iSixSigma

gage rr struggles

Six Sigma – iSixSigma Forums Old Forums General gage rr struggles

Viewing 25 posts - 1 through 25 (of 25 total)
  • Author
    Posts
  • #37446

    newb
    Participant

    how do you proceed if you cant get a gage r&r to pass?  I mean its not even close.  The defect we are trying to measure is a concern but its very subjective which is one reason that the r&r is failing.  There doesnt seem to be an effective measure technique that is reasonable cost.  We just cant afford to purchase digital image evaluation tools.  We are measuring printed ink defects.
    any ideas? 
    any of you have experience with projects that the r&r just wasnt acceptable with no other alternatives?  How did you proceed?
    thanks
    newb   

    0
    #110268

    Mike Carnell
    Participant

    newb,
    Attribute or variable data?

    0
    #110270

    newb
    Participant

    sorry.
    we are trying to measure it variable. 
     

    0
    #110272

    Mike Carnell
    Participant

    newb,
    Increase your sample size and use the average of the measurements. This is not a good long term plan. It will ultimately be more expensive than a better gage.
    If you honestly believe that you cannot cost justify a better gage then you don’t understand what the bad gage is costing you.
    Regards

    0
    #110295

    Whitehurst
    Participant

    Mike has the key.  I would add that you don’t have to increase all your testing though.  I suppose you know of the 10% acceptance limit on GR&Rs.  Are you also aware that 20% is acceptable and 30% may be acceptable if you are working to improve the gage and have other support measures (such as increasing your sample sizes.) Double check only the product falling outside a tightened tolerance.  Or, if your process spread is centered and  far less than the customer spec limits, forget the extra testing and support your case with the statistics.

    0
    #110318

    Ron Manubay
    Member

    newb,You did not mention what’s causing your unacceptable G&R. Is it due to the equipment or the appraiser? Anyway, here’s some general guidlines you can use.I would assume that you are measuring the ink size here. So my comment is based on this assumption.What I would do is to get repeated measurements and get the average as your reading. It will definitely increase your precision. The more samples, the better.If the above would still be not enough, why don’t you just make it a go/no-go decision instead of relying on actual measurements? You can use some sort of a template to see whether a particular ink is withing the limits you set or not. This would take out a lot of the subjectivity you mentioned.Regards,
    Ronnie

    0
    #110330

    Mikel
    Member

    Mike, Joe, and Ron all are giving good advice.
    You did not tell us if the failure was %Tolerance or %Study.
    For product acceptance, %Tolerance is all you should be concerned with. What was the number for that and how did that break out for the source of the issue?

    0
    #110387

    vineeth
    Member

    Hell Friend,
    so how your Query make me think that, you try to explain a situation which is very complex..and in getting in that complexity u missed the Basic principle of Gauge R&R..
    In your case, its visual and its most the simple one… You Collect 5 good Samples and 5 Bad Samples ( Bad samples can be again subgrouped in to Defect wise..) this is a set of Master Samples and done by the Expert.
    select the inspectors–give the samples 1 by 1 with out saying any thing about the sample and tell them to Give their Judgement..
    Compile the data from Operators and Match it with the Expert ( Master Sample..) say 10 samples , 5 samples all are agreeing the same ( Good / Bad.) then your Gauge R&R is 50 %..
    so you need to work with your Operators , Give them Training, Put some Displays, and come up with an Inspection Method to have a Gauge R&R above 70%..
    Its all about Discipline and Adhering to the Process especially in the case of attribue Gauge R&R ( Mainly in the case of Subjective – visual type..)
    regards
    vineeth

    0
    #110390

    Markert
    Participant

    There is a lot of hokum published about Gauge R&R studies. The usual acceptance criteria of 10% and 30% are purely arbitrary and do not tell you whether a gauge is good or bad for any purpose.
    Take a look at Don Wheelers paper “Good data, bad data, and process behaviour charts” on http://www.spcpress.com for a more common sense approach.
    Phil

    0
    #110394

    Gourishankar
    Participant

    newb
    if you have a poor gr & r result , there’s only one option – identify the root cause ( gauge or operator ) and fix it. There’s no way you can go to the next phase of improvement ( analyze) with unreliable data.
    From your problem description it is not very clear what the grr result is , but it could turn out to be simple case of operator training for better calibration.
     
     
     

    0
    #110405

    melvin
    Participant

    There are several items you may want to consider.  First, the measured variation is the sum of the process variation and the measurement system variation.  All the R&R study is telling you is the ratio of these variations.  That is why it is critical to base the study on test samples that are truly representation of the variation you see in your process, i.e. from different lots, machines, shifts, etc.  If you have chosen from too “good” of a range of samples, any gage will look poor.  Also as you improve your process, a greater proportion of the variation will be induced by the gage, thus you should do another R&R.  If this doesn’t help, you can try the brute force method.  Almost any measurement system can be improved by replication, but it can be very labor intensive.  To use this method with the R&R, let’s say that you will use five replications.  All you need to due is take five measurements and average them for EACH of the values required for your R&R.  (To make this work it is important to eliminate as much bias as possible, i.e. don’t ask the same inspector to take five readings of the same item in a row).  Using this technique, you can “dial-in” how “good” of a measure system you need.  Will it take three replications or ten to get where you want to be?  If you use this method, once you know how many replications you will need, calculate the cost and you may find that the high tech measurement system is not so expensive.
     
    This technique will also work to improve P to T ratio, however here you have another issue.  Where did the tolerance come from?  I don’t know anything about printing, but in metalworking I’ve found many designs with silly tolerances.  If a very tight tolerance is really required, then look at everything that can impact it, e.g., temperature and humidity in the test area may impact the substrate, does the test procedure specify lighting requirements?  If the design authority cannot provide good answers for the items that are relevant to your application, then it’s not a “real” tolerance and you need to develop the correct one.
     

    0
    #110431

    Mike Carnell
    Participant

    Stan,
    Good point. I don’t really care for %Study Variation. If you use a group of parts with no variation (or little difference in parts) it will force all the variation into the gage. If you get a process with a Cpk > 1 then it can help you sort out what is a contributor. Conversely if you have a Cpk >1 and you have a bad gage then you have to have some super capable process to off set the variation from the gage.
    The only measure that seems to really come across well is P/T ratio. It is something that people seem to understand easily. If you create the guardbands at the ends of the tolerance then you can do the repetitions on only those failures and reduce some of the work.
    Just my opinion.
    Regards

    0
    #110432

    Mike Carnell
    Participant

    vineeth,
    newb second post states it is a variable measurement.

    0
    #110433

    Mike Carnell
    Participant

    Gourishankar,
    There is more than one option and that is why we use the multiple measurements. If you are losing money daily because of a poor gage and can plug that hole by doing multiple measurements there is no reason to wait on the gage. You don’t forget doing it because the multiple measurements carries its own cost but it does allow you to move the project forward. It is about results.
    The multiple measurements lets you continue to work the process and you can turn the R&R study over toa GB. There is also the option that the entire process problem is also just the gage. Multiple measurements will let you confirm that.
    Your comment: “….but it could turn out to be simple case of operator training for better calibration.” R&R is a precision measurement. Calibration (bias) is part of accuracy and won’t show up in a R&R study.
    Good luck.

    0
    #110438

    newb
    Participant

    Thanks for all the responsses so far.
     
    This is where we are now.
     
    Repeatability-Equipment Variation = 66.72%
     
    Reproducibility-Operator Variation = 187.94%
     
    R&R = 199.44%
     
    We are using a magnification lupe with laser etched scale.  Our tolerance is +.020”.
    There is no (-) tolerance as the defect is only in one direction. 
     
    We considered making it an attribute measurement but we were under the impression that DoE’s were much more difficult with attribute data.  The process we are trying to improve has a lot of x’s and we really want to be able to perform some DoE’s. 
     
    We are currently looking for a higher magnification device.  We believe that with 2x the magnification we will be able to see the defect as well as the etched scale much better which should help eliminate at least some of the variation.
     
    thanks again
     
    newb

    0
    #110440

    Mike Carnell
    Participant

    newb,
    You guys are obviously not taking improvement seriously. Throw away the lupe, go back to attribute mesurements and flip a coin. You just reduced your R&R to 50%.

    0
    #110444

    newb
    Participant

    I dont feel like its fair to say we arent taking it serious? I would compare that to making fun of somebody that is having a hard time reading. Just because you can read well doesnt mean they dont take learning how to seriously even if they are going about it all the wrong way.
    The fact that we are doing an R&R and understand that it failed and that failed R&R’s are bad is at least evidence of taking it serious even if we arent certain how to proceed.  It’s a very large step for us to even think that way around here.
    Thanks for the help.

    0
    #110451

    Mike Carnell
    Participant

    newb,
    Your printing press cost how much? Your lupe cost how much? You don’t mind printing as long as you don’t have to spend money to set it up so it delivers a quality product (I am sure that is not your decision)?
    There is a saying about asking operators what they want. It is worse to ask and not do anything about it than to not ask at all. The R&R is 200% and you know it. If you are not going to do anything about it – figure out what it cost you and add a line to every invoice where you deduct the cost of inspection (stop inspecting) and tell them that is their discount for the poor R&R. At least the customer will feel there is some justice in the world – maybe.
    Good luck.

    0
    #110458

    newb
    Participant

    yes, i agree with you. 
    We are still in a “sales” phase of six sigma.  What we are trying to do is make certain that we are doing the best on GR&R before we go and say we have to have money for inspection equipment etc.  We have no intention of doing nothing about it.  We do, however; want to be certain that the GR&R we submit is the best we can get which is why i was asking for advice.  I want to be positive that we’ve done our homework.  I cant afford for us to discredit oursleves on one of the first projects we perform.
    See, at the moment, we actually use nothing to measure these parts.  The customer doesnt measure them either.  We still need to make improvements so our team has to find a way to measure.  The lupe was our first attempt because it was fast, cheap, and seemed simple. 
     

    0
    #110459

    Mike Carnell
    Participant

    newb,
    Ask the vendor of the expensive gage to bring in a demo unit and run a R&R study with the demo unit. It may not make it either. Big bucks won’t guarantee a good R&R.
    Good luck.

    0
    #110466

    Mikel
    Member

    I agree, but we need to get rid of the mushy rules that let most interpret that <30% is okay.

    0
    #110494

    Dreemr
    Participant

     
    Based on your last statements:
     
    See, at the moment, we actually use nothing to measure these parts.  The customer doesnt measure them either. 
     
    I have to ask were this spec coming from.  Why the concern?  If the customer is complaining the first thing to do is to define their desire in a measurable manner.  Otherwise you may be tossing perfectly acceptable product.
     
    If no one is measuring this, then who can really say that there is an issue.  Remember without data it is just another opinion.

    0
    #110497

    newb
    Participant

    yes, i absolutely agree; it is mostly opinion.  We are trying to make it measureable.  We created the spec based on customer feedback.  It’s purely internal.
    The fact is that we are not scrapping hardly any product at all.  It must improve though.  There is new competition that is taking a small piece of the business.  We have to improve to make sure we keep the business and out perform our competitor. 
    And yes, i realize that its late to be doing this.  I am the new guy here that is pushing continuous improvement.  I’m trying to help fix a broken system.  Its a tuff task and i apprecaite your help.
     
    newb
     

    0
    #110499

    Dreemr
    Participant

     
    I understand being the new guy and all the difficulties that go with it.
     
    Anyhow, is the business loss a direct result of quality issues or is it simply a business strategy by your customer.  It may be that they wanted a new supplier to help keep the current ones honest.  Or it could have been a cost issue and not quality driven.  In this case we will accept that it is quality driven if for nothing else but the sake of discussion. 
     
    Setting up a spec with the customer can be difficult.  Often times our customers do not truly understand the capabilities of a process.  Often times we do not understand our own processes.  Consequently, expectation can appear unreasonable from either side of the table.
     
    Trying to create a spec based on feed back can also be difficult and often times risky.  But, who wants to sit down with their customer and put it all out on the table and say, “Here is what I can do.  What do you want?”  
     
    It was explained to me like this:  A master black belt I once knew handed me a bowl of M&Ms.  He asked me to sort out all of the shiny ones.  I asked him “Shiny ones?”  He said “Yah, all the ones that shine more than the others.  It is quite simple actually.  Just give it a try.”  So, I did as asked.  I failed.  The sort I performed did not separate out the ones he wanted.  The problem was I had not defined the spec thoroughly enough.  He had said shiny but he wanted red.  If I had understood his definition or asked for examples or whatever, I may have performed to expectation the first time.
     
    If further definition of the spec in not the issue, then you can proceed with correcting the performance issues.

    0
    #110513

    Mike Carnell
    Participant

    Stan,
    Just people learning rules.
    Lets see if I am at 29% I am ok but I have lost the integrity of 58% of my tolerance but the rules say it is OK so it must be OK. If a customer came in and forced a 58% reduction in tolerance they would scream change of scope all the way to the bank but doing it to yourself is OK?
    More rules – keep those tool zombies comin’.
    What was that post a couple weeks ago – What is the Future of Six Sigma? Sum of squares, sum of squares, sum of squares, sum of squares,………….. I think I feel an aneurism coming on.
    Regards

    0
Viewing 25 posts - 1 through 25 (of 25 total)

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