iSixSigma

Gage RR and CMM

Six Sigma – iSixSigma Forums Old Forums General Gage RR and CMM

Viewing 15 posts - 1 through 15 (of 15 total)
  • Author
    Posts
  • #38192

    Ray in L.A.
    Participant

    My Six Sigma training stressed that you must have a valid GR&R involved with your project. 
    Well, my data is generated by our CMM.  We only have two trained personnel (one day, one night shift) for this machine, and it is almost in constant use.  How do I validate the output data?  Also, how do I justify “tying up” the machine with an R&R without angering the stupids (aka management).

    0
    #114033

    Mike Carnell
    Participant

    L.A. Ray,
    The second worst R&R I have ever seen came off a CMM. It was in the area of 1000%. For some reason people feel if they pay a lot of money for a machine it makes it a good measurement system. You can’t buy SS capability particularly in measurement equipment. As an industry they seem manage to no recognize precision as a necessary part of a measurement system.
    You run R&R on a CMM just like you do anything else. Setup on a CMM is critical particularly if the parts aren’t fixtured. If you are running a program the operator will rarely affect it. If you are running a touch probe you will probably fail the reprodcibility part because it is pretty technique sensitive.
    As far as tying up the machine. If you can’t sit them down and get them to understand the issues around not knowing the R&R makes the data suspect then you need to escalate it to the SS Program manager. If that person doesn’t get it done then change projects – a SS project without validating the measurement system is a waste of time. Find something productive to do get out of the SS program. If you can perform a R&R your SS program is just dressing you guys up with a bow tie and standing you in the window so you look pretty.
    Good luck.

    0
    #114050

    Dog Sxxt
    Participant

    Why you caanot run GRR for your case?

    0
    #114051

    Lee Gale
    Participant

    A quick and easy way to determine if there is a significant problem, without taking up too much time, is once the parts have been measured, take a copy the results and ask the operator to run the parts again.
    Obviously this is not an R&R, however it will provide a basic indication of some level of the confidence in the machine.
    Also if you find wildly varying results, that should convince management that the results they are recieving are of little value and work needs to be done to get confidence in your results. 

    0
    #114052

    Dog Sxxt
    Participant

    Repeatibility is also R&R, right? 

    0
    #114115

    Ray in L.A.
    Participant

    Any GR&R I’ve been involved with is set up with 3 inspectors, 10 parts, 3 trials.  For the CMM, I do not have 3 inspectors.  I suppose I could check the reproducability of the two operators we have…better than nothing.

    0
    #114141

    Schuette
    Participant

    Ray – I ran the gage department for a plant with 11 CMM’s.  We ran GRR on every CMM – the first time about 50% of the inspection programs failed, due to probe at wrong angle, wrong probe ball size, poor fixturing, etc.  Using two operators is fine – make sure they take each part out of the fixture and replace each time.
    In addition to the GRR, we also ran precision checks, where we ran the same program 10 times without removing from the fixture.  This gave us a good indication of the real precision of the measurement. 
    In production or part prove-out, whenever we had a very tight tolerance, we would run the measurement several times and take the average.

    0
    #114146

    Mike Carnell
    Participant

    L.A. Ray,
    I don’t know if you have ever seen one of those dog shock collars that has the wire buried around the yard and the first time the dog gets zapped he runs around the yard shaking his head. That is exactly how I feel when I read your post. I sat in front of the computer shaking my head hoping that if I did it enough the words would change to something that actually made some sense. Either you had a complete incompetant teach you R&R or you were not paying attention.
    Repeatability does not demand 3 operators. Did you actually believe there were exactly 3 inspectors involved in every measurement system in the entire world? and it only worked with 3?
    There will even be opportunities where there may only be one operator and different test stations will be substituted for the 2 or more operators. (just a hint – it actually does not work with only one)
    By the way it isn’t better than nothing. It is exactly what you have so you need to check it.
    Do you know what to do with the results?
    Good luck.

    0
    #114154

    Ray in L.A.
    Participant

    Mike,
    I really appreciate your input.  However, I never said repeatability demands 3 operators.  I was taught repeatability was equipment variation….however I’m not going to quibble with terms.  Yes, I was intructed the “recommended” method is ANOVA over X Bar & R, and is “recommended” 3-10-3…in the variable data world.  As I said, it is the only R&R method I’ve personally been involved with, hence my post.  Admittedly, I am only on my third SS project.
    Although, I think you hit upon something more relevant to my project.  The 2nd CMM guy is trained, but barely uses the equipment, so why get him involved?  Different test stations, (i.e. surface plate inspection by machine operator vs CMM at in-process inspection) could be getting different results.
    So, what do I do with the results?  Depends what the result are.  If we are using up less than 10% of the tolerance, generally I’d leave the system alone, but recheck it later to see if it is stable over time.  !0% to 30% is risky, and I’d look at possibly shoring up the system.  Above 30%, I’d reject the system and most likely improving the measurement system *becomes* the project.  I’d look at the main sources of measurement error.
    Heh heh, sorry about dog collar thing… 
     

    0
    #114155

    Ray in L.A.
    Participant

    Thanks Jim…some good things to watch for there.  I’ll keep those in mind.
    Ray

    0
    #114165

    Mike Carnell
    Participant

    L.A. Ray,
    You should be able to drop the CMM below 10%. They are good measurement systems when they are done properly. There are a lot of noise variables that affect them i.e. temp of parts (depends on the tolerance), fixturing the parts, training – if not programed, etc. One of the other posts had a good list.
    We have even done one of the big ones where you are walking around on the table and found out that where you stand affects the measurement. It sounds like yours is smaller so it is easier to control.
    Good luck.

    0
    #114596

    DrSeuss
    Participant

    Ray,
    Your delimma describes a battle I have many years ago in a former life.  I assume that your machine and touch probe have been calibrated and the table meets all the required flatness standards.  Here is an approach that might apease your leadership and give you what you need.  Look at the part/s involved with your project and catorgorize the type of check the CMM would be making.  For example, your part requires a flatness check, a curve profile check, hole true position, an angle measurement from a datum, etc.  By catagorinize these checks, you can pick a few of these and create a R&R inspection program that can be used for repetitive parts that routinely get inspected.  As long as each of these catagories get checked, you can use that information for your R&R.  You will probably see differences between manual versus automated inspections.  Big differences can occur between different operators approach to the same part inspection.  I have even seen significant differences between automated versus CAD generated program inspections.  At least the program will not have too many features to check, hence management should want to see these results too.  The other thing to remember, once your R&R has been done, it may not be necessary to repeat for subsequent projects as long as similar part characteristics are being checked.  Hope this approach helps.  Sorry for the lengthy reply…..Let me know how this approach goes?
     

    0
    #114604

    Jonathon Andell
    Participant

    I totally concur with Mike. I too experienced a CMM whose initial GRR was 150% of the tolerance, while the customer claimed to want “six sigma” capability – not that this customer was overly endowed with clues, either.By all means, do the R&R with however many opertors you have.For what it’s worth, here’s how I remember the difference between repeatability & reporducibility:”Repeatability” comes from a single operator saying, “I’m repeating myself.””Reproducibility” is something that takes place among onsenting adult operators.A possibly crude, but usually effective, memory device.

    0
    #176163

    DFSS guy
    Participant

    I share your experience. Gage R&R performed on some dimensions measured with CMM was great (<1%), however I quickly spot one which was really horrible (~50%). I was eager to do this as I did not believe explanation that "CMM measurement is 100% repeatible…". I showed that Gage R&R was very high, now I am struggling with improving it.

    0
    #176167

    Mikel
    Member

    Look at your fixturing and set-up – you will find the problem there.

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

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