0
    #156275

    Mike Carnell
    Participant

    Ray,
    This is where you find one of those college fresh outs ans ask them if they want some valuable real world experience.
    Regards

    0
    #156277

    CT
    Participant

    Mike
    screw those stat guys, they scare me.
    TGIF, I’m out of here.
    Ray-
    Mike has a very good suggestion about the college help!
     
    Cheer
    CT

    0
    #156278

    Mike Carnell
    Participant

    Summer interns work even better.
    Have a great weekend.

    0
    #156280

    Ray in L.A.
    Participant

      Great idea!
    “Psst…hey kid, wanna be a certified yellow belt?….”

    0
Viewing 19 posts - 1 through 19 (of 19 total)

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