Translating Machine Measurement Variation

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Translating Machine Measurement Variation

    Hey guys,

    I have a unique question. I am working with Cylindrical parts that have spherical centers. I am being told by engineering that there is a deflection variation in the probe when I measure the sphere in the middle of the part, thus giving me a variation in the result of the measurement. What they are saying is that because the ball does not contact the part on a 90° angle, DMIS is reading a variation from what it would on say, a flat part. I thought that DMIS was supposed to be able to take up any deflection error because it measures to the center of the ball probe, not to the edge, and because of this, it is not supposed to matter where on the ball probe the hit is taken, it should read the same.

    First, are they correct, and second, if the are correct and there is a diametrical error of .150 mm, how would I translate that in the program to automatically take up the .150mm slack?

  • #2
    PC-Dmis should be able to compensate but there might be some factors to consider.

    What probe system are you using?

    Are the sphere's internal or external, and how are you measuring the spheres (learnt hits or auto features, or constructed from points?)

    Also how much of the sphere (in °) can you access?


    To prove that it's doing its job correctly, do a program measuring the ref sphere.

    (Measure the full sphere with plenty of hits, then origin on it. Then measure it again using the same degree of arc available on your parts and see how the size and location compare)
    Last edited by NinjaBadger; 09-23-2014, 07:54 AM.
    Automettech - Automated Metrology Technology

    Comment


    • #3
      +1 !
      You can also construct a BFRE sphere.. It should be better if you don't use auto sphere.

      Comment


      • #4
        The engineers may be talking about Cosine error, if you are measuring points and you're not taking your hits normal to the surface.

        Comment


        • #5
          Even with probe compensation, snap, acceptable probe qualification, etc, you can still get sine/cosine error. As soon as your surface deviates in size or location, you are no longer hitting exactly where expected. It is an unfortunate inevitability of the measurement system. It should be quite small, at least nowhere near the magnitude of 0.15mm (of course, this varies based on ball size, size of measured feature, deviation of feature size/location, alignment inaccuracies, etc). Do you have a way to prove out your results manually, ie 'the old fashioned way'?
          "This is my word... and as such is beyond contestation."

          Comment


          • #6
            There is some truth in what the engineers are saying, but not what they think because they obviously don't know how the CMM takes hits.

            Originally posted by grandpaal56 View Post
            The engineers may be talking about Cosine error, if you are measuring points and you're not taking your hits normal to the surface.
            Originally posted by VinniUSMC View Post
            Even with probe compensation, snap, acceptable probe qualification, etc, you can still get sine/cosine error. As soon as your surface deviates in size or location, you are no longer hitting exactly where expected. It is an unfortunate inevitability of the measurement system. It should be quite small, at least nowhere near the magnitude of 0.15mm (of course, this varies based on ball size, size of measured feature, deviation of feature size/location, alignment inaccuracies, etc). Do you have a way to prove out your results manually, ie 'the old fashioned way'?

            This and that.

            In a perfect world with perfect parts and a perfect program, your probe would hit exactly normal to the surface regardless where the hit was taken (if I'm understaning you right).

            The short version, is the cmm figures out the approach, takes the hit, and compensates behind the scenes to deliver that 90 deg. contact to the answer we see on the screen... which is pretty much what you said above.

            But since parts are rarely perfect, etc etc there can be cosine error but not in the manner the engineers are talking about. So it is a good idea to back-check your results with other methods. There are also tricks and tools you can use to minimize this effect (alignments, remeasures, etc). GIGO applies here.

            The question eventually arises, "how close is close enough?"

            I'm guessing .15mm isn't close enough for the engineers



            TLDR: You're right, they're wrong but your measurement might need tweaking.
            Last edited by TheyCallMeWoody; 09-23-2014, 02:38 PM.

            Comment


            • #7
              ALSO -

              I'm guessing you have a qualification sphere - Have you tried your measurement technique on it as a baseline?

              I know it's an OD and sounds like you are measuring ID and you will probably need to adjust things.

              If your measurement technique doesn't deliver the exact diameter of your artifact (or within a couple microns), then you know your approach needs work, OR something else needs to be tracked down.

              If it does, then your technique *probably* isn't *blatantly* wrong

              Comment


              • #8
                another thing to keep in mind... watch and see if you are shanking when checking an internal sphere
                sigpic
                Originally posted by Ironhoe
                I got something under my sporran for you, take care of it and you got my vote.

                Comment


                • #9
                  Pcdmis handles spheres a little differently (unless they have changed it in newer versions).

                  NO probe comp is done on a sphere until ALL hits have been taken. None. Pcdmis makes the sphere from center-of-ball data, THEN comps (in or out) for the probe. You can measure a 10" sphere, with 2" prehit/retract, and if it is off 1" "SIDEWAYS" to the shank of the probe, it will still be good data. THIS is why there is no "RE-MEAS" option for a sphere (unless it also has been added in the newer versions of the software).

                  To 'eliminate' even this slight possibility, measure the sphere, the origin to it, then re-measure it, then recall the correct alignment. No problem.
                  sigpic
                  Originally posted by AndersI
                  I've got one from September 2006 (bug ticket) which has finally been fixed in 2013.

                  Comment


                  • #10
                    150 microns!!! That is a large number for either or both deflection and uncertainty.

                    You could dive off into the fantabulous world of task specific measurement uncertainty. You will need at least a couple of weeks to study up and digest the relevant standards and technical guides. ASME, NIST, BIPM, & NPL all have good info. Some is free and some costs $$. Once you have the knowledge you can set up data generating runs, crunch numbers and ultimately be able say with something like "the uncertainty is ______ with 95% confidence at an expanded (k=2) interval."
                    sigpic"Hated by Many, Loved by Few" _ A.B. - Stone brewery

                    Comment


                    • #11
                      Originally posted by Matthew D. Hoedeman View Post
                      Pcdmis handles spheres a little differently (unless they have changed it in newer versions).

                      NO probe comp is done on a sphere until ALL hits have been taken. None. Pcdmis makes the sphere from center-of-ball data, THEN comps (in or out) for the probe. You can measure a 10" sphere, with 2" prehit/retract, and if it is off 1" "SIDEWAYS" to the shank of the probe, it will still be good data. THIS is why there is no "RE-MEAS" option for a sphere (unless it also has been added in the newer versions of the software).

                      To 'eliminate' even this slight possibility, measure the sphere, the origin to it, then re-measure it, then recall the correct alignment. No problem.
                      I am not surprised to learn that this is the case. It makes sense conceptually. I am skeptical that your hypothetical sphere would still be "good", but it wouldn't surprise me to find that it was orders of magnitude more accurate than a circle of the same hypothetical dimensions (without find center).
                      "This is my word... and as such is beyond contestation."

                      Comment


                      • #12
                        Originally posted by VinniUSMC View Post
                        I am not surprised to learn that this is the case. It makes sense conceptually. I am skeptical that your hypothetical sphere would still be "good", but it wouldn't surprise me to find that it was orders of magnitude more accurate than a circle of the same hypothetical dimensions (without find center).
                        I've checked it. I've measured a sphere (manually) with all hits using a 0,0,1 vector (top hit & way out to the sides, but all moving up/down), and there wasn't even as much 'error' between that and a DCC sphere measured 'correctly' as the machine specs. AS LONG AS there are not shank hits.....
                        sigpic
                        Originally posted by AndersI
                        I've got one from September 2006 (bug ticket) which has finally been fixed in 2013.

                        Comment


                        • #13
                          Is this just with auto features we're talking about here?

                          I know this type of retrospective probe comp is how Measured features are calculated anyway.
                          Automettech - Automated Metrology Technology

                          Comment


                          • #14
                            For the shank problem, you can use the script "probing help" in the thread below. It works with excel, it's an "auto run" script, so you could have to autorize its execution...

                            http://www.pcdmisforum.com/showthrea...-damaged-probe

                            Comment


                            • #15
                              Originally posted by NinjaBadger View Post
                              Is this just with auto features we're talking about here?

                              I know this type of retrospective probe comp is how Measured features are calculated anyway.
                              I'm not exactly sure of all the auto-features, BUT, any that have the re-meas option are 'suspect' to me since it appears that they use 'theo' vectors for comp, otherwise, would there be a need to re-meas them? For example, a circle, the comp will be done in the plane of the feature (3 sample hits will 'update' the vector for comp and use that for that measurement of the hole), but I know that by using re-meas you can get big differences in the results from the re-meas for a circle. That makes me think that each hit is comped (in the plane, either nominal or sample hit plane) per the theo vectors of each hit, not as center-of-ball, THEN comped, so if it is off location a bunch, but not enough to not run, the results without re-meas can be 'bad'. Again, unless they have changed this in newer versions.....
                              sigpic
                              Originally posted by AndersI
                              I've got one from September 2006 (bug ticket) which has finally been fixed in 2013.

                              Comment

                              Related Topics

                              Collapse

                              Working...
                              X