1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.
Dismiss Notice
You must be a registered member in order to post messages and view/download attached files in this forum.
Click here to register.

Gage R&R Opinions: Are they effective ?

Discussion in 'Gage R&R and MSA - Measurement Systems Analysis' started by Kai Kidston, Sep 9, 2018.

  1. Kai Kidston

    Kai Kidston New Member

    Joined:
    Sep 8, 2018
    Messages:
    1
    Likes Received:
    0
    Trophy Points:
    1
    Location:
    Exeter, Devon
    It’s a small thread but hopefully get some good feedback;

    What is everyone’s professional and personal feeling on carrying out R&R’s?
    Are they effective?
     
  2. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    605
    Likes Received:
    663
    Trophy Points:
    92
    Location:
    Maine
    In my experience and using my knowledge of math, the AIAG Guage R&R is rarely useful. First it is used as a simplistic requirement: do it and get an acceptable score (that is mathematically incorrect and incorrect enough as to be very misleading) which results in a check the box approach.

    A well thought out and properly performed measurement systems analysis IS useful. In my organization we require MSAs for all tests, inspections and screens. The difference is that we realize that the ‘score’ mentality is counter productive. See my article in the resources section of this forum for details...we have used MSAs very effectively in every aspect of our business. From eliminating tests that are ineffective to helping us solve long term complex problems...
     
    Kai Kidston likes this.
  3. judegu

    judegu Well-Known Member

    Joined:
    Jul 11, 2018
    Messages:
    110
    Likes Received:
    17
    Trophy Points:
    17
    Location:
    China
    Least to say, Gage R&R is effective in addressing the customer requirements.:rolleyes:
    PS: Always wonder how they made the acceptance criteria. Such as <10% is good, 10%~30% is acceptable with connsideration of cost, >30% is unacceptable.
     
    Kai Kidston likes this.
  4. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    605
    Likes Received:
    663
    Trophy Points:
    92
    Location:
    Maine
    The acceptance criteria was atmospherically derived.
    Since the percentage tolerance ratios are mathematically invalid, the 10, 10-30 adn >30 lines could not have been empirically derived.
     
  5. Golfman25

    Golfman25 Well-Known Member

    Joined:
    Nov 6, 2015
    Messages:
    816
    Likes Received:
    402
    Trophy Points:
    62
    Depending on your manufacturing process it can be useful or completely useless. In our case, stamping, it is generally useless.
     
  6. judegu

    judegu Well-Known Member

    Joined:
    Jul 11, 2018
    Messages:
    110
    Likes Received:
    17
    Trophy Points:
    17
    Location:
    China
    @Golfman25
    Out of curiosity, could you please explain further why it is useless in the stamping process? Love to know more about other manufacturing process.
    @Bev D
    Sir, sometimes I really wonder the customers (auditers) really understand the essence of the Gage R&R. They just want to know whether you are doing it or not, and whether you do it in the AIAG`s way.
    What makes it worse is that the one who is responsible for MSA in my industry also just do the same without a real study of the interaction between the process and the MSA tool. He/She just follows the book, not questioning the book.
    When you got some kind of knowledge of how Gage R&R works, you can definitely use the excel to fabricate the data using some functions without trace.
    It is a shame.
     
  7. Golfman25

    Golfman25 Well-Known Member

    Joined:
    Nov 6, 2015
    Messages:
    816
    Likes Received:
    402
    Trophy Points:
    62
    Stamping just isn't a precise enough process to make measurements beyond .000 useful. We find we have to use over precise gages to find enough variation to make the calculations work. And then you're always .00000 "off" which you can't fix anyway. In our case, we do draw forming and there is just way too much within part variation to do anything. Unfortunately many expect the process to act like precision machining or something. It just doesn't work that way.
     
    judegu likes this.
  8. judegu

    judegu Well-Known Member

    Joined:
    Jul 11, 2018
    Messages:
    110
    Likes Received:
    17
    Trophy Points:
    17
    Location:
    China
    To me, "way too much within part variation" means each measurement on the same sample using the same gauge would be quite different from each other. Am I right? If so, how to ensure the validity of the inspection measurement even with over precise gauges? Just can`t think it through. Or there is another way available to ensure the dimension requirements have been met?
     
  9. Golfman25

    Golfman25 Well-Known Member

    Joined:
    Nov 6, 2015
    Messages:
    816
    Likes Received:
    402
    Trophy Points:
    62
    So think of a part shaped like a cup. Maybe 1 or 2 inch deep. And lets say you have an OD dim of 2.500 +/- .002. You literally have an infinite number of measurement points around the entire OD and up and down the wall. And part isn't perfectly round and the wall isn't perfectly straight. And then you have the radius blends into the bottom of the cup. You basically need to measure the same places at the same depth to have any chance. Very hard to do with a hand gage.

    How do you ensure the validity of the inspection measurement? Sometimes close enough, is good enough. You might be paying $0.10-$0.50 per part vs. machining them for over $1.00.
     
  10. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    605
    Likes Received:
    663
    Trophy Points:
    92
    Location:
    Maine
    These are drawbacks of the AIAG method. certainly purchasing or using a very high resolution measurement system when it's not needed is a waste. But not having the right gauging system is also a waste.


    If we do real MSAs they will actually help us in these situations. (See my paper in the Practical Quality Engineering Resources in the Resources tab "The Statistical Cracks in the Popular Gauge R&R Method" also “MSA Verification and Validation”.)


    The need to do an effective MSA should be related to the tolerances, the process capability and the severity of the functional failure that the characteristic has when it is out of tolerance.


    I had a part once that had caused a catastrophic shutdown of a large piece of equipment putting human lives at extreme risk. Upon performing an MSA ot begin the problem solving process, we discovered substantial within part variation and it required us to 'mark' the part on several locations along the length to ensure that we were measuring the same 'thickness' to separate measurement error form part variation. This was from a very high precision machining operation. we studied both the hand held gauge and the CMM. Both were lousy - about 50% of the measured part variation. The hand held gauge was susceptible to angle and 'grip' pressure and the CMM was susceptible to the angle of the part in the fixture and the approach angle of the ruby/shank of the probe. Once we improved the measurement error we were still left with substantial within part variation - we were also able at that point to understand how we had misapplied the testing so that we completely missed the out of tolerance situation: the hand held gauge was only used during in process testing and only measured the center of the part which was the best case condition for the part (the ends were always much larger than the middle and those locations were the cause of the equipment failure). The CMM was used for 100% inspection and piece by piece acceptance. It did measure several locations across the part BUT the original CMM Programmer had noticed the large within part variation and they just assumed that this was measurement error so they used the average of the within part values to ‘overcome’ the variation. Both of these applications allowed the manufacture and release of these parts into the field. A real MSA is a lot more than checking the AIAG R&R box and is a LOT more effective.
     
    judegu and Atul Khandekar like this.
  11. Golfman25

    Golfman25 Well-Known Member

    Joined:
    Nov 6, 2015
    Messages:
    816
    Likes Received:
    402
    Trophy Points:
    62
    Now that assumes you can do something about the within part variation or do 100% inspection. With stamped parts, neither is feasible in most circumstances.
     
  12. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    605
    Likes Received:
    663
    Trophy Points:
    92
    Location:
    Maine
    Golfman - unfortunately for you, your experience is not the same as most - I've been in automotive and stamping and many other industries, processes and technologies and I could always do something about the within part variation and/or 100% inspection if my process was not capable. so I still see MSA as highly effective if done correctly.
     
    Andy Nichols likes this.
  13. Andy Nichols

    Andy Nichols Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    5,086
    Likes Received:
    2,553
    Trophy Points:
    112
    Location:
    In the "Rust Belt"
    I agree! I think, like a lot of things in life, it's a matter of perspective. Some will push against having to do anything that's "required" and then will go to great lengths (and waste a lot of productive time) to justify why it wasn't necessary. On the other hand, others simply acknowledge the requirements and then use the result to confirm what was known or, in some cases, identify the need for action. It's a little like buying insurance...
     
    Last edited: Sep 13, 2018
    Bev D likes this.
  14. judegu

    judegu Well-Known Member

    Joined:
    Jul 11, 2018
    Messages:
    110
    Likes Received:
    17
    Trophy Points:
    17
    Location:
    China
    Can`t agree more on this sentence.
     
  15. judegu

    judegu Well-Known Member

    Joined:
    Jul 11, 2018
    Messages:
    110
    Likes Received:
    17
    Trophy Points:
    17
    Location:
    China
    Sir, I am quite curious about how the failed MSA caused the catastrophic shutdown, and how the improved the MSA prevented it from recurrence.
     
  16. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    605
    Likes Received:
    663
    Trophy Points:
    92
    Location:
    Maine
    To be clear, a failed or passing MSA never causes a failure - it enables defects to proceed undetected or enables the rejection and rework / scrap of acceptable material. You still have to have a process that is producing bad parts. without a good measurement system you can't detect bad parts, you can't detect process changes from good to bad parts and you will have a very difficult time getting to cause and a solution in a timely manner.

    There was no original MSA on the hand held gauge and the R&R performed on the CMM incorrectly attributed the within part variation to simple measurement error and the correction for this was to average the 3 measurements within the piece. Averaging the 3 measurements reduced the values to the average. Since the ends of the part were too think and the middle was too thin the average result ended up being 'in spec'. So out of tolerance parts were routinely accepted. in fact the process drifted so much over time that almost every part was out of spec, some grossly so. IF the CMM R&R had been properly performed and diagnosed, the relatively small measurement error would have been detected and corrected and it would have been obvious that the majority of the variation seen was actual within part variation. The manual system was extremely variable and was used only for measuring the middle thickness as the part was being machined. IF we had an effective measurement system we would have caught the non-conforming parts and not shipped them. We would also have had an opportunity to catch the drift of the process from just in spec to out of spec before a large portion of the install base was 'infected' with bad parts. This would have saved literally millions of dollars.

    Once we had an effective measurement system and could see - and believe - the within part variation we had the means to begin diagnosing the cause of the variation. if you don't have repeatable measurements it is very difficult to get to the cause and know you have fixed it. The very nature of the systematic pattern of within piece variation was a huge clue as to the cause.

    Doing this right is just good engineering. Shortcutting the process is not.
    Knowledge is power, ignorance is dangerous.
     
    tony s, judegu and Andy Nichols like this.
  17. Golfman25

    Golfman25 Well-Known Member

    Joined:
    Nov 6, 2015
    Messages:
    816
    Likes Received:
    402
    Trophy Points:
    62
    The averaging is less about the gage R&R and more about how you want to measure the part. Vary rarely have we gotten a "good" gage R&R with averaging. Our best bet is to pick a single location on the part to measure.

    We have dealt with this problem for 75 years. We frequently discuss with customers "how are you going to measure this?" Because how you do, will determine whether we even have a shot at making a "good" part. To use the example, our walls are never uniform thickness because we stretch metal. We have know that since fire was invented -- it's simple physics. So we can use an average, or pick a specific point, and send you parts. Or we can find the "bad" spot and never send you a part. Pick your poison. That is what a $0.10 ea. process gives you. If you need more, then you have to find a different process usually at significantly more expense. In most cases, you don't really need more. It's just the math and statistics that are telling you that you do.

    In the above case, if you can dial in your process to make the part consistent over it length, then great. In our case it would require a re-writing of the laws of physics -- simply not possible.
     
    tony s and judegu like this.
  18. judegu

    judegu Well-Known Member

    Joined:
    Jul 11, 2018
    Messages:
    110
    Likes Received:
    17
    Trophy Points:
    17
    Location:
    China
    @Golfman25

    With all these years experience in the LED industry, although lack of the real opportunities to contact the customers directly (I am not a CS and we have the oversea HQ to handle the important customer:(, in these cases, we are just the information provider and the instruction follower), regarding the product SPECs customer gave to us, sometimes some of them are quite important. If you failed to answer them, you would definitely get into a BIG REAL trouble. Sometimes some of them are just not THAT importance. When you failed to answer them, nothing would happen. :rolleyes:
     
  19. ncwalker

    ncwalker Well-Known Member

    Joined:
    Sep 21, 2015
    Messages:
    261
    Likes Received:
    168
    Trophy Points:
    42
    Location:
    North Carolina
    I think part of the problem, as everyone is saying, is that it has become a check box. The actual calculations are only part of the R&R and everyone seems to forget it is a DESIGNED experiment. They just grab 10 parts, measure them 3x3 and plug, chug and hope. Consider the car itself - MOST cars only see normal city/highway driving, with some weather effects. But when they TEST cars, they salt spray, desert, climb mountains. (Aside: Sometimes referred to Multiple Environment Over Stress Testing, or MEOST). Why do this when the intended use is normal, every day driving? The answer is you learn more from exercising the car throughout it's POSSIBLE operation range, which is usually larger than it's INTENDED operation range.

    An R&R is like this. For capability, we want a normal, in control process where we randomly take short subgroups. Hopefully, these cross shifts, etc, so we can see how stable the process is. But we are testing the process, so we set it up that way.

    An MSA (the Gage R&R is a piece of this) is testing the measurement system. So we want to verify it on more than random, but well centered parts from the process. In the simplest case, imaging a scale on an assembly line that's rejecting things by weight to make sure some container is full. Almost anyone, when asked, to confirm it works will come to the conclusion they need to see the scale reject some as well as pass some. Some for under weight, some for over. We ask "Where's your rabbit parts?" or you no-go master.

    An MSA is no different than this.

    To intrinsically ask "are they bad?" is like asking "do hot peppers taste bad?" The don't in scrambled eggs, but ice cream? It depends on how skillfully they are applied. This is why it's not sufficient to know how to put the number in Minitab (or whatever) and click the right things to spit out an R&R. You have to understand what it is actually trying to do, and feed it appropriate test subjects. Which may in fact change based on what you are trying to confirm.
     
  20. tony s

    tony s Well-Known Member

    Joined:
    Sep 10, 2015
    Messages:
    1,350
    Likes Received:
    1,054
    Trophy Points:
    112
    Location:
    Laguna Philippines
    Great discussion! Really pays to read from people who are passionate with their fields.:)