iSixSigma

GRR For Parts With Very Little Variation

Six Sigma – iSixSigma Forums Old Forums General GRR For Parts With Very Little Variation

Viewing 18 posts - 1 through 18 (of 18 total)
  • Author
    Posts
  • #52355

    walden
    Participant

    I have issues with GR&R studies that need to be conducted on parts that have very small part to part variation. These are typically  measurements where in 10 parts the total variation between the min and max part size may be up to .002″. The discrimination of the measuring instrument is .0001″.
    The issue in my mind is that due to the small amount of part-part variation the results typically show a total GR&R % more than 30% and only 1 or 2 distinct categories. Unfortunately, in a small run I can’t find parts with more variation for the GR&R study. Has anyone experienced similar issues, and if so, how did you proceed?

    0
    #184135

    Taylor
    Participant

    Chris
    Do you have any scrap parts which are close to the upper or lower spec. If not I have used a gage standard as a dumby part before, Assuming your goal is simply to varify that you can measure process variation.

    0
    #184137

    walden
    Participant

    Chad,
    I think we’re on the same page. My initial thought was to try and somehow introduce more variation into the parts by various methods, but some customers want the GR&R conducted on the actual parts from the capability study.
    Basically, I’m looking to confirm through others such as yourself with more experience, that using parts with very little variation will make it almost impossible to pass the GR&R criteria. If others agree that this is an issue, I can explain with confidence to the customers that we somehow need more part variation in the study, otherwise we have little chance of ever passing.
    Thanks for your help and comments.
    Chris
     

    0
    #184138

    Cone
    Participant

    Do you know your process capability?

    0
    #184143

    walden
    Participant

    Yes, I believe I do. Would you care to comment on my original query?

    0
    #184146

    Cone
    Participant

    You always have the option of substituting known variation into your
    calculations. In fact I prefer it because it gives you a better read on
    you measurement capability.

    0
    #184151

    walden
    Participant

    Gary,
    I’ll tell you what I’ve done to an existing study prior to my original post. Please confirm if this is what you’re referring to by “substituting known variation into the calculations”.
    The parts on original GR&R I conducted only exhibited the .002″ variation and the calculated total GR&R % was well above 30%. In an effort to see the effect of part variation only, I attempted to keep appraiser variation constant. I did this by increasing the values in all trials for part 1 by .007″, increasing part 2 by .003″, reducing part 3 by .002″, reducing part 4 by .008″ and not changing the values for part 5. This increased part variation from .002″ up to .015″. The recalculated total GR&R % was now less than 10%, which would be considered acceptable.
    Again, I’m trying to confirm that my line of reasoning is correct. If so, I can approach my customers and explain that using parts with very little variation will result in an unnaceptable GR&R % for an otherwise acceptable measurement system.
    I appreciate your kind response.
    Chris

    0
    #184152

    walden
    Participant

    Gary,
    I’ll tell you what I’ve done to an existing study prior to my original post. Please confirm if this is what you’re referring to by “substituting known variation into the calculations”.
    The parts on original GR&R I conducted only exhibited the .002″ variation and the calculated total GR&R % was well above 30%. In an effort to see the effect of part variation only, I attempted to keep appraiser variation constant. I did this by increasing the values in all trials for part 1 by .007″, increasing part 2 by .003″, reducing part 3 by .002″, reducing part 4 by .008″ and not changing the values for part 5. This increased part variation from .002″ up to .015″. The recalculated total GR&R % was now less than 10%, which would be considered acceptable.
    Again, I’m trying to confirm that my line of reasoning is correct. If so, I can approach my customers and explain that using parts with very little variation will result in an unnaceptable GR&R % for an otherwise acceptable measurement system.
    I appreciate your kind response.
    Chris

    0
    #184156

    Ryan.M
    Member

    Hi, what is the objective of this study? for process capability improvement? You may opt to % tolerance instead %study var. 

    0
    #184159

    Cone
    Participant

    What I had in mind is much simpler.If you know your historic long term variation (from Pp/Ppk
    calculations), simply substitute that for the variation seen in your
    study.If you use Minitab, there is the option of entering your historic
    standard deviation and it will be used in the calculations.The situation you have described is one of the most common
    mistakes make in MSA. If your study parts under represent your
    process, MSA can fail and there is nothing wrong. If your study
    parts over represent your process (quite likely with what you have
    done with the artificial variation you induced), your study might
    pass when it should not.The other poster who responded to this has the right question –
    why are you doing this study?

    0
    #184161

    walden
    Participant

    The customer is requiring a GR&R study be conducted on the parts, which they pulled and sent to us for the study. The problem is the parts have very little variation. I would propose using % of tolerance consumed, but some of the characteristics only have a minimum spec limit so there is no defined “tolerance band” to use.
    I am aware of not overestiminating the variation in order to make a study pass when it should not. The variation I introduced is what I might expect to see between several lots. The within lot variation is small, which is what prompted this thread. I will try your suggestion of using the historic standard deviation in Minitab and see what happens.
    I’m getting more and more customer requests to conduct GR&R studies and typically find that they don’t have a good understand of MSA. They’ve heard or read that GR&Rs should be conducted, but don’t have any practical knowledge or experience of how to properly conduct one. Quite simply the requirement is, “conduct a GR&R and the results must be less than 30%”. I’ve asked, 30% of what, and get dead silence on the other end of the phone.
    All I was originally looking for is confirmation that, as you said, “if your study parts under represent your process, MSA can fail and there is nothing wrong.” I didn’t want to use this rationale to educate my customers until I confirmed that this was correct. After these posts, I believe this has been confirmed.
    My thanks to all that have been kind enough to offer advice/comments.
    Chris

    0
    #184162

    Cone
    Participant

    Chris,MSA results and capability studies should always be offered to the
    customer together. I’ve written some fairly extensive explanations
    of this, I see if I can dig them up on some back up drives today.
    Dealing with one sided specs can be done, but is tricky in the case
    where the process mean shifts significantly batch to batch.The uneducated customer will probably have no clue if you use
    historical standard deviation but you should never present without
    an explanation.Send me an email to [email protected] and I’ll either find the
    write up or recreate it by the end of the weekend.

    0
    #184163

    walden
    Participant

    Gary,
    Thanks for the offer and your kind assistance.
    Chris

    0
    #184164

    MBBinWI
    Participant

    Chris: 
    1) Despite many good comments to this post, there is no such thing as “very small part variation.”  Rather, there is only measurement tools unable to measure the variation present.  It would be like trying to use a yardstick to measure microns.  Here, you indicate that you may have up to .002 range and discrimination level of .0001, so you have a possibility of 20 levels, so should be OK.  Do you actually observe this range or is that just the possible level?
    2) Your customer insisting on using the actual parts just goes to prove that they don’t understand what an MSA/GRR is meant to demonstrate.  You are evaluating the TOOL not the parts.  Does it matter if I measure a die, a cube of similar size, or two lines drawn a similar distance apart?  Only from the standpoint of how the operators use the tool.  This is a common issue with MSA – I often find that the operators want to use “good” parts, thus the GRR turns out poorly.  If needed, intentionally fabricate the parts to represent the range of allowable variation.  Be careful with this, though.  By creating these parts with larger variation, your customer may then come back and accuse you of producing “crap.”  You need to educate them.
    3) Gary makes a very good point on also looking at the capability.  In fact, this is a scenario that I often use in evaluating the skill set of folks – you have a GR&R with a study variation of 38%, the process which is being measured with this gage is currently exhibiting a Cpk of 1.67.  What actions do you take?  (since I will be using this in the future I won’t disclose the appropriate answer, but if it isn’t obvious to you, you can send me an e-mail at [email protected] and I’ll send you the answer).

    0
    #184165

    Taylor
    Participant

    Great advice, and even better example. I can’t count the number of times I’ve had this discussion with the shop floor.

    0
    #184166

    Cone
    Participant

    I know the answer to number 3 – Go find something that really needs improvement.

    0
    #184178

    MBBinWI
    Participant

    OK, Gary, you get a gold star.

    0
    #184191

    Craig
    Participant

    Makes sense to me!  No one seems to care about precision to tolerance ratio any more! I would look at that next.
    I have seen the other extreme where the parts chosen are 2X of the tolerance and the study variation is low as a result of that.

    0
Viewing 18 posts - 1 through 18 (of 18 total)

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