iSixSigma

Nested Gage RR for a destructive test

Six Sigma – iSixSigma Forums Old Forums General Nested Gage RR for a destructive test

Viewing 28 posts - 1 through 28 (of 28 total)
  • Author
    Posts
  • #47938

    Markert
    Participant

    Hey All,I’ve been given the task of performing MSA on a destructive measurement process (each part can only be measured once). I have no experience in this area, so correct me if anything I’m saying is wrong.From my searching, it seems that a Nested Gage R&R is the way to go for this. This assumes that all parts from the same lot are identical for the purpose of this test. I haven’t found a definitive answer on how to actually perform the GR&R.Here are some questions:
    – What sample size do I need?
    – How many operators? How many samples per operator?
    – What is the acceptance criteria for a Nested Gage R&R in destructive testing?
    – Should all the samples measured be from the same lot?I really appreciate any help you have to offer. Feel free to explain the basics of GR&R if you think that will help – I really have no experience at all.Thanks,
    Phil

    0
    #160413

    Chad Taylor
    Participant

    Phil
    You are on the right track.

    Sample size should be a total of 60 parts (10 batches of 6)
    Batches should represent the long term varibility of the process
    Use Three Operators
    Each operator will measure two parts per batch.
    Acceptance criteria is the same as Standard R&R
    By using Nested ANOVA method you are assuming that the parts are unique to the operator.
    If you have any other questions please ask
    Cheers TGIF
    Chad Taylor

    0
    #160414

    Dixit
    Member

    Can you please recommend a good text book to learn basics of Standard Gage R & R?Thanks

    0
    #160415

    Dixit
    Member

    I forgot to mention Chad in the greeting….Can you recommend me a text book or weblink for performing Gage r & r?Thanks

    0
    #160444

    Chad Taylor
    Participant

    Vijay
    There are several books that some information on the subject. My personal choice is Measurement System Analysis: Reference Manual, 1995. This has about all you need to know or ever wanted to know about performing R&R.
    This site has a ton of information as well if you perform a search. AIAG guidlines are really what most Quality Professionals reference in manufacturing processes.  Another source is Quality Engineering, 6 pp. 115-135
    Chad Taylor

    0
    #160450

    Dixit
    Member

    Thank you Chad.

    0
    #160451

    BTDT
    Participant

    Vijay:Have a look at the technical article on destructive Gauge R&R. The usual Gauge R&R is called a crossed design. Yours will be nested and the analysis is slightly different.Cheers, BTDThttp://www.minitab.com/resources/articles/MeasSysAnalysisDT.pdf

    0
    #160463

    Markert
    Participant

    Hey Chad,Tbanks for the quick response.I have access to only a limited number of parts (35) and they are all from the same lot (I guess we could consider them homogeneous). Is there an alternative way to run this GR&R, using only parts from a single lot? Or would it be better to get more parts from various lots?Thanks,
    Phil

    0
    #160465

    Chad Taylor
    Participant

    Phil
    35 parts maybe enough as long as they are true representive sample of the process. What may happen is the number of distinct categories will be too few to give accurate results. As long as you have 4 or more distinct categories you will be ok. But I would error on the safe side and try to get more product for the study; especially if you have access to a different lot.
    Another method of R&R for your situation is Precision to Tolerance ratio. Caution about this method: If your tolerance is inflated it can yield misleading results. Not sure what your trying to determine from your MSA, but this method is very useful in determining if product can be accepted or rejected.  Formula is  (5.15 * Sigma ms/tolerance)*100 Acceptance of this method is generally <20%.
    Chad Taylor
     

    0
    #160468

    Markert
    Participant

    Thanks Chad,It might be easier for me to describe the goals of my MSA so your efforts may be better directed towards my needs.I have a new machine which measures the torque transmission of a shaft by twisting it a set amount of degrees and then measuring the torque in the shaft. I hope to use this method to measure the torque transmission of a few different shafts which are relatively similar, but have varying length. Because the way the machine is set up can have a very strong effect on the measurement, I feel like some sort of MSA is necessary.This machine replaces an old method that was performed manually. What I am trying to establish with my MSA is 1) Is the variability due to operator acceptable for this particular measurement technique?; and 2) Is this method better than the old manual method (is there less variability due to operator/method) ?Thanks again Chad. Your help so far has been invaluable.-Phil

    0
    #160470

    Severino
    Participant

    Phil-
    While I do not have too much experience with rigid shafts, I have done quite a few similar studies on helically wound wire flexible shafts.  Before you even attempt a R & R you might want to ensure that you have explicit instructions for your operators to control the following parameters:
    A.)  Gripping force of chuck or collet
    B.)  Axial stretch (do you constrain the shaft at both ends or allow one to float free)
    C.)  The rate of twist/torque application
    D.)  Direction of twist
    E.)  Angle of twist to be achieved vs. shaft diameter (I assume you are not expecting the same angle for all shaft diameters unless you are in a very specific dimension range and using homogeneous materials)
    Also, I have seen calibrated standards which are able to apply torsional loads (they can also vary sinusoidally) which you might use to compare the results of your old test methods versus the new test method.  This might help you from having to rely solely on destructive testing for your R&R study.  I’m not sure if any of this information is helpful to you, but I figured I’d mention it just in case it was.

    0
    #160488

    Dixit
    Member

    Chad,
    Thanks for your message. Can you give me your email ID? I have a few other questions with is not related with this thread.
    Thanks

    0
    #160494

    Jim Shelor
    Participant

    Phil,
    According to Breyfogle, Implementing Six Sigma, in a destructive test it is impossible to separate Gage Variation from Part to Part Varation.
    Accordingly, I do not believe you can accomplish your goal using destructive testing.
    If I were you I would develop a non-destructive test for this part.  For example:

    Place a series of strain gages along the length of the shaft.
    Have the engineers calculate the strain gage readings corresponding to maximum specified torque for the shaft.
    Twist the shaft until the maximum specified torque is reached.
    Remove the twist force and measure the strain gages to determine that no plastic deformation has occurred.
    Use 10 shafts and 5 operators for the test.
    Reject any shaft that demonstrates plastic deformation.
    You now have a non-destructive crossed R&R and can obtain the data you seek.
    I hope this helps.
    Jim Shelor

    0
    #160499

    Markert
    Participant

    Jsev,I am indeed performing this test on helically wound wire shafts. The chuck force is controlled via torque wrench. An axial preload (tension) force is applied to the shafts. The rate and direction of twist are controlled by the software.Do you have any insight on gage R&R for this type of shaft?Thanks,
    -Phil

    0
    #160504

    Chad Taylor
    Participant

    Vijay
    [email protected]
    Chad Taylor

    0
    #160505

    Chad Taylor
    Participant

    Phil
    I have read your response from earlier and believe I have a simple solution to what you are trying to do.
    Since your 35 parts are from the same lot and you are really just trying to see if the new gage is better than the old. Randomally separate the lot into two different lots (you will have 3 extra). Then use two operators to run the test accross both gages. Each operator running 8 parts each on each gage. Compare the STD Deviations. Not rocket science but given the lack of availability of parts from different lots may be the only true test you have.
    By adding stress gages you are adding another variable which really doesn’t mean that much to the outcome your trying to achieve.
    just my 2 cents
    Chad Taylor

    0
    #160526

    Severino
    Participant

    Jim-
    Although I can’t speak for Phil, I can tell you that because he is testing flexible shafts and not a conventional shaft, as we might think of it, he won’t be able to torque without affecting the part permanately.  The layers of wire will move relative to each other as the shaft is exercised and so the shaft always have some hysterisis which will make repeatability for successive testing very low (hence why the test is considered destructive). 
    There are ways in which you might minimize that effect, mainly by testing a very large diameter shaft and applying a very small angle of twist, but I would still question that method. 
    In my experience the part to part variation for these items is very high due to winding parameters, cutting/welding parameters, etc.  My recommendation would be to develop some sort of  standard that approximates the manner in which you might test a flexible shaft (such as a solid shaft) that you can repeat the measurements on without altering the part.  This would allow you to determine how repeatable each device is without too much concern over part-to-part variation.  While this would not be exactly similar to testing a flexible shaft, I believe it would be close enough such that you could consider your R&R study valid for low deflections on high diameter flexible shafts.
    Not sure that helps, but that would be the direction I would be moving in.

    0
    #160546

    Jim Shelor
    Participant

    Jsev607,
     
    No matter how the shaft is constructed, if the purpose of the shaft is to transmit torque, then if I twist the shaft to the maximum torque specification and there is no plastic deformation of the shaft, I have proven that the shaft performs the function intended.
     
    Unless I do not understand this measuring device, the device is going to twist the shaft to a given torque that is high enough to make the shaft fail.  What is being measured after that torque is reached?  How is that telling me anything about the shaft, except that is does fail at the test torque?
     
    If the purpose of the shaft is to transmit a torque, testing the shaft to beyond the maximum torque specification seems to me to be a useless test.  If the shaft fails, what do you know about the shaft?  You know that it failed, but you know nothing useful, such as at what torque it failed.
     
    If the shaft is twisted to the maximum specified torque and it does not fail, now you know something about the shaft and you have a non-destructive test.  A non-destructive MSA will provide all the information Phil seeks.
     
    Respects,
     
    Jim Shelor

    0
    #160547

    Markert
    Participant

    The test fixes the rotation of the shaft at one end. At the other end, a motor rotates the shaft 270 degrees clockwise. A torque transducer then measures the torque across the shaft – the purpose of the test is to verify that a minimum amount of torque has been “transmitted.”Your suggestions have been very helpful thus far. Thanks all.-Phil

    0
    #160549

    Chad Taylor
    Participant

    Jim
    It is a destructive test because it cant be tested twice. Unless Phil is making transmission shafts from some new metal that a bunch of aliens gave him it will not “spring” back into its original form after the test is completed, whether it is twisted 1 degree or 270 degree.
    Chad Taylor

    0
    #160556

    Jim Shelor
    Participant

    Phil and Chad,
     
    If what I am hearing you say is true:
     

    The shaft is twisted 270 degrees and measured to confirm that a “minimum amount of torque has been transmitted”.
    “As little as 1 degree of flex constitutes a destructive test”.
     
    I am wondering how this shaft operates in the intended function and how much torque is required to be transmitted without deflection of the shaft.
     
    The answer to the minimum amount of torque being transmitted by the shaft, unless the shaft has physically failed (broken), in a static situation, i.e., no movement in progress, is the amount of torque being applied at the test end of the shaft.  How would you lose torque in a static system with torque being applied to a shaft?
     
    If indeed 1 degree of twist constitutes a destructive failure, why are we applying many, many times the torque that constitutes a destructive failure to the shaft?  What do we learn from doing that?
     
    There are many ways of testing this shaft in a non-destructive manner given the limits you have provided (i.e., no twist, minimum torque transmitted at many, many times the destructive torque point):
     

    Three times the normal operating torque without failure (i.e., no twist).
    Two times the maximum operating torque without failure (i.e., no twist).
    1.5 times the design torque without failure (i.e., no twist).
     
    Of course meeting those specifications would be nearly impossible because all shafts twist some amount when a torque is applied, the twist simply needs to be less than the onset of plastic deformation (usually designed at 1/10 the onset value).
     
    All of these are non-destructive tests that will provide you with the information you seek without having to make the very questionable assumption that all 35 parts from the same lot are uniform and have the same probability of failure.  In my opinion that assumption cannot be justified.
     
    Respects,
     
    Jim Shelor

    0
    #160559

    Chad Taylor
    Participant

    Jim
    The tolerance will be given that the shaft must produce “X” min. amount of torque when twisted to 270 degree rotation. 270 degree is a constant for the gage/part specifications. So when you test the part it is by all accounts destroyed. Most likely the part has already reached a maximum Torque and is either in “Flatline” or beginning to trend downward by this degree of rotation.
    I do understand your method, however tolerances on prints and thus gaging methods are not called out that way.

    0
    #160561

    Jim Shelor
    Participant

    Chad,
    Okay.  There must be something I do not understand about the physics of the shaft.
    There is definitely something I do not understand about the information gained by the test as it  is described.
    So be it.  The issue isn’t why the test is done that way, but how to statistically handle the results.
    Have a good day.
    Jim Shelor

    0
    #160566

    Severino
    Participant

    For this test you are either fixing the load and measuring the deflection or you are fixing the deflection and measuring the load.  I am used to doing the former, but from what you describe it sounds like the latter. 
    To clarify so that others understand, because the construction of this shaft is composed of layers of wire wound in alternating directions (clockwise-counterclockwise-clockwise-counterclockwise etc.) it does not behave like a normal solid shaft.  When you twist the shaft in one direction the outer layer will tend to unwind, picture twisting a spring opposite its wind and how the coils will actually become larger, and if you twist it in the other direction the outer layer will tend to tighten up.  It is precisely because of this phenomena that you will never get the same result from the same shaft and the effect happens at loads much lower than the elastic limit of the wires themselves.  Each layer moves relative to each other such that when you unload the shaft the deflection of the driven end versus the static end will never return to zero.  The effect of this phenomena will always be larger than the measurement system error unless you have chosen some very poor equipment or are very inconsistent in how you perform your testing.
    For the purposes of the Gage R & R though, you really don’t care about the shaft itself.  You want to know the error associated with the test method.  This is why I suggested that you use a much smaller deflection (and therefore a much smaller load) on a solid shaft (as opposed to a flexible shaft) such that the plastic limit is not exceeded.  Theoretically, under these conditions I believe that Jim is correct and a solid shaft would return to its free state when the load is removed or at least close enough that any resultant errors would be negligible.  You could then repeat this experiment with different operators and different equipment to get your R&R.
    What are the drawbacks?  Well there are quite a few.  For one thing, you are probably going to have to work with deflections that are orders of magnitude smaller than what you might achieve with a flexible shaft.  Therefore there may be some hidden variables that would not have a pronounced effect on the low deflection/low load that might occur in a high deflection/high load situation.  Also, there is the phenomena of axial load that results from twisting the shaft.  This can effect friction forces in your bearings, create thrust loads on your transducers, etc.  Also at high loads, the helically wound shaft tends to act like a screw and begins to unthread from the chucks if not tightened properly and result in a slip.  All of these might result in variation in measurements that would not be taken into account when using a solid shaft. 
    That being the case I still thing that a solid shaft will get you far closer to a useable result.  Otherwise, you are assuming that other parts in the same lot are homogeneous and therefore you can compare the results of testing one sample with one operator/machine combo versus the results of a completely different shaft using another operator/machine combo.  My experiences tells me that part to part variation is HUGE and therefore you will not get a meaningful result using this method because you will not be able to isolate what error is due to the parts and what error is due to the measurement system. 

    0
    #160569

    Jim Shelor
    Participant

    Jsev607,
    Nice, informative post.
    I was proposing fixing the load and measuring the deflection.
    Respects
    Jim Shelor

    0
    #160765

    Markert
    Participant

    Jsev,Thank you for the very informative post. One thing I should mention about the shafts I am testing is that they are coated in a rubbery laminate. This makes it difficult for me to find a non-wound shaft with similar properties that would be suitable for this test.In the end, I might just have to build some non-wound shafts and apply the rubbery laminate to those.-Phil

    0
    #160842

    howe
    Participant

    If quoting Breyfogle, please quote completely. In Implementing Six Sigma 2nd edition p326 he states;
    “When testing is destructive, it is impossible to separate the variation of the measurements themselves from the variation of the product being measured. However, it is often posssible to minimize the product variation between pairs of measurements through careful choice of the material that is to be measured.”
    Gage R&R’s are rountinely done on destructive tests.
     

    0
    #160866

    High School Teacher
    Participant

    Mike, we all never know if you missed the right job opportunity for you in life: High School Teacher. Instead, you have that sour-grape attitude about you when sharing your 2-binder knowledge of six sigma with the accompanying cliff note of Breyfogle III (at a minimum quote Breyfogle’s name correctly in the future when admonishing your fellow green belts. His name is: F.W. Breyfogle III. We collectively  appreciate the efforts that go into your imagined homework assignments. Even though, overall, and so far, D- (for effort).

    0
Viewing 28 posts - 1 through 28 (of 28 total)

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